The Netherlands

1st International Conference on Ageing of Materials & Structures

26-28 May, 2014

Delft The Netherlands

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands

Proceedings of the 1st International Conference on Ageing of Materials & Structures Delft University of Technology www.ams.tudelft.nl [email protected]

Editors: K. van Breugel, E.A.B. Koenders Cover: I.B. Design


Sponsors

TU Delft

www.tudelft.nl

DCMat

www.dcmat.tudelft.nl

RILEM

www.rilem.org

Ageing Center TU Delft


Table of contents Welcome address KEYNOTES

Page

Radiation Shielding Properties and Freeze-Thaw Durability of High-Density Concrete for Storage of Radioactive Contaminated Soil in Fukushima J. Pareek, Y. Suzuki, Ken-ichi Kimura,Y. Fujikura, Y. Araki

1

Stochastic Models for Risk and Failure under Ageing J. Hüsler

9

The Relevance of Aging for Civil Infrastructure: The Profession, The Politics, The Classroom D.A. Lange

17

Biomimetic materials: long-lasting and self-repairing T. Speck, M. Thielen, O. Speck

27

Forever young or ageing naturally? R.P.J. van Hees, S. Naldini

31

UNDERSTANDING AGEING

39

Development of a Database for the Restoration Mortars – von Konow DB E. Sistonen, P. Mutanen, F. Al-Neshawy

40

Damage assessment of early 20th century stone imitating mortars Y. Govaerts, A. Verdonck, W. Meulebroeck, M. de Bouw

48

Future-oriented building stock studies and the significance of values: addressing demolition behaviour with the Delphi method S. Huuhka

56

The Noble Patina of Age W.J. Quist, A.J. van Bommel

64

FUNDAMANTALS OF AGEING OF MATERIALS

72

Examining of weather resistance of ETICS with stresses which correspond to weather


conditions in Finland P.J. Annila, T.A. Pakkala, J. Suonketo

73

Ageing behaviour of neutron irradiated Eurofer97 I. Carvalho, A. Fedorov, M. Kolluri, N. Luzginova, H. Schut, J. Sietsma

81

Approach for the Investigation of Long-Term behaviour of elastomeric Seals for Transport und Storage Packages M. Jaunich, D. Wolff

87

Atomic density function modelling of atomic structure of grain boundaries O. Kapikranian, H. Zapolsky, C. Domain, R. Patte, C. Pareige, B. Radiguet,

94

P. Pareige Salt crystallization damage: how realistic are existing ageing tests? B. Lubelli, R.P.J. van Hees, T.G. Nijland

103

Postponing Crack Nucleation in Age Hardenable Aluminium Alloys M. Mahdavi Shahri, R. Alderliesten, S. van der Zwaag, H. Schut

112

Accelerated ageing protocols for (polymer modified) PA to obtain representative (rheological) properties, mimicking field aged materials S.D. Mookhoek, G. Liu, S.M. J.G. Erkens, C. Giezen, J.L.M. Voskuilen

118

Evaluation of test results with regard to ageing mechanisms of metal seals in casks for dry storage of spent nuclear fuel S. Nagelschmidt, U. Herbrich, U. Probst, D. Wolff

126

New test method for wind-driven rain penetration of ETICS J. Pikkuvirta, P.J. Annila, J. Suonketo

134

Chloride ingress in cracked concrete studied using Laser Induced Breakdown Spectroscopy B. Šavija, J. Pacheco, E. Schlangen, S. Millar, T. Eichler, G. Wilsch

140

PZT Patch Ageing Assessment Through Accelerated Testing V. Spitas, C. Spitas, P. Michelis

147


Modeling and analyzing autogenous shrinkage of hardening cement paste T. Lu, E.A.B. Koenders

155

Engineering ageing and end-of-life models F. Tolman

163

Ageing coefficient of fly ash concrete and its impact on durability G.J.L. Van Der Wegen, M.M.R. Boutz, A. J. Sarabèr, R.J. Van Eijk

171

Changing Model Properties with Time due to Corrosion of a Dynamically Loaded Reinforced Concrete Beam R.P. Veerman, E.A.B. Koenders

179

The influences of CSH on the carbonation resistance of cement blended with supplementary cementitious materials B. Wu, Y. Zhang, G. Ye

187

Ageing of Portland Cement Concrete Cured under Moist Conditions Z. Yu, G. Ye, K. van Breugel, W. Chen

195

Effect of Reactive Sandstone Powder on Suppressing ASR L. Yang, H. Zhen

203

Pore structure of blended cement paste by means of pressurization–depressurization cycling mercury intrusion porosimetry Y. Zhang, B. Wu, J. Zhou, G. Ye，Z. Shui

212

Using bio-based polymers for curing cement-based materials J. Zlopasa, E.A.B. Koenders, S.J. Picken

220

AGEING OF PRODUCTS AND STRUCTURES

227

Assessment of ageing in the population of power system components by means of statistical tools L.A. Chmura, P.H.F. Morshuis, J.J. Smit, A.L.J Jannsen

228


Energy Dissipation in Flexural for Two Stage Concrete H.S. Abdelgader, A.S. El-Baden

234

Aging Management for Extended Storage and Transportation Y.Y. Liu, M.C. Billone, D.R. Diercks, O.K. Chopra

242

Seismic Evaluation of Coastal RC Building Vulnerable to an Airborne Chloride Environment A.M.Y. Mohammed, A.Ahmed, Koichi Meakawa

250

Effect of Ageing and Cyclic Loading on the Strength of Gravity Dams A. Motsonelidze, V. Dvalishvili

256

Delayed deformations of segmental prestressed concrete bridges: the case J-P. Sellin, J-F. Barthélémy, J-M. Torrenti, G. Bondonet

266

Structural Longevity of FPSO Hulls - Extended Abstract M. Tammer, M.L. Kaminski

274

On the very long term delayed behaviour of concrete J.M. Torrenti, F. Benboudjema, F. Barré, E. Gallitre

281

Modeling Aging of Cementitious Pore Structure N. Ukrainczyk, E.A.B. Koenders

288

Protection of aged concrete structures: application of bio-based impregnation system V. Wiktor, H.M. Jonkers

295

Estimation of Carbonation and Service Life of Box Culvert for Power Transmission Line S-K Woo, Y Lee, Y-D Choi

302

Self healing of structural defects by Au precipitation in creep steels S. Zhang, H. Schut, E. Brück, S. van der Zwaag, N.H. van Dijk

308

MODELLING AGEING

312


Ageing mechanisms of building structures and their management F.Al-Neshawy, E. Sistonen

313

Prediction of Long term Deformation of Cementitious Materials A. Brahma

321

Bayesian modeling of bivariate extreme shocks P. Cirillo

330

Residual Fatigue Life Evaluation of Rail at Squats seeds using 3D Explicit Finite Element Anal X. Deng, M. Naeimi, Z. Li, Z. Qian, R. Dollevoet

337

Ageing effects of alkali-silica reaction in concrete structures R. Esposito, M.A.N. Hendriks

347

A Poro-mechanical Approach for Assessing the Structural Impacts of Corrosion in Reinforced Concrete Members E. Gebreyouhannes, T. Yuya, K. Maekawa

354

The Influence of Drying Shrinkage on the Fatigue Life of RC Slabs Y. Hiratsuka, K. Maekawa

362

Statistical models for interval censored time-to-event data G. Jongbloed

370

Long-Term Serviceability and Risk Assessment of Shallow Underground RC Culverts and Tunnels M. Kunieda, X. Zhu, Y. Nakajima, S. Tanabe, K. Maekawa

376

A level set model for computational modeling of fatigue-driven delamination in laminated composites M.Latifi, F.P. van der Meer, L.J. Sluys Introduction of structural ageing-specific functions for computational models based on synaptic networks

384


G. Mavrikas, V. Spitas, C. Spitas

392

Modeling ion transport and Alkali-Silica-reaction (ASR) induced damage in concrete based on micromechanics and phase field models M.N. Nguyen, J.J. Timothy, G. Meschke

400

Probabilistic corrosion forecasting of steel in concrete with potentialdependent chloride threshold A.N. Sánchez, A.A. Sagüés

406

Modelling of Environmental Action for Simulation of Long Term Variation of Moisture Content in Concrete Structures T. Shimomura, K. Onoya, H.T. Thynn

416

Chemo-Hygral Modeling of Structural Concrete damaged by Alkali Silica Reaction Y. Takahashi, K. Shibata, K. Maekawa

424

The influence of cement fineness on durability of cementitious materials Y. Zhang

432

CHARACTERIZATION AND MONITORING

439

Chloride Profiles in Concrete After 100 Year of Service in Panama Canal C. Andrade, N. Rebolledo, A. Castillo, R. Perez, M. Baz

440

The impact of UV radiation, carbonation, and hydration on hydrophobic treatments in concrete U. Antons, O. Weichold, M. Raupach

448

Research on Calculation Methods of Service Life Predication of Chinese Modern Reinforced Concrete Buildings Q. Chun, K. van Balen, W. Lv, J. Pan

454

Laser Induced Breakdown Spectroscopy (LIBS) as a tool for the investigation of aging of infrastructure T. Eichler, G. Wilsch, S. Millar, D. Schaurich

461


Residual Strength of Aging Concrete T-Girders Strengthened with Near-SurfaceMounted Composite System T. El-Maaddawy, A. Bouchair, A. Biddah, A. El-Dieb

469

Optical pH Imaging in Cementitious Materials E. Liu, M. Ghandehari, W. Jin, C. Bruckner, G. Khalil

477

Effects of accelerated ageing on the adhesive bond between concrete specimens and external GFRP reinforcements D. Ghosh, S. Saha, S. Karmakar

486

Assessment of Mechanical and Chemical Deterioration of Artworks R.M. Groves, C. Portalés, E. Ribes-Gómez

493

Improving the durability of prestressed concrete structures by employing high strength duplex stainless steels H. Mahmoud, M. Sánchez, M. C. Alonso, L. Bertolini

501

Internal Hydration Modulus and Related Viscoelastic Stresses in Cementitious Materials W. Hansen, Z. Liu, E.A.B. Koenders

509

Experimental study on the Corrosion Fatigue Performance of RC Beams Strengthened by Pre-stressed B/C HFRP Sheets J. Pan, Q. Chun, X. Zhou

516

Investigation of time-dependent deformation of RC beam with flexural crack generated in early age considering shrinkage property of concrete S. Komatsu, A. Hosoda

524

The Influence of Fiber-Matrix Adhesion on the Linear Viscoelastic Creep Behavior of CF/PPS Composites M.H. Motta Dias, K.M.B. Jansen, H. Luinge, K. Nayak, H.E.N Bersee The importance of chloride sensors stability in monitoring ageing phenomena in concrete structures: Ag/AgCl electrodes performance in simulated pore-water environment

532


F. Pargar, D.A. Koleva, E.A.B. Koenders, K. van Breugel

542

Analysis and Simulation of the Performance of a Sensing Array for Gear Tooth Cracks C. Spitas, V. Spitas, A. Nikolakakis

548

Fundamental research on the carbonation control effect by coating materials M. Sugiyama

556

DC current-induced curing and ageing phenomena in cement-based materials A. Susanto, D.A. Koleva, K. van Breugel

562

Novel Kelvin Probe electrode for non-intrusive corrosion rate evaluation of steel in aged concrete structures M.T. Walsh, A.A. Sagüés

569

Corrosion monitoring of a reinforcement steel using galvanostatically induced potential transient Y. Abbas, J. S. Nutma, W. Olthuis, A van den Berg

576

Salt Frost Scaling of High Strength Concrete Z. Liu, W. Hansen, B. Meng

584

SOLUTIONS AND DESIGN CONCEPTS

592

The Ageing Of The Creation L. Biçaçi, E. Schlangen, O. Copuroğlu, J.A. Poulis

593

Study on effective modifiers for damaging salts in mortar S. J.C. Granneman, E. Ruiz-Agudo, B. Lubelli, R.P.J van Hees, C. Rodriguez-Navarro

604

Densification and self-healing performance of mortar with clinker fine aggregate A. Hosoda, Y. Watanabe, T. Higuchi, M. Morioka

612

Study on Triage for Deteriorated Concrete Structures by JSCE-342 S. Miyazato, T. Yamamoto, J. Tomiyama, R. Takahashi, K. Watanabe Effect of different amounts of innovative self-healing additions on the microstructure of

620


cement pastes G. Perez, E. Erkizia, J.J. Gaitero, I. Jimenez, A. Guerrero

628

The effect of magnesia on the self-healing performance of Portland cement with increased curing time T. Qureshi, A. Al-Tabbaa

635

Colloidal nanosilica healing ability for reinforced concrete repair M. Sánchez, M.C. Alonso, I. Díaz, R. González

643

Bacteria-based self-healing concrete to increase liquid tightness of cracks E. Tziviloglou, H.M. Jonkers, E. Schlangen

650


WELCOME ADDRESS


Urgency and challenges of Ageing in Science and Engineering Klaas van Breugel Delft University of Technology, Delft, The Netherlands Abstract: Ageing is one of oldest phenomena observed in the real world. It is an inherent feature of nature. Yet it seems to be new topic in science and engineering. The main reason for this is our growing awareness that in industrialized countries ageing of their assets is a financial burden for the society and affect the overall sustainability of our planet. In this contribution the urgency and challenges for ageing are addressed, including costs aspects and research needs. Emphasis will be on ageing of infrastructure and plants and on justification of investments for research on ageing. Keywords: Ageing, Costs, Infrastructure, Risk, Sustainability

1 Introduction Ageing is everywhere. Huge mountains seem to keep their shape for ever. But, at a closer look, we see that the surfaces of rocks gradually change. Snow, rain, frost, light, wear, wind and sunshine are sufficiently powerful to crumble even the strongest rock. Mountains age! Tectonic action may fracture mountains, causing rigorous changes in the state of stress in the newly formed parts of the mountain. The fracture surfaces become exposed to climatic conditions and another cycle of ageing starts. Like rocks, also man-made infrastructure works are exposed to climate conditions. While exposed to the climate, structures have to carry life loads and deadweight in a safe way. Even the strongest structures exhibit decay of quality and function with elapse time. Roads and railways need continuous maintenance. If planned correctly the trouble maintenance will cause can be kept to a minimum. If too late, maintenance and repair will cause time and money consuming traffic jams, delays or even accidents. A proper functioning road network is vital for our economy. Roads are used by cars. The lifetime of cars has dramatically increased in past decades, but still a 20 year old car is an exception rather than a rule. Electronic systems and sensors may signal decay of functions, but cannot stop the car’s ageing. The electronic systems and sensor are subject to ageing themselves as well. Cars need fuel, either fossil fuel or electricity. Different types of petrol are produced in huge chemical plants and electricity for hybrid cars is produced in power plants. Fuel and electricity is transported to distribution points. For transport of energy carriers cars, pipelines and electricity grids are used. All these mobile and fixed assets are ageing. Even those with the highest quality sooner or later exhibit ageing and have to be replaced. If not replaced in time, catastrophic accidents are imminent. Power plants for generating electricity and energy transport grids have to function reliably for 24 hours day, the whole year round. Failing components constitute a risk for life and limb, as well as costly process interruption. The indirect costs of failure are often 5 to 10 times higher than the direct costs. Pro-active replacement of vital components of systems and structures is considered a safe strategy to prevent catastrophic failures. But do we really know how close we were to a catastrophic failure at the moment these components were replaced? Were we really at risk or did we spoil a lot of still perfectly operating components without improving safety substantially? Ageing is everywhere and unavoidable. But if the presence of ageing is a matter of fact, there are still questions as to whether the consequences of ageing are unavoidable as well and whether the rate of ageing is unchangeable for ever. For an answer on these questions we have to know what ageing really is. But let us first have a closer look at three different fields of interest where ageing is a hot topic today, i.e. infrastructure, chemical plants and power plants.


2 Ageing and society 2.1

Ageing infrastructure

Global infrastructure stock - In modern industrial countries the infrastructure makes out over 50% of the nation’s national wealth [1]. This infrastructure consists of roads and railway systems, water works, airports, power stations and electricity grids. Based on an inventory made in twelve countries, the value of infrastructure stock averages around 70% of the global gross domestic product (GDP). For a global GDP of € 53 trillion euros in 2012, this makes € 37 trillion euros. This infrastructure is vital for the mobility of people and for a country’s economy. Economic growth is inconceivable without growth of a country’s infrastructure, i.e. roads and railway systems, water works and electricity grids. To catch up with global economic growth McKinsey [2] estimates a required investment of € 42 trillion euros between 2013 and 2030. This means an annual investment of € 2.3 trillion euros, which is about 4.5% of the global GDP (GDP in 2012: € 53.4 trillion euros). The investment of € 42 trillion euros is needed for roads and railways, ports, airports, power stations, water works and telecommunication. Table 1 gives the breakdown of investments over these categories. These figures are (in part) based on the extrapolation from the data provided by 84 countries. These countries are responsible for 90% of the global GDP and are considered today’s best possible basis for estimating the investments needed for our infrastructure. Table 1 Estimated needs for global infrastructure in different categories in the period 2013-2030 [2] Catagory

1

Roads Rail Ports Airports Power Water Telecommunications Total

Source

OECD1) OECD OECD OECD IEA2) GWI3) OECD

) Organisation for Economic Co-operation and Development ) International Energy Agency 3 ) Global Water Intelligence

Required investment [ € 1,000,000,000,000] 12.2 3.3 0.5 1.4 8.8 8.4 6.8 41.4

2

Dutch infrastructure stock – In 2012 the GDP of The Netherlands was € 554 billion euros. If the aforementioned 70% rule of thumb also applies to The Netherlands, the value of its infrastructure stock would be € 388 billion euros. This figures matches quite well with the value of the Dutch infrastructure of € 312 billion euros provided by the Dutch Central Agency for Statistics (CBS) for the year 2009 (in [3]). The total national wealth of The Netherlands in 2009 has been estimated at € 3.8 trillion euros. Almost half of it consists of fixed assets, i.e. infrastructure, houses, industrial buildings and durable capital goods. From the data in Table 2 it can be inferred that Long’s statement that the infrastructure makes out about 50% of a nation’s national wealth indeed holds if the term infrastructure is used for all fixed capital goods. This is reasonable, since the total set of fixed capital goods provides the basic conditions for a society to function. Table 2 Value of fixed capital goods of The Netherlands, 2009. Total national wealth € 3.8 trillion [3] Fixed capital goods Infrastructure Houses Industrial buildings Permanent capital goods Total

Value [x € 100,000,000] 312 975 382 156 1,825

Percentage of national wealth 8% 25% 10% 4% 47%


From the figures in Table 2 we can also infer figures for the required future investments in the infrastructure in The Netherlands. For that we assume that the correlation between the global GDP and estimated investments for infrastructure mentioned in [2] also holds for The Netherlands. In 2012 the national GDP of The Netherlands was € 554 billion euros, which is about 1.06% of the global GDP of € 53.4 trillion euros. If the same percentage of 1.06% also applies to the estimated required global investment in infrastructure of € 42 trillion euros between 2013 and 2030, the Dutch investment in infrastructure in that period would be € 435 billion euros. This makes € 24 billion euros per year. This is the estimated amount of money to be spent on new built in the infrastructure to catch up with expected economic growth in The Netherlands. The yearly turnover of the Dutch building industry is about 10% of the GDP, roughly € 60 billion euros per year. About 50% of this amount is spent on new built and 50% on maintenance, repair and rehabilitation of existing structures, i.e. € 30 billion euros per year on new built and € 30 billion euros per year on existing structures. This figure matches well with the € 24 billion euros investment for new built derived above from the McKinsey report.

2.2

Ageing of chemical plants

Many chemical plants, particularly the larger ones, have got their present size through a process of gradual expansion. Subsequent parts have been built according to different codes, with different materials and designed while considering different technologies. This process of gradual growth results in a high heterogeneity of large plants. Whereas the typical life cycle of a plant has been estimated at 25 years, the ‘effective’ age of large plants varies a lot with respect to both its real age as well as its functional age. Because of this heterogeneity it is not easy the inspect and judge these plants with respect to their state of ageing. This holds for both onshore and offshore plants. The HSE Research Report RR 823 [4] “Plant ageing Study” gives an overview of ageing issues in chemical plants. It covers definitions of ageing and, more importantly, reveals the impact of ageing on plant safety. Based on three principal databases of incidents reports, i.e. RIDDOR, MARS and MHIDAS, the significance of ageing could be determined. The MARS study showed that approximately 60% of incidents were related to technical integrity and, of those, 50% have ageing as a contributory factor. From that it was concluded that ageing is a significant issue. The percentage of incidents with ageing as a contributing factor was considered likely to increase with time as assets age. Ageing may affect the installations, piping and containments as well as the Electrical, Control and Instrumentation (EC&I) equipment. Figure 1 shows the results of the MARS incidents with ageing as the cause of failure.

Figure 1 Proportion of incidents on chemical plants with ageing as the cause (Mars study, in [4]) In the UK 173 loss of containment incidents in the period from 1996-2008 have been attributed to ageing. This is 5.5% of all loss of containment events. Across Europe 96 incidents occurred due to ageing (MARS database, see [4]). These incidents represented 28% of all reported ‘major accident’


loss of containment events. In those accidents 11 people lost their lives and 183 were injured. The total direct financial losses amounted to € 170 million.

2.3

Nuclear power plants

More than in any other industrial sector ageing is a key-issue in the nuclear industry. In the 1990s the International Atomic Energy Agency (IAEA) began developing a comprehensive set of publications on ageing management. The urgency is given with the fact that most existing nuclear power plants (NPPs) were built 20-30 years ago. These plants are now old enough to discover new ageing phenomena that slowly developed during past decades. Moreover, new ageing phenomena are emerging as a result of more severe service conditions associated with increased plant performance, e.g. through implementation of the long-term operating experience obtained and/or the application of new technologies. Proactive ageing management - Because of the high consequences that may result from failure of components of a nuclear power plant emphasis is, and should be, on proactive management of ageing. This in contrast with a ‘run to failure’ strategy, where components are replaced once they fail. The primary aim of proactive ageing management is to help to ensure the availability or required safety functions throughout the service life of the NPP. Moreover, the effective management of ageing is also essential for achieving the desired plant performance and profitability of the plant. The aim of physical or materials ageing management of structures, systems and components (SSC’s), important for the safety of a plant, is to maintain their design safety margins above the SSC specific requirements, thus minimizing the risks to people and the environment. Based on experience it is known that many SSC-failures are the result of ageing mechanisms, such as general and local corrosion, erosion-corrosion, radiation and thermal embrittlement, fatigue, creep, vibration and wear. To ensure a high level of plant safety it is recommended to manage SSC-ageing effectively and proactively. This proactive ageing management strategy includes both physical ageing and nonphysical ageing that results from obsolescence, i.e. the plant’s being out of date with respect to current safety regulation, standards, practices and technology. Predictive models - The claim of being able to act proactively only holds if we know how far we stay away from the moment a material, structural component or system will fail. It presupposes that we have predictive models for the relevant ageing mechanisms and that we are able to monitor the progress of degradation with time reliably. According to the IAEA report Nr. 62 of 2009 [5], radiation embrittlement that leads to changes in bulk material properties has been successfully modelled. The predictability of embrittlement of stainless steels, and of irradiation and thermal degradation of polymers, used in cable insulation and seals and which also produce changes in bulk material properties, is also considered adequate. It is well known, however, that the Fukushima disaster has generated many new research questions and research projects, not only in Japan but also in Europe. The research focuses on durable solutions for nuclear waste through immobilization and containment. IAEA report 62 further states that the predictability of corrosion, wear and high cycle fatigue, which produce changes at material surfaces and interfaces, is generally low. The resulting uncertainty has caused significant nuclear power plant unavailability and increased costs for operation and maintenance. Important to mention is that operating experience in the nuclear industry has revealed degradation and failures of structures, systems and components caused by previously unrecognized ageing mechanisms. Given the potentially large consequences of accidents with NPP, this is an important observation. The IAEA, therefore, states that in addition to improving the understanding and predictability of known ageing mechanisms, there is a need to provide for early detection of new ageing mechanisms. This requires sensitive and reliable monitoring and control devices and thorough understanding of the running processes in a plant.


3 Ageing and science 3.1

Driving forces behind ageing

By definition ageing is a change of performance of a material, structure, system or plant with elapse of time. How time per se can result in a change of performance is not easy to understand at first sight. How can a material ‘at rest’ change its performance? It is better saying that time is the domain in which we describe observed changes in performance rather than the cause of these changes. The remaining question is then: what causes the changes of a material, structure or system that is, on the macroscale, in a status of ‘rest’. A closer look at any piece of matter ‘at rest’ tells us that the status of rest only applies to a certain reference scale. Going down to the atomic scale the world is in motion and fundamental entities, or basic building blocks, are continuous moving with a probability to leave their position for one that fits them better. This phenomenon takes place in the time domain. It is an inherent feature of matter and lies at the basis of ageing of materials. On top of this inherent feature we see, at different scales, a number of gradients, which may initiate basic building blocks of a materials to start moving. In general gradients are driving forces causing changes with elapse of time. At the boundary of any piece of material with its environment gradients are present. These gradients concern, for example, temperature, humidity and radiation and they may cause changes at the surface of the material. The chemistry of the environment may cause changes at the surface with elapse of time as well. Inside a material crystals are connected and build up a strong microstructure. But just at the contact points between crystals atoms are liable to leave their position causing changes in the performance of the material. In heterogeneous materials – and below a certain scale all materials are heterogeneous! – numerous interfaces offer sites for electrochemical activity, resulting in changes of the microstructure of a material with elapse of time and, hence, in ageing. Porous materials continuously communicate with their environment and never reach a condition of rest. This ongoing communication of porous materials with the environment induce alternating stresses and strains in the system, gradually changing the micro- and nanostructure of the material and hence its performance. The foregoing survey illustrates that a material ‘at rest’ is hardly conceivable. Al lower scales there is motion all the time and a variety of driving forces, many of them in the form of gradients, promote the basic building blocks of a material to change their position. This holds for all materials and systems. Basic building blocks search for a position where they feel more comfortable. Their search will be more intensive to more they are forced to leave their position for another one. This is the case particularly in made-made materials, where a lot of external energy has been applied to let nature do what people want. But what we want is not always what nature wants. Nature is dictated by entropy laws which we have to obey as well. By designing materials in a smart way, i.e. by minimizing internal gradients and concentrations of stress and strain, there will be less reason for basic building blocks to leave their position. Hence, the ageing process will slow down and the service life of materials, structures and systems will be enhanced.

3.2

Change of performance with time

The early lifetime of made-made material, structures and systems is often characterized by a high probability of failure. It takes some time to overcome the child diseases and to reach the required level of maturity and stability. Once that point is reached a ‘quiet’ period follows until we arrive again in a period of increasing probability of failure. Exceeding a certain predefined probability of failure marks the end of service life of a structure, system or plant. The high probability of failure in the beginning, the subsequent period of ‘rest’ and the next period of increasing probability of failure is generally presented with the bathtub curve (Figure 2). The length of the period in which the probability of failure is low is of crucial importance for the economic performance of a structure, system or plant. The bathtub curve suggests that this period is a period where ‘nothing happens’. In the previous section it was indicated, however, that this ‘macroscopic rest’ does not mean that there is no activity, motion or


Figure 2 Evolution of the probability of failure in a complex system change. Assuming that in the period of low probability of failure nothing happens is even highly misleading. This can easily be seen if we put the bathtub curve of figure 2 upside down, as shown in figure 3. On the vertical axis we now put ‘Performance’ instead of the ‘Frequency of failure’. After a short period of child diseases the structure, system of plant has reached the required level of maturity. That is the level at which the material should demonstrate its capacity to meet safety and functional criteria, if possible without intervention for maintenance or repair. It is the period for ‘top-level sport’ for all the atoms, molecules and interfaces of the material. When these basic building blocks give up and leave their position, the period of decay starts. The first tiny decay steps will most probably not be observed at the macroscale. The moment that the first basic building blocks give up to do their job can only be captured with comprehensive and appropriate material models at subsequent levels of observation. Here chemistry, physics, electrochemistry, mechanics and mathematics meet each other and need each other for developing tools for describing and predicting ageing processes at the fundamental level.

Figure 3 Evolution of the performance of ageing materials, structures and systems

4 Goals, benefits and investments 4.1

Mitigating ecological footprint

The products resulting from research on ageing are, among other things, tools for describing and predicting the change in performance of materials, structures and systems with elapse of time. Moreover, the use of these tools should result in savings! Savings because the use of these tool should


enable us to extend the service life of, for example, consumers goods. Savings also in the use of materials and energy and in mitigation of the ecological footprint associated with the production of man-made products. Whereas the development of man-made consumer goods generally result in an increase of the ecological footprint, ageing studies aim at reducing this footprint. Ageing studies are all about responsible stewardship and must be considered as studies for the future!

4.2

Benefits - The Dutch case

Savings by extending the service life of fixed capital goods – What are the achievable savings we can expect by increasing the service life of our fixed assets? For an indicative answer we concentrate on the Dutch situation. In The Netherlands the design service life of fixed capital goods varies between 50 years for houses and 80 to 150 years for large structures like tunnels, sluices, sewage systems and sea defence works. Assuming an overall service life of 50 years for all our fixed capital goods with a total replacement value of € 1.825 trillion euros (Table 2), the required investment for replacement in order to keep these assets on the required level of performance is € 36.5 billion euros per year. Huge saving are achievable if the average service life of these fixed capital goods could be extended with, for example, 1, 2, 5, 10, 20 or 50%. With increasing service life the required expenses per year will decrease, resulting in savings as indicated in the right column in Table 3. It must be borne in mind, however, that these calculated savings are not easy to realize. Several aspects have to be considered: - End of service life is often not determined by technical ageing, but functional ageing; - In case technical degradation and failures initiate the end of service life, ageing is not by definition the cause of these failures;. - Extension of the service life only results in savings if this extension is not accomplished as the result of intensive maintenance and repair. Whatever mechanism may have determined the end of service life, the potential savings realised by extending the service life can be huge. If only 5 to 10% of the reasons for service life termination could be attributed to ageing of materials, the attainable direct savings are still substantial. Table 3 Yearly savings for The Netherlands realised by extending the service life of fixed capital goods (inflation not considered) Estimated service life Increase service life Required expenses per Savings per year [years] [%] year [x 1,000,000] [x 1,000,000,000] 50 (reference) 36.5 51 2 35.8 700 52.5 5 34.8 1,700 55 10 33.2 3,300 60 20 30.4 6.100 75 50 24.3 12.300 100 100 18.3 18.250 Justification of investment in research on ageing - Let us try to make the foregoing reasoning on possible savings a bit more concrete. The huge costs a society has to pay for ageing of their assets and consumer goods reaches billions of euros per year (section 2). As mentioned above, the value of all fixed assets of The Netherlands is € 1.825 trillion euros. Let us assume, in line with the foregoing section, an average service life of these assets of 50 years. That would mean that each year € 36.5 billion euros has to be spent on new built. Let us further assume, as a research goal, that through focused research this service life can be increased by 10%, i.e. increases from 50 years to 55 year. The yearly replacement costs would then decrease to € 33.2 billion euros. This is a reduction of replacement costs of € 3.3 billion euros per year. Let us assume that for saving this amount of € 3.3 billion euros per year we have to invest an amount of 20% of it, i.e. € 660 million euros each year, in research. Let us further assume that 50% of the required research money has to be spent on management-oriented research and the other 50% on science-oriented research. Let us finally assume that 10% of the required science-oriented research, i.e. € 33 million euros per year, has to be spent on


fundamental research on ageing of materials. These € 33 million euros per year should be spent on all ageing-oriented materials research in The Netherlands. If we really have the ambition in The Netherlands to invest in a sustainable future, these € 33 million euros is considered a minimum yearly amount for fundamental ageing research. A similar reasoning regarding possible savings and investments required in order to harvest theses savings is expected to hold for other industrialized countries as well.

5 Concluding comments Ageing is a global issue. Hence, solving ageing problems and mitigating the consequences of ageing is a global responsibility. Increasing awareness and recognition of ageing as a (financial) burden for our society and for the environment forces scientists and engineers to look at their achievements differently. Different from the past. Ageing is not an unwanted guest, but an inherent feature of nature that requires our full attention right from the beginning of any design process. The omnipresence of ageing implies also its multiscaledness and multidisciplinarity. A precondition for promising and successful ageing-research is, therefore, interdisciplinary collaboration. Since ageing is a global issue, research projects should preferably be internal. International platforms for in-depth discussions on ageing are considered essential for making progress in this challenging field. Those are the places where science, engineering, industry and government should meet and define the research agenda on ageing.

6 References [1] Long, AE, (2007) Sustainable bridges through innovative advances. Institution of Civil Engineers, presented at Joint ICE and TRF Fellows Lecture. 23. [2] Dobbs R., et al. (2013) Infrastructure productivity: How to save $ 1 trillion a year. McKinsey Global Institute. 88 p. [3] De Haan, M. et al. (2009) The national capital of The Netherlands (in Dutch). In De Nederlandse economie 2009, pp 129-140 [4] Horrocks, P. et al. (2010) Plant Ageing Study – Phase 1. Health and Safety Executive. 144 p. [5] IAEA. (2009) Proactive management of ageing for nuclear power plants. Safety Report Series No. 62. International Atomic Energy Agency. 83 p.


Preface Ageing is an inherent feature of nature and of all man-made products as well. Small consumer good, electronic devices, power plants, electricity grids, infrastructural works and art treasures: they are all subjected to ageing. Although the consequences of ageing can be observed everywhere around us and many strategies have been developed to mitigate these consequences, no comprehensive scientific perception of ageing exists today. Moreover, the omnipresence of ageing and the often very slow progress of ageing phenomena are two reasons why we have got used to the obviously unavoidable presence of ageing. Ageing is a matter of fact and we have to adapt to it. This attitude, however, seems to be on its way back. More and more ageing is considered as a huge financial burden for modern industrialized societies. Ageing of our assets, assets in the widest sense of the word, is a trillion-dollars issue and a burden for the environment as well. But maybe ageing phenomena also offer new and unexpected possibilities! Ageing not only for worse, but also for better. Whether ageing occurs for either better or worse, in both cases we want to be in full control of ageing phenomena. That's where science and research comes in and why the 1st International Conference on Ageing of Materials and Structures, AMS’14, was organized. It is, of course, not for the first time that ageing is on the agenda. We have been faced with the consequences of ageing since the beginning of history and have learned, by trial and error, to deal with at least some of these problems. To be in control, however, a clear picture of the cause of ageing is needed, both qualitative and quantitative. AMS’14 wants to bring problem owners, scientists and engineers from different fields together with the aim to identify the fundamental causes of ageing, its influencing factors, its unavoidability as well as its preventability, its damaging potential and its recoverability, its charm and its risks, its multidisciplinarity and its multiscaledness. This all with the aim to come in control of ageing phenomena and to be able to support decision making processes where ageing play a role. The idea to adopt ageing as a specific theme for research and engineering was launched in the Delft Centre for Materials, DC-Mat. In 2012 it was decided to make a start with the Ageing Centre of the TU Delft, to some extent also inspired by the Materials Ageing Institute in Paris. In February 2013 the Ageing Centre of the TU Delft had its kick-off event and AMS’14 is the first international event organized by the Ageing Centre. This first international conference is considered a path-finding event, where we want to get ageing phenomena sharp and where we articulate the essence and complexity of ageing of materials and structures. Over 80 abstracts from 17 different countries were submitted, covering a variety of fields and also revealing the complexity and intangibles of ageing. The organizers do hope and believe that AMS’14 will contribute to a better understanding and increase of our knowledge of ageing. AMS’14 got the support from RILEM (International Union of Laboratories and Experts in Construction Materials, Systems and Structures). The organizers would like to thank all sponsors of AMS’14 and the respected delegates from all over the world to make this conference to a success. Prof.Dr. Klaas van Breugel Chairman Organizing Committee


KEYNOTES


Radiation Shielding Properties and Freeze-Thaw Durability of HighDensity Concrete for Storage of Radioactive Contaminated Soil in Fukushima Sanjay PAREEK1*, Yusuke SUZUKI2, Ken-ichi KIMURA3, Yusuke FUJIKURA3 and Yoshikazu ARAKI4 (1) College of Engineering, Nihon University, Koriyama, Fukushima-ken, Japan (2) International Research Institute of Disaster Science, Tohoku University, Sendai, Japan (3) Fujita Corporation, Tokyo, Japan (4) Graduate School of Engineering, Kyoto University, Kyoto, Japan Abstract: In this research work, high-density concrete (μ=4.71 g/cm3) using steel balls as aggregates and a normal concrete using recycled aggregates from the debris of the demolished concrete building of the earthquake affected region have been evaluated for the radiation shielding property for the radioactive contaminated soil in Fukushima. Two cylindrical model containers for storage of radioactive contaminated soil samples have been made using these two types of concretes. From the results of the experiments, it was demonstrated that the high-density cylindrical concrete container with a concrete shield thickness of 100mm can reduce radiation dose equivalents emitted from radioactive Cesium in contaminated soil up to as high as 90%. Good agreement was observed between the experimental and calculated dose rate using Monte Carlo simulation (MCNP4C2 code) for two types of concrete for various shield thickness and measurement distance. Also, the freezing and thawing durability of high-density concrete was found to be superior to normal concrete using recycled aggregates. Keywords: radioactive Cesium, high-density concrete, recycled aggregate, shielding design, freezing and thawing

1 Introduction After the disastrous accident at the Fukushima Daiichi Nuclear Power Plant in March, 2011, huge amount of radioactive material was dispersed into the air and contaminated a large land surface area of Fukushima prefecture with radioactive Cesium (134Cs and 137Cs). The government agencies have conducted vast decontamination activity throughout the region by scrapping the ground to a few centimetres and collecting the upper layer of contaminated soil along with organic waste from the surrounding vegetation[1]. The large volume of collected contaminated soil is temporarily stored in plastic bags (1-Ton Packs) and need to be stored safely for an unforeseeable time period. In order to solve this problem of storage of radioactive contaminated soil in a safe method, high-density concrete has been used for such applications[2]. High-density concrete is the most cost-effective material for shielding of radiations and has been popularly used for radiotherapy facilities, nuclear reactors and for spent fuel storage in nuclear power plants[3-5]. In this research work, high-density concrete (μ=4.71 g/cm3) using steel balls as aggregates and a normal concrete using recycled aggregates from the debris of the demolished concrete building of the earthquake affected region have been used to evaluate the radiation shielding property. Two cylindrical model storage containers for radioactive contaminated soil samples have been made using these two types of concretes. Concrete containers with high-density and normal concrete with recycled aggregates for comparison, have been designed for concrete thickness of 100mm and 200mm respectively for equivalent radiation *

Corresponding author. E-mail: [email protected]

1


shielding performance and tested for radiation shielding performance using Cesium radioactive contaminated soil samples as a volume source. Furthermore, in this study, Cesium(134Cs and 137Cs) contaminated soil samples have been used as volume radioactive source which cannot be simulated with point or line source from Cobalt (60Co) as usually used in the evaluation of shielding experiments. The radiation shielding performance of high-density concrete and normal concrete container against gamma(γ) rays emitted from soil volume source contaminated by radioactive Cesium, through experiments and analysis have been evaluated. In addition to this, long-term durability of such containers have been taken into consideration and tested for resistance of high-density concrete to freezing and thawing tests.

2 Materials and mix proportions 2.1 High-density concrete Table 1 gives the properties and mix proportions of high-density concrete used to produce cylindrical storage container. Tap water and ordinary Portland cement as a binder was used along with a shrinkage reducing agent (expansive agent) as an additive with a water-binder (W/B) ratio of 27% (water/cement ratio, W/C=28.7%). Steel balls with a particle size of >0.5mm was used as aggregate. In order to prevent segregation of steel balls, low water-cement ratio and high cement content with a combination of high-range water reducing agent was used to attain adequate flowability. The highdensity concrete tested had a flow of 220mm (JIS R 5201) and an air content of 3.0% in its fresh state. The high-density concrete in its fresh state had a unit weight of 4.71g/cm3. The 28d average compressive strength of moist-cured (20 ℃, 60%R.H.) Φ100x200mm cylinders was 69.7N/mm2. Table 1 Mix proportions and physical properties of materials for high-density concrete Water

Cement

Expansive Agent

Iron Ball and Powder

Chemical Admixture

Density (g/cm3)

1.00

3.16

2.94

7.80

－

Weight Ratio (%)

4.90

17.1

1.06

76.5

0.44

2.2 Normal concrete using recycled aggregates Table 2 shows the materials and mix proportions of the normal concrete using recycled aggregates from the demolished concrete building rubble, to investigate the application to recycle the huge amount of concrete debris from large number damaged and demolished buildings after Great East Japan Earthquake. This would help to solve another major social problem oriented to the recycling of debris from the earthquake disaster. Ordinary Portland cement and fly-ash was used as binder and tap water was used for the mix with a W/B=35.4%. The recycled coarse aggregate used was in accordance with JIS A 5021, from Miyagi prefecture concrete rubble and pretested for no radioactivity contamination by Ge-detector. Crushed granite sand was used as fine aggregate. The freshly mixed concrete with recycled aggregates had a slump-flow of 54cm (JIS A 1150) and an air-content of 5.8%. The 28d average compressive strength of moist-cured (20 ℃, 60%R.H.) Φ100x200mm cylinders was 38.4N/mm2. Table 2 Mix Proportions and physical properties of materials for normal concrete using recycled aggregates Water

Cement

Fly Ash

Fine Aggregate (Crushed Sand)

Coarse Aggregate (Recycled Aggregate)

Chemical Admixture

Density (g/cm3)

1.00

3.15

2.32

2.61

2.40

－

Weight Ratio (%)

7.56

17.6

3.78

34.7

36.2

0.26

2


3 Design and casting of concrete containers for storage of radioactive contaminated soil Figure 1 illustrates the shape and dimensions of the cylindrical model storage container for highdensity concrete. The cylindrical container with a cover-lid had a storage capacity of 0.028m3= 28 litres of contaminated soil in the centre. The high-density concrete shield thickness for all sides is t=100mm. The shape of the container made by using normal concrete with recycled aggregates was the same and with same storage capacity in the centre except for the concrete shield thickness which was t=200mm, twice the thickness of high-density concrete. The concrete shield thickness ratio of high-density and normal concrete of 1:2 had a total weight lower than normal concrete and was designed for an equivalent radiation shielding performance. Figure 2 and Table 3 gives the shape and dimensions of the concrete storage containers. Photo 1 shows the containers made from high-density concrete and normal concrete using recycled aggregates. Insert Bolt

2D Diagram

Insert Bolts

100

400 500

Cross-Section Diagram

100

100 100

300 500

100

300 500

100

Radioactive Source (Contaminated Soil)

(b) Cover-lid （Unit: mm）

(a) Container

Figure 1 Details of high-density concrete container and cover

φ

h’

t

Table 3 Dimensions of storage containers Identification

High-Density Concrete (mm)

Recycled Concrete (mm)

t

100

200

φ

500

700

h

600

800

h φ’

Radioactive Source (Contaminated Soil) Figure 2 Shape of container

3

φ’

300

h’

400


4 Radioactive contaminated soil samples The radioactive contaminated soil samples were taken from four different building sites in the surrounding area in Fukushima located 70km away from the nuclear power station accident site in Dec., 2011. Four samples each containing 20kg bags of contaminated soil were measured for radioactivity using a portable dose meter. In order to attain uniform radioactivity level all the four bags were mixed homogenously and 72 samples of each 50g were extracted from the mix and tested for radioactivity using a Ge detector. The average radioactivity of 72 samples of contaminated soil samples for 134Cs and 137Cs were 31.4+4.0 Bq/g and 48.0+6.2 Bq/g respectively.

5 Test methods for radioactivity shielding measurements Photo 2 and Figure 3 shows the measurement method and test conditions for radioactive contaminated soil samples with and without concrete shielding. The environmental background radioactivity of the test site was 0.5μSv/h. The radioactive contaminated samples were packed in 1mm thick polyethylene bag in a cylindrical shape with a diameter of 300mm and height of 400mm and measurements were carried out for the following three conditions: (1) Radioactivity contaminated Bare soil samples in polyethylene bags with no shielding (Bare) (2) Shielding of radioactive contaminated soil samples by high-density concrete (H-Con) (3) Shielding of radioactive contaminated soil samples by normal concrete with recycled aggregates (N-Con) All the measurements were carried out in such a position that the centre of soil remained at a ground height of 450mm as shown in Figure 4. The measurements for radioactivity shielding by high-density concrete were conducted for 3 points at a distance of L= 0.1, 0.2 and 0.3m from the centre point of the sample and for normal concrete with recycled aggregates for 2 points at a distance of L= 0.2 and 0.3m since the concrete shield thickness was 200mm. Survey Instrument Model 5000 was used for the measurements. Each measurement was carried out for 3mins and was an average of 6~8 measurements for each point. The influence of the environmental background radiation was deducted from the readings for shielding effect by concrete containers.

Cover

Cover

Cover

Container

Container

High-Density Concrete Container

Recycled Concrete Container

Photo 1 A view of concrete containers

Container

NaI Scintillation Detector

Photo 2 A view of measurement set-up

4


150

Contaminated Soil

Distance, L

100 100 100

450

Wooden Pallet

NaI Scintillation Detector GL

(Unit: mm) （a）Bare (No shield)

150

High-Density Concrete Container

Contaminated Soil

Distance, L 450

Wooden Pallet

NaI Scintillation Detector

100 100

GL (Unit: mm) (b) High-Density Concrete Shielding Figure 3 Measuring methods for radiation shielding

6 Test results and discussions 6.1 Radiation shielding performance of concrete Table 4 gives the radiation shielding performance by high-density concrete and normal concrete with recycled aggregates in comparison to the measurements for the contaminated Bare soil samples without shielding for respective measurement distance L. The shielding ratio or attenuation of high-density concrete and normal concrete with recycled aggregates is a ratio of respective measurement of concrete to the Bare contaminated soil sample. Figure 4 illustrates the measurements of radiation dose rates of the contaminated soil sample for L=0.1, 0.2 and 0.3m respectively. It is clearly evident that the radiation dose rate is inversely proportional to the distance (L) and shows a drastic decrease with an increase in distance of measurement point irrespective of shielding concrete thickness or type of concrete. Remarkable attenuation of 94% was observed for high-density concrete which measured 0.18μSv/h at the distance of L=0.1m in comparison to the contaminated Bare soil samples which measured 3.15μSv/h. Similarly, the measurements of radioactive dosage at a distance of L=0.2m for contaminated Bare soil sample, high-density(H-Con) and normal (N-Con) concrete was 1.71μSv/h, 0.14μSv/h, and 0.15μSv/h respectively. Furthermore, at a distance of L=0.3m for Bare soil sample, high-density and normal concrete was 1.07μSv/h, 0.07μSv/h, and 0.09μSv/h respectively showing a remarkable decrease in radioactivity dosage. The shielding performance by containers made with high-density concrete with a thickness of 100mm and normal concrete using recycled aggregates

5


with a thickness of 200mm showed more than 90% attenuation or reduction of radioactivity dosage. The radioactivity attenuation rates by concrete containers made with high-density concrete (t=100mm) in comparison to normal concrete with recycled aggregates (t=200mm) was 1.01 times at L=0.2m and 1.02 times at L=0.3m showing equivalent or higher shielding performance Table 4 Radiation shielding performance by high-density concrete and normal concrete Bare Soil (μSv/h)

H-Con Shielding (μSv/h)

N-Con Shielding (μSv/h)

Shielding or Attenuation by H-Con (%)

Shielding or Attenuation by N-Con (%)

H-Con / N-Con

0.1

3.15

0.18

‐

94.3

‐

‐

0.2

1.71

0.14

0.15

91.8

91.2

1.01

0.3

1.07

0.07

0.09

93.5

91.6

1.02

Radioactive Dose Rate (μSv/h)

Measurement Distance, L (m)

4.0

Bare High-Density Concrete Shielding Normal Concrete Shielding

3.5

3.15

3.0 2.5

2.0

1.71

1.5

1.07

1.0 0.5 0.0

0

0.1

0.2

0.3

0.4

Distance from Contaminated Soil Surface, L(m)

Figure 4 Effect of concrete shield thickness or measurement distance on radioactive dose rates

6.2 Analysis and simulation of experimental results Analysis of radiation shielding was performed using Monte Carlo N-Particle Transport Code System (MCNP42C) using photo library (MCPLIB02) with “Radiation dose conversion coefficients fpr radiation shielding calculations based on the AESJ-SC-R002:2010 and the results were compared to those of experimental measurements[6-7]. The measured densities of the contaminated soil samples μ=1.25g/cm3 , high-density concrete μ=4.57 g/cm3 and normal concrete with recycled aggregates μ=2.14g/cm3 was used for calculations. For the simulations, respective concrete was assumed to be uniform and isotropic. The contaminated soil sample was also assumed to be uniform and the atomic distribution of 134Cs and 137Cs as isotropic photon volume source. As the shielding from gamma (γ) rays is strongly influenced by the density and concrete shielding thickness, the equivalent number of atoms for each material was assumed for analysis purpose to constitute the same density as measured. Figure 5 shows the experimental and analytical results of attenuation curves for radiation dose rates for high-density concrete, normal concrete and contaminated Bare soil sample in as a function of measurement distance (L) or concrete shield thickness (t). The analysis results were simulated for two types of concretes with various shield thicknesses. The experimental results were only for a

6


single concrete shield thickness (0.1m for H-Con and 0.2m for N-Con) and fitted well on the analytical curves for high-density concrete and normal concrete with recycled aggregates. The results of the analysis matched well with the experimental results of the contaminated Bare sample and the variation of data was below 15%. The results of the MCNP analysis for simulated conditions are in good agreement with the experimentally measured results. In addition, highdensity concrete shows the same shielding performance as the normal concrete with twice the thickness of concrete shield as compared to the high-density concrete.

6.3 Design for optimization of concrete shield thickness

Ratio of Dose Transmiission (%)

Radioactive Dose Rate (μSv/h)

Figure 6 illustrates the concrete shield thickness vs. ratio of dose transmission. The curves represent intensity of radiation transmission through high-density concrete (μ=4.57 g/cm3) and normal concrete (μ=2.14g/cm3). The calculated values from the analysis were below the intensity of radiation transmission curves and the difference is higher with the increase in shield thickness. These differences in values are due to the use of point radiation source for analysis which is different from the volume radiation source from the contaminated soil. Therefore, for the optimization of design shield thickness of concrete for containers, analysis methods using volume radiation source need to be taken into consideration [7]. 1.E+01

1.E+00

Bare (Exp.)

1.E-01

Bare (Ana.) H-Con Shielding (Exp.) H-Con Shielding (Ana.)

1.E-02

N-Con Shielding (Exp.) N-Con Shielding (Ana.)

1.E-03

1.E-04

1.E+00 H-Con Shielding (Volume Source)

1.E-01

N-Con Shielding (Volume Source)

1.E-02 Normal-Con (2.14g/cm3) Shielding （Point Isotropic Source）

1.E-03

1.E-04 H-Con (4.57g/cm3) Shielding （Point Isotropic Source）

1.E-05

0.0

0.2

0.4

0.6

0.8

1.0

Thickness of Shielding Concrete or Distance, L(m)

Figure 5 Experimental and analytical results for concrete shield thickness vs. radioactive dose rates

1.E-06

0.0

0.2

0.4

0.6

0.8

1.0

Thickness of Shielding Concrete，L(m)

Figure 6 Analytical results for shield thickness vs. radioactive dose transmission

6.4 Freezing and thawing resistance of high-density concrete The freezing and thawing resistance of high-density concrete is an important factor for using it as storage containers for radioactive contaminated wet soil samples. Figure 7 gives the relative dynamic modulus of elasticity of high-density concrete and normal concrete for 0 to 150 freez-thaw cycles. A dramatic loss of relative dynamic modulus of elasticity for normal concrete is observed after 50-60 cycles in comparison to high-density concrete. At 100 freeze-thaw cycles, the relative modulus of elasticity of normal concrete was hardly 40% of the initial stage in comparison to highdensity concrete which remained unchanged, showing a superior freeze-thaw resistance. Figure 8 shows the weight change of high-density and normal concrete specimens subjected to freezing and thawing cycles. The weight change of specimens is directly proportional to the deterioration of relative dynamic modulus of elasticity of specimens. A sharp weight loss due to scaling is observed for normal concrete specimens after 50 cycles and subsequent cycles, whereas the weight change of high-densisty concrete does not show a distinct loss in weight of the specimens.

7


1

60 40

N-Con

20

H-Con

0

30

60

90

120

150

0

80

Weight Change (%)

Relative Dynamic Modulus of Elasticity (%)

100

0

-1 -2 N-Con -3

H-Con

-4 0

30

60 90 No. of Cycles

120

150

Figure 7 Number of freeze-thaw cycles vs.

relative dynamic modulus elasticity of concretes

of

-5

No. of Cycles

Figure 8 Effect of freeze-thaw on weight

change of concrete specimens

7 Conclusions Two types of model cylindrical containers for storage of radioactive 134Cs and 137Cs contaminated soil in Fukushima were made using high density concrete (μ=4.57g/cm3) and normal concrete with recycled aggregates (μ=2.14g/cm3) and were evaluated for shielding performance. From the results of the experiments, it was demonstrated that the cylindrical container using high-density concrete of 100mm shield thickness had attenuation as high as 90% for radiation dosage rates emitted from radioactive Cesium in contaminated soil. Good agreement was observed between the experimental and analytical values using Monte Carlo MCNP4C2 simulation for shielding performance. Also, the freezing and thawing durability of high-density concrete was found to be superior to normal concrete using recycled aggregates. Therefore, the proposed high-density concrete with adequate strength and durability is suitable for shielding radioactive contaminated soil of Fukushima.

8 Acknowledgements This research was partially supported by Grant-in-Aid for Young Scientists (B) No. 25820265 provided by Japan Society for the Promotion of Science (JSPS).

9 References [1] The Society for Remediation of Radioactive Contamination in Environment (2012) ： Abstracts of 1st Research Presentations on Remediation of Radioactive Contamination in Environment (In Japanese) [2] H. M. Cheyrezy(1996): Development of HPC in France: Recent Achievement and Future Trends, American Concrete Institute, Special Publication, Vol.159, pp.145-158. [3] Joseph Davidovits (1994): Recent Progresses in Concretes for Nuclear Waste and Uranium Waste Containment, American Concrete Institute, Concrete International, Vol.16, No.12, pp.53-58. [4] Yu-Chu Peng and Chao-Lung Hwang(2011): Development of High Performance and High Strength Heavy Concrete for Radiation Shielding Structures, International Journal of Minerals, Metallurgy and Materials, Vol.18 No.1, pp.89-93. [5] S. M. J. Mortazavi, M. A. Mosleh-Shirazi, P. Roshan-Shomal, N. Raadpey and M. BaradaranGhahfarokhi(2010): High-Performance Heavy Concrete as a Multi-Purpose Shield, Radiation Protection Dosimetry , Vol.142, No.2-4, pp.120-124. [6] Los Alamos National Laboratory (2003): MCNP4C Monte Carlo N-Particle Transport Code System, CCC700. [7] Yusuke SUZUKI, Ken-ichi KIMURA, Yusuke FUJIKURA, Yujin LEE, Sanjay PAREEK and Yoshikazu ARAKI(2013): Concrete Research and Technology, Vol.24, No.2 (In Japanese), pp. 43-52.

8


Stochastic Models for Risk and Failure under Ageing J¨ urg H¨ usler∗ University of Bern Abstract: We will discuss several stochastic models for the failure of a material or for risk event. Typically, the failure of a material occurs if the load of the material is higher than its designed maximal load value. However, the material can become weaker by age or use or slow deformation by near critical loads. Such stochastic models will be treated. More realistic models will be also discussed, where the critical load level of a material is not fixed and can change in time. Also a Bayesian approach will be mentioned with possible applications. Keywords: Risk model, cumulative shock, extreme shock, urn model, ageing model.

1

Introduction

A material or a structure is constructed based on a certain design. It is expected that the material will function well as long as it is not overloaded or over-stressed by some impact factors. For bridges it is the load, for wind turbines it is the strong wind or storm, for a steel plate it is the force or load, for a aircraft it is the acceleration in flight programs. The material is at the beginning possibly well constructed and in good shape which is certainly checked before use. Unfortunately, with time and use the construction is getting weaker and weaker. The steel of a bridge or tower will be weakened by the weather impact, the aircraft is less robust by the many training flights with many large accelerations depending on the flight programs. It means that with time, the material cannot be loaded or stressed as when new. Each material has a designed life time during which no failure should occur. For example an aircraft has a given number of flight hours which denotes its (technical) life time. Since failure can always happen, one should know the probability of such an event during the designed life time. However, it would be better to know the distribution of the time until failure or the so-called survival distribution. Typically, one does not have many data at the beginning when designing the material and later when using the material. Often, one has physical models and engineering methods which allow to determine the construction of the material for an intended maximal load or strength. Other models are used in accelerated statistical testing. In some cases one can collect data on the survival time or time to failure, but the data are related only to some conditions. For other situations one applies generalization or extensions which might be debatable. ∗

Dept. of math. Statistics, E-mail: [email protected]

9


Ageing has an impact on the material and on its maximal load below which no failure will happen. However, it is possible that the large loads have an impact on this crucial load, not only the more or less natural ageing. We think that some loads can change the crucial load value, in particular if a load is close to the crucial one. For instance, a large stress or acceleration may produce a crack in the neighborhood of a bolt in an aircraft which weakens the strength of the aircraft. In many cases the material (an aircraft, a bridge, a steel construction) is controlled in regular time periods to prevent a failure during the inspection periods because of small or moderate damages, use or ageing. Such controls have an impact on the behaviour and the failure probabilities. This should be included also somehow in a stochastic model. We will deal with such probabilistic models for failure events and discuss more advanced models which are motivated by realistic scenarios. Risk and failure models occur in many different fields. We mention for instance the insurance business where risk has to be modeled in their applications. A simple risk model consists of the premiums and the claims. The claims occur at random time, the premiums are paid in a more or less regular way. The claim sizes are random with a certain underlying distribution. The premiums should cover the claims at any time. With a certain reserve the risk process should not become negative which means that the last claim is no more covered. This is the so-called ruin event. There exists several different models for the risk process in the insurance topic. Some may be used also as a model for the ageing process.

2

Simple failure models

A failure of a structure can happen in different ways. It might be that many small or moderate shocks sum up to a certain critical level when the load becomes too large and the structure breaks down or fails at the next load. It might be also that the structure is not harmed as long as the load is lower than the designed level, but it collapses if the load is larger than the critical value. These two scenarios are called cumulative shock and extreme shock model, respectively. The shocks are arriving irregular in time. These arrival times are best modeled as random variables. Mathematically, it is described as follows. Let the shock sizes or load values be denoted by Xi , i ≥ 1, which happen at certain random time points Ti . Let u denote the designed critical value. In the cumulative shock model a failure occurs the first time at Tk if the partial sum ∑ Sk = Xi > u, but Si ≤ u for i < k. 1≤i≤k

In the extreme shock model this happens the first time at Tk if Xk > u, but Xi ≤ u for all i < k. Then the number of shocks τ until failure is defined as τ = min{k : Xk > u} in the extreme shock model, or τ = min{k : Sk > u} in the cumulative shock

10


model. Let us call τ simply the failure time. The life time of the system of failure time is denoted by Tτ , since no failure occurs before Tτ . One can also combine both failure events into one, i.e. a failure occurs at Tk if either Sk > u or Xk > u happens. Such models are called mixed shock models. For these simple models one can derive the distribution of the failure time τ under further assumptions on the random variables Xi and Yi = Ti − Ti−1 , the interarrival time (see [11-13] and references in these papers). It is known that the failure time τ of the extreme shock model has a geometric distribution in the simplest model with independent, identically distributed shocks. The failure time distribution converges to an exponential distribution, as the probability of a failure is small, tending to 0, hence the threshold u tends to the endpoint of the underlying distribution F of the shocks Xi . It is also known that the life time has an asymptotic normal distribution in the cumulative shock model. This is also true in the case that the shock Xi and the inter-arrival times Yi are not independent. For instance, the financial models do work with the cumulative approach. The cumulative shock model can be viewed as a risk model under ageing. The shocks during the life time are summed up until the sum is larger than the designed critical value. The life is spent. A small shock can ruin the structure. If the shock Xi is not directly harming the material, one may take a function g of Xi which denotes the impact on the structure. But this can be rewritten by putting g(Xi ) as Xi which changes only the distribution of the random variables in the sum Si . Mathematically this does not introduce any essential change. Such models consider a certain random ageing effect, because the designed critical value u is varying from shock to shock.∑ With age the level u is decreasing in time. One observes a failure if Xk > u − i u(k) or an extreme shock Xk > u(k) depending on the model. The function u(k) can be modeled for example as u(k) = u − βYk for just a linear ageing effect, or ∑ u(k) = u − 1≤i 0 and δ > 0. If we let λ = 0, hence δ = θ. Thus, a red ball is not replaced together with another black ball. The urn is always reinforced by additional balls of the same color as the drawn one. The urn content is changing depending on the randomly selected balls. The number of white and red balls are increasing before a failure. Hence the structure becomes safer. A red ball has no impact on the failure and is therefore not needed in the urn. Thus one should let λ > 0, then the process shows some dependence between the red and black balls. A drawn red ball changes the failure probability because of the added black balls. The probability of a failure

13


(drawing a black ball) is after the k-th drawing with j(j ≤ k) selected red balls: (b + jλ)/(w + r + b + kθ) where w, r, b denote the numbers of white, red and black balls, as the prior distribution, before drawing any ball. Note that the urn is changing randomly, so a random ageing effect applies here. Such models can be easily simulated to receive results on the number of drawings until failure or on the failure time and the life time. One can use prior knowledge from the design and material properties to select w, r, b and θ, λ. Simulations allow also to investigate the impact of selecting the prior w, r, b, θ, λ on the failure times. If one has an inspection plan, one can say that after m drawings, the urn is replaced by the same urn content as at the beginning (’repaired is as new’) or by a slightly changed content (to model the ageing with inspection and repairing). Depending on the assumptions of the reinforcement matrix R, [1-8] derived theoretical results on the failure distribution for several special and more general urn models and showed some applications.

4

Modelling with extreme value distribution

Let us consider the following corrosion example for this third approach. This is a statistical approach. The deepest hole in a tube or a metal plate should be measured in an inspection, e.g. in the inspection of the tube for oil transport or pumping from ground. One samples the tube at several sites following an inspection plan and measures the thickness of the tube material. Based on the inspection data, one should estimate the distribution of the smallest thickness value of the tube at any site. This is one of the many applications of extreme value theory. It says that the distribution of the smallest tube thickness (based on a large sample of measured thickness of the tube) can be approximated by the generalized extreme value distributions. If one uses all the smallest values lower than a given threshold, one can approximate the tail of this (conditional) distribution by a so-called generalized Pareto distribution which are related to the generalized extreme value distributions. The class of extreme value distributions Gγ for maxima is well-known (see e.g. [9-10], [14-16]): ) ( x−λ x − λ −1/γ )) , where 1 + γ( )>0 Gγ (x) = exp −(1 + γ( σ σ with λ and γ the location and shape parameters (real values), and σ > 0 the scale parameter. For γ = 0 we have by continuity G0 (x) = exp(− exp(−

x−λ )). σ

For γ < 0, the distribution has a finite upper endpoint λ − σ/γ. ˜γ : For minima we get the class of (minima) extreme value distributions G

14


˜ γ (x) = 1 − Gγ (−x) G using the relation min(Xi , i ≤ n) = − max(−Xi , i ≤ n). The parameters λ, σ and γ are estimated using one of the many estimators as moment estimator, maximum likelihood estimator, probability weighted estimator, Hill estimator and further variants (see e.g. [14-15]). These three parameters may change because of age. The ageing process can be modeled into the parameters. After several inspections one may have sufficient data to model or estimate the ageing process, by assuming that the parameters γ(t), σ(t) and λ(t) depend on the (inspection) time or age t in a certain way as in the former sections. If one uses in the data analyses all data above a threshold, or as in our example of the tube thickness, the data below a threshold, one applies the so-called peaks over threshold approach (POT) where the tail of the distribution can be approximated by a generalized Pareto distribution. This family of distributions Wγ for POT is related to the extreme value distributions by Wγ (x) = 1 + log Gγ (x) = 1 − (1 + γ

x − λ −1/γ ) if log G(x) > −1 σ

which depend on the three parameters γ (shape), σ (scale) and λ (location). Again, the three parameters can be modeled depending on time t for ageing. For our tube thickness example one has to use the approach for peaks lower threshold (PLT) or the mentioned transformation for minima. The class of distributions for the PLT approach for minima is ˜ γ (x) = (1 + γ x − λ )−1/γ W σ

where 1 + γ(

x−λ ) > 0. σ

For our tube example, we think that the lower boundary is naturally 0. Hence, the shape parameter γ should be negative or 0. This is used when modeling the ageing applying the POT method. Since the boundary of the generalized Pareto distribution for negative γ is depending on γ we may use another parametrization of this class of distribution which is simply a Beta distribution for γ < 0. We have that the thickness data below a threshold can be modeled with the distribution ( x )α Wα∗ (x) = σ with α = −1/γ > 0 and 0 ≤ x ≤ σ. With this PLT approach we can again introduce an ageing effect in the two parameters α and σ as above. Sometimes other distributions are used for the approximation of the empirical distribution because they fit possibly slightly better, but they would have no theoretical background as the GPD distributions. With this statistical approach one can estimate the probability of a thickness value below a certain critical value. Or one can also estimate the quantile (percentile) for a very rare event, say 10−5 or 10−6 , which is related to the possible number of sites of the tube which are not sampled during the inspection.

15


Such extreme quantiles are needed to discuss the safety of the tube or the material.

References: [1] Cirillo P and H¨ usler J (2009) An urn-based approach to generalized extreme shock models, Stat. Probab. Letters 79:969-976. [2] Cirillo P and H¨ usler J (2009) On the upper tail of Italian firms size distribution, Physica A: Stat. Mechanics Appl. 388:1546-1554. [3] Cirillo P and H¨ usler J (2010) Shock models and firms’ default: parametric and nonparametric models. In: Festschrift T. Hettmansperger, WSPC Proceedings. [4] Cirillo P, H¨ usler J and Muliere P (2010) A nonparametric approach to interacting failing systems with an application to credit risk modeling, Intern. J. Theor. Appl. Finance 13:1223-1240. [5] Cirillo P and H¨ usler J (2011) Generalized extreme shock models with a possibly increasing threshold, Probab. Engin. Inform. Sci. 25:1-16. [6] Cirillo P and H¨ usler J (2012) An urn model for cascading failures on a lattice, Probab. Engin. Inform. Sci. 26:509-534. [7] Cirillo P, H¨ usler J and Muliere P (2013) Alarm systems and catastrophes from a diverse point of view, Method. Comp. Appl. Probab. 15:821-839. [8] Cirillo P, Gallegati M and H¨ usler J (2012) A Polya lattice model to study leverage dynamics and contagious financial fragility, Adv. Compl. Systems 15 supp. 02, 1250069 (26 pages). [9] Embrechts P, Kl¨ uppelberg C and Mikosch T (1997) Modelling extremal events for insurance and finance. Springer, Berlin. [10] Falk M, H¨ usler J and Reiss RD (2010) Laws of small numbers: Extremes and rare events. DMV Seminar Band 23 3rd edn. Birkhäuser, Basel. [11] Gut A and H¨ usler J (2005) Realistic variation of shock models, Stat. Probab. Letters 74:187-204. [12] Gut A and H¨ usler J. (1999) Extreme shock models, Extremes 2:293-305. [13] Gut A and H¨ usler J (2009) Shock models. In: Nikulin, M.S. et al. (eds) Advances in Degradation Modeling: Applications to Reliability, Survival Analysis, and Finance, Series Statistics for Industry and Technology, Birkhäuser, Basel, pp. 59-76. [14] Haan L de and Ferreira A (2006) Extreme value theory. Springer Series in Operations Research and Financial Engineering. Springer, New York. [15] Reiss RD and Thomas M(2007) Statistical analysis of extreme values. 3rd edn. Birkhäuser, Basel. [16] Resnick S (2007) Heavy-tail phenomena. Springer, New York.

16


The Relevance of Aging for Civil Infrastructure: The Profession, The Politics, The Classroom David A. Lange University of Illinois, Urbana, IL, USA Abstract: The work of civil engineers and researchers has strong impact on public policy for infrastructure funding. This talk considers several professional, political, and education-related thrusts. First is the challenge of communicating about intrinsically qualitative matters with metrics that support rational decision-making for spending public funds. Second, it is necessary to confront political realities when trying to influence infrastructure investment policy in free and democratic societies. Third, aging of infrastructure is complex and it demands new knowledge to address technical, economic, and political aspects. Fourth, the new knowledge and expanded skills expected of future engineers are driving reform of curriculum and pedagogy. This talk revels in the opportunity to marry technical fact with political commentary in the spirit of provoking thoughtful debate and discussion. Keywords: public policy, infrastructure funding, sustainability, education reform

1 Introduction This conference is entitled The First International Conference on Aging of Materials and Structures. It is the first of what we hope will be a long and lasting conference series. Most of us in attendance are academic researchers and engineers who study a wide variety of technical issues related to material deterioration. Indeed, we are part of a long tradition of scientific research that aims to expand our knowledge of materials so we can improve the quality of infrastructure, achieve longer service life, and possibly do it all at lower cost. So it is right and fitting that the bulk of our conference is focused on technical topics and the latest experimental research results. This talk is a bit different. I was asked to think about the relevance and urgency of ageing issues regarding the American infrastructure. What is the extent the aging infrastructure problem, the money involved, and the means to solve the problems? The relevance is huge and expansive as our society spends billions of dollars (or euros) on infrastructure. It is a goal of a free and democratic society that we work on behalf of the citizens who fund those expenditures so as to achieve the highest possible value for the benefit of all. The public policy question is simple at its core: “How can we best spend our resources on infrastructure to achieve the highest benefit?” The implementation, of course, is complicated and controversial because everyone has a different answer to that question. We as civil engineers and researchers have a unique and essential role. It is a role we do not often realize we play, as we may be quite content to stay within our closed academic community. But make no mistake – others are looking to us for answers. Our research sponsors are curious about our research for a reason! They are often trying to answer that very basic question – How can we best spend public money on infrastructure? My strategy in this talk is to develop four lines of argument about the relevance of aging infrastructure. First, I want to make a few comments about how we face very hard problems in describing the condition of infrastructure because it is a complex matter affected by material deterioration and structural performance that can be redefined as we learn, for example, more about seismic loads. It is never easy to transition from neatly quantitative parameters of scientific work to complex qualitative matters such as infrastructure condition, value of infrastructure for national competitiveness, costs of inadequate infrastructure, and cost for repair of complex systems. Second, I want to comment how infrastructure spending decisions are made in the United States, the level of infrastructure funding over the years, and how infrastructure is a hot potato in US politics. How one communicates and persuades requires effective communication of infrastructure priorities targeted at

17


the layman (i.e. politician) who may not be as interested in the technical story as he is in the political ramifications of infrastructure spending. Third, for the past 10 years we have seen the rise of Sustainability as a powerful buzzword across all aspects of civil engineering, and indeed across the global economy. Sustainability is a critical theme that touches on repair, rehabilitation, and reuse of infrastructure. Civil engineers naturally embrace sustainability because it is a central principle of responsible engineering practice. And fourth, the technical challenges of aging infrastructure represent an opportunity for engineering educators. Education reform – happening in both North America and in Europe – can improve the effectiveness of higher education and elevate the technical skill set associated with infrastructure. Curriculum needs reform to address emerging topics such as forensic studies, repair materials, and service life prediction.

2 The State of Infrastructure 2.1

Condition survey and infrastructure inventory

Translating infrastructure condition into meaningful quantities is an ongoing challenge for agencies responsible for highways, airports, railroads, and ship transportation. The practice of conducting condition surveys emerged as a matter of necessity for agencies overseeing diverse and geographically distributed facilities. Today, all US States have established practices for condition survey of infrastructure, although the uniformity among the States is not perfect. The Federal government has made great strides in building national inventories for infrastructure. There is a myriad of inventories maintained at the national level: Briges, roads, railroad grade crossings, airport runways, dams, pipelines, etc. The purpose of infrastructure inventories is slowly maturing as a tool for scheduled maintenance activities, and as such, they gain value for budgeting activities. The procedures for inventories are increasingly defined by higher levels of government, and today the federal level is often specifying which parameters are to be measured and how they are to be measured in a consistent way. Repair professionals have developed methodology for condition surveys that can be applied to private and public infrastructure. Consultants are often engaged to give advice about repair or replacement of structures and pavements, and there are now ample tools for standardizing the initial surveys required to assess options for repair.

2.2

ASCE Infrastructure Report Card – “The bridge is falling!”

The American Society of Civil Engineers (ASCE) has been an advocate for the profession since its founding in 1852. With a mission to advance professional knowledge and improve the practice of civil engineering, ASCE strives to be a focal point for the transfer of research results and technical policy. In 1998, ASCE published its first Infrastructure Report Card (Figure 1), taking the somewhat controversial approach for communicating dire need for repair and maintenance. The idea of grading our nation's infrastructure did not originate with the ASCE, but came in 1988 by a presidential commission created to report on the state of the US infrastructure. The presidential commission assigned an overall grade of C, and the title of their report "Fragile Foundations: A Report on America's Infrastructure" hinted at the shaky state of US infrastructure. ASCE has re-worked their analysis every few years, always arriving at a grade close to D. The latest report in 2013 observes a slight rise to D+. The report card comments on a wide swath of infrastructure categories, observing, for example, that 1/9 of the nation’s 600,000 bridges are structurally deficient.

18


Figure 1 The ASCE Infrastructure Report Card [1]

The impact of the ASCE report card has been striking. With every renewal of the report, the U.S. media gives fresh visibility to infrastructure quality. The report resonates well with the public as every citizen has their own complaint about poor quality roads. Tragic accidents have been attributed to decaying infrastructure. Perhaps one of the most publicized failure occurred in 2007 when the dramatic rush-hour collapse of the I-35W bridge over the Mississippi River in Minneapolis caused the death of 13 people. The ASCE report card is now familiar to all state and federal legislators, and over time it seems to have gained credibility as a touchpoint for political posturing. President Obama made famous reference to the report card in his 2014 State of the Union address, using the speech as a launching point for new spending on infrastructure. Since the economic recession in 2008-9, the U.S. government has asserted a variety of economic stimulus initiatives, increasing the flow of federal dollars in ways designed to promote economic stability. The 2014 emphasis of infrastructure is positioned as another spending plan that increases economic activity in a broad way while achieving a broad and popular goal of improving infrastructure quality.

2.3

Contrary viewpoints – “The bridge is not falling”

Not everyone thinks the ASCE report card is a good thing [2]. Many observers believe ASCE is playing politics with their report card. The methods for measuring infrastructure condition are extracted from national inventories, but the manner of interpreting the data is thought to be selective for the purpose of drawing attention to infrastructure spending. Infrastructure spending has been relatively steady for many years. Total public construction spending has varied between 1.7% and 2.3% of GDP for the last 20 years, according to the U.S. Census Bureau. By the Congressional Budget Office's slightly different measure, infrastructure spending has been between 2.3 percent and 3.1 percent of GDP since 1956. [3]. Another source estimated that the U.S spends about 3.3% of its federal spending on infrastructure while in Europe the comparable figure is 3.1% [4]. The fiscal conservative flinches when ASCE trots out its final conclusion that $2.7B is needed to bring the infrastructure up to par when the current annual spending on infrastructure is in the $700M range. My goal today is not to parse out the debate between ASCE and its detractors, but to illustrate the clear point that civil engineers DO make a difference in the setting of public policy. Whether we want the role or not, we are a profession that holds responsibility for infrastructure and seeks to deliver high value for maximum public benefit.

19


3 Making the case for strong investment in infrastructure 3.1

Infrastructure spending in the United States

In 2013, the U.S. federal government spent $3.7 trillion. It is not easy to dissect the budget [I’ll leave that job to the professionals!], but there are some pertinent statistics available to the common man. One measure of infrastructure investment is the budget of the Department of Transportation (although, admittedly, not all DOT money is spent on new construction or repair). The DOT budget has been reasonably stable since the 1980s at 2.0% of the federal budget. For 2013, the DOT budget was $750M. The allocations within the DOT budget vary from year to year, and certainly there is a political element in how certain themes are featured as programs de jure. For example, in 2010 the U.S. made a $1B initiative for high speed rail in its annual spending plan above the previous year’s spending level. In another case, a $4B Infrastructure Innovation Fund was established to stimulate projects with high priority. Other sources suggest that actual expenditures on infrastructure have lagged in recent years [6]. Tracking expenditures in different ways suggests that actual spending on construction has fallen as shown in Figure 2. The broad explanation for the contraction of spending even as federal budgets are maintained is that States and local governments are the biggest part of the story. Even federal dollars funnel through the States for the vast majority of spending for roads, highways and bridges, and the States have pulled back on spending since 2008 as a result of the economic downturn and requirements to balance State budgets. For example, California’s transportation spending declined by 31% from 2007 to 2009 while Texas's fell by 8 percent.

Figure 2 U.S. infrastructure spending [6]

3.2

National competitiveness is dependent on infrastructure

High quality infrastructure is a requirement of an efficient and effective modern society. The quality of infrastructure is one of the clearest differentiators between Third World and First World economies. The World Economic Forum developed an index system for global competitiveness [7]. The U.S. comes out as #7 on their list of 144 nations in terms of competitiveness, and the nations ahead of the U.S. include European countries such as Switzerland, Finland, Sweden, Netherlands, and Germany, along with Singapore. More interesting than the simple national ranks is the methodology used by the WEF. The competitiveness index includes 12 pillars, of which Infrastructure is the second. [The 12 pillars are Institutions, Infrastructure, Macro economic environment, Health and primary education,

20


Higher education and training, Goods market efficiency, Labor market efficiency, financial market development, technological readiness, Market size, Business sophistication, Innovation.] The Infrastructure Pillar represents fully 25% of the WEF global competitiveness index – a major statement about its importance. Half the Infrastructure score is comprised of traditional infrastructure categories like roads, railroads, airports, and ports. The other half is comprised of electric and telephone infrastructure. Every country is scored and problems are identified. For India (59th on the index ranking), the greatest problem is the inadequate supply of infrastructure. For the U.S. (7th on the index ranking), the greatest problem is inefficient government bureaucracy, and infrastructure is near the bottom of the list.

4 Sustainability – a guiding principle 4.1

The Rise of Sustainability

Sustainability has been a popular emphasis across the civil engineering profession for more than 10 years. Many conferences have been held with the word “Sustainable” prominently featured in their titles. In the U.S., the American Concrete Institute reflected this trend when they established a new technical committee on sustainability in 2008 that quickly drew more than 100 members. In 2009, the Concrete Joint Sustainability Initiative was established, bringing more than 30 construction and material trade associations together in a common cause. Sustainability, from the concrete industry viewpoint, represents a broad set of issues including environment, economy, and society. Our industry has emphasized recycling and reuse of concrete materials as one of the most sustainable practices. Aging is relevant to sustainability because long service life is the best way to make concrete construction more sustainable. Aging has many attributes, but materials durability is among the most threatening. Durability of concrete materials is an attractive focus because our current knowledge allows us to achieve very long service lives – perhaps beyond 100 years – if materials, construction technology, and reasonable maintenance practices are employed. Concrete is inherently a durable material unless economic compromise, poor workmanship or inadequate design leads to high permeability, cracking, or overloading. Obsolescence is another factor that is hard to predict because needs and requirements can change in unpredictable ways, rendering worthless an otherwise serviceable structure.

4.2

The Future of Sustainability

Sustainable infrastructure demands commitment to principles of recycling, rehabilitation, and reuse. New construction gets the headlines, but Repair Engineering is perhaps even more important for improving sustainable practices. The Repair community has made tremendous strides over the past 20 years. The profession has elevated the visibility of repair, expanded the professional organizations, and advanced the state of the art by creating codes and standards to ensure the use of best practices across the industry. For example, members of the International Concrete Repair Institute (ICRI) and ACI have worked together to create the ACI 562-13 Code Requirements for Evaluation, Repair, and Rehabilitation of Concrete Buildings. ACI, RILEM and other organizations have supported committees to address repair-related issues such as NDT, strengthening of existing structures, and repair methodology. Further development of repair methods, standards and codes will come in the future to better equip the profession to respond to challenges of aging infrastructure. Sustainability as a rallying call seems a bit “old in the tooth,” although these kinds of trends have long-lasting tails. I sense there is a rising enthusiasm for the word “Resiliency” as the next big theme across the construction industry. Resiliency captures a sense of responsiveness, readiness, and recovery from disaster. Aging is relevant also to the new theme of resiliency because infrastructure has to deliver performance at a level defined by the original design, and aging is a threat that degrades performance over time. So whether the focus is Sustainability or Resiliency, attention must be paid to aging of infrastructure.

21


5 Engineering Education Reform 5.1

How did our education model become stagnant?

I have been a faculty member at the University of Illinois for more than 20 years. For six years of that time, I served as the Associate Head for Undergraduate Affairs, and in that role I oversaw an undergraduate program with 800 students. During those years, I studied how our curriculum had evolved over the many years, and it was sobering to observe just how little change had occurred within my lifetime. I went to college in the 1970s, and our curriculum (course titles, credit, and sequencing) at the time was actually quite similar to the University of Illinois in the 2010s. Almost course for course the same! My study drew me to look back as far as the archives would take me. The civil engineering curriculum was very different 100 years ago. In the late 1800s, about a third of the curriculum was some aspect of surveying, whereas in 1996 we ceased teaching the single remaining surveying course at UIUC. There was a large shakeup in engineering education in the years around WWII. After WWII, it became clear that technical preeminence was essential to national security, and there was a sense of urgency to build a strong educational system with a new commitment to the research enterprise. From 1940-1960 we saw adoption of the “Engineering Science Approach” that increased science and math content considerably, and decreased technician-training aspects of the curriculum. But if you compare the today’s curriculum with that of the 1960s, you’ll see relatively little change. I believe that one of the strongest factors that led to stability and uniformity (and stagnancy) of U.S. engineering curriculum is the accreditation processes used by universities and colleges. The most common accreditation body used by U.S. civil engineering programs is ABET. While the expressed attitudes at ABET seem open to curriculum reform and innovation, the highly structured accreditation process remains an imposing barrier to change.

5.2

Education reform in the US and Europe

Engineering educators are in the early stages of a wave of change that may emerge as the greatest since WWII. There is broad awareness that it is time to revisit the old education model [8]. Bold innovations are percolating through higher education, and much of it is driven by the desire to more effectively reach today’s students who have grown up with different expectations shaped by their experience with computers. These Internet Age students have higher expectations with regard to immediacy, interactiveness, and impact of course materials. Educators have been adopting innovations for course management, video, models, and powerful computational tools like MatLab. Online education is becoming mainstream, and the possibilities for archiving lecture material, pacing delivery to meet student expectations, and self-study exercises has affected on-campus instruction as much as off-campus. Pedagogy is being taken more seriously, and educators are trying to understand learning processes so that teaching can be more efficient with better outcomes. Taken in sum, the wave of innovation is profound, and engineering educators are eagerly challenging assumptions about traditional curricular design. For example, Design is traditionally an activity for upper-level students who have taken all the science prerequisites. No longer. Design is being introduced at freshman level as a motivator and as a way to convey context for the detailed course material yet to come. As another example, educators seek ways to better embrace the liberal arts. Feedback from the profession trumpets the need for broad communication skills and appreciation for the human condition in a global economy. Certainly our coming future demands different skills and knowledge than 50 years ago. Merely mentioning the advances of information technology and biological engineering is enough to persuade one that engineering curriculum content needs to be responsive to changing needs. ASCE has fostered debate about engineering education through its Raising the Bar committee reports. The premise is that engineering, as a profession, needs to exert greater expectations for training as the needs of the profession rise. As shown in Figure 3, other professions have “raised the bar” for training to enter their profession and earn licensure. States that hold responsibility

22


for licensure are considered to hold the key to institutional change, and over time, I think we will see the Masters Degree rather than the Bachelors Degree used as the standard entry-level preparation. Europe has wrestled with their own set of engineering education challenges. The discussion is focused on uniformity across the EU and transferability of educational credentials. There is a direction embodied in the Bologna Process that seeks to ensure comparability in the standards and quality of higher education qualifications. To date 47 nations are participating, indicating that wide consensus is established, and important steps are being taken to examine engineering education and its future. While the emphasis of the Bologna Process is international cooperation, the future of such dialog will reach curriculum, pedagogy, and professional preparedness [9]. Topics like Aging beg for greater professional training and tools. Aging is not a simple onedimensional controlled lab experiment, fully explained by simple math, simple chemistry, simple physics. Rather, Aging is a complex area of study that requires synthesis, multidisciplinarity, design and problem-solving skills, economic sophistication, political savvy, and skills of communication and persuasion. These requirement have profound implications for education of future engineers. This conference on Aging of Materials and Structures, to me, is a call for professionals who may have seen themselves comfortable in “scholarly silos” to rise up to greater awareness of and enthusiasm for their role in impacting public policy. We may enjoy the technical inquiry of the laboratory, but we are called to have broader impact outside the laboratory.

Figure 2 Education Requirements for Professions [10]

6 Summary Aging of infrastructure involves technical challenges, but the engineer and researcher is also called to impact public policy that supports high quality infrastructure that benefits society. In addition, I cannot help but highlight the need for education reform to better meet the changing needs of the profession. My talk today seeks to make several points: •

Communicating ideas about Aging of Infrastructure requires the ability to convey qualitative information (e.g. infrastructure condition) into quantitative metrics. The scientific knowledge underpinning such metrics needs to advance.

•

Infrastructure quality contributes to national competitiveness, and as such, investment in infrastructure is critical to economic vitality.

•

Infrastructure spending on infrastructure is relatively stable over time, but may shift topical focus under political pressure. Structural and long-lasting increases in

23


infrastructure spending have been elusive, even as professional groups lobby in more sophisticated ways. •

Scientists and engineers have a role to play by informing the decision-making process that leads to public policy for infrastructure spending.

•

Aging is a complex topic that involves multidisciplinarity, design and problem-solving skills, economic sophistication, political savvy, and the ability to synthesize and communicate all of these factors.

•

Aging connects well to other current themes in civil engineering: sustainability, resiliency, and repair. All of these themes have implications for educating the next generation of engineers.

•

Aging and other rising needs of the engineering profession call for educators to update curriculum and reform pedagogy. Today’s challenges cannot be met by educational approaches that have changed little in the last 50 years.

7 Acknowledgements I am grateful to the VTT-Fulbright Program for its support of my activities during my sabbatical leave in Finland. I appreciate the collaboration and discussions with my colleagues at VTT, Espoo, Finland that have contributed insight for this paper.

8 References [1]

ASCE, “Infrastructure Report Card,” http://www.infrastructurereportcard.org/grades/, 2013.

[2]

E. Soltas. “Bloomberg View: The Myth of the Falling Bridge” http://www.bloombergview.com/articles/2013-04-08/the-myth-of-the-fallingbridgettp://webmineral.com/data/Cryptomelane.shtml. April 8, 2013.

[3]

S.J. Peterson, “U.S. Infrastructure Spending: How Much Is Enough?”, Urban Land, April 2009.

[4]

P.R. Gregory, “Infrastructure Gap? Look at the Facts. We Spend More Than Europe,” Forbes, http://www.forbes.com/sites/paulroderickgregory/2013/04/01/infrastructure-gap-look-at-thefacts-we-spend-more-than-europe/, April 2013.

[5]

B. Plomer. “U.S. infrastructure spending has plummeted since 2008”, http://www.washingtonpost.com/blogs/wonkblog/wp/2013/05/24/u-s-infrastructure-spendinghas-plummeted-since-2008. May 24, 2013.

[6]

James Pethokoukis, “Actually, America doesn’t have a trillion-dollar infrastructure crisis,” American Enterprise Institute, http://www.aei-ideas.org/2013/11/actually-america-doesnthave-a-trillion-dollar-infrastructure-crisis/, May 6, 2014.

[7]

K. Schwab, The Global Competitiveness Report 2012-2013, World Economic Forum, 2012.

[8]

National Academy of Engineering, Educating the Engineer of 2020: Adapting Engineering Education to the New Century, 2005.

[9]

Engineering Council, “Bologna Process,” http://www.engc.org.uk/education--skills/bolognadeclaration, 2014.

24


[10]

American Society of Civil Engineers, “Raise the Bar,” http://www.raisethebarforengineering.org, 2014.

25


26


Biomimetic materials: long-lasting and self-repairing Thomas Speck1,2,3*, Marc Thielen1, Olga Speck1,3 1

University of Freiburg, Freiburg, Germany Freiburg Materials Research Centre (FMF) and Freiburg Institute for Interactive Materials and Bioinspired Technologies (FIT) 3 Networks of Competence Biomimetics and BIOKON e.V.

2

Abstract: Using R&D-projects of the Plant Biomechanics Group Freiburg as examples, the interdisciplinary approach and the development of long-lasting, self-healing, (self-)adaptive and energy-dissipating biomimetic materials and structures is presented. Keywords: Biomimetic materials, self-healing, long-lasting, energy-dissipating

Introduction

Biological ‘constructions’ often possess outstanding mechanical properties that are mainly based on a complex hierarchical structuring including a multitude of interfaces on the various structural levels. They are not built of a huge variety of constitutive materials as typically used in traditional engineering but are characterized by a limited number of basic chemical components and show a large variety of micro- and nanostructures [1,4]. The extremely efficient biological ‘materials design’ is brought about by the evolution of hierarchical structures covering more than ten orders of magnitude and being well adapted to the requirements at each level of hierarchy. Many biological materials possess in addition to their fascinating mechanical functions ‘self-x-properties’ (self-organization, self-cleaning, self-healing...). The combination of the respective properties allows them to interact very efficiently with their respective environment. These biological solutions are cost- and energy-efficient, multi-functional, longlasting and environmentally friendly, and with several billion test runs, they have surely stood the test of time.

Figure 1 Process sequence in Technical Biology, Biomimetics and Reverse Biomimetics [from 1, 4]

*

Corresponding author affiliation and e-mail address

27


During the last decade, novel sophisticated methods for quantitatively analysing and simulating the form-structure-functions-relationship on various hierarchical levels made possible new fascination insights in multi-scale mechanics and other functions of biological materials. On the other hand, new production methods allow for the first time the transfer of many outstanding properties of the biological role models into innovative biomimetic products. Especially polymers proved to be very suitable materials for the production of innovative biomimetic materials and structures [4, 8, 9].

Results and Discussion

Based on current R&D-projects of the Plant Biomechanics Group Freiburg as examples, the interdisciplinary approach in the development of hierarchically organized biomimetic materials and structures is presented. These examples also demonstrate the process-sequences of biomimetic research including the quantitative analysis of the biological concept generators, the abstraction process and – finally – the implementation in novel bio-inspired materials and structures (see Figure 1) [1, 4].

Figure 2 Bird-of-paradise flower (Strelitzia reginae, left) the biological role model for the elastic façade shading system Flectofin® of which a functional demonstrator is shown in the right (build by ITKE Stuttgart, ITV Denkendorf in collaboration with Clauss Markisen).

The presented examples include four bio-inspired developments:

(1) Biomimetic façade shading systems, as for example the Flectofin®, which is inspired by pollinator induced elastic deformations during the bird-mediated pollination process in the bird-of-paradise flower (Strelitzia reginae). These biomimetic façade shading systems are long-lasting and need no or very little maintenance as their function is based on elastic deformation processes avoiding localized hinges (see Figure 2) [2-4].

(2) Self-repairing coatings for pneumatic structures, drastically reducing or entirely stopping air leakages after puncturing inspired by self-sealing and self-healing processes in lianas of the genus Aristolochia (see Figure 3) [4-6, 8, 9], and self-healing elastic polymers for sealings exposed to high cyclic loading inspired by self-healing processes occurring in latex containing rubber plants, as e.g. the weeping fig (Ficus benjamina) [3, 7-9]. 28


Figure 3 Habitus and stem cross-section with repaired fissures in the outer strengthening tissue ring of the twining liana Aristolochia marcophylla which served as a role model for self-repairing foam coatings for pneumatic systems. Custom-made set-up for testing the repair efficacy of coated and uncoated membranes.

(3) Biomimetic shock-absorbing fibre-reinforced gradient foams inspired by pomelo peels (Pomelo maxima) (see Figure 4) [4, 10]. These foam structures can be produced by using various material groups, as e.g. metals or polymers, and can be used for different types of technical protection structures including crash boxes for cars, transport vessels for hazardous goods, or protection wear and helmets [4, 11].

(4) Bio-inspired shock-absorbing pallets for the transportation of sensible goods inspired by hedgehog and porcupine spines and bamboo culms [4, 12].

Figure 4 Different hierarchical levels of the pomelo fruit and fruit peel (Citrus maxima).

29


Acknowledgements We thank our colleagues and their research groups for excellent cooperation within the projects presented above, especially Andreas Bührig-Polaczek (RWTH Aachen), Claudia Fleck (TU Berlin), Rolf Luchsinger (EMPA Dübendorf), Rolf Mülhaupt (University of Freiburg and FMF) and Anke Nellessen (Fraunhofer Institute UMSICHT Oberhausen). For financial support we are grateful to (1) the German Research Foundation (DFG) for funding within the priority program 1420 “Biomimetic Materials Research: Functionality by Hierarchical Structuring of Materials”, (2) to the German Federal Ministry of Education and Research for funding within the scope of the programme BIONA and the within the scope of the funding programme ‘Ideenwettbewerb: Bionik – Innovationen aus der Natur’, and (3) the EMPA Dübendorf for additional funding.

References [1] Speck T and Speck O (2008) Process sequences in biomimetic research. In: Brebbia, CA (ed) Design and Nature IV. WIT Press, Southampton, pp 3-11. [2] Lienhard J, Schleicher S, Poppinga S, Masselter T, Milwich M, Speck T and Knippers J (2011) Flectofin: a nature based hinge-less flapping mechanism. – Bioinspiration and Biomimetics 6: DOI:10.1088/17483182/6/4/045001 [3] J. Knippers & T. Speck (2012): Design and construction principles in Nature and Architecture. Bioinspiration and Biomimetics 7: DOI:10.1088/1748-3182/7/1/015002 [4] Masselter T, Barthlott W, Bauer G, Bertling J, Cichy F, Ditsche-Kuru P, Gallenmüller M, Gude M, Haushahn T, Hermann M, Immink H, Knippers J, Lienhard J, Luchsinger R, Lunz K, Mattheck C, Milwich M, Mölders N, Neinhuis C, Nellesen A, Poppinga S, Rechberger M, Schleicher S, Schmitt C, Schwager H, Seidel R, Speck O, Stegmaier T, Tesari I, Thielen M and Speck T (2012) Biomimetic products. In: BarCohen Y (ed) Biomimetics: nature-based innovation. CRC Press / Taylor & Francis Group, Boca Raton, London, New York, pp 377-429. [5] Busch S, Seidel R, Speck O and Speck T (2010) Morphological aspects of self-repair of lesions caused by internal growth stresses in stems of Aristolochia macrophylla and Aristolochia ringens. Proceedings of the Royal Society London B 277: 2113-2120. [6] Rampf M, Speck O, Speck T and Luchsinger R (2013) Investigation of a fast mechanical self-repair mechanism for inflatable structures. International Journal of Engineering Science 63: 61-70. [7] Schüssele AC, Nübling F, Thomann Y, Carstensen O, Bauer G, Speck T and Mülhaupt R (2012) Selfhealing rubbers based on NBR blends with hyperbranched polyethylenimines. Macromolecular Materials and Engineering 297: 411-419. [8] Speck T, Bauer G, Flues F, Oelker K, Rampf M, Schüssele AC, v. Tapavicza M, Bertling J, Luchsinger R, Nellesen A, Schmidt AM, Mülhaupt R and Speck O (2013) Bio-inspired self-healing materials. In: Fratzl P, Dunlop JWC and Weinkamer R (eds) Materials Design Inspired by Nature: Function through Inner Architecture, RSC Smart Materials No. 4, The Royal Chemical Society, London, pp 359-389. [9] Speck T, Mülhaupt R and Speck O (2013) Self-healing in plants as bio-inspiration for self-repairing polymers. In: W. Binder (ed) Self-Healing Materials. Wiley-VCH, Weinheim, pp 69-97. [10] Thielen M, Schmitt CNZ, Eckert S, Speck T and Seidel R (2013) Structure-function relationship of the foam-like pomelo peel (Citrus maxima) - an inspiration for the development of biomimetic damping materials with high energy dissipation. Bioinspiration and Biomimetics 8: DOI:10.1088/17483182/8/2/025001 [11] Fischer SF, Thielen M, Weiß P, Seidel R, Speck T, Bührig-Polaczek A and Bünck M (2014) Production and properties of a precision-cast bio-inspired composite. Journal of Materials Science 49: 43-51. [12] Masselter T, Milwich M, Monnerat H, Scharf U, Hartel M and Speck T (2008) Bio-inspired solutions for technical problems: biomimetic cable entries and schock-absorbing pallets. In: Brebbia CA (ed) Design and Nature IV. WIT Press, Southampton, pp 51-58.

30


Forever young or ageing naturally? Rob P.J. van Hees 1,2, Silvia Naldini 2 (1) TNO Technical Sciences, Delft, The Netherlands (2) Delft University of Technology, Delft, The Netherlands Abstract: Age and ageing can be felt as negative occurrences. For monuments however, old age is traditionally considered to be a positive quality. Without a certain age the nomination of monument hardly applies. Ageing can be seen as the work of time, which has always been valued: ageing was sometimes even artificially induced in the past. In this paper we will discuss the meaning of ageing in monumental buildings. The fact that in the case of interventions in monuments a perpetual service life is strived for, but restoration ethics clearly put limitations on what can be done, can lead to dilemmas and can make it difficult to take decisions. Cases will be discussed to derive at some criteria to base interventions upon, seeking a balance between mere preventive conservation and rejuvenating practices. Keywords: monuments, ageing, preventive conservation

1 Introduction Ageing is often felt as a scaring and negative process. For monuments, however, old age is traditionally a positive quality: without being of a certain age, a building will hardly be listed as a monument. Ageing and patina are concepts whereupon many discussions in the field of architectural conservation focus. One of the first questions to be answered concerns the meaning of ageing and patina. Ageing of monuments is a natural process, which may result in the formation of patina. A definition of ageing is ‘the process of growing old or developing the appearance and characteristics of old age’ [1]. Patina can be also considered the result of a chemical transformation of the surface (only) of the material. As a general term, patina refers to the change in an object's surface resulting from natural aging [2]. The removal of the patina needs careful consideration, as it can lead to the obliteration of the traces of time. Ageing could be accelerated by exogenous decay processes – like salt or frost damaging mechanisms - affecting the cohesion of the materials of a monument. A decay process can also develop inside the ancient materials, enhancing their ageing dramatically, without any environmental aggression. This is the case of creep, affecting the structural strength of the building. Ageing due to damaging processes is obviously unwanted, and should be hindered. One could further wonder whether an aged look could even be desired, and patina could be artificially created. Talking about paintings, the discussion on the wanted patina is centred on the procedures assumed to be used by old masters to give their colours homogeneity, to start with Apelle’s ‘atramentum’ reported by Plinius [3]. Stone statues and façades could be treated as well, which should be considered when intending to clean. In the case of a painted masonry, the colour will also alter as a result of ageing, which means that different materials will show different signs of ageing. Some cases of ageing will be commented aiming at discussing the meaning of ageing and extracting some values to refer to when confronted with the dilemma: allow ageing or intervene? and also with the extension of the intervention. We will start with a ruin, an extreme case of ageing, to further tackle the problems of the extent the thoroughness, and the envisaged result of cleaning and restoration actions. How to approach problems of restoration is a complex matter. In his Teoria del Restauro, first published in 1963, Cesare Brandi, director of ICR (Central Institute of Restoration) founded in 1939 in Rome, faced most problems, which have been debated up to the present day. It is therefore interesting to recall some of his principles in our discussion of the cases, to see whether they can be referred to, in the light of the present quest for a new balance between young and old.

31


2 The castle ruin of Asten (The Netherlands) A ruin is any object witnessing human history, even though not having preserved its original form, which may be even hard to recognize. In the case of the Castle of Asten (the Netherlands) [4], only a ruin remains, which however shows interventions made in different times. The castle dates back to 1430 ca. and went through transformations in the course of time, the latest concerning the fantasy interventions of baron Clemens van Hövell tot Westerflier in the XX cent.. The current attitude is to keep each addition to the original body, even the most recent one, dating to the first decennia of the XX cent., because it is perfectly recognizable as such and provides for added historic value to the object. The initial form of architecture at the moment of its creation, should be no longer re-created, but all its components should be maintained as expression of different techniques and craftsmanship belonging to different historic phases. From the hand moulded brick to the concrete, from the lime mortar to mortars containing cement, all materials and techniques are individual expressions of historic periods, and all have withstood time in a different way. The orange bricks, originally meant for interior spaces and presently exposed to the environmental agents, suffer from the exposure, other than other than the darker bricks. The masonry of the XV cent. construction still shows a rather good state of preservation. Some maintenance needs to be done making re-pointing and solving the problems related to water penetration. Rather than the preservation of the original form and aesthetical value of the castle, the preservation of its historical value can be achieved by preventing restoration, that is to say trying to maintain the status quo [5]. Also the environment, the countryside the castle lies in, if kept untouched, will provide the monument with the natural space, which is necessary for its appreciation. A new aesthetic value will derive.

Figure 1 In the ruin of the castle the action of time can be seen, as well as the traces of historical events, including the grenade which hit the construction in 1944, starting the final phase of the castle, now no longer residence of the owner

The Foundation in charge of the building has decided to leave the ruin as it is, without integrations, and limiting the interventions to the essential, like substituting the pointing where necessary, which can be described as the conservation of the materials in their current state. Even the vegetation is meant to be maintained, that is to say to be left growing onto the walls, only preventing it from becoming too heavy and inducing local collapse. This is an interesting approach, which is not aimed at guaranteeing the ruins eternal life and can be even called romantic. We would maybe act differently if the ruin was an old Roman excavation. The point would be then, how to preserve it? Preserving the history involved in the ruin could also suggest a solution like the creation of a structure enclosing the

32


ruin, as was done in the case of the Museum of the Roman Baths in Heerlen (the Netherlands). The ruin would be separated from its historic environment and inserted in a new, artificial one. This is the alternative to Brandi’s ‘preventive restoration’, as pointed out by Urbani [6], who concludes noting that Brandi’s theory is rather than a conclusive work, an opening to the future of restoration. The case of the ruin is an extreme one and we are often confronted with the problems concerning the preservation of the historical and aesthetical value of a monument still having its unity, even though its original form could have been altered by interventions. Keeping the building with the traces of the passage of time is a matter of careful analysis of each case. Only a thorough study of the monument and the message attached to its forms, could guide a good intervention.

3 San Marco’s Campanile in Venice (Italy) As mentioned above, certain damage mechanisms can enhance ageing dramatically. The following case of creep is pointed to as an example of ageing, finally leading to loss of the original form of the object. The campanile (bell tower) of San Marco in Venice was constructed between 1156 and 1173. If we compare the sudden collapse of the San Marco bell tower on 14th July 1902, with more recent disasters like the Pavia tower, for which creep has been assessed as the major cause of failure, we may assume that a similar phenomenon may have occurred in Venice with San Marco’s bell tower. Creep is a form of ageing. Within a few years after the collapse it was decided to reconstruct the tower, and a new one was erected with the same proportions and materials as the one that collapsed (‘as it was’ and ‘where it was’) in 1912.The order of magnitude of creep of historic masonry, can be considerable and may therefore be a serious problem [7]. Creep in compression, due to dead loads, generally leads to (deep or trans-sectional) vertical cracks. This type of damage (passing through cracks) is typical of slender structural elements like stone or brickwork columns and piers and of heavy but tall structures like towers (and heavy structures as to be found for example in ancient churches). It may develop in a relatively short or very long time, depending on the brittleness of the material and is due to the creep behaviour of the material when stressed beyond the elastic limit. Cracks can propagate very slowly for decades or even centuries, but in the end, if the phenomenon is not stopped, the element or structure can suddenly collapse.

33

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands Figure 2 San Marco’s bell tower, assumed behaviour during collapse. Photo taken from a shop window in Venice, 2004

Figure 3 San Marco, Venice, view of the church with the reconstructed bell tower San Marco’s tower can be considered as a fundamental part of a historic site. The loss of this tower meant therefore that something similar had to come in its place. Also Brandi points at this reconstruction as a way of re-establishing a unity, however arguing that what was originally there and was lost cannot be revived: a copy is both a historical and an esthetical falsification, unless made for didactical purposes [5]. In the case of Venice the essential missing element was a vertical body. Nowadays’ tower is, except for the materials, maybe too much a copy of the lost one, which points at a re-creation. One could however argue that the time elapsed between collapse and re-erection was so short and the shock on the involved community so big, that this reconstruction might be justified.

4 The colour of the façade: the restoration of St. Peter’s in Rome (Italy) and the Royal Palace in Amsterdam (the Netherlands) The problem of where to stop in a restoration process was faced in the following cases of cleaning and restoration. The relevance of the buildings and their representative function explain the broad discussion originated. Within the framework of the activities around the (approaching) Jubilee Year 2000 having as object Rome and its buildings, the façade of St. Peter’s was restored. The surfaces were not only cleaned but also painted in the colours originally chosen by Maderno, the architect who extended the basilica designing nave and façade. Differently from many other cases, the colours were the means to almost theatrically create a depth in the façade, letting the white columns emerge, whereas the surface behind, in the nuances of tobacco - ochre, recedes. Maderno solved thus the problem of the deviation from the design of Michelangelo’s church and façade re-establishing the form of the temple in an almost Baroque way. In this case, the restitution was not merely meant to go back to the original colour given to the materials in contrast with the ageing process, but aimed at finding back the unity and meaning of the creation. Being such a prominent and symbolic building, the restoration of St. Peter’s divided scientists and general public into two fronts of supporters and opponents of the chosen approach. The

34


party against the restoration would use the argumentation that going back in time is not only historically but also aesthetically incorrect, even though it could be demonstrated that traces of brown colour had emerged after removing the dirt layer, brushes had been bought at the time of Maderno, and a painter had been paid for colouring the wooden model of the façade to show how the final effect would be. Too many façades in Rome, was said, have been restored in the last decennia going back to the original light, pastel colours, instead of keeping the ochre-red shade, which has become the dominant colour of Rome in the course of time. Moreover the opponents found the context important, and that the colour of the façade should be matching the colours of the buildings of the area. A last problem was psychological or maybe emotional, and consisted in the difficulty people have to adjust to changes. All mentioned elements needed to be taken into account and weighed against the elements in favour to reach a well-balanced and well-argued decision.

Figure 4 Façade of St. Peter's before restoration, 1985 (source: http://www.kofc.org)

Figure 5 Front façade St Peters after intervention (April 2000)

35


The Dutch Government Building Agency aims at achieving well-conceived restorations based on the respect for the traces of time [8]. Facing the problem of the removal of the patina formed on natural stone, an interesting approach was chosen for the restoration of the façade including marble reliefs of the main pediment, of the Royal Palace in Amsterdam, originally built as the town hall of the city (1648). Like St. Peter’s for Rome, the Palace is a prominent building for Amsterdam and the Netherlands, as testified by the genuine amazement of contemporaries like the poet Vondel, who named it ‘the eighth world wonder’. The masterpiece of the great Dutch architect, Jacob van Campen, the building showed his interpretation of the Classical architecture, and was the symbol of the power of the city of Amsterdam during the Dutch Golden Age, and is presently the most important Royal Palace of the Netherlands, which explains why its conservation has become the subject of a widespread controversy. The final result can be described as a well-argued restoration, aware of the importance of the perception [9]. One important characteristic of the Palace is that its façade is made of sandstone, in a country where, due to the scarce availability of natural stone, the traditional building material is brick and only the most relevant buildings are cladded in natural stone. In Carrara marble were carved the sculptures of the pediments of both front and rear façade. Both natural stone types had undergone ageing, resulting in a patina on the stone and local staining. As far as the relief of the front façade is concerned, this was considered disturbing and hindering a thorough appreciation of the features of the work of art. The condition of the façade was perceived as shabby and neglected, within the context of the Dam square and the surrounding buildings. The main problem was how to clean the stone. Going back to the original colour of the stone, removing all traces of time, would mean to consider time as reversible, forgetting the historical instance [5]. Besides, it was not clear how Van Campen had intended to make the stone blocks look more homogenous, whether he wanted them to be painted, following a common practice, or oiled, as it was proven to have been the case in 1689. The blocks had originally been tooled and this was a relevant aspect to be maintained, because all parts originally formed a unity. Also the cleaning technique was a matter of discussion, considering that the materials used were different, even the sandstone was of two types, Obernkirchen and Bentheim, which have different (ageing) characteristics, and the technical state of conservation of the stone could be critical [9]. The most suitable technique was chosen on the basis of a study and tryouts.

Figure 6 Royal Palace Amsterdam, front façade and surrounding buildings (2006 - before restoration)

36


Figure 7 Royal Palace Amsterdam, front façade after restoration (Feb. 2014) It was decided to limit the intervention to a minimum, and not to remove all traces of patina, as not only the colour but also the texture would have been otherwise affected, and look artificially and awkwardly ‘young’. Therefore, it was considered necessary. Guiding principle was that any intervention needed to be reversible and its effect should not vary in relation to the stone it was applied upon [8]. In the case of poor quality of some of the stone blocks, repair work with a repair mortar was carried out; re-pointing was done when necessary. Back to the colour, the contrast was sought between sand stone walls and marble decoration of the main pediment, that had been regularly painted with a limewash until the XVIII cent.. The cleaning and local consolidation have been preceded and followed by a thorough documentation campaign aiming at testifying the state of the materials before and after the interventions. Concluding, the idea of undisturbing was introduced, concerning the extent of the cleaning process. Leaving the patina was unacceptable, however too light a colour of the stone would have been disturbing. The removal of patina was therefore not integral, leaving some traces of its respectable age, and stone parts being too light (for example due to later replacement) were made a bit darker, making the overall aspect of the façade better readable, but still ancient in aspect and homogenous in colour. Most outspoken was the cleaning of the pediment in Carrara marble, for which readability was the main aim and which is now again looking ‘as new’. Still the problem was not to go back to the origin of the creation, but to make it readable. Each case is different, still we can say that a general well-considered approach should not be aimed at going back to the moment of the creation of the object, cancelling the time which separates us from then, but should also not neglect the unicity and significance of the monument. There will be cases, like the façade of the St. Peter’s, where the final result could be felt as rejuvenating, even though, the restitution of colour means restitution of depth, of a vision on architecture as modern, for the time, as it could be. Already in 1978, Brandi agreed upon the necessity of cleaning the façade of St. Petronius in Bologna [10] not only justified by the damage to the materials, but especially by the fact that the deposition layer on the surface hindered the appreciation of the two main colours of the stone used in the façade, which had become monochrome.

5 Discussion and conclusions Ageing and service life are closely connected with all buildings and building materials. Whenever a building does not fulfil its function anymore, or does no longer possess esthetical qualities, the end of

37


the service life approaches and demolition is about to come. A monument is a different case, we want to keep it and we tend to strive for a kind of perpetual service life [11]. The question is then: may the monumental building show its age? And the answer to that is not ‘yes’ or ‘no’, it is not black or white. If all buildings and all building materials age, then also monuments do. The difference is that in ordinary buildings it is easy to decide on either maintenance and face-lifting or in the most extreme case on demolition, depending on the taste and the wish of the owner. In case of (listed) monuments, instead, there are many more people that may decide on a suitable approach, and there is no rule, at least no fixed rule; only demolition is normally impossible or at least only allowed when safety is severely endangered. The opposite, that is to say ‘reconstruction’ or rather a complete ‘re-creation’, however, is also possible, even though severely debated, as mentioned in the case of San Marco’s bell tower. ‘Don’t touch any of my wrinkles it took me ages to get them’ actress Anna Magnani told her makeup artist. Growing old gracefully [12]. This has been the attitude shown by Dutch and other EU heritage authorities towards ageing of monuments for a long time. Nowadays things tend to change as awareness has grown that a monument is at least not fully comparable with the human body. The most important restoration of the past ten years in the Netherlands, that of the Royal Palace in Amsterdam, introduced concepts like undisturbing, for making too light parts a bit darker and too dark parts a bit lighter, thus leaving some traces of its respectable age, while making the overall aspect of the façade more equilibrated and better understandable and consequently enjoyable. More outspoken was the cleaning of the pediment in Carrara marble, for which readability was the priority and which is now looking ‘as new’, showing that even within the same monument, different parts could require a different approach The choice needs to be supported by a qualified and well balanced judgement, allowing to act respectfully towards old age, because, as Brandi pointed out, no universally applicable solutions exist.

6 References [1] http://www.thefreedictionary.com/ageing, accessed Dec. 2014 [2] http://www.cartage.org.lb/en/themes/arts/scultpureplastic/UnderstandingSculpture/Patina/ WhatisPatina/WhatisPatina.htm, accessed Dec. 2014 [3] Plinius, Naturalis Historia, XXXV, 41-43 [4] http://kasteelasten.nl/historie/ (accessed 21-01-2014) [5] Brandi C. (1977, 1963), Teoria del restauro, Torino, Einaudi, p. 30-37 [6] Urbani G. (2000), Intorno al restauro, ed. Zanardi B., Genova-Milano, Skira, pp. 69-75 [7] van Hees R.P.J., Binda L., Papayanni I. & Toumbakari E. (2007), Damage assessment as a step towards compatible repair mortars, in: C. Groot, G. Ashall & J. Hughes eds, Characterisation of Old Mortars with Respect to their Repair – RILEM report 28, pp.105-150, isbn: 978-2-91214356-3 [8] Bommel van B. (2008), De gevels van het Koninklijk Paleis Amsterdam, P.C.E., 4, 13, SDU Publ. Sept. 2008 [9] Bommel van B. (2013), Terugblik op een geslaagd project. De restauratie van het Koninklijk Paleis Amsterdam, KNOB, 1, pp. 12-23 [10] Brandi C. (1979), Intorno a due restauri eccezionali: la facciata di San Petronio e la “Santa Cecilia” di Rafaello, ‘Bologna incontri’, X, 1979, n.10, pp. 24-27 [11] Herdis A. Heinemann, Rob P.J. van Hees, Timo G. Nijland, Hielkje Zijlstra, (2010), The challenge of a perpetual service life: conservation of concrete heritage, in: K. van Breugel, Guang Ye, Yong Yuan eds., Proceedings of 2nd International Symposium on Service Life Design for Infrastructures, pp. 1067 – 1074, RILEM Publications SARL [12] Symposium ‘Growing old gracefully’ (Gracieus verouderen), RDMZ, Nov. 1999

38


UNDERSTANDING AGEING

39


Development of a Database for the Restoration Mortars – von Konow DB Esko Sistonen, Petri Mutanen, Fahim Al-Neshawy Department of Civil and Structural Engineering, Aalto University, Espoo, Finland Abstract: Before the 1950s lime mortars were the primarily mortars used with the stone buildings. In traditional construction, lime was a fundamental material like timber, stone or brick. Lime was used for making building mortar, external render, internal plaster, lime wash and solid floors. Stone architecture was based on lime mortar and its visual aspect made up of the signs of the professional hand work. It is reported that the damage of renovated historical structures has been intensified in recent decades. The choice of suitable mortar is influenced by the need to maintain the historic interest of the building, its structural integrity, and its appearance from the historical perspective. From a practical point of view, there is a lack of information on repair mortars properties and the use of suitable mortars for the restoration. For this reason, a research is carried out in order to produce fundamental guidelines for the design and implementation of repairing mortars for the restoration. This research includes (i) research knowledge upon behaviour of mortars in Nordic exposures, (ii) research for different deterioration mechanisms of mortars and substructure, (iii) continuity of the on-going field studies of mortars for deeper understanding and knowledge, (iv) development of new mortars for restoration, and (v) development of guidelines combining appropriate materials with different architectural and historical periods. The objective of this paper is to develop a database based on the accurate field research and laboratory studies of mortars initiated by PhD Thorborg von Konow. The goals of the “von Konow DB” database are to (i) collect the essential data of the condition and the performance of the historic mortars, (ii) store and update these data effectively, (iii) allow sophisticated search strategies, (iv) produce detailed reports automatically for the historic mortars and (v) enable data transfer to other software for further analysis. The design process of the “von Konow DB” database includes the system analysis, the logical and physical design, and then the final system implementation ant testing. This database will allow access to the data needed for the selecting of suitable repairing mortars for the restoration purposes. The results are directly utilized in guidelines combining appropriate lime mortars with different architectural and historical periods. Keywords: Mortars, Deterioration mechanisms, Ageing, Restoration, Database

Introduction

Stone architecture was based on lime mortar and its visual aspect made up of the signs of the professional hand work. It is reported that the damage of renovated historical structures has been intensified in recent decades. The choice of suitable mortar is influenced by the need to maintain the historic interest of the building, its structural integrity, and its appearance from the historical perspective. In this paper presents an overview of a research project which is carried out in order to produce fundamental guidelines for the design and implementation of repairing mortars for the restoration.

Basic requirements for restoration mortars

Many of the basic requirements for restoration mortars are similar to other materials used in restoration. But on the other hand the exact methods and recipes are constantly debated among the experts. Furthermore it must always be remembered that every case is different and therefore decisions should be made in situ after surveys on the characteristics of the case. However there are some general principles that should be taken into account. In real situations all 40


of them cannot be completely fulfilled, because they may be partly contradictory with each other. [1]

Repair is always a compromise especially when there are not enough resources, but in restoration there are also different opinions on the basic philosophies of restoration. Different experts may have the same information, but they recommend different repair actions. Sometimes it is better to do nothing and let the structure age in a natural way. In addition decisions made today are based on valid arguments, but in the future they are not necessarily approved. For example replacing old mortar with new materials is always a risk culturally and sometimes also technically. The most important requirement is that the new mortar should function technically with the old mortar. It should also look like the old mortar for aesthetic reasons. New mortar should not be harder or weaker than the original in order to maintain mechanical properties of the structure. Otherwise the restored structure may be damaged. This happens rather frequently when cement is used in restoration. This concerns has led to favouring more traditional mortars with relatively low Young's-modulus and other properties similar to mortars in old structures. [1].

New kinds of mortars can be used, when there is enough knowledge of their behaviour with the historic mortar. Because mortar is usually the most vulnerable part of the structure, it is likely to go through several repairs during the buildings service life. Future restorations should be taken into consideration. That is why removing the new mortar must be possible without harming the structure. In other words the restoration should be reversible and also the service life should not shorten because of the restoration itself. Repair mortar should be flexible and form a bond, which can deal with the loads and movements of the masonry. [2]

Main philosophical principle behind the restoration work is the authenticity, which is a layered concept. It includes ideas such as: form, feeling, material and workmanship. Restoration mortars must fulfil these requirements in a way that ensures compatibility between repair material and the old one. [3] Even if the recipe of the mortar is adequate for the task, the execution of the work itself or the mixing of material may result in failure. Mortars require craftsmanship, which is different than with modern materials and is thus hard to find. The craftsmen should be trained to master the specific techniques of restoration or otherwise work might be even harmful for the old structure. [4]

The causes of deterioration on restoration mortars

Although mortars are usually the most easily deteriorating materials in the structure they may last for centuries, but eventually the will deteriorate. The causes can be divided in five categories: environment, materials, design, workmanship (or construction) and maintenance. Moisture is perhaps the most common reason for deterioration, but there are others like salt supply, air pollution, variation and extremes in temperature, exposure to fire, dynamic loads and soil settlement. Especially freeze-thaw cycles are dangerous, if there is too much moisture in the mortar. [5]

Deterioration may be caused by the material itself. Composition of the mortar is not always optimal or the structural material may harm the mortar. If at least one of the materials contains salts, problems are likely to occur. In the design phase wrong kinds of materials might have been chosen and the detailing drawn poorly. Also wrong kind of design of the restoration can cause minor or major deterioration. Then there are factors, which are result of the construction phase or maintenance. Curing of the mortar and maintenance are neglected quite often. Sometimes they are caused by insufficient knowledge or maintenance programme. [5] Deterioration is unavoidable, because the environment will in the long run have harmful effect on the mortar. That is why von Konow has highlighted the viewpoint that all deterioration is basically humans fault, because structures should be designed and maintained in a way that preserves them from severe damage. Sometimes the owners just have not had the knowledge or the resources to protect mortars from environment or from their own effect. Nowadays new 41


research is starting to find out how to make as good mortar as the old masters did, but it should be reminded that centuries old mortars are exceptions –– most old mortars have already been repaired before modern times. [6] and [11]

In general, It is possible to list the causes of damage in mortars and plasters as atmospheric effects, effects of use and production conditions and destructive effects of restoration. However it is possible to group the most commonly observed causes of damage in two basic groups as shown in Table 1. [8] Table 1. Some factors that cause damage on original mortar and damage types.

The destructive effects of the environment Factor

Effect

Damage Type

Acidic waters (with CO2, SO2) that come with rain or snow The continuity of freezing/thawing cycles.

Dissolve the carbonates of lime binder.

Adhesion and Cohesion features of the mortar is decreased. Aggregates are decomposed. Leads to the dissolution of the mortar.

Exposion to extreme amount of water vapour (in case of fire). Sea water, air pollution, use of dirty material. Formation of plants

Existance of organic growth Factor Using more cement than lime

Salts that may come from the cement Adding synthetic resin, (if it is too much)

The bonds of the mortar among the binding aggregates are dissolved. The critical water vapour content the mortar can carry is exceeded. Anionic salt crystals i.e. Chlorur, Sulphates and Nitrates are formed. Especially some plant roots lead to the dissolution of the mortar. With the formation of insects, the binding quality of the mortar is reduced.

Leads to the hanging of the mortars in folds through decomposition.

Decomposition of the mortar, deep cracks and draping of the mortar are observed. Biological decay, colouring of the mortar and dissolution. Microbiological decay and dissolution of the mortar.

The destructive effects of the repair mortars Effect

Damage Type

Formation of highly stiff mortar, cracking

Shrinkage cracks and diffusion of water through cracks, drapings due to different work The salts cause the efflourescence and lead to internal stresses. Dissolution in the form of shells on the surface of the mortar

Efflourescence on the surface of the mortar. The water and vapour permeability regime of the original mortar is deteriorated.

Behaviour of restoration mortars in Nordic exposures The climate is one of the factors that determine what kinds of mortars are used. In the Nordic countries the winter conditions are usually dangerous to the mortars, particularly because there are occasionally freeze and thaw cycles. Only the mortars that are used inside do not have to be frost resistant, but practically in all restoration sites there is a need to produce freeze-proof materials. Old mortars have already survived several winters in a good condition and thus their characteristics are a good hint for contemporary restorers. [6]

Testing the old mortars scientifically has given a lot of useful information on frost resistance. It depends mainly on the pore structure, but also on the cracks in the mortar, because they increase the capillary transportation. When the pores and cracks reach the critical degree of water saturation, freezing stars to deteriorate the structure. Freezing and thawing makes this more detrimental. It is most common in the south and west facades due to higher level of solar radiation during the short winter days. [4] 42


When freezing water starts to expand it causes pressure and eventually cracks will appear. After that even more water can penetrate the structure and accelerate the deterioration process. In the 19th century it was rather common to add crushed brick to mortar. It made the material more porous and frost-resistant. Moisture problems and insulation are in our climate quite often reasons for structural problems. Mortar should be breathing material, especially when it is applied in wooden buildings [4]. If mortar is based on cement permeability is usually poor, which is yet another reason to favour more traditional mortars. Hydraulic mortars are in many cases good enough, but not as good as non-hydraulic ones. [7]

Condition assessment of the restoration mortars

Restoration mortars used in historical buildings provide important helpful information about the building technology of their historical period and they are as important as historical documents. Therefore, the evaluation analyses of original mortar and plaster during the restoration should be made based on a scientific base. The work may require including experts with various professions such as art historians, restorators, physicians, chemists, biologists, engineers and architects within the same team during the analyses process.

The purpose of condition assessment mortar is to determine their physical, chemical and mechanical properties. Therefore, it is required to provide the information that explains the current condition of the material as well as the factors that have led to the formation of the current situation of the material used in the building. An experimental method that can be used in evaluating the analyses of restoration mortars is shown in Figure 1. This method is composed of four basic phases, (i) visual analysis and documentation, (ii) experimental research, (iii) evaluation of experiments made in experimental research in order to produce the repair mortar, (iv) decision making on the appropriateness of the repair mortar.

Figure 1. Overview of the experimental method used in evaluating the restoration mortars.

The main objective during the visual analysis and documentation phase would be to map and define areas where degradation occurs by gathering readily available information of the mortars. All of the mortars damage could easily be detected visually by inspecting the mortars. After mapping and determining the causes of mortars damage, their pictures should be taken in the visual analyses.

Further investigations can then be divided into non-destructive and destructive (laboratory) tests. Non-destructive (in-situ) tests are conducted to provide information about the physical and mechanical properties of the mortar in order to determine the level of deterioration. Nondestructive tests also constitute a sub-knowledge accumulation for the required laboratory tests. Examples of the non-destructive methods could be used for the determination of (i) the amount of 43


water absorption, (ii) the amount of humidity and (iii) the hardness of mortars. Ultrasound and endoscopic examination of the mortars could be used to determine the homogeneity, micro-cracks and the thickness of the mortars layer.

Destructive test methods are used in order to design the mixture ratios of the original mortar and to find its current physico-mechanical properties. This information will help producing the appropriate repair mortar. In order to conduct these experiments in a laboratory, samples should be taken from certain parts of the building in adequate amount and dimension. Those samples are collected from different parts of the building which are thought to be constructed or repaired in different time periods.

In the evaluation phase of the visual and experimental analyses, a new mortar which is similar to the original mortar in appearance is designed base on the quantitative and qualitative results of the mixture ratios and the physico-mechanical properties of the original mortar. The repair mortar produced is evaluated and whether it is an appropriate mortar or not is decided in the decision-making phase. Finally, the mixture ratios are prescribed for the application. [8], [9] and [12]

Design of the restoration mortars database – von Konow DB

In this paper the authors present an overview of the on-going development of the restoration mortars database – von Konow DB. The database is developed for assembling and systematically organizing the information gathered from the condition assessment of the restoration mortars. The goals and objectives of the database are to: •

• • •

collect the essential data about the restoration mortars, their types, mixture design, properties of their raw materials, their damage and their performance, store and update these data effectively,

allow sophisticated search strategies, and

produce detailed reports automatically for different types of restoration mortars

The design of the von Konow database involves two phases, the definition phase and the implementation phase. In the definition phase, the structure of the database is established. The implementation phase involves raw data collection, validation and harmonization for general use. The development of the von Konow database begins with the requirement analysis and the conceptual design, then logical and physical design, and final system implementation [10].

Requirement analysis of the von Konow database

Ph. D. Thorborg von Konow was an appreciated researcher in the field of restoration and an expert in the ageing and analysis of older mortars. Having worked the last years as a private consultant, her research was left without a natural continuation. Her material from over 20 years of research was left behind with three restoration architects. The database will be built from this material. The objectives of the database are (i) to classify the von Konow research materials into an easily accessible and distributable form, (ii) serve as a tool of information during the future research and (iii) serve as a source of information for architects and engineers working with older mortars. The database materials include authentic, documentary materials, information from restoration sites, photos, mortar recipes etc.

Conceptual design of the von Konow database

The data modelling is one part of the conceptual design process. The data modelling focuses on what data should be stored in the database. To put this in the context of the relational database, the data model is used to design the relational tables. The data collected for the von Konow research materials is analysed and arranged into data collections for each element of the database. 44


The conceptual data model identifies the highest-level relationships between the different tables of the von Konow database. A flow chart of the conceptual design of the von Konow database is shown in Figure 1.

Figure 1. Conceptual design of the von Konow database .

Logical design of the von Konow database The logical designing of the von Konow database involves two processes. The first process is to understand the requirements of the end users such as the need for the database, the achievement of the database and what kind of the real-world process the database is designed to emulate. The second process is to create a technical solution – a set of tables, complete with columns, each of which has the correct data type. Example of the logical design of the von Konow database is shown in Figure 2.

Figure 2. Example of the logical design of the von Konow database .

Physical design of the von Konow database The physical design involves the creation of the entity/relation (E/R) model and the determination of data storage techniques. Physical data model of the von Konow database represents how the model will be built in the database management system. The physical database model describes all table structures, including column name, column data type, column constraints, primary key, foreign key, and relationships between tables. The physical data model of the von Konow database is still under progress. An example of the physical design of the von Konow database using MS ACCESS 2010 database application is shown in Figure 3. 45


Figure 3. Example of the physical design of the von Konow database.

The von Konow database graphical user interface (GUI) is a user-computer interface that uses windows, icons and menus to input the data, search data and print reports from the database. This research work is aimed at designing of a simple GUI to enable users to interactively use the database and affords anybody with little or no prior knowledge of von Konow research materials to be able to use the database. The development of the GUI is still under progress. A screenshot of the documentary material window is shown in Figure 4.

Figure 4. The graphical user interface of the von Konow database .

Conclusions This paper presents development of the von Konow database which include material from over 20 years of research on restoration mortars by Ph. D. Thorborg von Konow. The design of the von Konow database involves two phases: the definition phase where the structure of the database is established and the implementation phase that involves the raw data collection, validation and harmonization for general use. The development of the von Konow database is still under progress. Once it is developed, the platform will be used as a tool for designing of restoration mortars and serve as a source of information for architects and engineers working with older mortars. 46


This database will allow access to the data needed for the selecting of suitable repairing mortars for the restoration purposes. The results are directly utilized in guidelines combining appropriate lime mortars with different architectural and historical periods.

References

[1] von Konow, Thorborg, (1997). Restaurering och reparation med puis- och murnruk. Turku: Åbo Akademis Förslag. [2] Groot, Caspar. (2012). Repair mortars for historic masonry. RILEM Publications SARL. e-ISBN: 978-235158-112-4 [3] van Balen, K., Papayianni, I., van Hees, R., Binda, L. & Waldum, A., (2004). Introduction to requirements for and functions and properties of repair mortars. Rilem publications. RILEM TC 167-COM. e-ISBN: 2912143675 [4] Balksten, Kristin, (2007). Traditional Lime Mortar and Plaster. Reconstruction with Emphasis on Durability. Göteborg.: Chalmers University of Technology. ISBN 978-91-7291-990-7. [5] van Hees, R. P. J., Binda, L., Papayianni, I., & Toumbakari, E. (2004) Characterisation and damage analysis of old mortars. Materials and Structures, Vol. 37, November 2004, pp 644-648. [6] von Konow, Thorborg, (2006). Mortars for older buildings. (in Finnish: Laastit vanhoissa rakenteissa.) The Governing Body of Suomenlinna. Helsinki. ISBN: 13:978-951-9437-31-6 [7] Bartos, P, Groot, C. & Hughes, J.J., (2000). Historic Mortars: Characteristics and Tests. Paris: Rilem Publications. 459 s. ISBN: 2-912143-15-2. [8] Acun S., Arıoglu N. (2005) The evaluation of lime mortars and plasters with the purpose of conservation and Restoration. http://cipa.icomos.org/fileadmin/template/doc/antalya/149.pdf. Accessed 26 January 2014. [9] Hughes, J.J. and Callebaut K. (2002) In-situ visual analysis and practical sampling of historic mortars. Materials and Structures, Vol. 35, March 2002, pp 70-75 [10] Moodi, F., Development of a Knowledge Based Expert System for the Repair and Maintenance of Concrete Structures, PhD. Thesis, Newcastle upon Tyne University, Newcastle upon Tyne, UK. (2001). [11] von Konow, T. (2002) The Study of Salt Deterioration Mechanisms Decay of Brick Walls influenced by interior Climate Changes. The Governing Body of Suomenlinna, Finland. [12] von Konow, T. (2007) Practical and Analytical Methods for Evaluating Deterioration in Brick Walls. http://www.arcchip.cz/w09/w09_konow.pdf. Accessed 26 January 2014

47


Damage assessment of early 20th century stone imitating mortars Y. Govaerts1*, A. Verdonck1, W. Meulebroeck1, M. de Bouw2,3 (1) Vrije Universiteit Brussel, Brussels, Belgium (2) Belgian Building Research Institute, Brussels, Belgium (3)University of Antwerp, Antwerp, Belgium Abstract: The present research evaluates damage problems of historic stone imitating renders. Such decorative mortars were applied on façades to create the illusion of a French sandstone or limestone masonry. Visual analysis of three listed case studies from the early 20th century sheds light on the lifespan of these materials by assessing their characteristic damage patterns and helps to understand the causes of their degradation. Soiling, cracks, adhesion problems, biological growth and surface erosion are observed and discussed. Although the renders are in relatively good condition, discolorations disrupt the monumental value and deterioration often increases after improper conservation interventions. Keywords: Damage assessment, Stone imitation, Rendering mortars, Young heritage, Case study

1 Introduction

At the beginning of the 20th century a new type of finishing materials was developed in central Europe. Implementing modern Portland cement in mortar compositions enabled to obtain a façade rendering with the appearance of sandstone masonry. Especially the German industry focused on the development of these materials, but their products were initially only limited appreciated [1, 2]. However, with the introduction of the Art Deco Movement around 1920, in which the emphasis in architecture changed from ornamental Neo styles towards clean geometries with sober finishes, the artificial sandstone plasters gained popularity and were exported to neighbouring countries such as Belgium, The Netherlands and Italy. Because an imitation render was a qualitative and less expensive alternative for natural stone, this finish was applied at many construction sites during the interwar period [2-4].

Nowadays a lot of these buildings are listed as young heritage, making their conservation highly important. After nearly a century, these artificial sandstone mortars have stood the test of time properly, but not without any damage. Their restoration is still a challenge in practice, because the ageing and related deterioration process of cement mortars is not well explored [5, 6]. Professionals have to take into account three criteria whilst restoring. The first challenge is to achieve a visual uniformity of the façade, using modern cleaning techniques and repair renders which match with the original appearance. To avoid a patchwork of different colours, contractors use a time-consuming trial-and-error methodology adjusting continuously the mortar composition until the target colour is obtained. In conservation practice, aesthetical aspects are usually more important than other properties. A second criterion imposes to repair current damage patterns, which need to prevent returning degradation. The third challenge is related to the previous criteria and includes avoiding future damage caused by usage of incompatible products or improper technologies. Unfortunately the prediction of the effects of restoration campaigns is rather complex, given the interaction between unidentified parameters. To meet the upper challenges it is important to learn how these historic imitating renders behave in time and interact with their environment. Only if we are able to understand how historic renders age and degrade, we can try to counteract the damage causes. This *

Department of Architectural Engineering, [email protected]

48


research is a first step towards the identification and assessment of typical damage typologies for these peculiar historic stone imitating renders. Degradation patterns were characterized by using a damage-atlas for brick masonry [7].

1.1

Stone imitation

A stone imitation or simili-pierre can be considered as a piece of architectural art. Few compositions consist of the exact same ingredients/ratio’s and yet all these render layers resemble real stone. These variations are due to the on-site mixing of binder and aggregates by skilled craftsmen, according to their own experiences and material suppliers. In general, a stone imitation can be obtained by using white cement or a combination of white cement and slaked lime (bad mortar) as binder with quartz sand and crushed natural stone aggregates. The addition of muscovite minerals ‘mica’ enables a sparkling visual value and other additives like starch and oils were sometimes incorporated to improve the workability [3]. Also premixed renders with brand names were launched just after World War I, such as Terranova, Dura and Simili Euville. Adding clean water to a bag of powdery substance was sufficient to obtain a uniform mortar of high quality. In most cases, the simili-pierre top layer has a thickness of about 5 millimetres and is always applied on a rough base layer. Ranging from 1 to a few centimetres, the bottom layer is composed of a mixture of Portland cement and sand. In order to resist tensile stresses the undercoating sometimes contains pieces of reinforcement such as cow hair, straw or wire mesh [8]. After spreading the render layers on the wall surface, a finishing technique was chosen. A plasterer commonly used a sharp steel lath to scratch the top layer, in order to create a rough stone appearance. Brushing, cutting or chiselling the mortar were alternative convincing treatments for a perfect stone imitation. A masonry appearance could be achieved by drawing simulated joints into the surface. Painting these grooves with grey cement slurry or lime paint was common practise.

1.2

Case studies

Three Belgian cases, all listed buildings, were selected to perform a damage assessment. Apart from their finishing material, they have not much in common. Because of these very different boundary conditions, the influences of the environment can be studied properly. Case 1 (C1) is a cinema complex in Antwerp from 1928 (Figure 1). Being situated in the middle of a city block, the building has entrances at three streets. These street façades are almost completely finished with a sand stone render. According to the building specifications the commercial ready-mix plaster Terranova was used. This German product was available in different appearances, varying from coarse to fine grained aggregates. A sample from the west façade was examined by the Royal Institute for Cultural Heritage. Optical observation of the sample, combined with SEM and EDX analysis, allows us to identify the constituent components of the mortar [9]. Terranova’s formula includes limestone fragments derived from crushed Euville stone and crinoids, but also quartz sand as main aggregates. The binder is presumably a mixture of lime and white cement and also mica particles with a length of a few millimetres are found. Especially the north façade is more deteriorated than the others, possibly influenced by the dense traffic, orientation and geometry. Case 2 (C2) is the Gothic revival Bank building and 18th century Eynatten mansion in Louvain (Figure 1). The courtyard between these historic buildings has a simili-pierre cladding, dating from 1913. The rendering is composed of 4 volume parts of aggregates to 1 part of lime and 1 part of early cement. Very fine colourless sand with black, yellow and red particles, including mica’s is identified as a typical component of the aggregate [10]. Case 3 (C3) is Breivelde Castle in Zottegem (Figure 2). Unlike the other case studies, this monument is situated in rural area at the highest point of a park with pond. Around 1904 the building was transformed into a Neo Flemish Renaissance Style with dominating brick masonry. Striking horizontal strips and coping stones in light grey imitation render improve the monumental value of the castle. Two different renders were identified and analysed in the laboratory. The first sample is taken from the south tower and has a greenish grey appearance. It 49


is a strong hydraulic mixture of cement with crushed Belgian porphyry (ratio 2:1). The second sample is part of a coping stone and contains 5 volume parts of Portland cement with pozzolanic components like furnace slags or fly ash and 4 parts of coarse sand with feldspars [11].

Figure 1 Case 1: North façade of De Roma cinema theatre (left), Case 2: East façade of the Eynatten mansion in Louvain (right)

Figure 1 South façade of Breivelde Castle

2 Damage assessment Historic rendering mortars age in different ways, according to a complex interaction of degradation processes [12, 13]. Renders are designed to protect façades, making them prone to weather conditions. Damage assessment has to determine key degradation patterns for sand stone imitations and indicate which factors lead to increased ageing. Visual observation of the three cases highlights substantial problems like soiling, cracks, loss of adhesion, biological growth and erosion, which are discussed in detail.

2.1

Soiling

Contrasts generated by appearance of grey and black crusts on a light plaster layer have a major impact on a building’s image and aesthetics. Façades in urban context (C1, C2) are more sensitive to soiling than walls in rural, forested surroundings (C3). It is well known that air pollution, the mortar composition and façade geometry are responsible for soiling [5, 14]. Air pollution can be defined as a mixture of contaminating gasses and particulate matter like dust and soot. Influenced by gravity, rain and wind, these particles are deposited at any location on 50


the façade. When a particle moves close along the mortar surface, friction or electrostatic forces can retain it. Due to gravity, the pollution concentration is about twice as high near ground level than at a height of 10 meters [15]. Observation shows indeed a uniform dust layer, which is darker on the lower floors, possibly because of the traffic emissions (Figure 1). This is not the case for Breivelde castle (C3) presumably due to the limited air pollution (see §2.5). C3 only has black stains beneath window sills close to ground level and among the tower’s cornice. Also in urban environment, dirt concentrations are situated on surfaces sheltered from rain runoff (C1, C2). Formation of black crusts increases roughness and retention of particulate matter [5]. Dark crusts arise when limestone fragments interact with atmospheric SO 2 , forming gypsum efflorescence. Higher concentrations of sulphate are linked to a greater amount of gypsum, leading to thicker crusts over time [5]. Dirt particles are imbedded within the gypsum layer usually derived from coal or oil combustion. Gypsum is observed on almost all hydraulic cement and pozzolana mortars [14]. Since gypsum arises from mortar components, cleaning campaigns can be effective, but these imply removal of authentic material. However it is important to get rid of these crusts to avoid a second damaging process. Literature shows that hydraulic mortars contain calcium aluminate hydrates [14]. If gypsum is formed at the surface, it can be dissolved by rain and migrate into the mortar. Reaction between calcium aluminate hydrates and sulphate solution creates ettringite precipitation, which expands and gives rise to severe stress within the pores of the mortar structure. Accumulation of ettringite may lead to total destruction of the material [14, 16]. This degradation process is not observed. It seems that mica particles at the surface are able to withstand the dirt layer, creating a shiny contrast to the surrounding crust.

2.2

Cracks

Most cracks are not caused by the render itself, but by failure of the underlying support. Cracks appear when materials with a different thermal expansion coefficient lead to stresses in the structure (C1). Also corrosion of reinforcement in coping stones or building settlement (C3) result in a cracked finishing aspect (Figure 3). Hair cracks arise as a result of dry shrinkage just after application. Renders with a large amount of Portland cement are more likely to develop small cracks because of its brittle character. Although hair cracks are harmless, water infiltration is possible, which is a plausible cause for corrosion (C3) and blistering (C2).

Figure 3 Cracks due to: thermal expansion coefficients (left), corrosion (middle) and building settlements (right)

2.3

Loss of adhesion

Lack of adhesion frequently leads to removal of the simili top layer. Adhesion between the two layers is achieved by roughening the base layer to increase its adhesion surface. Only plastered window sills, coping stones and other ornamental cantilever elements contain reinforcement bars. Because these bars are too close to the outer surface, damage is a logical consequence. Small cracks in the mortar layer make way for water infiltration and exposure to air. Such conditions interact with steel, causing corrosion and involve a material expansion. As a result, large pieces may lose adhesion with their support (Figure 4). Blistering only appears on the courtyard in Louvain (C2). Due to a poor adhesion, water can infiltrate and accumulate between 51


two mortar layers, creating a local swelling of the surface. Particularly the effects of frost may cause an expansion behind the top layer, making detachment inevitable (Figure 4). Sometimes accidental damage due to the impact of an instrument leads to broken coatings. Especially edges of rendered ornaments at street level are prone to detach.

Figure 4 Adhesion through: mechanical impact (left), corrosion (middle) and local swelling (right)

2.4

Biological growth

Every site has to deal with some kind of biological attack. Algae, lichens and higher plants principally cause aesthetic problems to the monument, such as colour changes (Figure 5, 6). Algae are plant-like microorganisms which develop on wet areas with sufficient light where they form a green deposit. In literature, northern oriented façades are considered as the ideal habitat for these organisms since they are least exposed to solar heat [17]. C1 and C3 confirm the presence of green algae growth on vertical surfaces of northern window sills. Also imitation strips protruding from the wall surface, show green stains at the top because water is stuck on the horizontal plane. Considering the fairly new imitation strips at the north side, C3 was probably seriously damaged and overwhelmed by vegetation. Southern façades are less suitable for algae, but certain circumstances can facilitate their development. Consider for example a broken or clogged rain drain (C1), which ensures a permanent humidity, surfaces covered by branches or the lower side of consoles, which spend most of the time in shadow (C3). Dryer crusts are found on horizontal surfaces and areas close to ground level. Furthermore, an increasing mortar roughness seems to stimulate algae growth. These are situated in the deeper eroded areas between pebble particles, mainly on the west side (C3). A rough texture affects rainwater runoff and increases adhesion of organic compounds. Algae are harmless, but make way for mosses, lichens and plants which could cause mechanical damage. Lichens are solely identified on the south tower and coping stones of Breivelde castle (C3). Such mosses arise from a symbiosis of fungi and algae [17]. Coping stones are always subjected to rain as they are roof elements, yet exposure to sun and wind provide a relatively dry favourable environment for lichens. Different types can be distinguished, such as the dominant yellow Xanthoria parietina, white Diploicia canescens and Lecanora chlarotera, which alternately grow alongside each other. Some locations contain more nutrients, which influence the growth behaviour of the Xanthoria family, for example next to trees or roof ridges where bird droppings are deposited (C3). Lichen degradation is shallow (0,1-1,5 mm), depending on the plaster’s hardness. Since most lichens produce (ascorbic) acid, porous and calcareous plasters will be very sensitive to lichens, but this is not the case here. Portland cement is the main component and interacts with water to Ca(OH) 2 , NaOH and KOH, which provide a highly alkaline environment (pH ≈ 13). However, adding pozzolanic fly ashes (see §1.2) provides a more acidic mortar with pH < 11, but still slightly alkaline. As C3 illustrates, Xanthoria parietina grows well on less alkaline substrates. Presumably, the present stone imitation mortars did not suffer from lichens during the first years. After all, new cement mortar is too alkaline and it takes at least five years before the acid content becomes more attractive. Lichens serve as indicators for clean air. They are very sensitive to air pollution, especially in areas where high SO 2 concentrations are measured. This explains why no lichens were observed in urban climates. [17-19] The Western façades in C2 and C3 are partly covered with Virginia 52


creeper, but the roots are located below ground level. Despite the mortar layers are used for support and not for water supply, stems can cause visual and mechanical damage after removal. Since Virginia creeper is securely attached to the surface, removal can imply detachment of the top layer (C2). Moreover, areas behind wooden stems are better protected against pollution than uncovered areas, creating a distracting pattern (C2, C3). Biological damage can be avoided by monitoring algae evolution. If algae are already removed in an early stage, lichens and plants have no chances to develop.

Figure 5 Colour changes by biological degradation: algae C1 (left), lichens C3 (middle) and plants C3 (right)

Figure 6 Closer look at the damage patterns: algae C1 (left), lichens C3 (middle) and vegetation C2 (right)

2.5

Erosion of the surface

Since façades are always subjected to the weathering conditions, time is a key factor in the abrasion process of surface material. Flowing or dripping water causes a rougher texture by dissolving certain components in the upper layer or by deposition. Also heavy wind carrying sand particles can have an abrasive effect on the renders [7]. C1 does not show exceptional surface erosion. The top layer of C2 is washed out. Mica particles and other fine aggregates are loose on the wall surface in most areas and in other zones the simili top layer is completely removed. C3 illustrates the effect of orientation on surface excavation. Western zones are generally more weather worn. A small surface is affected by a local white salt efflorescence, creating a rough powdery surface (Figure 7). Simulated joints are also very susceptible to weathering. Only a few of these thin coatings are still intact (Figure 8). C2 shows a small line, drawn with a whitish slightly transparent paint on top of the grey mortar surface. The remaining white paint contains calcium carbonate and calcium sulphate [10]. Calcium sulphate or gypsum is formed by interaction between acid rain and the lime paint. A part of calcium carbonate is transformed into gypsum, making the joint rougher and easier subjected to rain and wind. Grey cement joints characterize C1 and C3, but the exact composition is unknown. Table 1 gives an overview of the quantity of damage types in function of orientation.

53


Figure 7 Surface erosion C3: limited affected (left), washed out (middle) and salt reaction (right)

Figure 8 Decay of imitation joints: C1 (left), C2 (middle) and C3 (right) Table 1 Damage patterns in function of orientation Case 1 (Antwerp) Damage type North South-East West Algae + + + Mosses Lichens Higher plants Dust layer + + Black crust ++ + ++ Cracks + + + Hair cracks + + ++ Corrosion problems + Poor adhesion Surface erosion Simulated joints + + Improper repair

Case 2 (Louvain) East South West + +

++ +

+ ++

+ + ++

++ ++ + +

++ ++

++ +

+ ++

+ +

North +

Case 3 (Zottegem) East South West + + ++ ++ + +

++ ++ +

+ +

+ +

++ + ++

++

+ +

+

+

+

+ +

++

+

3 Conclusions Early 20th century renders for stone imitation age and have a lifespan that can be drastically shortened by degradation phenomena. Despite the presence of severely affected façade areas, upper case studies show that many render surfaces are still intact after nearly a century. Thus imitation plasters may be considered as relatively durable materials. Research indicates that the underlying structure with or without corrosion problems often leads to cracks and loosening of the render layers. Considering orientation, wind and rain being fixed parameters for every case, the location and related air pollution has a major effect on the development of damage. Air pollution in urban areas causes black gypsum crusts through chemical reactions with the 54


hydraulic renders. On the other hand, renders in rural regions are prone to develop lichens due to a good air quality if the mortar is not too alkaline. Subsequently, discoloration is often considered as a threat within heritage conservation. In order to restore the monument to its former glory, cleaning techniques are frequently applied. However, cleaning methods usually imply a loss of authentic material, especially if a uniform clean aspect is required. After all it is difficult to remove every dirty particle without using abrasive techniques or inserting a lot of water into the mortar, making it more sensitive to frost. If there is optimal adhesion between top layer and base layer, and the above-mentioned damage is absent, erosion through abrasion, salt accumulation or frost conditions would lead to failure in the end. Abrasion of a simili top layer can be observed in case 2. In order to start an in-depth study on the ageing processes, damage assessment of more cases is essential, including analysis of cleaning treatments.

4 References

[1] Stenvert R (2010) Mooier voor minder: cementlagen en betonafwerking. In English: More beautiful for less: cement layers and concrete finish. In: Stuc, kunst en techniek. Waanders, Zwolle, pp 412-425 [2] Garda E (2003) Terranova, history of a modern plaster: Smooth, hard, clean, perfect, in Huerta S & de Herrera J (Eds.) Proceedings of the 1st International Congress on Construction History, Madrid 20-24 January 2003, Madrid, pp. 970-977 [3] Govaerts Y, Verdonck A, de Bouw M & Meulebroeck W (2013) Terranova, a popular pierre-simili cladding: Strategies and techniques for restoration. Proceedings of the 3rd Historic Mortars Conference, Glasgow 11-14 September 2013, University of the West of Scotland, Glasgow [4] Poptie A (1948) Handboek voor den stucadoor (deel 2). In English: Handbook for the plasterer (part 2) N.V. De Technische Uitgeverij H. Stam, Haarlem [5] Ozga I, Bonazza A, Bernardi E et al. (2011) Diagnosis of surface damage induced by air pollution on 20thcentury concrete buildings, Atmospheric Environment 45: 4986-4995. [6] Klisińska-Kopacz A, Tišlova R (2013) The effect of composition of Roman cement repair mortars on their salt crystallization resistance and adhesion, Modern Building Materials, Structures and Techniques 57: 565-571. [7] Franke L et al. (1998) Damage atlas: Classification and analysis of damage patterns found in brick masonry, Research n° 8, volume 2. Fraunhofer IRB Verlag, Stuttgart [8] Verdonck A & Dekeyser L (2010) Schijn of werkelijkheid? Steenimitaties en sierpleisters. In English: Illusion or reality? Stone imitations and decorative plasters. In: Steen & co. Monumenten en Landschappen, Brussels, pp 142-166 [9] KIK-IRPA, Royal Institute for Cultural Heritage (2011) Decorative renders, material analysis, Sample M8, pierre-simili render in Borgerhout (Scientific report), Brussels, 11 pages [10] Jägers E (2011) Hôtel d’Eynatten and the Helleputte building – Courtyard: Scientific examination of the plaster (Scientific report), Bornheim, 3 pages [11] KIK-IRPA, Royal Institute for Cultural Heritage (2013) Zottegem, Breivelde Castle, reporting mortar analysis (Scientific report), Brussels, 20 pages [12] Van Hees R et al. (2004) Damage analysis as a step towards compatible repair mortars. In: Groot C, Ashall G, Hughes J (eds) Characterisation of Old Mortars with respect to their Repair – Final Report of RILEM TC 167-COM. RILEM, pp 107-152 [13] Smith BJ & Přikryl R (2007), Building stone decay: From diagnosis to conservation, Geological Society Special Publications 271. The Geological Society Publishing House, Bath [14] Sabbioni C, Zappia G et al. (2001) Atmospheric deterioration of ancient and modern hydraulic mortars, Atmospheric Environment 35: 539-548. [15] Herremans T & Vangheel T (2003) Biologisch herstel van schade aan gebouwen: biomineralisatie. In English: Biological repair of damage on buildings: biomineralization. Thesis, University of Louvain. [16] Sabbioni C, Zappia G et al. (1998) Black crusts on ancient mortars, Atmospheric Environment 32: 215-223. [17] Rijksdienst voor archeologie, cultuurlandschap en monumenten (2008) Algen, mossen en korstmossen. http://www.cultureelerfgoed.nl/sites/default/files/u4/racm_brochure_techniek_16.pdf. Accessed 05 February 2014 [18] Anonymous (2014) Air quality and lichens. http://www.air-quality.org.uk/19.php. Accessed 07 February 2014 [19] Loutz S & Dinne K (2000) Vervuiling en verwering van steenachtige materialen door micro-organismen. In English: Soiling and decay of stonelike materials by microorganisms. BBRI-journal n° 2: 3-13

55


Future-oriented building stock studies and the significance of values: addressing demolition behaviour with the Delphi method S. Huuhka1* (1) Tampere University of Technology, Tampere, Finland

Abstract: This paper presents a research in progress that applies Delphi, a futures research method, to ageing building stocks. In the empirical part, the method is used to predict scenarios future demolition behaviour. Special weight is given to reflecting how the construction material and its ageing affect the decision between demolition and life cycle extension. Delphi survey respondents argued for and against the survival of materials from their professional backgrounds. Two major lines were revealed: one that found the demolition of concrete buildings likely, and other that argued that timber buildings would face removal. Based on the respondents' answers and counter-arguments provided by the author, the relationship between material properties and survival of structures comes across anything but straightforward. Keywords: building stock, demolition, values, futures research, Delphi method

1 Introduction The life and death of buildings are profoundly characterized by intentionality. Although the ageing of materials follows the laws of physics, the length of a building's life cycle is not predetermined by nature. Because buildings do not die naturally, deliberate demolition is the final endpoint for their ageing process. Nonetheless, it is not the age of the structure that has been found to determine demolition, but how people feel about it: this phenomenon is called obsolescence [1]. In fact, the oldest layers of the building stock have been found to be likely to survive better than the younger ones [2]. Although related to ageing, the obsolescence of a material or structure is apparently defined by more than just physical deterioration. Buildings or their parts may be deemed obsolete — not because they would be broken, unfit or even performing poorly — but simply because norms and technologies or aesthetic aspirations have changed. The label of obsolescence is, clearly, value-infused. In preparing for the climate change, Europe is setting policies that aim at resource and energy intelligent construction sector. This is no surprise, as buildings occupy land, consume raw materials and produce emissions from construction to their demolition. These policies rely largely on replacing the existing stock with new construction; often assuming that new is superior to the existing from the environmental point of view [3]. However, in most of Western Europe, annual new construction accounts only for 1% of the building stock [2]. Despite the long duration of habitation in Europe, this situation is unprecedented in history in many countries. Finland is among those who lack the experience to deal with this shift from new construction to sustainable stock management. Setting a target for service life is easy in the design stage of a new building, but how to treat consciously an ageing stock that is already there? Not to mention that this stock is still rather young, unprecedented in size and built of contemporary materials that may only have been use for more or less the same time as the buildings themselves. Under these circumstances, past behaviour appears to lack the explanatory or predictive power for future development. Therefore, the method of mathematical extrapolation seems unreliable for building stocks studies in countries like Finland. Trend extrapolation also ignores the fundamental significance that free choice and changing values have for the survival of the research object. The dynamic behavior of building stocks may, surely, appear to resemble natural forces, as the behavioral patterns arise from a huge number of individual choices made by singular building owners. Nonetheless, the decision between life cycle extension and demolition is driven by values, just like any human behaviour. It is also governed by national and *

Tampere University of Technology, [email protected]

56


international policies as well as corporate cultures in construction and real estate industries, which, for their part, reflect the shared values in human communities. Extrapolation demonstrates how things end up if circumstances stay the same, but it does not take a stance on if this is the way things should be. Futures research, on the other hand, bases on the idea that future is not predetermined but can be affected or even chosen [4]. When dealing with a research object fundamentally dependent on the actions of humans such as the building stock, this perspective must not be bypassed.

1.1

Goal of research

The paper presents a research in progress. It experiments with Delphi, a futures research method, in pursuing target-oriented management for ageing building stocks. The method is used to predict future demolition behaviour, which is both age- and obsolescence-related. The work has been inspired by an application of the Delphi method on traffic volume growth, which reflected on shifting from determinist business-as-usual predicts to proactive future-making in the transportation sector [5].

1.2

Research questions

The research questions were as follows: how does the ageing of the different layers of the building stock affect the amount of demolition in future? Is demolition increased or is life cycle extension favoured instead? Is demolition emphasized in certain layers of the stock and life cycle extension in others? Does the construction material affect the decision between demolition and life cycle extension? Which materials are present in the stock to be removed? How are the demolished materials and components to be processed - landfilled, incinerated, recycled or reused?

2 Research material and methods 2.1

The Delphi method

Delphi is a futures research method that uses the expertise of interest groups to generate understanding about possible future developments. It is an expert method in which participants predict prospective development. As no one can truly know the future, the respondents' task is to present an educated guess about possible development on the question researched. As they are experts in their fields, their educated guesses are expected to be more relevant than those of randomly selected people. A large number of participants is not necessary or even ideal in Delphi, because statistical representativeness is not an aim. Instead, it is the expertise and involvement of the participants that counts. [4] In the field of ageing and materials, the International Federation for Structural Concrete (fib) has used the Delphi method to define the propagation model for steel corrosion in carbonated concrete. Experts estimated the duration of corrosion propagation period until concrete cracks or spalls in given conditions based on their practical experience. The cumulative frequency curves drawn based on Delphi results suggested that the mean value for the occurrence of cracking would be 4,5 years and 9,0 years for spalling. A similar survey was performed for estimating corrosion penetration depth, the results of which were used to calculate the duration of the propagation period until collapse. [6]. The fib application of Delphi, linking observations to the mathematic calculation of a likely future, represents a fusion of conventional empirical science and an older type usage of the Delphi method: endeavouring a consensus about the likely future. Predicting one future is how the method has been used in the past (but without a base on a mathematical model), and for this it has also been criticized as unscientific. More recently, Delphi has been mostly employed to survey the variety of possible futures, defined by the different interest groups based on their own distinct values. When applied this way, the method can reveal attitudes and varying interests among power players. [4]. It also involves the decision-makers into the research process and carries the potential to commit them in pursuing a value-driven change. This kind of application of Delphi is popular in human sciences and it is also represented in this paper. A combination of modelling and a Delphi survey would have also fitted the purposes of the study, but it was not possible given the time frame and the available resources.

57


2.2

Performing the survey and the analysis

In this study, a Delphi survey was formulated and sent to a panel of experts. This panel consisted of partners of the research project Repetitive Utilization of Structural Elements (ReUSE) as well as selected persons from interest groups that have previously participated expert workshops on material efficiency in the construction sector, arranged by the Ministry of Environment. They include people from construction industry, waste management, public administration as well as research and development. The survey was sent to 35 respondents, 17 of which returned it. The questionnaire consisted of three graphs, which the respondents were asked to supplement. The first graph depicted the amount of construction and demolition (C&D) waste; the second showed distribution of materials in the waste flow and the third one portrayed processing of that waste. Each fulfilled graph was based on 15-20 years of statistical data; in addition, minor background information was provided on the size and composition of the existing building stock in the form of keywords. The participants were requested to draw continuation for the trend lines from the 2010's to years 2020 and 2050. In addition to predicting the probable development, the respondents were asked to also define a preferable future for graphs 1 (amount of waste) and 3 (processing of waste) in the same manner, keeping in mind the organization they represent. After drawing their probable and preferable scenarios, the respondents argued in writing which assumptions their scenarios were based on. The purpose of defining probable and desirable scenarios is to act as a platform for discussion on future alternatives. These scenarios can also be used for backcasting. Backcasting is a procedure that starts from a defined future endpoint and works its way back in order to define what measures would be required to achieve the proposed scenario. However, backcasting was not performed in the 1st phase of this study. The method of analysis was qualitative. The curves the respondents drew were examined visually. Written answers were clustered to groups manually. This paper describes results from the 1st round of the Delphi survey. The method includes at least two rounds, but the second one has not been performed. Therefore, the results are preliminary. As the significance that used materials have for demolition is of special interest for the conference, the paper will focus on this issue, dealing briefly with the other aspects as well.

3 Results and discussion 3.1

Total amount of waste: overview

The total amount of waste consists of waste from demolition as well as masses from renovation. Up to 2020, most respondents predict extrapolated growth and only few foresee levelling off. Most curves remain upwards until 2050, but two respondents out of 17 suggest a decrease since the 2030's that would result in going below current waste amounts by 2050. In addition to them, two respondents foresee levelling off that would take place between 2020 and 2050. However, the majority of curves suggest 1,25 to 2,25 times the current amount of C&D waste in 2050. These answers depict probable futures. When asked about preferable futures, the curves are generally lower. Half the respondents would prefer to go below the current level by 2030 as well as 2050. These estimates vary from 0,25 to 0,75 times the current amount. (For comparison: 0,5 times the current amount has occurred for the last time during the recession of 1990's, when almost one fifth of all Finns were unemployed.) Five respondents prefer a steady development, one goes for moderate growth and two would settle for 1,5 times the current amount. To summarize, while most of the experts preferred a decrease in waste, only few of them believed that it would actually happen. This suggests that current incentives to decrease C&D waste are not powerful enough.

3.2

Materials in the waste: overview

The graph for materials distinguishes four categories: mineral materials, wood-based materials, metals and other materials. Mineral materials consist of bricks and concrete. Bricks are rather insignificant in the Finnish building tradition, although they have been used as a structural material in the beginning of the 20th century and later as a façade material. Concrete has been very common in structures and

58


facades from the 1960’s on [7]. At the moment, mineral materials and wood are equally present in the waste flow: both are accountable for approximately 1/3 of all demolition waste. The remaining 1/3 is divided between metals and other materials that rarely are structural, such as glass, gypsum, etc. In the 1990's, mineral materials dominated C&D waste. The tide was turned in the beginning of the 2000's, when wood increased its share, eventually exceeding minerals for a brief while. The future estimates the respondents gave were mostly more or less linear, with only minor curvatures. Only few respondents foresaw a steady development. Instead, the answers were polarized: most respondents took a stance that either mineral materials (in practice, concrete) or wood would start to dominate over the other. The shares of metals or other materials were most often seen to remain at the current level.

3.3

Materials: concrete

The top 5 respondents who believed in the increase of concrete all shared the view that the building to face demolition would be prefabricated blocks of flats in high-rise neighbourhoods from the 1970's and later. From other answers mentioning concrete, several respondents formulated that decreased use of wood in contemporary construction would lead to the increased share of concrete in C&D waste. Vacating industrial areas were also mentioned. Respondents believing in the demolition of mass housing repeat a conception that has popped up frequently in the Finnish press during the last ten years, at least. So far there has been a lot of talk but little action. Yet, extensive demolition of public high-rise housing has taken place in the Netherlands, Britain and Germany, and Finland has been historically prone to follow the European trail just slightly behind. A special feature of the Finnish discussion is that these buildings are commonly believed to be designed to be temporary and to last only for 30 - 40 years - a time that has already passed. The view has been shared by professionals and laymen alike. There is some evidence that supports the belief about the short design life [7,8]. The concrete industry has admitted to have used chlorides (which contribute to corrosion) and excessive heat treatment (which may cause internal cracking and contribute to frost damage by building up ettringite in the pore network) to speed up hardening of concrete in the early years, which may have added to the negative impressions. Also, the concrete used those days was not frost resistant (non-air-entrained) and mild steel was used for reinforcing with too small cover depths [9]. In addition, carbonisation of concrete has progressed, meaning that the alkalinity of concrete can no longer protect reinforcement from corrosion. Consequently, one respondent brought up the need for heavy repairs in this stock, while it seemed that the others did not even bother to mention 'the R-word'. However, despite the obvious deficiencies in material properties as well as structural and productional solutions, envelopes of the era have been recently found to be in an unexpectedly good and still repairable condition [9]. Indeed, ageing and damaging of concrete has received extensive attention in research lately, which has resulted in elaborate knowledge in its behaviour, investigation of structures and repair methods [10]. Studies have even covered prediction of future performance. In addition, many of the apartments of the era were built for owner-occupation [11]. This fact may effectively prevent demolition of the buildings, no matter what their condition or material properties are. Namely, the Finnish legislation for these 'limited liability housing companies' requires unanimity in decision-making in such a drastic issue as demolition. It has been noted that the difficulty in decision-making has prevented or slowed down even indispensable repairs. While housing companies own the building (and possibly the land), shareholders own shares that entitle to use a certain apartment in the building. Thus, the buildings (and land) are the only possessions housing companies have. The value of the shares is, thus, mainly bound to the value of the house. Should the company decide to demolish the house, it renders the shares practically valueless. To the author's knowledge, only one limited liability housing company has decided to demolish its building (due to severely damaged foundations). This project was enabled by re-zoning that raised remarkably the permitted building volume (and consequently, the value) for the plot, located centrally in the metropolitan area of Helsinki. This kind of solution is only possible in growth centres, where the price of land and the demand for new construction are high. The author is not aware of any applicable models for limited liability housing companies in declining communities, where the aforementioned conditions are absent. It seems that buildings with unsatisfactory material properties and structural solutions may

59


have become undemolishable because of tenure. Yet, it should be finally remarked that none of the survey respondents touch upon this issue.

3.4

Materials: wood

Respondents believing in the increase of wood had versatile grounds for their opinions. One respondent named moisture and mould damage and considered it to affect especially wooden buildings. Moisture problems are, indeed, frequently in the news. However, based on the news coverage, large-scale, health-damaging problems seem to localize in public buildings (schools and hospitals) built with mineral materials in the contemporary era, and not solid-timber buildings. Having said that, there is no denying that timber is naturally vulnerable to moisture and microbial action that can damage both human health and the bearing capacity of timber. This behaviour, which is mostly linked to unfit structural solutions or details and not so much with ageing of the material itself, is well known and thoroughly described in literature (e.g. [12,13,14]). In fact, mould problems in concrete or brick buildings also originate from interior materials that were not expected to receive moisture [15]. Problems can be caused by, for example, timber casting moulds erroneously left inside the structure. Remarkably, solid timber is also simple to repair. Timber buildings can have extensive life spans if they are protected from moisture with structural solutions and receive appropriate maintenance. The oldest wooden building in Finland, the church of Vöyri, dates back to 1620's, which makes it approximately 390 years old. Even older timber buildings have survived at the same latitudes in Norway, where up to 900 years old stave churches are still standing. The fact that timber requires maintenance (painting) may have an impact on its survival, as the appreciation of carefree housing is said to be increasing. Younger generations are thought not to possess the necessary craftsmanship or interest, and the generations that do are growing increasingly older. Nevertheless, the respondents did not bring up this viewpoint. And remarkably, buildings constructed of mineral materials do not survive the Nordic climate without regular upkeep, either. Another factor characteristic of wood, although not mentioned in this context by the respondents, is the fact that wood burns. For example, the CEO of VTT Technical Research Centre of Finland, Erkki Leppävuori, has argued publicly in 2011 that the Great Fire of Turku in 1827 would still affect Finns' attitudes towards timber construction [16]. One respondent believed that efforts to densify urban structure in cities would lead to demolition of older timber housing, while another respondent brought up the other side of the urbanization coin: vacating countryside. Two people specifically named detached reconstruction era timber houses as the yielding type. These so-called veteran houses can be found in both urbanity and rurality, and they are the result of the resettlement of nearly half a million people that lost their homes in the Winter War. While these houses are appreciated as built heritage, they are also do-it-yourself constructions from the scarcity of the post-war era. Whatever materials available were utilized with whatever skills the builder happened to have. This combination of cultural significance and imperfect materials, skills and structures might explain why veteran houses appear to be disrespected among structural engineers but appreciated by their layman residents and architects. From urban design point of view, densifying urban structure does not have to denote demolition. Especially the aforementioned post-war era residential areas have excellent premises for infilling as the plots are large (the extra land was reserved for domestic-use cultivation). Recent studies on socalled 'dense-low' design have proved that plot ratios as high as 0,5 can be achieved with max 2-storey buildings [17]. Also, re-zoning private land with existing buildings may prove difficult, as the Finnish legislation guarantees that landowners' opinions are heard. This may effectively hinder any densificating aims if they are not shared by the residents. Even if re-zoning would be successfully pulled through, it does not oblige the proprietor to demolish. Legislation allows existing buildings to be used as such despite operative urban plans. In addition, detached housing is the desired form of housing for more than half of Finns [18]. With all this in mind, it seems unlikely that existing lowdensity residential areas could grow significantly higher. However, in the 1960's and 70's residential areas were zoned without any respect to the residents' opinion. These urban plans, often still operative, may possess remarkable buildings rights that could turn low-density areas into medium- or highdensity quarters. Re-zoning is difficult in these cases, too, because some landowners prefer to stick to oversized building rights due to the speculatively higher value of the plot. Consequently, there are even heritage-listed post-war neighbourhoods that still have outdated urban plans allowing demolition.

60


If tenure was the barrier for concrete buildings, the question is much simpler with wood. So far, timber has been used for small constructions that rarely have more than one or two owners. However, this may change in the future if multi-family timber housing gains more ground, as one respondent suggests. This respondent predicts that the Finnish forest industry shifts from paper production to construction, which would eventually lead to the increase of wood in demolition waste. Another respondent mentions the same point, but argues that it would not have an effect in demolition waste in the examination period (40 years). A third one foresees European recycling targets as a restrictive factor for the use of timber, "if the targets are taken seriously". The respondent refers to the fact that Finland is currently not recycling timber, and the 70% recycling target announced in EU's Waste Framework Directive will not be reached without it. If this thought is turned upside down, the 70% target could also restrict the demolition of timber buildings: because they cannot be recycled, they must be maintained. Nevertheless, one respondent argues that wooden buildings will be demolished simply because they are so much less laborious to demolish than concrete buildings. As a result of their easy demolition, plots with wooden buildings can be vacated for new construction quickly.

3.5

Materials: metals and others

Very few respondents comment on steel buildings. However, one respondent names commercial, industrial and public buildings as a future source of metal waste. Another respondent working within the recycling industry argues that steel reinforcing from concrete structures would form the majority of the metal waste in construction, not actual steel structures. The share of steel in C&D waste would, thus, parallel the share of concrete. A respondent brings up the scarcity of metals and argues that they would be needed for other purposes than construction in future, which is why their share in waste would eventually decrease. A couple of respondents mention new composite materials as a factor that would eventually increase the share of other materials in the waste. One respondent considers recycled materials as other materials and predicts their increase. Contemporary glass façades are also mentioned as a trend that would eventually show in the waste flow, but the respondent considers their significance to be, nonetheless, minor.

3.6

Materials: appreciation and values

Although the respondents were asked to evaluate only the probable share of materials (as the question of preferable share of materials seemed too complex to ask), the answers can be interpreted to reflect their personal appreciations, equally. Some arguments support the preservation of concrete buildings and others the preservation of timber buildings. Values affect what weight these arguments are assigned with. The perspectives include cultural history, architecture, structures and materials, economy, tenure and urban planning, among others. The answers also reflect the respondents' professional backgrounds. For example, respondents who believe in the demolition of concrete mass housing come from recycling industry, housing construction and timber industry. The recycling industry is currently making business from the challenging demolitions of large concrete buildings and recycling that concrete. Timber buildings are, perhaps, typically demolished by smaller companies. Concrete is, thus, what the respondent deals with in the daily work. Companies in housing construction, on the other hand, are dependent on demand for new apartments. It is therefore not surprising that it would like to see old mass housing go. Timber industry, then again, is likely to be more appreciative of old timber construction than old concrete construction. Then again, respondents who believe in the demolition of older timber housing are either researchers in the field of construction or practical actors in fields related to heritage conservation. While professionals in construction might often not value older timber construction, conservationists are likely to witness demolition of culturally significant timber buildings in their work.

3.7

Utilization of the waste: overview

Currently, little less than 30% of C&D waste is landfilled, over 30% is incinerated and less than 40% is recycled. Practically all the respondents consider that landfilling is decreased and recycling increased already by 2030 and most definitely by 2050. Although the respondents were asked to

61


distinguish reuse of components from recycling of materials, almost none of them did so. Incineration, on the other hand, divides the respondents' opinions. Half of them suggest that energy usage is increased while the other half believes in its decrease. Energy usage would appear to have a natural connection to the share of wood in C&D waste, as concrete does not burn. However, the respondents do not seem very logical in this sense. The ones predicting most incineration are not the same people who predict the largest increases in the share of wood. When asked for preferable futures, none of the respondents prefer a noticeable increase in incineration. Surprisingly, the shares for landfilling are higher (and shares of recycling consequently lower) in the preferable scenarios than in the probable scenarios. The design of the graph may explain this peculiarity as some of the respondents seemed to have had difficulty in fulfilling the figure. The fact that there were several cases in which arguments and curves appeared to conflict each other suggest the same. Later iteration rounds with face-to-face interviews can help out to eliminate the inconsistencies.

4 Concluding remarks This paper has dealt with the ageing of materials from the point of view of demolition and values. Although the profile of materials used in Finland differs significantly from those utilized in Central Europe by a high share of wood and a near absence of bricks, the wider significance of the paper lies in focusing the attention to values: What is seen worthy or preservation and what is not, and why? What are the practical boundary conditions? As the building stock survives or disappears as the result of decisions made by a vast number of individual building owners, these issues can and should be discussed openly in the society. Owners' decisions can be encouraged or discouraged with legislative instruments. Governments endeavour target-oriented decisions in other complex issues, such as the climate change. There is no reason why conscious and collective decision-making could not be pursued in the management of building stocks, too. Currently, practices exist only for protecting heritage buildings. Although buildings of all kinds have valuable materials and work embedded in them, there are virtually no procedures to conserve these resources. Futures research methods, such as Delphi surveys, scenario formation and backcasting can be valuable tools in this societal discussion. When it comes to the conference theme, the question of interest is: What is the significance of material properties in this decision-making? In this study, still in progress, respondents argued for and against the survival of materials from their personal, professional backgrounds. Two major lines were revealed: one that found the demolition of concrete buildings likely, and other that argued that timber buildings would face demolition. The paper examined these arguments and provided counterarguments. The counterparts were found to be polarized in many senses: large, hard-to-demolish, hardto-repair concrete buildings with a complex ownership structure versus small, easy-to-demolish, easyto-repair timber houses with a simple ownership structure. It was outlined that the relationship between material properties and appreciation is not straightforward. Material properties may not be decisive for the survival of structures when they are juxtaposed with human factors.

5 Acknowledgements The work has been conducted in the research project ReUSE, funded by the Finnish Ministry of Environment, Finnish Wood Research and Ekokem Ltd. Research assistant Jani Hakanen helped in creating and processing the Delphi questionnaires. He also collected the statistical data from several sources and produced the graphs. Several co-operative partners for project ReUSE helped by testing the questionnaire.

6 References [1] Thomsen A and Van der Flier K (2011) Understanding obsolescence: a conceptual model for buildings, Build Res Inf 39:649-659. [2] Hassler U (2009) Long-term building stock survival and intergenerational management: the role of institutional regimes, Build Res Inf 37: 552-568

62


[3] Power A (2008) Does demolition or refurbishment of old and inefficient homes help to increase our environmental, social and economic viability? Energy Policy 36: 4487-4501 [4] Kamppinen M, Kuusi O, Söderlund S (eds) (2003) Tulevaisuudentutkimus [Futures research]. Suomalaisen kirjallisuuden seura, Helsinki [5] Tapio P (2002) The Limits to Traffic Volume Growth. The Content and Procedure of Administrative Futures Studies on Finnish Transport CO 2 Policy. PhD Dissertation, University of Helsinki [6] fib Bulletin No 34 (2006) Model Code for Service Life Design. International Federation for Structural Concrete, Lausanne [7] Seppänen M, Hytönen Y (2009) Tehdään elementeistä. Suomalaisen betonielementtirakentemisen historiaa. [Let's prefabricate it. History of prefabricated concrete construction in Finland.] 1st edn. SBK-säätiö, Helsinki [8] Hankonen J (1994) Lähiöt ja tehokkuuden yhteiskunta [Mass housing neighbourhoods and the society of efficiency]. PhD Dissertation, Tampere University of Technology [9] Lahdensivu J (2012) Durability Properties and Actual Deterioration of Finnish Concrete Facades and Balconies. PhD Dissertation, Tampere University of Technology [10] Tampere University of Technology, Research group for the service life engineering of structures (2014). Publications.http://www.tut.fi/en/about-tut/departments/civil-engineering/research/structural-engineering /the-service-life-engineering-of-structures/publications/index.htm Accessed 30 January 2014 [11] Lankinen, M (1998) Lähiöt muuttuvat ja erilaistuvat. 36 lähiön tilastollinen seuranta 1980-1995 [Largepanel neighbourhoods are changing and differentiating. A statistical follow-up of 36 neighbourhoods between 1980 and1995]. Ministry of Environment, Helsinki [12] Kaila, P (2008) Talotohtori [The building medic]. 15th edn. WSOY, Helsinki [13] Pirinen, J (2006) Pientalojen mikrobivauriot: lähtökohtana asukkaiden kokemat terveyshaitat [Microbial damages in small houses: from the premises of health detriments experienced by the residents]. PhD Dissertation. Tampere University of Technology [14] Ministry of Environment (s.a.) Hometalkoot - Ohjeita kosteusvaurioiden kartoitukseen - Homevauriot ja kosteusvauriot [Mould bee - Instructions for mapping out moisture damages - Mould damages and Moisture damages]. http://www.hometalkoot.fi. Accessed 30 January 2014 [15] Annila, PJ, Suonketo, J, Pentti M. (2014). Kosteus- ja mikrobivauriot koulurakennuksissa TTY:n suorittamien kosteusteknisten kuntotukimusten perusteella [Moisture and mould damage in school buildings based on condition assessments conducted by TUT]. In Säteri, J, Backman H (eds) Sisäilmastoseminaari 2014. Sisäilmastoyhdistys, raportti 32, pp. 301-306 [16] Kankare M (2011) Unohdetaan jo Turun palo, rakennetaan puukerrostaloja [Let's already forget the Great Fire of Turku and build blocks of flats in timber]. http://www.talouselama.fi/uutiset/unohdetaan +jo+turun+palo+rakennetaan+puukerrostaloja/a2026580. Accessed 22 January 2014 [17] Kuismanen K (2005) Tiivis-matala puurakentaminen - suunnittelu ja toteuttaminen [Dense-low timber construction - design and realization]. http://www.kuismanen.fi/plansu.pdf. Accessed 22 January 2014 [18] Juntto A (2007) Suomalaisten asumistoiveet ja -mahdollisuudet. Tulot ja kulutus 2007. [Housing preferences and possibilities for Finns. Incomes and consumption 2007.] 1st edn. Tilastokeskus, Helsinki

63


The Noble Patina of Age W.J. Quist1*, A.J. van Bommel1,2 (1) Delft University of Technology, Delft, The Netherlands (2)Atelier Rijksbouwmeester, The Hague, The Netherlands Abstract: This paper discusses the aesthetical aspects of ageing and focusses on the use of the terms ‘patina’ and ‘damage’ to decide on the cleaning of historical facades and the application of artificial ageing. Conservation campaigns can be characterized by the wish to preserve an object, building or building complex as a coherent piece of Cultural Heritage. This paper discusses the professional debate on balancing between preserving values, i.e. represented by patina and the need to intervene from a technical point of view. Cases on cleaning of limestone and sandstone together with replacement of natural stone and the application of artificial ageing are used to illustrate the debate. The perception of professionals is compared with the perception of laymen. Keywords: Building Conservation, Cultural Heritage, Patina, Damage, Aesthetics

1 Introduction

In daily practice and in most fields of science ageing is seen as a negative aspect, because of the gradual decrease of the properties of the base material. In the field of architecture, buildings suffer from all kinds of ageing phenomena of which the weathering of exposed materials and the wear of interior materials are the most visible ones. In preservation of Cultural Heritage, either be it tangible or intangible, ageing is often referred to as a positive aspect because it makes history visible and therefore is an important part of the values to be preserved. The wish to strive for preservation of authenticity, visual unity and technical functionality leads to a discussion on the value of patina compared to the trouble of damage, both related to the architectural design and the question on what to preserve. Decisions about conservation of historical buildings often depend on technical considerations, but also arguments regarding the artistic value of the object, sometimes including the need of a reconstruction of a situation gone since long, and, in practice, nontechnical arguments relating to public appreciation, tourist concerns, or even political purposes are taken into account [1, 2]. This paper explores the thin line between ‘patina’ and ‘damage’ and how both concepts are dealt with in a technical and esthetical way.

2 Terminology 2.1

Historical value

Ageing of building materials implies a contradiction. On the one hand materials are degrading over time and become les functional and less attractive. On the other hand the ageing of buildings in general and building materials in particular can lead to an increase of ‘values’ due to an addressed cultural significance and historical importance. Since the mid-nineteenth century, when people became - in a romantic way - interested in the past, we started to preserve historical buildings. Several nineteenth-century scholars such as John Ruskin (1819-1900), William Morris (1834-1896), and Alois Riegl (1858-1905) tried, all in their own way, to describe what the essence was of historical buildings and why (how) those buildings had to be preserved [3]. Ruskin for example writes in his Lamp of Memory: ‘… some mysterious suggestion of what it had been, and of what it had lost ; some sweetness in the gentle lines which rain and sun had wrought’ [4]. The Manifesto by Morris, issued in 1877, was a clear pamphlet against the common *

[email protected]

64


nineteenth-century behaviour of restoring buildings in a way they could have looked before without paying attention to the traces of time visible [5]. Riegl, from an art historical point of view, brought ‘alteswert’, the value of age, into the discussion [6]. Although people and society have changed over the years, the sentiments regarding the value and meaning of old buildings in general and built Cultural Heritage in particular did not change much. The age of a building is up till now present in many official documents on the formal protection of historical buildings such as the Dutch Law on Monuments (Monumentenwet 1988) [7].

2.2

Nature of materials

How does a building become liked in a historical way and when will visible signs of ageing be accepted? When a building is just finished damages to materials and constructions are not accepted and even damages or failures that occur within the first decennium are not accepted. The confronting articles by Hendriks in the Dutch periodical Detail in Architectuur painfully illustrated damages to modern materials and constructions in recent buildings due to badly designed or executed details [8]. People are willing to accept a certain degree of degradation due to ageing of materials when historical buildings are concerned or even tend to value buildings or building parts higher when the traces of time are visible. Research by Andrew [9, 10] indicated that blackening of sandstone facades sometimes added to the appreciation of buildings. Research on the perception of small scale damage and repairs of natural stone [1, 2] confirms this hypothesis. When do buildings start to be liked for its age value and become ‘monumental’ and how long does it take before people start accepting damages and does it have any relation with the nature of materials? The nature of materials and how people perceive materials is probably of high influence on the visible signs of ageing that are accepted. Studies from the nineteen-seventies indicate already that concrete is perceived negatively when compared to i.e. brick and wood [11, 12, 13, 14]. The capability of materials and the way they are combined in a building to age graciously is probably the most important factor for an old building to become liked [e.g. 15]. Traditional, natural building materials like natural stone, brick and wood have a certain robustness that are capable of withstanding ‘the tooth of time’ and therefor liked. Materials that could be called modern from an architectural point of view like concrete have an image problem regarding its ageing. Most people tend to dislike concrete structures because of its greyish image, illustrated by the debate on the appointment of modern buildings as a listed building. The qualities of concrete structures from an engineering point of view and from an esthetical point of view are often not recognised [16].

2.3

Damage

Assuming that considerations on preservation of historical buildings are often made by persons with different interpretative frameworks, depending on their education and professional experience, a clear definition of damage is would be helpful, as starting point for any decision on interventions. The definition of damage should be objective and commonly accepted. Several glossaries combined with damage atlases have been developed over the years, involving a wide range of specialists in conservation [e.g. 17, 18, 19, 20, 21]. Still the definition of damage is rather subjective. Damages to buildings that are technically identical are often handled in different ways, depending on the building concerned and the people involved in the conservation process. Apparently, not only the technical arguments are decisive. Research in The Netherlands and Belgium on the perception of interventions in historical buildings indicated that situations that do not clearly show an intervention are appreciated higher. Whether such situations were ‘authentic’ or not did not seem to have any influence on the outcome [22]. The harmony or esthetical compatibility of old and new building materials in historical buildings is highly appreciated. Although the parameters are difficult to define and most probable differ according to the cultural and professional background of people it is clear that ageing phenomena can contribute in a positive way to the valuation of historical buildings. 65


2.4

Patina

The Oxford Dictionaries define patina as ‘a green or brown film on the surface of bronze or similar metals, produced by oxidation over a long period’, but in the context of Cultural Heritage, patina is often referred to as the visible traces of time on the surface of a material or object [e.g. 3, 15]. The term ‘patina’ is often used to distinguish between the often highly valued (and to be preserved) traces of time and the undesirable (and to be repaired) ‘damages’. This use of the term ‘patina’ goes back to the nineteen-fifties and –sixties when there was an intensive scholarly debate among art historians about the cleaning and restorations of paintings, i.e. published in a series of articles in the Burlington Magazine [23] The famous Austrian Ernst Gombrich (19092001) and the Italian Cesare Brandi (1906-1988) took active part in the discussion. The discussion was about the value of patina, that on the one hand shows the age of the painting but on the other hand, due to darkening in time, hides the bright original colours. Gombrich, who was not in favour of cleaning, also brought into the discussion that some painters already took into account the ageing of the varnish by colouring it when applying and by doing that anticipating on the ‘patina’. The discussion on patina in the field of conservation of built cultural heritage has always been polarized. On the one hand the romantic scholars and architects that argue in favour of preserving as much authentic substance as possible and on the other hand the ones that plea either from a theoretical or pragmatic point of view - for restoration and reconstruction of the historical architecture. How the (Dutch) layman exactly fits in is not studied, but it is assumed that – looking at many neatly cleaned and restored houses – the majority of the Dutch fits into the second category (see Figure 1).

Figure 1 Although technically not necessary, visible traces of ageing are often removed, i.e. when cleaning brickwork facades (photo: W.J. Quist).

66


3 Decisions on intervention 3.1

Man-made traces versus natural weathering

Article 11 of the Venice Charter states that “The valid contributions of all periods to the building of a monument must be respected, since unity of style is not the aim of a restoration. When a building includes the superimposed work of different periods, the revealing of the underlying state can only be justified in exceptional circumstances and when what is removed is of little interest and the material which is brought to light is of great historical, archaeological or aesthetic value, and its state of preservation good enough to justify the action. Evaluation of the importance of the elements involved and the decision as to what may be destroyed cannot rest solely on the individual in charge of the work” [24]. This article is often referred to in many ways to advocate the preservation of the ‘architectural layers’ in a historical building, but can it be used in favour of preserving a ‘patina’? Figure 2 shows two different ‘traces of time’. Carvings, used to sharpen a knife or to gather stone powder with addressed healing capacities can be called ‘mechanical damage’, but due to its historical meaning the stone will never be repaired or replaced. The right image shows a black crust on sandy limestone. Is this ‘damage’ as safe as the first one? Although it is a comparable trace of time, the latter will often be cleaned away based on a hazy mix of technical and aesthetical reasons.

Figure 2 Left: Highly valued man-made traces at St. Bavo Church Aardenburg, right: black crust due to weathering at Old Church Delft (photos; W.J. Quist).

3.2

Complete cleaning of facades

The weathering of (sandy) limestone has been studied intensively in the nineteen-eighties and nineties (for example in the case of the Church of Our Lady in Breda [25, 26]). The general conclusion on the damaging mechanism resulting in the black gypsum crust was that this process was stable in many places, but it was also concluded that the dense and stiff cement repointing (from an earlier conservation campaign) had a big influence on the hygric behaviour of the façade, leading to los of material around this pointing [27]. The architect of the conservation campaign in the nineteen-nineties – who was in favour of the complete cleaning of the facades – used the problems of the cement pointing as a technical argument for cleaning: 67


complete removal of the gypsum crust would reveal all damages due to the repointing and would enable to replace the repointing and repair the stone. The necessity of complete cleaning was disputed by several experts, but in the end, the hasty mix of technical and aesthetical arguments resulted in cleaning [28]. On the one hand this conservation campaign resulted in loss of ‘traces of time’, but on the other hand the church was made ready for a new chapter in its long history (see figure 3).

Figure 3 Church of Our Lady in Breda, left situation before cleaning (photo: Architectenburo J. van Stigt bv), right situation after cleaning (photo: W.J. Quist).

Figure 4 Application of an artificial patina on new sandstone to visually match old sandstone at St. Lawrence Church Rotterdam (photo left: W.J. Quist, photo right: T.G. Nijland)

3.3

Artificial patina

Although only recently applied on several important historic Dutch buildings, the application of an artificial patina on restored facades has a long history. Among (stone)masons it was common to use dirt from gutters or ink to darken newly constructed masonry or pointing next to old masonry [29], but since a few decades the use of artificial products to colour stone became in use. Together with careful partial cleaning of old blocks of sandstone, the application of an artificial patina on the brightly yellowish looking new blocks of Bentheim sandstone helped to unify the architecture of the façade of the Royal Palace in Amsterdam [30, 31, 32, 33]. Figure 4 68


shows the tower of the St. Lawrence Church in Rotterdam that has recently been restored in sandstone. Partially the new blocks have been aged artificially to aesthetically match the greyish old stones. The new sandstone for the balustrade and the pinnacles on the corners of the towers were seen as an architectural unity on its own and therefore not aged artificially. The technical possibilities of applying artificial ageing (i.e. by pigments fixated by ethyl silicate) imposes an ethical question. Can it also be used to colour a different type of stone to match its surroundings? Figure 5 shows two examples of types of stone that were coloured to match its surroundings, not by anticipating on its change of colour when ageing but by altering its natural colour. Figure 5 (left) shows a pinnacle in Peperino Duro coloured like it is Weibern tuff stone. A highly durable type of stone has been used and afterwards coloured to match. Is it fake, or is it an example of ‘making use of all possibilities there are’? Figure 5 (right) shows the effect of ageing of artificial ageing. Due to an almost complete ban on the use of sandstone from 1954 onwards it was in many cases necessary to choose another type of stone for replacement of degraded sandstone [28, 34]. In the case of the Central Station of Amsterdam white limestone with an artificial colouring was chosen to replace brownish sandstone. After 25 years the colour on the limestone has been washed away and both types of stone can easily be distinguished. This example clearly shows that aesthetics were once a reason to apply colour, but is this treatment repeatable, both technically and financially?

Figure 5 Left: colour on Peperino Duro to visually match Weiberner tuff stone (St. Johns’ Cathedral, ‘sHertogenbosch), right: the applied artificial patina on limestone has been washed away by rain (Central Station Amsterdam), photo’s W.J. Quist.

4 Discussion and conclusions The past decades theory of architectural conservation has brought forward several concepts on how to handle in case of historical buildings. The most important and well known ones are ‘reversiblity’ as a follow up on the Charter of Venice and ‘minimal intervention’ as advocated by Brandi [28]. Since the nineties of last century, ‘compatiblity’ became more and more the leading principle for decisions on intervention in existing fabric. Teutonico et al. defined it as: “A treated 69


material should have mechanical, physical and chemical compatibility with the untreated historic materials under consideration. Simply stated, compatibility means that introduced treatment materials will not have negative consequences” [35]. This definition is workable, but represents a technocratic view on materials and is missing any link to the aesthetical representation of those materials and structures. Any ‘aged’ situation can be evaluated according to uniform technical standards, but when it comes to evaluating the ‘visible traces of time on the surface of a material or object’ it often comes down to the individual perception of the people in charge. Therefor we would like to plea for the development of criteria for intervention or non-intervention that take aesthetical aspects into consideration. To set the criteria, it is necessary to study in-depth how the ageing of materials and structures is perceived by professionals in the field of conservation technology, but also by laymen. Based on the fragmentary research presented in this paper it is suggested that this could lead to less, but precisely targeted, intervention. This would make the ‘patina’ not only the result of neglect and undesired ageing but the result of careful technical, ethical en aesthetical considerations and therefor truly noble.

5 References [1] Quist, W.J., Van Hees, R.P.J., Naldini, S. and Nijland, T.G. (2007) De beleving van schade en reparaties aan natuursteen. In: Praktijkboek Instandhouding Monumenten, (30) : 15 pp. [2] Quist, W.J, Van Hees, R.P.J., Naldini, S. and Nijland, T.G. (2008) The perception of small scale damage and repairs of natural stone. Proceedings of the 11th International Conference on Durability of Materials and Components, Istanbul, paper T15. [3] Jokilehto, J. (1999) A history of Architectural Conservation, Oxfort [4] Ruskin, J. (1989) The seven lamps of architecture, 1849. Reprint: Dover Publications, Mineola [5] Morris, W. (1877) The manifesto of the Society for the Protection of Ancient Buildings. http://www.spab.org.uk/what-is-spab-/the-manifesto/ Accessed 23 February 2014 [6] Iversen, M. (1993) Alois Riegl: Art History and Theory, The MIT Press, Cambridge, Massachusetts, London [7] Monumentenwet 1988. http://wetten.overheid.nl/BWBR0004471/geldigheidsdatum_25-02-2014 Accessed 23 February 2014 [8] Hendriks, N. (2001-2004), several aricles, in Detail in architectuur [9] Andrew, C., et al (1994) Stonecleaning – a guide for practitioners, Historic Scotland & The Robert Gordon University [10] Andrew, C. (2002) Perception and aesthetics of weathered stone façades. In: Přikryl, R. & Viles, H.A., red., Understanding and managing stone decay. Karolinum Press, Praag, 331-339. [11] Van Wegen, H.B.R. (1970) Onderzoek naar de belevingswaarde van vier bouwmaterialen met behulp van de semantische differentiaal – techniek. Centrum voor Architectuuronderzoek, TH Delft, Delft. [12] De Jonge, D. (1971) Over de belevingswaarde van enige bouwmaterialen. Centrum voor Architectuuronderzoek. TH Delft, Delft. [13] Brunsman, P. (1976) Beleving van monumenten. Dr. E. Broekmanstichting, Amsterdam. [14] Steffen, C. (1983) De beleving van gevelvervuiling. Centrum voor Architectuuronderzoek, TH Delft, Delft. [15] Denslagen, W., Querido, J., Vries, A.D. (1978) De tand des tijds; The tooth of time, RV bijdrage 07, SDU Uitgeverij [16] Heinemann, H.A. (2013) Historic Concrete: From Concrete Repair to Concrete Conservation, dissertation TU Delft [17] Fitzner, B., Heinrichs, K. and Kownatzki, R. (1995) Weathering forms- classification and mapping, Verwitterungsformen - Klassifizierung und Kartierung. Denkmalpflege und Naturwissenschaft, Natursteinkonservierung 1. Ernst & Sohn, Berlijn, 41-88. [18] ICOMOS-ISCS (2008) Illustrated glossary on stone deterioration patterns / Glossaire illustré sur les formes d’altération de la pierre. ICOMOS, Champigny. [19] Löfvendahl, R., Andersson, T., Åberg, G. and Lundberg, B.A. (1994) Natursten i byggnader. Svensk byggnadssten & skadebilder. Riksantikvarieämbetet, Stockholm. [20] Naldini, S., Hees, R.P.J. van and Nijland, T.G. (2006) Definitie van schade aan metselwerk. Praktijkboek Instandhouding Monumenten, 28(19): 19 pp. [21] Van Hees, R.P.J. and Naldini, S. (1995) Masonry Damage Diagnostic System. International Journal for Restoration of Buildings and Monuments, 1: 461-473.

70


[22] Quist, W.J., De Kock, T., Nijland, T.G., Van Hees, R.P.J. and Cnudde, V. (2014), Conservering van witte steen: verbetering of verspilde moeite? De beleving van interventies in Vlaanderen en Nederland , in Geologica Belgica (in press) [23] Several articles by different authors, in Burlington Magazine, 1949-1967 [24] International Charter for the Conservation and Restoration of Monuments and Sites (the Venice Charter 1964), http://www.international.icomos.org/charters/venice_e.pdf Accessed 23 February 2014 [25] Naldini, S. and Van Hees, R.P.J. (1993) Monitoring the decay of monuments - the Church of Our Lady in Breda – part a, TNO Building and Construction Research, Delft [26] Verhoef. L.G.W. and Koopman, F.W.A. (1993) Monitoring the decay of monuments – the church of our lady in Breda – part B,Delft [ 27] Quist, W.J. and Van Hees, R.P.J. (2006) De reiniging van de Grote Kerk in Breda tien jaar later. In M.L. Stokroos et al. (red.), Praktijkboek instandhouding monumenten, Den Haag [28] Quist, W.J., (2011) Vervanging van witte Belgische steen. Materiaalkeuze bij restauratie. PhD thesis, TU Delft, Delft. [29] Quist, W.J., Nijland, T.G. and Van Hees, R.P.J. (2013) Replacement of Eocene white sandy limestone in historic buildings. Over 100 years of practice in The Netherlands. Quarterly Journal of Engineering Geology and Hydrology, dx.doi.org/10.1144/qjegh2013-023. [30] Van Bommel, B. (2013) Terugblik op een geslaagd project. De restauratie van het Koninklijk Paleis Amsterdam, in Bulletin KNOB 112, p. 68-79 [31] Nijland, T.G. (2013) Reinigen en retoucheren. De restauratie van de zandsteengevels van het Koninklijk Paleis Amsterdam, Bulletin KNOB 112, p. 112-121 [32] Van Bommel, B. (2005) Assessment van ingrepen bij vergrijsde gevels. In: Lagrou, D. & Dreesen, R., red., Belgische natuursteen in historische monumenten en hun vervangproducten bij restauratie in België en Nederland. 1e Vlaams-Nederlandse Natuursteendag, Leuven. VITO, Mol. [33] Van Bommel, B. (2009) Uitgangspunten gevelreiniging gereviseerd. Aantekeningen bij het artikel Assessment van ingrepen bij vergrijsde gevels van 2005. In: Nijland, T.G., red., Schoonheidsbehandeling of make-over: Hoe gaan we met de monumentenhuid om ? Syllabus TNO-NVMz studiedag, Delft, 69-92. [35] J.M. Teutonico, A.E. Charola, E. de Witte, G. Grasegger, R.J. Koestler, M. Laurenzi Tabasso, H.R. Sasse and R. Snethlage (1997) ‘Group Report How Can We Ensure the Responsible and Effective Use of Treatments (Cleaning, Consolidation, Protection)?’. In: N.S. Baer, and R. Snethlage, (eds.), Dahlem Workshop on Saving Our Architectural Heritage: Conservation of Historic Stone Structures, Chichester, p. 293-313

71


FUNDAMENTALS OF AGEING OF MATERIALS

72


Examining of weather resistance of ETICS with stresses which correspond to weather conditions in Finland Petri J. Annila1*, Toni A. Pakkala1, Jommi Suonketo1, Jukka Lahdensivu1 (1) Tampere University of Technology, Tampere, Finland Abstract: The Department of Civil Engineering at Tampere University of Technology (TUT) has performed 37 weathering resistance tests according to the national test procedure since the year 2006. 31 of these structures tested have been ETICS with thin-layers of exterior plaster and without mechanical anchors. The stress cycle, which is defined in a national guidelines, has been designed to correspond to the Finnish weather conditions as much as possible. These 31 tests have included 395 bond strength tests and 804 impact resistance tests. Possible cracking and other deterioration have also been visually followed during the weathering resistance tests. The tests have shown that bond strength and impact resistance are highly dependent on the thermal insulation which is used in ETICS. Keywords: ETICS, accelerated ageing, weather resistance, rendering, thermal insulation

1 Introduction External Thermal Insulation Composite Systems (ETICS) are used globally. These structures consist of a thin-layer of exterior plaster (typically 6-10 mm in Finland) which is plastered to exterior surface of thermal insulation layer. The thermal insulation used can be a mineral wool (MW), expanded polystyrene (EPS), extruded polystyrene (XPS) or polyurethane (PUR). The insulation layer has been fixed with mechanical anchors or by adhesive mortar to the concrete or masonry wall. This structure is also referred to by the names External Wall Insulation Systems (EWIS) in Ireland and the United Kingdom and Exterior Insulation and Finish Systems (EIFS) in North America. ETICS are popularly used to improve energy efficiency and weather resistance of old façade structures [1]. Nowadays ETICS are also used in prefabricated elements in Finland. In these elements there is typically only a thin-layer of plaster in exterior surface of thermal insulation, which is made in element factory. The main task of this thin layer is to protect the thermal insulation from the weather stresses, primarily from the UV radiation. Finishing coats and glass fibre reinforce mesh is installed in construction site. Accelerated climate ageing in laboratory is test method which gives information about possible long-term durability issues and deterioration mechanisms of structure at short time period. Accelerated climate ageing tests are also sometimes called weather resistance testing or hygrothermal behaviour testing. Daniotti et al. [2] and Griciutė et al. [3] present the principles how to create reliable accelerated ageing cycle which would be as similar as natural ageing. The basic idea, in this planning process, is to find out climate conditions under which façade structures deteriorate. Climate conditions differ significantly depending on the location which is examined. Therefore, there is multiple different accelerated climate ageing test methods. Next step in planning process is to determine frequencies of these extreme conditions. After the analysis of extreme conditions and their frequencies in natural ageing these phenomena are connected into stress cycle, which is repeated several times during the accelerated climate ageing test. ETAG004 External Thermal Insulation Composite Systems (ETICS) with rendering guideline is published by European Organisation for Technical Approvals (EOTA) [4]. ETAG004 introduce laboratory tests for ETICS for example accelerated climate ageing test, bond strength test, hard body impact resistance test and water absorption (capillarity) test. In spite of a general guideline *

Tampere University of Technology, P.O.B. 600, 33101 Tampere, Finland. [email protected]

73


in Europe there are numerous different accelerated climate ageing test methods for ETICS. Some examples of used tests are presented in article of Griciutė et al. [3]. For accelerated ageing of ETICS have been used a couple of different test methods also in Finland [5, 6]. Also test which is presented in ETAG004 have been used. Test which is used nowadays is presented in paper by Pakkala and Suonketo [7]. The same test method can be found also in national guideline by57 Insulation and panel render systems 2011 [8]. TUT exposes the test wall to the weathering test cycle recommended in the by57 consisting of three phases: 1) Spray irrigation for 60 minutes: 1 l/min/m², 2) Freezing for 240 minutes: temperature drops quickly to -20 °C and 3) Radiation heating for 180 minutes: temperature increases quickly to +60 °C. Test cycle is repeated 100 times during the accelerated ageing tests. The cycle is based on ETAG004 but revised to meet better Finnish climate conditions. The main difference is the freezing phase right after spraying water, i.e. the surface of a test wall and ambient conditions are wet. That is often the situation in Finnish outdoor conditions during autumn and spring time when the daytime temperature is slightly above 0 °C and by night below it. At the same time, especially in autumn and winter time, the relative humidity is commonly very high and level of solar radiation low which makes the drying of pore structure slow. The Finnish Meteorological Institute (FMI) has studied the total amount of freeze-thaw cycles in different locations across Finland. Between 1961 to 2009 the average amount of annual freeze-thaw cycles, i.e. temperature has dropped below 0 °C has been e.g. in south coastal area (Vantaa) 90,4 and in inland (Jyväskylä) 92,2 [9]. However, the pore water does not freeze immediately as temperature reaches 0 °C. For example on concrete it has been shown that the temperature needs to reach approx. -5 °C for water to freeze in capillary pores [10]. The annual amount of freeze-thaw cycles when temperature has dropped below -5 °C or lower has been on average 21,5 in Vantaa and 29,6 in Jyväskylä. The amount of freeze-thaw cycles itself is not the most essential indicator for studying possibility of frost damage because frost damage needs also water on pore structures to occur. As the water in the pore structure freezes, it expands about 9 % by volume creating hydraulic pressure in the system. If at the same time the level of water saturation of the system is high, the overpressure cannot escape into air-filled pores which causes damage in internal structures of the cement based materials. With facades water enters the pore structure e.g. as a consequence of driving rain or sleet. That is why the FMI also calculated the amount of freeze-thaw cycles at most 3 days after rain or sleet. The amount of such cycles differ depending on the location and also when we take into account the temperature needed for water to freeze in the pore structure, see table 1 [11]. In figure 1 can be seen annual changes in Vantaa. Table 1. Number of annual freeze-thaw cycles at four different observation stations at most 2 days after rain or sleet events at present climate [11]. Location Vantaa Jyväskylä

Temperature under (at most 3 days after rain or sleet) -2 °C -5 °C -10 °C 22,0 13,6 6,1 23,5 16,4 8,2

As table 1 and figure 1 show the amount of freeze-thaw cycles when ambient conditions are wet can be remarkably high during one year which has to be taken into account on testing materials which are facing such harsh climatic conditions.

74


40

35

30 Min. T max. 3 days after rain

Number of cycles

25

-0 °C

20

-2 °C -5 °C -10 °C

15

10

5

0

Figure 1. Number of annual freeze-thaw cycles from 1961 to 2006 at most three days after rain of sleet in southern Finland.

Testing the weathering resistance of ETICS in Finland consists of build test and unstressed reference walls in TUT laboratory facilities, visual observation during and after the test, bond strength tests and impact resistance tests. Tests actions are closely described in by57 publication [8]. Test wall is examined visually each working day during the test and the progress of visual cracks or other defects are listed. After the test possible damages are also investigated by surface moisture detector. After the 100 stress cycle is carried out 3 and 10 Joules hard body impact tests and bond strength tests. Tests were conducted according to the standard ISO-7892 [12]. Tests procedure is shortly describe in ETAG004. ETICS are ranked to three categories on the basis of the results. Categories are described in section 3.2. Bond strength tests after stress cycles determine the bond between a base coat and an insulation material. Bond strength tests are performed regular tensioning speed. Ultimate bond strengths are recorded and typically expressed in MPa (N/mm²).

2 Materials and methods

TUT has performed 37 weathering resistance tests according the national guideline by57 [8] since the year 2006. 31 of these tested structures have been ETICS. In the examined ETICS the insulation layer has been 15 times mineral wool lamellas (MW-L), 4 times mineral wool boards (MW-B), 6 times expanded polystyrene (EPS), 3 times extruded polystyrene (XPS) and 3 times polyurethane (PUR). Test structures are typically made on the surface of old concrete wall, because one has wanted to study the operation of ETICS that have been used to improve energy efficiency of old buildings. These old concrete walls have been built in connection with previous studies and they have dried in the laboratory conditions for several years. A few times have also been tested prefabricated elements that have been done in element factory. In these cases the used concrete cover is casted during the manufacturing. Thickness of insulation layer is typically 150-200 mm in test walls. At the end of year 2013 the results of weathering resistance tests have been collected in a database. Database consist of basic information of examined structures (materials, layer 75


thicknesses); visual observations (e.g. cracks and other damage that have occur during the tests); the results and failure modes of bond strength tests; and the failures of hard body impact resistance tests. The database is updated after every new test and now the database consists of 395 bond strength test results and 804 impact resistance test results.

3 Results and discussion

In the weathering resistance tests of TUT the following issues have been found to be main weaknessess of ETICS [13]: - Bond strenght is below 0,08 MPa, which is requirement of Finnish national guideline by57. Futhermore, high variations is found in the results. - The impact resistance is weak and none of examined ETICS is not suitable for use class I (following ETAG004 instructions). One reason for weak impact resistance is that the thicknesses of render is not built by following manufactures on instruction. - In spite of the fact that the professionals, determined by the subscriber of study, has made the test walls, there could be found a lot of work mistakes. The weakening of the insulation layer is the biggest shortcoming that has been perceived in TUT weathering resistance tests. However, bond strenght of ETICS is in several cases high enough that the problem would not exist in the building stock. The suction pressure of the wind is usually below 0,003 MPa in the most Finnish buildings [14]. The most general failure mode in bond strength test, is the breaking in the thermal insulation layer. This failure mode occured in 74,1 % bond strength tests of actual test wall and 78,7 % bond strenght tests of reference wall. Other failure modes are failure occured in the middle of the render, or in the interface of render ja insulation material. Failures in bond strength tests are classified into these 3 category depending on where at least 50 % of the fracture area is located. On the basis of this, thermal insulation is the weakest layer, when bond strength of the structure is examined. Freeze-thaw durability of porous facade materials is important because they are exposed for over 90 freeze-thaw cycles in every year [9]. Freeze-thaw durability of the render products has been often studied separately before the accelerated ageing test. The typical failure mode in bond strength tests and good freeze-thaw resistance of used renders are main reason why next chapters focus on influences of different insulation materials. The weak impact resistance could be improved by using a render thicknessess that have been recommended or by adding extra glass fiber reinforced mesh in base coat. The results of TUT have shown that the extra reinforced mesh has a clear effect on the impact resistance of the structure.

3.1

Effect of insulation material on the bond strengt of ETICS

Figure 2 shows the cumulative distribution of all ETICS bond strengt tests that have been made at TUT. The red curve presents results of test walls subjected to weathering action and the blue curve presents results of reference wall. The average bond strength in weathered test walls was 0,0995 MPa and 0,0798 MPa in reference walls. Standard deviation was 0,0613 MPa and 0,0526 MPa respectively. Figure 2 shows that 56 % of bond strengt results in weathered test wall and 42 % in reference wall is below 0,08 MPa. The Finnish requirement for bond strength tests is 0,08 MPa [8]. This limit has been described with a thick black vertical line in figure 2. The space between red and blue lines shows the weakening of bond strength during the weather resistance test. However, the thermal insulation that has been used in ETICS affects significantly the bond strength. Table 2 describes the differences of the thermal insulation that have been used in ETICS. The table shows number of the bond strength tests (n); average of bond strenght tests for both test and reference wall; and residual bond strength.

76


100 % 90 % 80 % 70 % 60 % 50 % 40 % 30 % 20 % 10 % 0% 0

0,05

0,1

0,15 Bond strength (MPa)

Test wall subjected to weathering action

0,2

0,25

Reference test specimen

0,3

0,08 MPa

Figure 2 Distribution of all bond strength test results. Table 2 Bond strength results and influence of used thermal insulation.

n 24 92 36 19 18

MW-B MW-L EPS XPS PUR

Test wall average (MPa) 0,011 0,068 0,117 0,169 0,061

n 21 72 30 18 18

Reference structure average (MPa) 0,018 0,093 0,125 0,215 0,063

Residual bond strength 63 % 74 % 93 % 79 % 96 %

Figure 3 demostrates visually the same results that have been presented in table 2. The continuous lines present results of actual test wall and the dashed lines results of reference structures. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0

0,04 0,08 MPa PUR, R MW-B, R EPS, R

0,08 MW-B EPS MW-L XPS

0,12 0,16 0,2 Bond strength of structure (MPa) MW-B, R EPS, R MW-L, R XPS, R

MW-L XPS PUR

0,24

MW-L, R XPS, R PUR, R

0,28 PUR MW-B EPS

Figure 3 Distribution of bond strength test results

The results show (table 2 and figure 3) that the bond strength weakens most in mineral wool boards (MW-B); and almost as much in mineral wool lamellas (MW-L) and XPS boards during the weather resistance test. The residual bond strength was 63 % in mineral wool boards, 74 % in mineral wool lamellas and 79 % in XPS. 77


The average bond strength of mineral wool boards was only 0,011 MPa. However, the low bond strength compared to the Finnish requirement (0,08 MPa) is not a problem, because mineral wool boards are typically fixed with mechanical anchors. In the mineral wool lamellas and XPS products the weakening was considerable, but however the average strength (0,068 MPa and 0,169 MPa respectively) is still multiple compared to stresses in natural ageing (e.g. wind pressure and gravity). The ETICS with EPS and PUR insulation withstand the stresses best; the weakening during the weather resistance tests was only 7 % and 4 % respectively. These results represent durability of ETICS only in laboratory tests. Due to this it is important to study durability of ETICS in natural climate conditions and compared those results for these laboratory test results. This is the next and significant step when developing weathering resistance tests. Zirkelbach et al. [15] have compared the differences between laboratory tests and natural ageing. These test have been performed only with mineral wool boards (MW-B) and lamells (MW-L). When compairing Zirkelbach et al. natural ageing tests results and TUT weathering test results, the weakening of bond strenght is approximately the same level.

3.2

Effect of insulation material on the Impact resistance of ETICS

Based on the damage caused by hard body impact tests, the systems are placed in the following three different use categories according to the ETAG004 Guideline [4]: I Ground-level walls with unhindered public access (street level) II Walls at impact height in buildings which people take care of themselves (courtyards, attached houses, etc.) or areas that can have objects thrown at them (e.g. second storey street façade) III Areas unlikely to be subject to impacts or thrown objects (upper floors). In Finland is also used category X, which means that ETICS is not suitable for categories I-III. Structure will be classified to the category X, if the 3 Joule hard body impact causes circular cracking which penetrated the rendered surface. The example of circular cracking is presented in figure 4; on the left circular cracking in exterior surface of render, and on the right interior surface of render.

Figure 4 The circular cracking in hard body impact tests.

Figure 5 represents how ETICS have been distributed into different use categories depending on used thermal insulation. Impact resistance of ETICS with PUR products is weakest; every structure is classified in category X. Results show also that 30-40 % of ETICS with EPS or mineral wool lamellas, cannot withstand stresses caused by 3 Joules hard body impact. It has not been possible to classify even one of ETICS to use category I, irrespective of used insulation material. 65 % of ETICS with XPS insulation could be classified in category II on the basis of weather resistance tests of TUT. 78


100 % 90 % 80 % 70 % 60 % 50 % 40 % 30 % 20 % 10 % 0% I

II MW-B

III

MW-L

PUR

EPS

X XPS

Figure 5 Effect of insulation material to use category of ETICS.

In some of the tests is also examined the effect of additional glass fibre reinforced mesh. Figure 6 represent how additional mesh (e.g. EPS, DM) is influenced on the impact resistance of ETICS. In figure 6 solid columns represents structures with additional mesh and oblique stroke pattern structures with one mesh. The classification into different use categories has been made only on the basis of 10 Joules tests. Results show that the additional reinforced mesh improves impact resistance of ETICS; in which has been used mineral wool boards, mineral wool lamellas, polyurethane or extruded polystyrene. However, impact resistance of ETICS with expanded polystyrene, is not improved due to additional mesh. The improving of the impact resistance is a partly consequence of thicker base coat layer and partly additional glass fibre reinforced mesh. 100 % 90 % 80 % 70 % 60 % 50 % 40 % 30 % 20 % 10 % 0% I MW-B, DM PUR

II MW-B EPS, DM

MW-L, DM EPS

III MW-L XPS, DM

PUR, DM XPS

Figure 6 Effect of insulation material and additional mesh (DM) to use category of ETICS. The classification has been made only on the basis of 10 Joules tests.

4 Conclusion The properties of the thermal insulation materials vary between the different products. These variables are such as price, long-term durability, hygrothermal behaviour, deterioration mechanisms, strength properties and fire safety. The whole complexity of structure must be examined when thermal insulation product is chosen to ETICS. 79


TUT results have shown that for example bond strength of ETICS with mineral wool boards weakens most during the weather resistance test (table 2 and figure 3). However, impact resistance of ETICS with mineral wool boars is the best: almost 70 % structures can be used in ground level with additional mesh (figure 6). Another example is polyurethane structures: bond strength weakens only 4 % during the weather resistance test (table 2 and figure 3), but the impact resistance of the structure is weak (figure 5). In the future, the next step will be to compare the results of accelerated ageing tests with results which are collected from building stock.

5 References

[1] Künzel, H., Künzel, H.M. and Sedlbauer, K. (2006) Long-Term Performance of External Thermal Insulation Systems (ETICS). Architectura 5 (1) 2006, 11-24. [2] Daniotti, B. (2008). Climatic Data Analysis to Define Accelerated Ageing for Reference Service Life Evaluation. Proceedings of 11th International Conference on Durability of Building Materials and Components, Istanbul 11-14 May 2008. [3] Griciutė, G., Bliūdžius, R. and Norvaišienė, R. (2013) The Durability Test Method for External Thermal Insulation Composite System (ETICS) used in Cold and Wet Climate Countries. Journal of Sustainable Architecture and Civil Engineering 2013. No. 1(2) pp 50-56. [4] European Organisation for Technical Approvals (2013) Guideline for European Technical Approval of External Thermal Insulation Composite Systems (ETICS) with Rendering. Edition 2000. Amended Februay 2013. Brussels. 143 p. [5] NT BUILD 495 (2000) Nordtest Method. Building materials and components in the Vertical position: Exposure to accelerated climatic strains. Finland. 2000. 4 p. [6] Nieminen, J. (1997) Case Studies on External Wall Insulation Systems Faced with Thin Render. Proceedings of the International Conference on Building Envelope Systems and Technologies, Bath 1997. [7] Pakkala, T. and Suonketo, J. (2011) Hygrothermal Behaviour Testing of External Thermal Insulation Composite Systems with Rendering in Nordic Climate. Proceedings of the 12th International Conference on Durability of Building Materials and Components, Porto 12-15 April 2011. [8] Concrete Association of Finland (2011) by57 Insulation and panel render systems 2011. Helsinki, Finland. 2011. 196 p. (in Finnish) [9] Jylhä, J., Ruosteenoja, K., Tietäväinen, H., et al. (2011) Rakennusfysiikan ilmastollisten testivuosien sääaineistot nykyisessä ilmastossa ja arviot tulevaisuuden muutoksista. Väliraportti. Finnish Meteorological Institute. Helsinki. 6 p. 20 app. (in Finnish) [10] Pigeon, M., Pleau, R. (1995). Durability of concrete in cold climates. Suffolk. E & FN Spon. 244 p. [11] Lahdensivu, J. (2012) Durability Properties and Actual Deterioration of Finnish Facades and Balconies. Tampere University of Technology. Faculty of Built Environment. Tampere. Publication 1028. 119 p. 37 app. [12] ISO 7892 (1988) Vertical building elements – Impact resistance tests – Impact bodies and general test procedures. [13] Annila, P. J. (2013) TTY:n kokemukset eriste- ja levyrappausten säänkestävyystutkimuksista. Proceedings of Rakennusfysiikka 2013. 22-24 October 2013. Tampere, Finland. pp. 113-120. (in Finnish). [14] Finnish Standards Association SFS (2011) SFS-EN 1991-1-4 + AC + A1 Eurocode 1: Actions on structures. Part 1-4: General actions. Wind actions. Helsinki, Finland 2011. 254 p. [15] Zirkelbach, D., Holm, A. and Künzel, H.M. (2005) Influence of temperature and relative humidity on the durability of mineral wool in ETICS. 10DBMC International Conference On Durability of Building Materials and Components. Lyon, France. 17-20 April 2005. 8 p.

80


Ageing behaviour of neutron irradiated Eurofer97 I. Carvalho1, 2, 3*, A. Fedorov2, M. Kolluri2, N. Luzginova2, H. Schut3, J. Sietsma4 (1) Materials innovation institute (M2i) Delft, The Netherlands (2) Nuclear Research and consultancy Group (NRG), Petten, The Netherlands (3) Delft University of Technology (TUDelft), Faculty of Applied Sciences, Delft, The Netherlands (4) Delft University of Technology (TUDelft), Faculty of Mechanical, Maritime and Materials Engineering, Delft, The Netherlands Abstract: Eurofer97 is candidate structural material for nuclear fusion reactors. To better understand the ageing effects due to radiation during a long term use in real fusion conditions, Eurofer97 was neutron irradiated at the High Flux Reactor in The Netherlands. TEM images of post-irradiation Eurofer97 reveal a high density of irradiation induced dislocation loops, fine precipitates and agglomeration of point defects. Microscopy results are related to post-irradiation tensile mechanical tests done at room temperature and at 300 °C. An increase of yield and ultimate tensile strength combined with a decrease of total elongation is observed in both tests and correlated with the presence of radiation damage. Keywords: Eurofer Chromium steel, radiation damage, nuclear fusion, High Flux Reactor, mechanical properties

1 Introduction

Reduced Activation Ferritic Martensitic (RAFM) Eurofer97 steel is a primary candidate to be used as a structural material in fusion reactors [1, 2]. The chemical composition of this steel contains Fe, Cr, W, V, Ti, Ta and C, which are low activation elements that allow recycling of the waste in 100 years time [1]. As a structural material, Eurofer97 will not only be exposed to a neutron environment but also to thermo-mechanical loading. Both conditions cause microscopic alterations. In a fusion 14 MeV neutron spectrum, helium and hydrogen will be produced via transmutation reactions. Although Eurofer97 has a low sensitivity to radiation-induced swelling and helium embrittlement under neutron irradiation [1], a detailed understanding of the effects of irradiation and temperature is crucial for a correct design and application of Eurofer97 in a fusion reactor. Analysis of Eurofer97 mechanical properties has been previously reported [3, 4] but a correlation with the type of defects found in the material after irradiation has not yet been clearly established. This work is among the first that aim to establish the relation between radiation damage and deterioration of mechanical properties. Aiming at a better understanding of the ageing effects of the long term use of Eurofer97 in a fusion reactor, this steel was neutron irradiated at the High Flux Reactor (HFR) in Petten, The Netherlands. Irradiation programs with temperatures of 60 and 300 °C and neutron doses of 2.5 and 10 displacements per atom (dpa) [3, 4] were completed. This paper focus on the 10 dpa and 300 °C irradiation. To comprehend the consequences of the irradiation damage on the mechanical properties of the Eurofer97, Transmission Electron Microscopy (TEM) investigations are correlated with post-irradiation mechanical tensile tests.

*

M2i, NRG, TUDelft, [email protected]

81


2 Experimental 2.1

The material

European Union batches of Eurofer97 were produced in Böhler, Austria. The nominal composition of Eurofer97 is Fe-9Cr-1W-0.2V-0.1Ta-0.1C (wt.%). After fabrication the material was subject to austenitization at 980 °C for 30 min, followed by annealing at 760 °C for 90 min to achieve a tempered martensitic structure [5]. Both steps were followed by air cooling.

2.2

Irradiation program at HFR

Eurofer97 material was irradiated in multiple campaigns at HFR [4]. The irradiation program of the Eurofer97 piece that investigated by TEM is designated ‘in-SodiUm steel Mixed specimens irradiation 04’ (SUMO-04). In this program a set of different steels, already cut to the correct specimen shape to perform mechanical post irradiation examination, were irradiated with 10 dpa at a temperature of 300 °C. The calculated helium content is 12.6 appm. The specimens are placed in irradiation rigs which are filled with sodium to ensure good heat conductivity. The irradiation temperature is determined by the balance between the gamma heating and the heat dissipation via the gas gaps introduced in the sample holder [3]. The temperature of the irradiation is controlled with 20 thermocouples mounted on the specimens. As for the neutron monitoring, 13 detectors are placed in key positions within the irradiation rig. The uncertainty of the irradiation dose in terms of dpa is 14 % [3].

2.3

Sample preparation

TEM discs were cut from broken pieces of a fracture toughness specimen that was postirradiation mechanically tested at RT, in a region away from the fracture zone. Discs with a diameter of 3 mm and a thickness of approximately 100 µm were manufactured by a sequence of grinding and polishing steps in hot cells. The final thinning was done by electro-polishing with a solution of 135 ml of acetic acid and 15 ml of per-chloric acid.

2.4

TEM examination

The TEM investigation reported here was done at NRG using a JEOL 1200ex STEM/TEM microscope with an accelerating voltage of 120 kV.

3 Results 3.1

Microstructure of unirradiated Eurofer97

TEM images of unirradiated Eurofer97 are shown in Figure 1 [6]. On the left, the overview of the sample reveals the typical lath martensitic structure expected for Eurofer97 [7]. According to specifications [1] the Eurofer97 used in this work has a grain size in the range 9 – 23 µm. Coarse spheroidal shaped particles are present mostly along the grain boundaries and have a size of ~ 100 nm. Although at a lower density, platelet particles are observed with sizes in the range of 100 – 200 nm [6]. The composition of the particles was not studied but the observed big (~ 100 nm) spheroidal shape particles can be recognised as M 23 C 6 or MX – type phases (where M is a Fe, Cr or W and X is Ta or V) based on literature [7, 8].

82


Figure 1 Unirradiated Eurofer97. On the left is an overview of the lath microstructure expected for Eurofer97. On the right, a detailed image reveals spheroidal shaped particles present close to grain boundaries [6]

3.2

Microstructure of 10 dpa, 300 °C neutron irradiated Eurofer97

Figure 2 shows a low magnification TEM image of 10 dpa, 300 °C neutron irradiated Eurofer97. In the figure the pre-irradiation lath structure [6, 7] is observed. Precipitates are observed mostly around the grain boundaries but also inside the gains. The composition of the precipitates was not investigated but, as noted for unirradiated Eurofer97, these defects are expected to be M 23 C 6 or MX – type precipitates [8]. Furthermore, other authors identified the formation of small α’- (chromium-rich precipitates) and M 6 C-phases inside the grain as a consequence of irradiation enhanced diffusion [9, 10]. The dimensions of the precipitates range between 20 and 200 nm. Radiation damage, characterized by a high density of uniformly distributed “black dots” (identified as an agglomeration of point defects or fine precipitates [11]) is found throughout the material. On Figure 3 an image of a Eurofer97 grain damaged by neutrons is shown. Two types of radiation damage are distinguishable: black dots and dislocation loops. The black dots reach dimensions of approximately 10 nm and are dispersed uniformly in the grain. The dislocation loops reach dimensions of 15 nm and are also located throughout the grain.

Figure 2 Overview of the microstructure of 10 dpa 300 °C neutron irradiated Eurofer97 irradiated at HFR. The pre-irradiation lath can be observed. Radiation damage is spread throughout the sample

83


Figure 3 Detail of the microstructure of 10 dpa 300 °C neutron irradiated Eurofer97 showing the irradiation induced defects. Black dots are marked by arrows and dislocation loops are encircled

4 Discussion Comparing the TEM investigations of pre- and post-neutron irradiation Eurofer97, significant microstructural changes are observed to be induced in the latter condition. These changes appear as dislocation loops and “black dots”, definition given to irradiation induced defects such as small Frank loops or defect clusters observed with TEM but not clearly identifiable [11]. Post-irradiation mechanical tensile tests of Eurofer97 are published elsewhere [3, 4]. The stress-strain curves are shown in Figure 4. Three curves are plotted: one of Eurofer97 before irradiation, used as a reference curve, and two regarding post-irradiation tensile tests done at different temperatures – one at room temperature (RT) and the other at 300 °C, the same temperature as used for the neutron irradiation. In the figure the curves are plotted starting at the onset of the plastic region. The elastic region of the curves is not shown as it was affected by the tensile testing machine compliance used for the measurements and is not relevant for this analysis. For the three curves the same strain rate and specimen cross section were used: 5x10-4 s-1 and 12.57 mm2, respectively. The yield strength (YS) and the ultimate tensile strength (UTS) values of unirradiated Eurofer are 550 and 692 MPa, respectively. After irradiation, the YS and UTS for each tensile test temperature have the same value, i.e.: 1066 MPa for the RT test and 883 MPa for the 300 ° test. The reference curve of Eurofer97 shows a long plastic region before necking. In the irradiated samples’ curves the uniform elongation is practically zero and the total elongation has significantly decreased. The black dots and dislocation loops observed in Figure 3 are envisaged to be responsible for the increase in YS and UTS after irradiation in both conditions. In the tensile test performed at RT the black dots and precipitates act as obstacles for dislocation movement. In the test done at 300 °C, the same temperature used for irradiation, the YS is lower because of thermally activated dislocation gliding promoted by the increased temperature. Annihilation of defects is unlikely to be related to the lower YS and UTS of this test curve as the test did not take longer than 200 s at 300 °C.

84


1200

Engineering Stress (MPa)

1000

800

600

400

200

0

Eurofer97 unirradiated, Ttest = RT Eurofer97 irradiated, Ttest = RT 0

5

Eurofer97 irradiated, Ttest = Tirrad.

10

15

20

25

Plastic Engineering Strain (%)

Figure 4 Plastic region stress-strain curves of unirradiated and neutron irradiated 10 dpa, 300 °C Eurofer97. Test temperatures are indicated in the legend. The curves are plotted from the onset of the plastic region

5 Conclusions A comparison of Eurofer97 TEM investigations before and after 10 dpa and 300 °C neutron irradiation reveals that a significant amount of lattice damage is created by the neutron exposure. For both conditions, precipitates are observed mainly along the grain boundaries. The radiation damage induced by neutrons appears as black dots and dislocation loops. The black dots are very small and unresolved dislocation loops, defect clusters or small α’- and M 6 C-phases [9, 10]. The resolved dislocation loops can reach sizes of 15 nm. The TEM images are correlated to post-irradiation tensile tests published elsewhere [3, 4]. In comparison to values of reference unirradiated Eurofer97, an increase in yield strength and ultimate tensile strength after neutron irradiation is observed for the tensile tests performed at room temperature and at the irradiation temperature (300 °C). Both irradiated Eurofer97 samples show significant decrease in uniform and total elongation. The lattice damage caused by irradiation and observed with TEM is held responsible for the irradiation induced hardening. Dislocation gliding is promoted at 300 °C tensile test temperature, manifested as a decrease in YS and UTS in comparison to the values obtained for the neutron irradiated specimen tested at room temperature.

6 Acknowledgements This research was carried out under project number M74.5.10393 in the framework of the Research Program of the Materials innovation institute M2i (www.m2i.nl). The authors thank F. v.d. Berg for his support in the TEM measurements and M. Klimenkov for discussions concerning the TEM images.

85


7 References [1] Lindau R, Moslang A, Rieth M, Klimiankou M, Materna-Morris E, Alamo A, Tavassoli F, Cayron C, Lancha AM, Fernandez P, Baluc N, Schaublin R, Diegele E, Filacchioni G, Rensman JW, vd Schaaf B, Lucon E and Dietz W (2005) Present development status of EUROFER and ODS-EUROFER for application in blanket concepts, Fusion Eng Des, 75-79:989-996. [2] vd Schaaf B, Tavassoli F, Fazio C, Rigal E, Diegele E, Lindau R, LeMarois G (2003) The development of EUROFER reduced activation steel, Fusion Eng Des 69:197-203. [3] Rensman J (2005) NRG irradiation testing: report on 300 and 60 ºC irradiated RAFM Steels, NRG 20023/05.68497/P, Petten. [4] Luzginova N, Rensman J, Jong M, t Pierick P, Bakker T and Nolles H (2014) Overview of 10 years of irradiation projects on Eurofer97 steel at High Flux Reactor in Petten, in press: J Nucl Mater. [5] Rieth M, Schirra M, Falkenstein A, Graf P, Heger S, Kempe H, Lindau R and Zimmermman H (2003) Eurofer 97: tensile, charpy, creep and structural tests, FZKA, report 6911. [6] Kolluri M, Edmondson PD, Luzginova N and vd Berg F (2014) A structure-property correlation study of neutron irradiation induced damage in EU batch of ODS Eurofer97 steel, Mater Sci Eng A, 597:111-116. [7] Klimenkov M, Lindau R, Materna-Morris E and Moslang A (2012) TEM characterization of precipitates in EUROFER 97, Prog Nucl Energ, 57:8-13. [8] Fernandez P, Lancha AM, Lapena J, Serrano M and Hernandez-Mayoral M (2002) Metallurgical properties of reduced activation martensitic steel Eurofer'97 in the as-received condition and after thermal ageing, J Nucl Mater, 307:495-499. [9] Gaganidze E, Petersen C, Materna-Morris E, Dethloff C, Weiss OJ, Aktaa J, Povstyanko A, Fedoseev A, Makaroc O and Prokhorov V (2011) Mechanical properties and TEM examination of RAFM steels irradiated up to 70 dpa in BOR-60, J Nucl Mater, 417:93-98. [10] Materna-Morris E, Moslang A, Rolli R and Schneider H (2009) Effect of helium on tensile properties and microstructure in 9%Cr/WVTa/steel after neutron irradiation up to 15 dpa between 250 and 450 ºC, J Nucl Mater, 386-388:422-425. [11] Koning R (2012) Comprehensive Nuclear Materials 1st edn. Elsevier, United Kingdom.

86


Approach for the Investigation of Long-Term behaviour of elastomeric Seals for Transport und Storage Packages M. Jaunich1*, D. Wolff1 (1) BAM Federal Institute for Materials Research and Testing, Berlin, Germany

Abstract: Elastomers are widely used as main sealing materials for containers for low and intermediate level radioactive waste and as additional component to metal seals in spent fuel and high active waste containers. According to appropriate guidelines and regulations safe enclosure of the radioactive container contents has to be guaranteed for long storage periods as well as down to temperatures of -40 °C for transportation. Therefore the understanding of seal behaviour in general is of high importance and ageing of elastomeric seals has to be considered with regard to possible dynamic events taking possibly place during transport after storage. Keywords: O-Ring, seal, ageing, elastomer, compression set

1 Introduction

As elastomers are widely used as main sealing materials for containers for low and intermediate level radioactive waste and as additional component to metal seals in spent fuel and high active waste containers their required service life lies in the range of several decades. According to appropriate guidelines and regulations safe enclosure of the radioactive container contents has to be guaranteed for long storage periods as well as down to temperatures of -40 °C for transportation. Therefore the understanding of seal behaviour in general is of high importance and especially ageing of elastomeric seals has to be considered. Possible dynamic loads may occur during the whole interim storage period (so far approved in Germany for up to 40 years) and during transportation after storage. To fulfil their purpose, the seals have to remain in good shape with an adequate resilience to ensure a certain contact pressure and without e.g. cracks crossing the sealing surface which would act as leakage path. A typical ageing effect on a rubber material under the influence e.g. of ozone is shown as an example in Figure 1. This shows the importance of ageing for tightness relevant applications.

Figure 1 Ozone cracking of an natural rubber sheet.

Ageing of materials is an undesirable but unavoidable process which can lead to (non-reversible) changes in e.g. mechanical properties, thermal properties, colour and chemical composition [1]. The origin of these property changes can be innate to the material or is caused by influences from the environment. Typical innate effects are e.g. trapped stresses or remaining crosslinking agents. Effects *

Division 3.4 Safety of Storage containers BAM Federal Institute for Materials Research and Testing, Berlin, Germany [email protected]

87


caused by the environment are e.g. oxidation processes, influences from light and other radiation sources or even heat. The changes of the material properties can be caused by purely physical or chemical processes. Physical ageing can take place as molecular rearrangement, demixing or crystal growth. Usually it is driven by thermodynamic non equilibrium conditions which are often innate effects. Chemical ageing results in changes of the chemical composition of the material by chemical reactions.The ageing of polymeric materials is generally considered to consist of two pathways. One is the formation of additional crosslinks and the other is chain scission [2]. Generally, these two processes take place in parallel whereas one process is dominant. This means, that some materials tend primarily to crosslinking and others to chain scission. This predominant mechanism must not remain the same over the whole ageing period. Additionally, temperature has an important influence on the ageing process. It is possible that a material forms a high amount of crosslinks at one temperature and at e.g. a higher temperature chain scission becomes dominant. To describe the effect of ageing on the performance of a material it is important to test the relevant properties but for long-term applications an accelerated test is required to ensure long-term safety. This approach is described by standards, e.g. ISO 188 [3]. Often these standards address only the general approach for accelerated ageing and assume an Arrhenius-like behaviour of the ageing process, whereas the necessary prerequisites and the data analysis are often not described in detail. Moreover, often these standards describe ageing of standard samples and not of components. The samples are often unstressed and therefore the effect of a continuous deformation of e.g. an elastomeric seal cannot be evaluated even if such permanent stress might lead to different ageing effects as e.g. predominantly chain scission of stressed chain molecules. In literature there are several examples that show non-Arrhenius ageing effects and, therefore, the need for a more detailed analysis of the test results and a careful selection of test parameters increases [4, 5]. Especially for complex technical elastomer compounds consisting of a multitude of different components their interactions are hard to predict. It is possible that some material properties are much more sensitive to slight changes then others. Therefore, for a comprehensive investigation it is important to measure relevant properties which are also sensitive to the occurring changes. For elastomeric seals there are several properties that can be measured as e.g. hardness or tear properties that have typically very little or highly complex correlation with sealing performance [6]. Other standard methods such as compression set or stress relaxation measurement are more related to seal applications providing an indirect value for seal performance [5]. An overall goal of ageing investigation of seal materials is to determine which properties are sensitive to occurring changes and how they are correlated with the safety relevant performance of the seal. In the past we have investigated the compression set measurement in detail and we have developed a new method to determine this value by use of a measurement device for Dynamic Mechanical Analysis [7, 8]. The mathematical fit of continuously measured compression set values allows e.g. to determine discrete retardation times and retardation strengths. For this purpose, we have applied an equation composed of three exponential decay functions:

y = y 0 + A1e

−

x T1

+ A2 e

−

x T2

+ A3 e

−

x T3

(1) where y is the compression set value, y 0 the value reached after long times, A 1 to A 3 the calculated retardation strengths and t 1 to t 3 the corresponding retardation times. In Figure 2 an example of compression set values for an FKM material manufactured at BAM is shown with the representation over temperature and the fit parameters.

88


Figure 2 Compression set values measured on an FKM rubber over time (upper picture) with isochronal representation over temperature and fit results (lower pictures).

The results show a clear temperature dependence of the compression set values; The compression set increases with decreasing temperature. The fit of the data shows that this behaviour can mostly be attributed to the parameter A 1 , the retardation strength of the fastest retardation process which is clearly temperature dependent. In this paper we focus on our approach to investigate the behaviour of elastomeric seals over extended periods of time at elevated temperatures. We describe our approach to investigate the occurring changes of material properties of elastomeric seals by ageing performed at different temperatures and under assembly conditions and first results are presented. It is one goal to determine the influence of compression during the ageing period and therefore compressed and uncompressed samples will be stored and later analysed simultaneously. The temperature influence must also be analysed to ensure an appropriate acceleration of the ageing processes (e.g. exclusion of additional sample degradation as a result of high temperatures for accelerated ageing).

2 Material selection and ageing conditions As mentioned, the use of elastomeric seals in containers for radioactive waste is our motivation to start the described research and hence the investigation parameters and the material selection are related to this application. The material selection focuses on material classes that are at least internationally used for this application. Therefore a fluorocarbon rubber (FKM) and an Ethylene-Propylene-Diene Rubber where selected. The seal dimensions were chosen with a rather high torus diameter of 10 mm and an inner diameter of 190 mm to have sufficient amount of sample material for different analytical methods to be performed. Mechanical ageing condition should also be varied to investigate

89


whether a compressed seal shows a different ageing behaviour in comparison to an uncompressed sample. Ageing temperatures are selected in the range from 75 °C up to 150 °C and are therefore spread broad enough to have a significant acceleration but due to the spacing of 25 K the results from different temperatures should still be comparable. With an overall of four test temperatures a sufficient amount of data should be available to judge the applicability of typical time-temperature-superposition approaches as e.g. Arrhenius. During ageing, samples are stored in unstressed condition as complete O-rings and in compressed condition with a degree of compression of 25 %. To compress the sample materials they are positioned between two metal plates that are screwed together until both plates are pressed against distance pieces. The used devices are schematically shown in Figure 3.

Figure 3 Used equipment for ageing of samples in uncompressed (upper picture) and compressed condition (lower picture).

Beside these two ageing conditions it is important to allow a correlation of measured changes in material properties with seal function which is possible with seals mounted in flanges. A schematic picture is given in Figure 4. This setup allows for leakage rate measurement of mounted seals by a pressure rise measurement after defined ageing times. The principle setup for such a measurement is described in [6, 9].

90


Figure 4 Schematic cut through image of the test flange construction for leakage rate measurements.

The samples are removed from the high temperature chamber at defined times to be investigated by several methods. The ageing times are given in Table 1. Table 1 Ageing times for different samples

Storage time

0h

3h

10 h

1d

3d

10 d

30 d

100 d

1a

2a

uncompressed samples

3

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

compressed samples

3

samples in flange

3 Planned investigation methods and first results For the investigation of the ageing effects several methods are helpful to describe the occurring changes and to quantify the impact on the seal performance. For the investigation of structural/morphological changes e.g.: - FT-IR is a suitable method that enables the measurement of the chemical structure of polymeric materials. As the covalent bonds are detected it is possible to detect time dependent changes of the samples under investigation. - Differential Scanning Calorimetry (DSC) allows the detection of phase transitions over temperature and therefore to characterise occurring changes - Thermal Gravimetric Analysis (TGA) measures the mass of a sample during a heating program to detect occurring mass losses as result of decomposition processes.

For investigation of mechanical properties several methods can be applied as e.g.: - Dynamic Mechanical Analysis (DMA): measuring the modulus in dependence of temperature to characterise the viscoelastic properties of the material and occurring transitions. - Hardness: at standard test temperature - Tensile testing: stress and strain at break at standard test temperature - Compression Set (CS): measurement of the rebound behaviour after compression (at different temperatures)

The observed changes will be correlated with the performance of the aged seals in the components tests for which the leakage rate is measured. First results of the ongoing investigation exemplified by TGA results are shown in Figure 5.

91


Figure 5 TGA measurements performed under Nitrogen up to 800°C and under synthetic air at higher temperatures.

The TGA data shows clear differences between the three materials. The NBR shows a first weight loss between 250 and 400 °C with about 10 %. Between 400 and 500 °C all materials show a strong weight loss. The temperature range of this steplike reduction has the following order: NBR 0

convection

translation wave (dynamic)

.bcd..

hyperbolic

conv. - diff.

sand transport along flume bottom

..cd..

parabolic, D = 0

diffusion

dispersion, ion or heat diffusion

..cde.

parabolic

conv. - diff.

surface water wave

...de.

-

convection

translation wave (kinematic)

...d.f

-

simple growth model

a..d.f

parabolic

classical mechanics, systems theory

example

2.4 Continuum Damage Mechanics Kachanov proposed a general model of (mechanical) damage growth, which may be extended to other mechanisms. Kachanov's CDM has despite its speculatively and simplicity some attractive practical merits: mathematical simple, suits measurements, wide applicability The twofold basis is •

a set of three equations describing geometry, driving force or energy and constitution

•

definition of gradient power law of the material constitution

166


Local damage D is defined as the relative reduction of a state of interest A, for instance the cross section of a tensile rod

D=

As − A As

(17)

The index s indicates a reference state of interest, which is usually a global state. The resistance x to a load is assumed to be conservative

xA = x s As

(18)

A power law growth model is assumed

dD  x = dt  x r

  

n

(19)

The index r indicates a reference state of the material. The resulting differential equation in D is

dD  x s = dt  x r

n

  (1 − D )− n 

(20)

Integration with boundary condition D (t0 ) = D0 , 0 indicates the initial state 1

n  n +1   xs  n +1  D = 1 − (1 − D0 ) − (n + 1)  (t − t0 )    xr    Set t0 = 0 and by definition D t f = 1 , where f indicates the failure state, so

(21)

( )

  1 (1 − D0 )n +1  xs  tf = n +1  xr 

−n

(22)

The exponent n has to be known anyway and the result t f and the imposed x s are measurable, so

 t depending on the availability of information on x r or on D 0 either D = 1 − (1 − D0 )1 −  t f  1− m

x  D = 1 − m  s   xr  −m

Q  t = 1− Q0  t f

   

(t

f

m − t ) with m =

   

m

or

1 can be used. In terms of quality Q = 1 − D n +1

m

(23)

167


1

Q / Q0

100

0,5

10 3 1 0 0 0

0,5

t / tf

figure 1: loss of quality in time for different values of n

1

figure 2: results of cyclic load tests on porous asphalt [Tolman, Gorkum, 1997]

An application of CDM is shown in figure 2. Specimens are from the real structure, laboratory made prepared from the real material and fully laboratory made.

3 Data 3.1 Curve fitting As damage phenomena are essential non-linear the usual least square methods to fit measurement data mostly lead to instable solutions (Tolman, Kotte 1994). Two obvious ways to solve this problem are transformation to linearity before fitting and separating the dataset in subsets that are relevant for each model parameter. The first approach requires attention in data sampling. Some authors (a.o. Jacobs 1994) report a relation between the exponent and coefficient of power law models and suggest a physical law, however without giving a convincing explanation. Inspection of the dataset reveals that all data sets share rather narrow one centre, so the least square fitting procedure may be the more obvious cause. If a and b are the offset and coefficient of the regression line in an x-y plane, a = m y –b m x , with m the average value. At first sight a promising alternative seems to determine the 3 CDM-parameters in a more direct manner. However •

the offset Q 0 , apart from usually being small, is hard to separate from test set up effects; this parameter is the hardest to determine

•

the slope, which can be used to derive the exponent m is usually not a maximum or minimum, but a bending tangent

•

the failure time t f is probably the easiest to determine

Probably for that reason data is often presented as a t f - ∂c/∂t plot.

3.2 Probabilistic modelling As degradation and life time phenomena show a large scatter in data, a probabilistic approach is useful. A first estimate may be obtained by linearized models, though large errors may be introduced

168


due to the strong non-linearity. Set τ = 1 −

t dτ τ ⇒ = . The life time model Q = Q0τ m can be tf dt f t f

linearized as

tf Q Q m = −m* (1 + ln (τ * )) + 0 + m* + m* ln (τ * ) Q* Q0* t f* m* n The end of life model z = xs t f =

(24)

1 Q0n +1 xrn can be linearized as n +1

 n n + 1 Q0  ln (e(n + 1)z* x r* )   1 x z  −  + r + ln (Q0* x r* ) + * = − 2 − n* z* n* + 1 n* Q0* x r*  n*   n* + 1 

(25)

4 Evaluation of engineering models Some obvious criteria for engineer knowledge and models are •

handy, preferably even useable as rules of thumb; this requires insight in the relative importance of parameters and variables

•

coherent, both in comparison with related aspects, changes in data, and limit cases (limit cases may also be useful to single out variables); the ultimate test is how well the model suits reality and rejected should be the symbolic acceptance by agreement

•

availability of data

•

accurate (correct, true)

•

precise (close together)

Apart from the content, criteria regarding cost of modelling, presentation and communication should be considered. It does not help much to have models which are too expensive to use or poorly unacceptable by stakeholders in projects. Some considerations are: Engineering knowledge is at a junction of way of understanding, freedom (nature – interests) and planning and control (design – research). This complex situation requires practical and coherent models. Practicality means handy - rules of thumb like – and availability of data. Coherence implies consistency – for which limit cases may serve as check – and sufficiently accurate (correct, true) and precise (close together). Initial state, degradation and failure phenomena may in principal be based on TT, but they require additional formulations to become applicable for engineers. RT allows distinguishing mechanisms. CDM reduces RT to 3 abstract parameters for initial state, degradation and failure. For particular damage phenomena these 3 parameters may be expressed in comparatively quick and simple experimentally determinable parameters. RT and CDM may be applied on many time dependent dissipative phenomena, Three broad classes are atomic or molecular reactions, including heat, deformation (flow, creep) and formation of new surface (surface tension, fracture) of both solids and fluids. Applications of RT or CDM on conservative phenomena like elasticity and inertia are not part of this investigation.

169


5 Conclusions Engineering models pose other requirements than research models. An established engineering methodology is not available. Some aspects regarding degradation and failure, coherence, simplicity and variability of data have been touched. Four levels of modelling are discerned 1. basis may be physical laws for transformation of matter [Atkins 1986], in particular the Kinetic Theory (Arrhenius, Eyringh, Krausz) 2. Kinetic Theory with imaginable descriptions of geometry and material laws 3. Kachanov’s power law damage grow, CDM 4. derivation of rules-of-thumb-like models, which should be more coherent than empirical rulesof-fist Data manipulation deserves attention, as well before collection as in application in models. Past data may need some consideration. Criteria for quality of models, in particular engineering models, are poor. Improvement is needed to contract, design and control projects, as well as for evaluation.

6 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Simon, H.; The sciences of the artificial; M.I.T. Press Cambridge USA; 1969 Young, H.D.; Fundamentals of mechanics and heat; Mac Graw-Hill Book Company; 1964 Atkins, P.W.; Physical chemistry; Oxford University Press; (1978) 1986 Allaart, A.P.; Design principles for flexible pavements, a computational model for granular bases; PhD thesis TU Delft 1992 Krausz, A.S.; Krausz, K.; Fracture kinetics of crack growth; Kluwer Academic Publishers; 1988 Christensen, R.M.; Glaser, R.E.; The application of kinetic fracture mechanics to life prediction for polymeric materials; Journal of applied mechanics; vol. 52/1 March 1985 Luijerink, J.; Creep of concrete during load-controlled experiments; note TU Delft 1982? Put, T.A.C.M van der; Deformation and damage processes in wood; PhD thesis TU Delft 1989 Jacobs, M.M.J.; Crack growth in asphalt mixes; PhD Thesis TU Delft 1995 Kachanov, L.M.; Introduction to continuum damage mechanics; Martinus Nijhoff; 1986 Paas, M.H.J.W.; Continuum damage mechanics with an application to fatigue; PhD thesis TU Eindhoven; 1990 Kotte, J. F. A. K.; Tolman, F.; Rafeling of drain asphalt; SHRP and Traffic Safety on Two Continents; The Hague 1993 Tolman, F.; Gorkum, F. van; European conference on porous asphalt; Madrid 1997 Tolman, F.; Ebels, L. J.; Permeability of bound granular covers; Euromat, Muenchen; 1999 Gehlen, C.; Probabilistische Lebensdauerbemssung von Stahlbetonwerken; Beuth Verlag GmbH; 2000

170


Ageing coefficient of fly ash concrete and its impact on durability G.J.L. Van Der Wegen (1), M.M.R. Boutz (1), A. J. Sarabèr (2), R.J. Van Eijk (3) (1) SGS INTRON, Sittard, The Netherlands (2) Vliegasunie, Culemborg, The Netherlands (3) KWR Watercycle Research Institute, Nieuwegein, The Netherlands Abstract: The time dependency of the chloride diffusion coefficient, the so-called ageing coefficient, of fly ash concrete has been investigated. The effect of origin of the fly ash as well as the origin and strength level of the cement combined with the fly ash, has been determined. Results obtained using the RCM-method on samples aged up to 1 year under ambient conditions in The Netherlands are presented in detail. Our initial research shows that the ageing coefficient of the fly ash concrete is much higher than the values mentioned in the Dutch guideline for service life design of structural concrete. The impact on the service life design of reinforced concrete structures is discussed. Keywords: chloride diffusion, ageing coefficient, concrete resistivity, fly ash concrete, durability

1 Introduction The service life of reinforced concrete structures exposed to high concentrations of chlorides is mainly determined by the depth of concrete cover and its resistance to chloride ingress [1]. The cumulative ingress of chlorides in time is of course related to the initial value of the chloride diffusion coefficient, but it is much more affected by the so-called ageing coefficient of the concrete [2]. Concrete in which part of the cement has been replaced by fly ash is known to have an initial (28 days) chloride diffusion coefficient significantly higher than the same concrete without fly ash. However, due to the pozzolanic behaviour of the fly ash the chloride diffusion coefficient will decrease more rapidly in time than the concrete without fly ash. Hence, the ageing coefficient of fly ash concrete is much higher [3]. Based on the principle of equivalent concrete performance, concrete compositions with very low cement content (down to 200 kg/m3) and very high fly ash content (up to 100 kg/m3) have been used successfully in The Netherlands on a large scale for more than 20 years now [4]. Although the benefits of such fly ash concretes are generally well recognized, there is some discussion on the actual value of their ageing coefficient [2]. Mangat [5] was the first to include the effect of ageing of concrete on the chloride diffusion coefficient and later improved by Maage [6]: 𝑡

𝑛

𝐷𝑡 = 𝐷0 � 𝑡0 �

(1)

in which D t is the chloride diffusion coefficient at time t, D 0 is the chloride diffusion coefficient at reference time t 0 (usually 28 days) and n is the ageing coefficient. Based on this model, ageing coefficients have been determined from time dependency of chloride diffusion coefficients calculated from chloride concentration profiles in (long term, continuously) exposed concretes as well as from instantaneous measurements (e.g. Rapid Chloride Migration (RCM) test according to NT Build 492 [7]) on aged but not chloride exposed concrete samples. One is not always aware that both methods require different equations for calculating the ageing coefficient [8]. In general, ordinary Portland cement concrete possesses the lowest value for the ageing coefficient compared to concrete with blended cements. Concretes containing components reactive at early age (like granulated blast furnace slag or silica fume) as part of the cement or as an additive exhibit a higher ageing coefficient as well as a lower initial chloride diffusion coefficient compared to the ordinary Portland cement concrete. The use of pozzolans (like fly ash), which react much slower in time than granulated blast furnace slag, will result in a higher initial chloride diffusion coefficient as well as in a higher ageing coefficient [1, 3, 8-18]. Based on the evaluation of a database containing more than 500 Dutch RCM measurements, comprising different types of binder and water cement

171


ratios, ageing coefficients for different types of binder and exposure conditions were calculated as shown in Table 1 [9]. Table 1 Ageing coefficient for different types of binder and exposure conditions Type of binder (S=slag; FA=fly ash)

Exposure conditions Very humid, wet

Atmospheric

CEM l

0,40

0,60

CEM I+ 25-50% S, CEM II/B-S,

0,45

0,65

CEM III with 50-80% S

0,50

0,70

CEM I with >20% FA

0,70

0,80

CEM V/A (25% S + 25% FA)

0,60

0,70

CEM III/A with 20% FA binder shown in Table 1 were questioned. Therefore, a long term research programme was started to examine the effect of different origins of fly ash and different origins as well as strength levels of ordinary Portland cement on the value of the ageing coefficient of fly ash concrete. A typical Dutch CEM I 52.5N cement was combined with 3 fly ashes originating from different Dutch coal fired power stations; one of the fly ashes was additionally combined with 2 other CEM I 52.5N cements of different origin and strength level. As a reference Dutch CEM III/B cement was used, which is known for its high resistance against chloride ingress. The chloride migration coefficient and the electrical resistivity of concrete specimens prepared with these binders will be tested at different ages up to 5 years. The initial, up to 1 year results are presented and discussed in this paper.

2 Materials and methods 2.1

Materials and properties

The chemical composition of the fly ashes and compressive strength of the cements used, are shown in Table 2. Dutch river sand (0-4 mm) and gravel (4-16 mm) were used as aggregate. The maximum grain size is limited to 16 mm in order to lower the variation in chloride migration coefficient

Figure 1 Grain size distribution of aggregates in concrete mixtures Table 2 Relevant properties of fly ashes and cements used

172


Property (unit)

Code

Fly ashes

Compressive

SMZ

Hemweg

Amer9

Strength

FA1

FA2

FA3

(MPa)

Composition (M-%):

SiO 2

58,6

61,0

61,0

Al 2 O 3

20,7

21,1

20,6

Fe 2 O 3

7,6

6,8

6,8

CaO

4,5

3,5

MgO

2,0

Na 2 O-eq

Cements CEM

CEM I

CEM I

III/B

52,5N

52,5N

I

42,5N

ENCI

CBR

52,5N

ENCI Code

CEM

Seibel

Cref

C1

C2

C3

2 days

11,8

30,1

23,2

30,0

4,7

7 days

33,3

47,9

38,7

51,1

2,2

1,8

28 days

56,4

61,8

55,0

63,0

2,5

2,7

3,0

TiO 2

0,88

0,94

0,86

P2O5

0,55

0,34

0,38

LOI

2,7

2,0

0,72

0.5 um), according to Young and Sidney Mindess [4]. Fig. 4 shows the results of the pore size distribution of six samples at the curing age of 182 days obtained by PDC-MIP method.

Figure 4 Pore size distribution of cement pastes at 182 days obtained by PDC-MIP tests

As shown in Fig. 4, the volume of gel pores is almost unchanged with the increasing w/b ratio. At the meantime, capillary pores, especially the medium capillary pores (0.05-0.5 um), are increasing greatly as w/b ratio increases. This implies the dilution effect (increasing water-tocement ratio) mainly works on medium capillary pores (0.05-0.5 um). The addition of fly ash increases the volume of gel pores (< 0.01 um). This observation is characteristic for materials containing SCMs owing to the smaller pore size resulting from the pozzolanic reaction. Increasing replacement level of fly ash from 10% to 30% leads to the formation of more large capillary pores (> 0.5 um). Limestone powder is considered as an inert SCM, and its addition in fly-ash cement dilutes the hydrating cementitious system, resulting in the increase of the volume of small and medium capillary pores (0.01-0.5 um). This also provides evidence that the dilution effect plays an important role in the formation of small and medium capillary pores (0.01-0.5 um). 4.2.2 Evolution of pore sizes

The changes of pore size distribution resulting from continuous hydration are displayed in Fig. 5. All samples show pore size distribution towards a finer one, as time elapses. As the hydration proceeds, the volume of gel pores (0.5 um) decrease with curing age. The dilution effect (increasing water-to-cement ratio) has influence mainly on the pore sizes range of 0.05-0.5 um.

The addition of fly ash refines the pore size of cement paste by increasing the volume of gel pores and small capillary pores (0.01-0.05 um). The addition of limestone powder dilutes the cementitious system and results in volume increment of the pores (0.05-0.5 um). The more throat pores exist, the more ink-bottled pores form.

REFERENCES Journal article:

[2] Jenn ings H.M. (2008), Refinements to colloid model of C-S-H in cement: CM-II, Cem.Concr. Res. 38 (2008) 275–289. [3] Powers T.C., Brownyard T.L., Studies of the physical properties of hardened Portland cement paste. Part 2. Studies of water fixation, J. Am. Concr. Inst. 18 (3) (1946) 249–303. [7] Washburn E.W.(1921), Note on a method of determining the distribution of pore sizes in a porous material, Porc. Natl. Acad. Sci. USA 7 (1921) 115–116 [8] Willis K.L., Abell A.B., Lange D.A.(1998), Image-based characterization of cement pore structure using Wood's metal intrusion, Cem Concr Res 28 (12) (1998) 1695-1706 [9] Radim VocÏka, Christophe GalleÂ, Marc Dubois, Patrick Lovera (2000). Mercury intrusion porosimetry and hierarchical structure of cement pastes Theory and experiment. Cement and Concrete Research 30 (2000) 521-527 [11] Zeng Qiang, Li Kefei, Teddy Fen-chong, Patrick Dangla (2012). Pore structure characterization of cement pastes blended with high-volume fly-ash. Cement and Concrete Research 42 (2012) 194–204 [12] Sidney Diamond. Review-Mercury porosimetry An inappropriate method for the measurement of pore size distributions in cement-based materials. Cement and Concrete Research 30 (2000) 1517-1525 [13] Moro F. and Bőhni H. (2002). Ink-Bottle Effect in Mercury Intrusion Porosimetry of Cement-Based Materials. Journal of Colloid and Interface Science 246, 135–149 (2002) [14] Jian Zhou, Guang Ye (2010), Klaas van Breugel. Characterization of pore structure in cement-based materials using pressurization–depressurization cycling mercury intrusion porosimetry (PDC-MIP). Cement and Concrete Research 40 (2010) 1120–1128 [15] Luke K. and Glasser F.P (1988). Internal chemical evolution of the constitution of blended cements. Cement and Concrete Research. Vol. 18, pp 495-502, 1988 [16] Li Z, Ding Z (2003). Property improvement of Portland cement by incorporating with metakaolin and slag. Cem Concr Res 2003;33:579–84. [17] Cook RA, Hover KC (1993). Mercury porosimetry of cement-based materials and associated correction factors. Constructure and Build Material 1993;7(4):231–40. Dissertation [10] Ye, G. (2003) Experimental Study and Numerical Simulation of the Development of the Microstructure and Permeability of Cementitious Materials, Ph.D. Thesis, Delft.

218


Conference proceedings [1] Karen, L. S. (2012). “Impact of microstructure on the durability of concrete.” Second International Conference on Microstructural-related Durability of Cementitious Composites, 11-13 April 2012, Amsterdam, The Netherlands. Book [4] Young J. Francis and Sidney Mindess (1981). Prentice-Hall, 1981 - Technology & Engineering. ISBN-13: 9780131671065 [5] P.K. Metha, P.J.M. Monterio (2006), Concrete, Microstructure, Properties and Materials, McGraw-Hill, London, 2006. [6] Han Young Moon, Hong Sam Kim, Doo Sun Choi (2006). Relationship between average pore diameter and chloride diffusivity in various concretes. Construction and Building Materials 20 (2006) 725–732

219


Using bio-based polymers for curing cement-based materials J. Zlopasa1*, E.A.B. Koenders1,2, S.J. Picken1 (1) Delft University of Technology, Delft, The Netherlands (2) COPPE-UFRJ, Rio de Janeiro, Brazil Abstract: Curing is the process of controlling the rate and extent of moisture loss from the surface of cement based materials. It is the final stage in the production of cement-based materials and it is the essential part for achieving continuous hydration of cement, while avoiding cracking due to drying shrinkage. Continuous cement hydration also guarantees a strong bond between aggregate, fewer voids, and depercoliation of capillary pores. Thus, a properly cured cement-based material is prepared for a long service life. Using environmentally friendly, water based bio-polymers could help to achieve more durable cement-based materials, and, therefore preventing a premature end of service life of building materials. Rapid Chloride Migration tests and Environmental Scanning Microscope are employed to investigate the functional properties, e.g. transport property, and microstructure properties, respectively. Mortar samples were cured in air and applied by water-based curing compound, made of sodium alginate. We observed strong beneficial effects of applying sodium alginate as a curing compound in terms of microstructure and hydration development. Based on these results, a less porous microstructure and an improved durable cement-based material was achieved that was prepared for longer service life. Keywords: curing, bio-based polymer, transport property, microstructure

1 Introduction

Cement-based materials are composite materials that consist of a continuous phase of cement paste and a discontinuous phase of aggregates. The aggregates phase is viewed as an inert filler that contributes to volume stability and higher durability, and are bonded by hydration products of cement clinker in cement paste [1]. Cement clinker is composed of four mineral phases Ca 3 SiO 5 Ca 2 SiO 4 , Ca 3 Al 2 O 6 and Ca 4 Al 2 Fe 2 O 10 . Two major phases (around 80wt.% for most cements) of cement clinker are tri- and di-calcium silicate. Tri-calcium silicate reacts faster and is responsible for strength development during the first weeks, whereas di-calcium silicate reacts slower and contributes to the long-term strength of cement-based materials. In general, the reactions of both silicate phases are presented as follows: (CaO) b (SiO 2 ) + (b-Ca/Si+y)(H 2 O)  (CaO) (Ca/Si) (SiO 2 )×y(H2O) + (b-Ca/Si)Ca(OH) 2 (1)

where b=2 or 3 (di- or tri calcium silicate, respectively), Ca/Si=1.7-1.8 [2], and y=4. Since water is an essential component of hydration, and cement hydration will only proceed in a water-filled space, sustaining the hydration process requires that inter-particle voids remain filled with water [3]. Powers et al. showed that hydration of cement is greatly slowed down when the relative humidity in cement capillary pores goes below 80% [4], which most likely happens at the surfaces of e.g. concrete elements. Curing is a name given to procedures that aim to avoid this. They are used for promoting the hydration of cement, and consists of a control of moisture movement (and temperature) from and into the cement-based materials [1] through the exposed surfaces. By preventing the loss of water from cement-based materials, continuous hydration could be achieved and drying shrinkage be avoided, leading to a minimum of surface cracks, a stronger bond between aggregates, fewer voids, and lower connectivity of pores. Such a microstructure is denser and can prevent slower penetration of aggressive fluids that can be *

Corresponding author. [email protected] (J. Zlopasa)

220


harmful, e.g. for corrosion of steel reinforcement [5]. Therefore, a properly cured cement-based material is better prepared for a long service life. Generally, curing can be performed by adding water or by hindering water escape from the cement-based material’s surface. Continuing adding water by means of water ponding, water spraying, and/or by the use of wet burlaps usually gives the best end results. However, this technique requires workers on site that keep the concrete moist, which can be costly. In addition, this method can be especially costly in places where there is a scarcity of water. As said, the second way of curing cement-based materials is by preventing it from water evaporation. This can be accomplished by covering the surface with a plastic sheets or by spraying it with a curing compound (polymer solutions/emulsions) that creates a film that hinders water evaporation. Curing compounds can be water based or organic-solvent based [6]. In general, curing compounds based on organicsolvents show better performance when compared with water-based ones [7]. As a drawback, there can be an environmental impact, especially when using it in low ventilated environments. In this study, we propose the use of a water-soluble bio-based polymer, sodium alginate, as potential curing compound for cement-based materials. Alginates are linear water-soluble polysaccharides comprising of (1— 4) linked units of α-ᴅ-mannuronate (M) and β-L-guluronate (G) at different proportions and different distributions within the chains. Functional properties are strongly correlated with composition (M/G ratio) and with the sequence of the uronic acids. They are present in brown algae and can also be found in metabolic products of bacteria, e.g. pseudomonas and azotobacter [8-11]. Alginates are commonly used as food additives, gelling agents, wound dressings and for drug delivery[12,13]. The gelling property of this polymer gives a unique opportunity to be applied in various fields. Alginates gel either by lowering the pH below the pK a value of the uronic residue or in the presence of polyvalent ions [14,15]. Polyvalent ions act as bridges between different G units of chains, as shown in Figure 1.

Figure 1. Structure of alginic acid (left) and schematic representation of crosslinking of alginate by polyvalent ions (right), from [15] Following Equation 1, a large amount of Ca2+ is produced by cement hydration, which is freely available (usually for carbonation). Since sodium alginate reacts rapidly with Ca2+ by crosslinking, forming non-water soluble calcium alginate, we have investigated the possibility to use sodium alginate as external curing compound. For this purpose, we compared air-cured and alginate-cured mortar samples. We have examined functional and microstructure properties of mortar samples with and without application of sodium alginate solution, by the use of Environmental Scanning Microscope (ESEM) and Rapid Chloride Migration (RCM) Tests.

221


2 Experimental plan 2.1

Materials and mortar preparation

Sodium alginate with an average molecular weight of 150,000 and M/G ratio of 1.56 was obtained from Sigma Aldrich Co. Water solution of 3wt.% sodium alginate was prepared using demineralised water under vigorous stirring. Mortar samples were prepared using commercial Portland cement Cem I 52.5R (produced by ENCI Heidelberg Cement group) with w/c of 0,5 and aggregate particles that follow Fuller distribution. Table 1 gives an overview of the chemical composition of the cement used. The mineral content of the cement calculated, using the Bogue method [16], is 68,3% Ca 3 SiO 5 , 6,2% Ca 2 SiO 4 , 6,3% Ca 3 Al 2 O 6 and 10,0% Ca 4 Al 2 Fe 2 O 10 . Table 2 gives an exact mix design of the mortar. Table 1 Chemical composition (X-ray Flouresence Analysis) of Cem I 52,5R Chemical composition

Weight fraction

-

%

CaO

64,9

SiO 2

20,1

Al 2 O 3

4,5

Fe 2 O 3

3,3

K2O

0,46

SO 3

3,3

MgO

1,4

P2O5

0.4

TiO 2

0.2

Loss on ignition (900°C)

1,1

Samples were cast in cylindrical moulds with 100mm diameter which was followed by compaction using a vibrational table. After compaction the sodium alginate solution was poured on top of the three mortar samples while other three sample were air-cured (surface of mortar was left uncovered). Furthermore, samples were placed in a lab with constant environmental conditions of 50% RH and 20°C for 28 days, after which they were de-moulded analysed. Table 2 Mortar mix design

Cem I 52,5 R [kg/m3] 3

Water [kg/m ]

253.48 3

Aggregates [kg/m ]

2.2

506.96

1520.88

Characterization methods

In order to evaluate differences in functional properties of the mortar samples, the chloride migration coefficient is determined by use of the Rapid Chloride Migration, RCM, test. The method is described in NT Build 492 [17]. The principle behind this method is in the application of external electrical current axially across the mortar specimen which forces the chloride ions 222


to migrate into the specimen at higher rate. Since curing of cement-based materials mostly affects the near surface area, we have modified the standard RCM test slightly by not removing the first 10-20 mm of the mortar surface from the samples. A sketch of the RCM test is shown in Figure 2. After the samples were de-moulded they were preconditioned by placing them in vacuum for 3 hours and then, with the vacuum pump still running, a saturated Ca(OH) 2 solution was introduced and samples were completely submerged in it. After 1 hour the vacuum pump was turned off and air was allowed to enter the container and the specimens were further keep in the saturated Ca(OH) 2 solution for 18 hours. Once the preconditioning was completed the samples were placed in the RCM setup, which consists of a rubber sleeve in which the samples are placed in. Electrodes are immersed in the anolyte (0,3 M NaOH) and catholyte (2 M NaCl) solutions and connected to the power supply unit. Initial current through the sample at 30 V is recorded and the voltage was adjusted according to the standard which also states the duration of the test. The voltage is adjusted in order for the chlorides to penetrate through about half of the sample. Initial and final temperatures of both anolyte and chatolyte were also recorded. After the described test duration specimens were removed from the RCM setup and split axially into two pieces. On the freshly split surface 0,1 M AgNO 3 was sprayed and after few minutes white silver chloride started to precipitate (see Figure 3) and change colour. The precipitated silver chloride represents the chloride penetration depth, from which the migration coefficient is calculated using Equation 2: D nssm =(0.0239×(273+T) ×L)/((U-2) ×t) ×(x d -0.0238×√((273+T) ×L×x d )/(U-2))

(2)

where T is the temperature, L thickness of the sample, t test duration, U applied voltage and x d is the chloride penetration depth.

Figure 2 Schematic representation of RCM test

Difference in developed microstructure of mortar samples were investigated using environmental scanning microscope (ESEM) in backscattered electron mode (BSE), in conjunction with energy-dispersive spectroscopy for qualitative analysis (Ca, Si, C, Al, Mg, Fe, Na, K). Polished sections for ESEM and energy-dispersive X-ray spectroscopy (EDS) is done as follows. The sampling was done 30 mm from the surface, cement hydration was stopped by submerging the samples in liquid nitrogen and subsequently moved into a freeze-dryer (sublimation) with temperature of -24°C and under vacuum for 21 day. The dried samples were 223


impregnated under a vacuum with epoxy resin. After the resin hardened, the surface of specimens were grounded and polished.

3 Results and discussion

From the penetration depths, measured after silver chloride precipitation, the chloride nonsteady state migration coefficient, Equation 2, was calculated. It can be seen in Figure 3 (right) that the migration coefficient decreased substantially when applying Sodium Alginate as an external curing compound. Chloride migration coefficient for alginate-cured mortar was 1,6 times lower than for the air-cure mortar sample.

Figure 3 Chloride penetration (left) and chloride migration coefficient obtained from RCM test results (right)

It has also to be noted that, as mentioned, in this study we used Cem I, which has a low amount of supplementary cementitious materials that go through slower secondary hydration reactions. This allows more time for the water, that is intended for hydration, to evaporate. And since in marine environments, e.g. The Netherlands, cements with supplementary cementitious materials are frequently used.

Figure 4 ESEM-BSE micrograph of near surface area of alginate-cured mortar sample

In Figure 4 and 5 visual observation of near surface area by means of electron microscopy was done. Microstructure of the mortar sample that was covered with alginate exhibited a clearly denser microstructure when compared to the microstructure that was air-cured only. 224


Also, the microstructure that was cured with alginate as an external curing compound exhibited fewer cracks which is a usual phenomena when there is excessive drying of young cement-based material at the surface, e.g. drying or plastic shrinkage.

Figure 5 ESEM-BSE micrograph of near surface area of air-cured mortar sample

Figure 6 shows a near surface area analysed by element mapping, where the elements of interest are calcium and carbon. Calcium is the most abundant polyvalent cation. We analysed calcium leaching due to carbonation or due to the reaction with alginate. In contrast to what was expected, due to the calcium reaction with alginate, a lower calcium content in near surface area was observed for the air-cured sample. The calcium is leached due to bleed water that comes on the surface of the sample from the interior, subsequently Ca2+ reacts with CO 2 and forms CaCO 3 , which leads to a concentration gradient and more calcium will be drawn to surface. In [18] sodium alginate was mixed with gypsum in order to reduce calcium leaching. Authors observed chelation of Mg2+ and Ca2+, which depends on the initial concentration of alginate mixed in with gypsum. Since the samples were impregnated with epoxy resin, analysis of carbon content will indicate the porous and cracked regions in the sample. Air-cured mortar sample showed a higher intensity for carbon, which is in agreement with chloride migration results and observed microstructure, e.g. more porous regions.

Figure 6 ESEM-BSE image with EDS mapping of Ca (yellow) and C (red) of near surface area of A) alginate cured mortar sample and B) air cured mortar sample.

4 Conclusions In this work the influence of alginate as external curing compound for cement-based materials was studied and compared with air-cured samples. Objective was to see whether a watersoluble bio-based polymer, sodium alginate, can be used as an external curing compound for 225


cement-based materials. Rapid cross-linking was observed once the sodium alginate solution arrives at the surface of mortar sample. This indicated that sodium alginate was cross-linking to calcium alginate by means of rapid ion exchange with excess of calcium produced by cement hydration. After 28 days RCM test were performed and significant differences were observed between sodium alginate and air-cured mortar samples. Sodium alginate cured samples had a 1,6 times lower migration coefficient, which gives an indication that the mortar samples cured with sodium alginate could have a potentially longer service life than air-cured. This was confirmed by the use of ESEM, where a more porous and cracked microstructure was observed. Element mapping of near surface are showed more porous regions for air-cured samples compared to sodium alginate cured mortar samples. Future work will focus on the use of sodium alginate as an external curing compound for Cem III/B, which consist of a high mass replacement of cement by supplementary cementitious materials and is therefore more sensitive to curing conditions. Also, an improvement of alginate barrier properties will be investigated by addition of natural clays.

Acknowledgements The authors like to acknowledge the Dutch National Science foundation STW. The research conducted within this project is financed by STW as a part of the IS2C program (www.is2c.nl), number 10962. Arjan Thijsen assistance in ESEM measurement is acknowledged

5 References

[1]Neville AM (1995), Properties of concrete 4th edn. Wiley, New York [2] Richardson IG (2000), The nature of the hydration products in hardened cement pastes, Cem. Concr. Compos. 22(2) 97–113. [3]Hover KC (2011), The influence of water on the performance of concrete, Constr. Build. Mater 25(7):30033013. [4]Powers TC, Brownyard TL (1946-1947), Studies of the physical properties of hardened Portland cement paste, J. Am. Concrete I. 18:1-9. [5]Meeks KW and Carino NJ (1999), Curing of high-performance concrete: Report of the state-of the art. NISTIR 6298, US Department of Commerce. [6]Wang J, Dhir RK, Levitt M (1994), Membrane curing of concrete: moisture loss, Cement Concrete Res. 24(8):1463:1474 [7]Al-Gahtani AS (2010), Effect of curing methods on the properties of plain and blended cement concretes, Constr. Build. Mater. 24:308-314. [8]Grasdalen H, Larsen B, Smidsrød (1981), 13C-n.m.r. studies of monomeric composition and sequence in alginate, Carbohyd. Res. 89(2):179-191. [9]Draget KI, Skjåk-Bræk G, Smidsrød O (1997), Alginate based new materials, Int. J. Biol. Macromol. 21(1):47-55. [10]Linker JB, Jones RS (1966), A new polysaccharide resembling alginic acid isolated from Pseudomonads, J. Bio Chem 241:3845-3851 [11]Gorin PAJ, Spencer JFT (1966), Exocellular alginic acid from Azotobacter vinelandii, Can. J. Chem 44:993998. [12]Laurienzo P (2010), Marine polysaccharides in pharmaceutical application: an overview, Marine drugs 8(9):2435-2465. [13]Matthew IR, Browne RM, Frame JW, Millar BG (1995), Subperiosteal behaviour of alginate and cellulose wound dressing materials, Biomaterials 16(4):275-278. [14]Russo R, Malinconico M, Santagata G (2007), Effect of cross-linking with calcium ions on the physical properties of alginate films, Biomacromolecules 8:3193-3197. [15]Narayanan RP, Melman G, Letourneau NJ, Mendelson NL, Melman A (2012), Photodegradable iron (III) cross-linked alginate gels, Biomacromolecules 13(8):2465-2471. [16]Taylor HFW (1997), Cement Chemistry 2nd edn.Thomas Telford publishing, London [17]NT Build 492 (1999), Concrete, mortar and cement-based repair materials: chloride migration coefficient from non-steady-state migration experiments, Nordtest Finland. [18]Belcarz A, Janczarek M, Kolacz K, Urbanik-Sypniewska T, Ginalska G, (2013), Do Ca2+-chelating polysaccharides reduce calcium ion release from gypsum-based materials?, Cent. Eur. J. Biol. 8(8):735746.

226


AGEING OF PRODUCTS AND STRUCTURES

227


Assessment of ageing in the population of power system components by means of statistical tools Lukasz A. Chmura1*, P.H.F. Morshuis1, J.J. Smit1, A.L.J Janssen2 (1) Delft University of Technology, Delft, The Netherlands (2) Liander, Network Operator, Arnhem, The Netherlands

Abstract: The power system consists of different types of physical assets such as generators, transformers, cables, switchgears etc. Once the component is installed in the power network, it starts to age. The ageing of the component does not mean the passage of time only. Different types of stresses occurring during operation, cause the gradual and irreversible decrease of component operational abilities. The continuous deterioration of the component properties often result in failure occurrence. In addition, the ageing of the population might be seen as increase in failure occurrence over time. One of the useful methods for the modelling of future failure behaviour is statistical analysis of a life-data. In our paper we will present a case study of the life-stage analysis based on the life-data characterizing the population of high-voltage switchgears, obtained from a Dutch utility. Keywords: Ageing, life-data, life-stage, high-voltage components

1 Introduction

The power system is one of the most complex infrastructures that has been ever built. Its primary roles are to continuously generate and deliver the electrical energy to the customers. Nowadays, the utilities are facing the populations of components reaching, or even exceeding their designed life. Although failures occurring within power system are relatively rare, it must be assured that the future failure occurrence will not endanger continuous energy delivery to the users [1]. Despite the certain level of redundancy, the high reliability of the power system components operated by utilities, is required to provide continuous and uninterrupted energy flow [3]. For that reason it is necessary to recognize the number and trend of failures expected to occur in the future, within the given population of components. Particularly to organize or change spare parts and/or the replacement policy.

1.1

Concepts of components life-cycle and ageing

Each component installed in the power network ages and deteriorated during its service-life. The properties of the subcomponents change irreversibly, what finally results in the decrease of operational abilities. This gradual decrease leads finally to the component failure. The moment of failure occurrence determines the end of assets’ technical life-time. However, if the reparation is possible and is made, then the asset is brought into operation and the technical life-time continues. Each type of an asset, has its characteristic technical life-time – service time when stressed nominally. Deterministic approach

Technical

Life-time

Economic Stochastic approach

Strategic

*

Łukasz Chmura, Intelligent Electrical Power Grids, TU Delft, [email protected]

228

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands Figure 1 Different aspect of the life-time [2][4]

In addition, the life of the given asset has to be assessed also from the economic and strategic point of view, using deterministic and stochastic approaches [2, 7]. The details are presented in Figure 1. For the asset, the period of operation during which all the costs related to operation (such as spare parts, installment, operation, maintenance, modifications, dismantling and scrapping) can still be accepted by the asset owner, refers to the economic life-time. Prior the asset installation, a number of requirements for the asset is set from the system point of view. The period of asset operation during which all these requirements are fulfilled by the asset, refers to the strategic life-time. Further on, the life-time can be assessed by means of deterministic or stochastic approaches. This brings the possibility of assessing the life-time on the level of individual component, but also on the level of the whole population. In the first case, the physical condition of the asset is assessed by means of diagnostic tools and inspections, what might show possible deterioration of the asset. For the population of assets where hundreds of assets are in operation, it is difficult to monitor all units. In the second case, the mathematical techniques are employed to estimate the failure behavior within the given population [3]. One of the methodologies that is often used for such needs, is the applicability of the statistical models to model the failure rate behavior.

1.2

Applicability of statistics for ageing assessment

For the needs of statistical analysis, failures that occurred within the population during operation, have to be listed. More specifically, the information about the number and ages-tofailure is to be collected. In addition, the information about the number and ages of in-service population is to be gathered. The failure and in-service data represents life-data of the investigated population. Once the data is collected, the statistical model that represent data best is searched. The example of the distribution, together with accompanying functions such as probability density function f(t), reliability R(t), is presented in Figure 2 - left. Once the parameters of the distribution are estimated, the time-dependent failure rate function λ(t) can be derived. The behavior of this function provides the information about the life-stage of the population. In Figure 2 - right, the exemplary bath-tub curve is presented.

Figure 2 Relation between different functions of the statistical distributions (left) and the bath-tub curve (right).

The bath-tub curve is constructed with three independent failure rate functions, where each depicts different period of the population operation. The periods are:

1) Decreasing failure rate that is often referred to “infant mortality”. In this period, the failures of the assets occur due to production, transportation and installation errors. 2) Constant failure rate – “useful life”. Failures occurring within the population are of random character. 3) Increasing failure rate – “ageing of the population”. The number of failures increases over time. 229


For the ageing and end-of-life assessment, not only the fact of increasing failure rate function is interesting. The critical level of failure rate function which cannot be tolerated from the perspective of economic and strategic requirements is of interest (indicated with red color in Figure 2 – right).

1.3

Weibull distribution

There is a number of statistical distributions allowing the analysis of continuous random variables (e.g. time-to-failure). Among others, these are: the Normal, the Log-Normal, the Gamma and the Gumbel distributions [5]. Special attention has to be paid to the Weibull distribution published first in 1951 [8]. The Weibull distribution characterized by 2 or 3 parameters, is very flexible and can easily imitate other distributions. Additionally, due to its inherent properties, is applicable to the analysis of failures of electrical insulation. The formulas presenting 3 parameter Weibull distribution becomes (1): = f (t )

β  t −γ  η  η 

β −1

  t − γ β  exp  −      η  

(1)

Where: f(t)≥0, t≥0, β˃0 - shape parameter, η>0 – scale parameter, -∞80 years. The term in the initial or renewal license is important and indicates a finite period of operation and, although not mentioned specifically in the current regulations, does not rule out license renewal for multiple terms, as long as aging effects are adequately managed.

4 Conclusion Managing aging effects on DCSSs for extended long-term storage and transportation of used fuel requires knowledge and understanding of the various aging degradation mechanisms for the materials of the SSCs and their environmental exposure conditions for the intended period of operation. The operating experience involving the AMPs, including the past corrective actions resulting in program enhancements or additional programs, provides objective evidence to support a determination that the effects of aging will be adequately managed so that the intended functions of the SSCs will be maintained during the period of extended operation. Compared to nuclear power plants, the operating experience of the DCSSs and ISFSIs is not as extensive; however, evaluations have been performed on the NRC’s requests for additional information (RAIs) on applications for renewal of licenses for ISFSIs and DCSSs, as well as the applicant’s responses to the RAIs, to assess their relevance to the TLAAs and AMPs described in Chapters III and IV of the Rev. 1 aging report, respectively. Those found relevant have been incorporated into the generic AMPs and TLAAs in the report. Managing aging effects on DCSSs for extended long-term storage and transportation of used fuel depends on AMPs for the prevention, mitigation, and early detection of aging effects on the SSCs through condition and/or performance monitoring. Detection of aging effects should occur before there is a loss of any structure’s or component’s intended function. Among the important aspects of detection are the method or technique employed (i.e., visual, volumetric, or surface inspection),

248

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands frequency, sample size, data collection, and timing of new/one-time inspections to ensure timely detection of aging effects. The challenges in the detection of aging effects include (1) detection of aging in regions that are inaccessible for inspection and monitoring, and (2) determining the frequency of inspection and monitoring (i.e., periodic versus continuous). DOE, the NRC, and industry personnel worldwide are actively engaged in various R&D and demonstration activities and meet regularly to exchange information through EPRI’s Extended Storage Collaboration Program and other venues such as International Atomic Energy Agency meetings and international conferences.

Acknowledgment This work is supported by the U.S. Department of Energy’s Used Fuel Disposition Research and Development, Office of Nuclear Energy (NE-53) under Contract DE-AC02-06CH11357. The authors wish to acknowledge their colleagues David Ma, Zenghu Han, Vik Shah, Shiu-wing Tam, and Bud Fabian of Argonne National Laboratory for their contributions to the Rev. 1 report. The authors would like to thank Robert Einziger of the U.S. Nuclear Regulatory Commission for helpful discussions during the course of this work.

References

[1] Chopra OK (2013) “Managing Aging Effects on Dry Cask Storage Systems for Extended Long-Term Storage and Transportation of Used Fuel, Rev. 1,” FCRD-UFD-2013-000294, September 30, 2013, http://www.dis.anl.gov/UFDC/FuelCycle.html. Accessed 6 February 2014. [2] NUREG-1801 (2010) Generic Aging Lessons Learned (GALL) Report, Rev. 2, U.S. Nuclear Regulatory Commission, Washington, DC, December 2010. [3] Einziger RE (2013) An aging management for spent fuel dry storage and transportation, Radwaste Solutions, July–August 2013. [4] EPRI (2014) High Burnup Dry Storage Cask Research and Development Project, Final Test Plan, Electric Power Research Institute, Contract No.: DE-NE-0000593, February 27, 2014. [5] Billone MC, Burtseva TA, and Einziger RE (2013) Ductile-to-brittle transition temperature for high-burnup cladding alloys exposed to simulated drying-storage conditions, J. Nucl. Mater. 433: 431–448. [6] Billone MC, Burtseva TA, and Liu YY (2013) Baseline properties and DBTT of high-burnup PWR cladding alloys, Proceedings of The 17th International Symposium on Packaging and Transportation of Radioactive Materials, PATRAM 2013, San Francisco, CA, August 18–23, 2013. [7] NUREG-1927, Rev. 1 (2011) Standard Review Plan for Renewal of Spent Fuel Dry Cask Storage System Licenses and Certificates of Compliance, Nuclear Regulatory Commission, Washington, DC, March 2011. [8] Takeda H et al. (2008) Development of the detecting method of helium gas leak from canister, Nuclear Engineering and Design, Vol. 238, Issue 5, May, pp. 1220–1226. [9] Tsai HC, Liu YY, Nutt M, and Shuler JM (2011) Advanced surveillance technologies for used fuel long-term storage and transportation, Proceedings 14th International Conference on Environmental Remediation and Radioactive Waste Management, Reims, France, September 25–29, 2011. [10] Tsai HC, Liu YY, and Shuler JM (2013) Monitoring critical facilities by using advanced RF devices, Proceedings of the 15th International Conference on Environmental Remediation and Radioactive Waste Management, Brussels, Belgium, September 8–12, 2013.

249


Seismic Evaluation of Coastal RC Building Vulnerable to an Airborne Chloride Environment Ahmed Mohammed Youssef MOHAMMED1*, Ali AHMED2, Koichi MAEKAWA3 (1) Post-doctoral fellow, The University of Tokyo, Japan. Assistant professor, Department of structural engineering, Cairo University, Giza, Egypt (2) PhD candidate, The University of Tokyo, Japan. (3) Professor, The University of Tokyo, Japan. Abstract: Coastal RC structures for residential or industrial facilities are generally vulnerable to airborne chloride attack. The accumulated chloride upon the faces of these RC structures may cause severe deterioration to the concrete cover and corrosion to the steel reinforcing bars. In this paper, the authors present an analytical investigation on the varying seismic performance of aged RC mid-rise building according to the progress of steel corrosion and spalling of concrete cover. A three stories RC frame is considered and its seismic performance against severe earthquake event is investigated using a 3D nonlinear program DuCOM-COM3 which is verified by a number of experiments. The results clearly show the negative impact of heavy gravitational loads on the lateral resistance of coastal RC structures especially when the probability of corrosion of steel bars is potentially high. The corrosion rate has a significant impact on the lateral resistance capacity of RC structures. Keywords: RC structure, rebar corrosion, seismic ductility

1 Introduction

RC Structures located in coastal cities especially those facing coastal lines are more vulnerable to airborne chloride attack which may cause significant corrosion of reinforcement inside concrete. The vertical and lateral elements of RC structures might be significantly deteriorated due to corrosion, so the analytical evaluation of these structures is needed to predict the future seismic performance. Proper counter measures would be needed to protect these structures from sudden collapse. By monitoring the present corrosion and its effect upon the whole performance of the structure, the future performance will be easier to control. Figure 1 shows a real case of heavily corroded structures located in The Alexandria city (Egypt). These structures are facing the north coastal line and located just several meters from the salted water. It shows a corroded RC slab of a shed from bottom side, heavily corroded RC elements including a slab, beam, column and the connections in a residential building and a beam-column connection of another RC building. Accordingly, a number of research works have been conducted during the last three decades to understand the corrosion mechanism [1-4]. The major effective parameters like cover/diameter ratio, crack propagation, corrosion gel product formation and migration are experimentally and analytically studied as serious trials to propose reliable numerical models to predict mechanism of damage by corrosion [4, 5]. Though the proposed corrosion models have been experimentally verified according to certain aspects but they may not be satisfactory alone to evaluate the overall structural performance with corrosion based on existing conditions. As the material-structure interactions gets complex with corrosion progress and concrete hysteresis consideration so an analytical integrated system is needed to predict the nonlinear *

Post-doctoral fellow, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan. Assistant professor, Department of structural engineering, Cairo University, Giza, Egypt. [email protected]

250


performance of corroded RC structures in future under sever loading conditions like earthquakes. Pertaining to the discussed issues, the thermo-hygro model of concrete composites is linked with both the corrosion rate modelling and the structural dynamic analysis to simulate the structural aging associated with steel corrosion. Here, the accelerated drying and corrosion affected by cracking caused by drying shrinkage and other service loads is taken into account. With this framework, time-dependent performance change of building against severe seismic actions is reported. In some cases, the corrosion induced reduction of structural rigidity may lead to the upgraded safety and story drift modes. It is high-lightened that the localized material based damage is not necessarily the damage of the global structural systems. The experimental verification of the coupled analysis program DuCOM-COM3 [6, 7] is conducted in use of the real scale building columns exposed to the real environment at some semi-tropical regions. As for the building structural system, the shaking table test (e-defense) with a real-scale 7-story RC building is used for verification and the impact of drying shrinkage is also simulated as well [8]. Therefore, a three stories RC Frame as an analytical model for a typical RC residential building is considered to investigate the effect of corrosion on the nonlinear performance. The lateral load resistance and whole structure stability of the building model is clearly illustrated under severe earthquake shaking. Variant rate of steel corrosion is considered to predict the significant change in the structure’s stiffness and ductility. The RC columns considered in this analysis were subjected to two different levels of gravitational loads to investigate the effect of the axial load level on the lateral performance of a corroded RC column.

Figure 1 Existing RC residential-structures suffering from heavy corrosion caused by airborne chloride attack, from Mediterranean sea in Alexandria city, Egypt. 2013

251


2 Constitutive Model DuCOM-COM3 is an analytical platform that couples the thermodynamic integrated computational system (DuCOM) with the structural mechanics modeling (COM3). A reinforced concrete material model has been constructed by combining constitutive laws for cracked concrete and those for reinforcement. The fixed multi-directional smeared crack constitutive equations [6-7] are used as summarized in figure 2. Crack spacing and diameter of reinforcing bars are implicitly taken into account in smeared and joint interface elements, no matter how large they are. The constitutive equations of structural concrete satisfy uniqueness for compression, tension and shear transfer along crack planes. The bond between concrete and reinforcing bars is taken into account in the form of a tension stiffening model, and the space-averaged stress-strain relation of reinforcement is assumed to represent the localized plasticity of steel around concrete cracks. This RC in-plane constitutive modeling has been verified by member-based and structural-oriented experiments. As the corrosion develops, the cross-sectional area of the steel rebar increases because of the accumulation of the expanded rust on its outer surface. The concrete surrounding the steel rebar comes under tensile stress and might be cracked if the stress passed its tensile strength (see figure 2). Herein, the authors skip the details of the RC material modeling by referring to Maekawa et al [6-7].

Figure 2 Constitutive model of RC and experimental verification of a 7-story building

252


3 Analytical Model A three story RC frame 24m in width and 10.5m in height is considered. It is supported by four columns (100cm x100cm) extending through 16.5m of soft sandy soil to bear on a rigid raft foundation. For simplicity, all concrete slabs and foundation mate are considered as 3D elastic solid elements {density=12.83(30%A.L) and 4.28(10%A.L) t f /m3; E c = 30 GPa; and Poisson’s ratio = 0.2} 50 cm in thickness, and the column elements are considered as 3D solid elements with full nonlinearity features of RC (see figure.3). The RC column has been considered under two levels (high: 30% A.L, and low: 10% A.L) of gravitational load by changing the slabs density. The compressive strength of concrete and yield strength of reinforcing bars are assumed to be 30 MPa and 400 MPa respectively. Scaled Kobe earthquake with a PGA of 0.5g is used in the current analysis as shown in Figure 4. R

R

P

P

R

R

Figure 3 FEM model of three stories RC frame

Figure 4 The earthquake (Scaled Kobe EQ) record and its acceleration spectrum

4 Analytical Results

The axially loaded members like RC columns/piles show less ductile performance as the axial load increases [9]. Figure 5 shows the average nominal shear stress per column (for first floor) corresponding to the understory drift of the same floor. The nominal shear stress was calculated by dividing the transmitted base shear force by the columns’ cross section area. With increase in the corrosion rate the structural performance becomes more ductile and the lateral resistance and stiffness reduces. As the axial load becomes higher, the possibility of increasing the base shear transmitted from the foundation to superstructure might increase as shown in figures 5 and 6 by comparing cases of 10% axial load (left side) with 30% axial load (right side). 253


Figure 5 The lateral nominal shear stress as average value per columns of first floor w.r.t the interstory drift; 10% axial load (left side) with 30% axial load (right side)

254


Figure 6 The transmitted shear stress per floor considering levels of axial load and corrosion

Figure 7 Principal strain profile Figure 8 show the vertical movement of the first floor slab against the horizontal one. For the cases of low level of axial load (10% A.L) it shows an increase in the vertical upward movement with increasing corrosion rate due to yielding of reinforcement as the behavior is mainly bending mode. This sort of movement is usually acceptable and comfortable for people living under such slabs. The opposite behavior is noticed for the other cases of high axial load (30%A.L), as the vertical downward movement is significantly large. Corrosion rate has a significant effect on the lateral and vertical movement in both cases of load levels. Low level axial load is a more likely condition for the coastal RC structures.

255


Figure 8 The vertical movement of the first floor slab versus the horizontal one

256


5 Conclusion RC structures in coastal cities and seaside are more vulnerable to airborne chloride attack which may deteriorate the reinforcement by accumulating steel corrosion inside concrete. As the corrosion rate increases the overall building’s lateral capacity reduces and the ductility increases. The lesser the axial load applied to RC columns suffering corrosion environment, the safer the RC structures will be against earthquakes.

Figure 9 the ultimate transmitted shear stress (force), levels of axial load and corrosion

Acknowledgments

This study was financially supported by JSPS KAKENHI Grant No. 23226011

6 References

[1] Takewaka K, Matsumoto S (1984) Behaviors of reinforced concrete members deteriorated by corrosion of reinforcement, Proceedings of JCI 6:177-180 [2] Okada K, Kobayashi K, Miyagawa T (1988) Influence of longitudinal cracking due to reinforcement corrosion on characteristics of reinforced concrete members, ACI Structural Journal 85(2):134-140 [3] Toongoenthong K, Maekawa K (2005) Multi-mechanical approach to structural performance assessment of corroded RC members in shear, Journal of Advanced Concrete Technology 3(1):107-122 [4] Toongoenthong K, Maekawa K (2005) Simulation of coupled corrosive product formation, migration into crack and propagation in reinforced concrete sections, Journal of Advanced Concrete Technology 3(2):253-265 [5] Shang F, An X, Mishima T, Maekawa K (2011) Three-dimensional nonlinear bond model incorporating transverse action in corroded RC members, Journal of Advanced Concrete Technology 9(1):89-102 [6] Maekawa K, Pimanmas A, Okamura H (2003) Nonlinear Mechanics of Reinforced Concrete. Spon Press, London [7] Maekawa K, Ishida T, Kishi T (2003) Multi-scale modeling of concrete performance -Integrated material and structural mechanics, Journal of Advanced Concrete Technology 1(2):91-126 [8] Shibata K, Homma S, Chijiwa N, Maekawa, K (2013) Seismic evaluation of RC columns considering equivalency of circular and square cross-sections, 11-13 September 2013, The thirteenth East AsiaPacific Conference on Structural Engineering and Construction D:1-5, Sapporo [9] Mohammed AMY, Maekawa K (2012) Global and local Impacts of Soil Confinement on RC Pile Nonlinearity, Journal of Advanced Concrete Technology 10:375-388

257


Effect of Ageing and Cyclic Loading on the Strength of Gravity Dams Archil Motsonelidze1*, Vitaly Dvalishvili2 (1) Georgian Technical University, Tbilisi, Georgia (2) Georgian Technical University, Tbilisi, Georgia

Abstract: The finite element based approach for the analysis of existing old and “tired” concrete dams takes the structure’s age and past loading history into account. The following factors are included in the model: a) fatigue of the dam material under slow cyclic loading corresponding to its performance history and b) elastic-plastic constitutive relationship for the dam concrete and the interfaces/joints in accordance with the dam loading history. Keywords: concrete dam, ageing of concrete, cyclic loading, modulus of elasticity

1 Introduction The proposed approach is based on the assumption that the current mode of deformation of dams is formed throughout the history of their operation. Hence, the analysis should take into account all of the major factors acting during their performance. The main factors that are taken into account in this approach are cyclic-induced degradation of stiffness and strength for concrete (fatigue of concrete) and aging-induced change of stiffness and strength for concrete (aging of concrete). Besides these factors the technique enables to analyse creep strains by means of modified Boltzmann-Volterra theory of linear hereditary creep. A nonlinear-elastic fracture constitutive model for concrete in the plane strain condition is used. Special interface/joint elements are employed to model the relative movement of adjacent boundaries at the dam-foundation interface. Interface/joint elements can also be used to simulate the behaviour of interfaces and joints which may generally exist in the body of a gravity dam. Analysis of the future state of the dam is performed by simulating the future number of loading/unloading cycles and time-period of dam operation. The main steps of the analysis are summarized below (see Table 1). Table 1 Summary of the Analysis Framework Type of Deformation Concrete in plane strain condition

Creep strain

Crack occurrence and propagation

*

Retrospective Analysis Stage R-1 Input: Initial values of stiffness and strength for concrete and interfaces. Analysis method: nonlinear elastic fracture constitutive model for concrete in the plane strain condition; constitutive model for interfaces. Output: initial mode of deformation of a dam - (σ,ε) in Stage R-2 Input: initial values of stiffness and strength for concrete (from Stage R-1); Initial mode of deformation of a dam - (σ, ε) in (from Stage R-1). Analysis method: analysis of creep strain ε cr by means of modified Boltzmann – Volterra theory of linear hereditary creep; modified nonlinear-elastic fracture constitutive model for dam concrete using calculated creep strain ε cr . Output: modified stiffness and strength of concrete; modified mode of deformation of a dam - (σ, ε) mod . Stage R-3 Input: modified stiffness and strength for concrete (from Stage R-2); modified mode of deformation of a dam - (σ, ε) mod (from Stage R-2).

Member of the scientific research society “Sigma Xi” and the Georgian National Academy of Power Engineering. E-mail: [email protected]

258


Analysis method: failure criterion for dam concrete in plane strain condition; failure criterion for interfaces; stress degradation effect near crack tips. Output: crack occurrence and propagation in concrete and/or interfaces; modified mode of deformation of a dam - (σ, ε) mod. Stage R-4 Input: initial values of stiffness and strength for concrete and interfaces (from Stage R-1); values of n p (number of loading/unloading cycles) and t p (time-period of dam operation). Analysis method: empirical relationships based on the experimental findings on cyclic behavior and ageing of concrete; calculation of cyclic-induced plastic (residual) strain ε p ; cyclic-induced degradation of stiffness and strength for concrete (fatigue of concrete); ageing-induced change of stiffness and strength for concrete. Output: cyclic-induced plastic (residual) strain ε p ; modified values of stiffness and strength for concrete and interfaces. Stage P-1 Input: cyclic-induced plastic (residual) strain ε p (from Stage R-4); modified values of stiffness and strength for concrete and interfaces (from Stage R-4). Analysis method: nonlinear-elastic fracture) constitutive model for concrete in the plane strain condition; constitutive model for interfaces. Output: modified mode of deformation of a dam - (σ, ε) mod . Stage P-2 Input: modified values of stiffness and strength for concrete (from Stage R-4); modified mode of deformation of a dam - (σ, ε) mod (from Stage P-1). Analysis method: analysis of creep strain ε cr by means of modified BoltzmannVolterra theory of linear hereditary creep; modified nonlinear-elastic fracture constitutive model for dam concrete using calculated creep strain ε cr . Output: modified stiffness and strength of concrete; modified mode of deformation of a dam - (σ, ε) mod . Stage P-3 Input: modified stiffness and strength for concrete (from Stage P-2); modified mode of deformation of a dam - (σ, ε) mod (from Stage P-2). Analysis method: failure criterion for dam concrete in plane strain condition; failure criterion for interfaces; stress degradation effect near crack tips. Output: crack occurrence and propagation in concrete and/or interfaces; modiied mode of deformation of a dam - (σ, ε) mod. Stage F-1 Input: modified mode of deformation of a dam - (σ, ε) mod (from Stage P-3). Analysis method: future performance of a dam simulated by assuming hypothetical values of the future number of loading/ unloading cycles n f and time-period of dam operation t f , and repeating analysis from Stage R-1 through Stage P-3. Output: forecasting the future performance of a dam and drafting the relevant recommendations.

Cyclic-induced plastic (residual) strain and degradation of material properties of concrete. Ageing of concrete

Concrete in plane strain condition

Creep strain

Crack occurrence and propagation

Cyclic-induced plastic (residual) strain and degradation of material properties of concrete. Ageing of concrete

Definition: R=Retrospective analysis; P=Analysis of the Present state; F=analysis of the Future performance

1.1

Stage R-1 and Stage P-1

On the analysis Stages R-1 and P-1as a constitutive model, the four-parameter failure criterion [1], modified for plane strain problems [2], is employed:

a

J2p

σ c2

−b

J2p

σc

−c

I1 p σ 1p −d −1 = 0 σc σc

with:

259

(1)


I1 p = σ 1 p + σ 2 p + σ 3 p ;

J2p =

[

]

1 (σ 1 p − σ 2 p )2 + (σ 2 p − σ 3 p )2 + (σ 3 p − σ 1 p )2 ; 6

(2)

(3)

where σ 1p ≥ σ 2p ≥ σ 3p = principal stresses corresponding to the peak state; σ c = uniaxial compressive strength of concrete; and a, b, c, d = parameters which are to be determined experimentally (a=2,018; b=0,9714; c=9,1421; d=0,2312). The general secant stress-strain relation for transversely isotropic materials in the principal stress space [3], modified for plane strain problems [2] is employed as well.

1.2 Stage R-2 and Stage P-2 In order to take the concrete creep into account, one should know the initial mode of deformation and material properties of the dam-foundation system. The calculations are carried out in three steps [4]: Step 1 - the initial mode of deformation of the dam-foundation system is calculated; Step 2 - the damfoundation system is analyzed corresponding to the time when the strength of the dam concrete is anticipated to have reached its maximum value: this can happen within a period of 8-10 years, following the placing of concrete in the body of a dam, by which time the creep process may reasonably be assumed to have been completed. Within this period, if the cracks occur in the body of the dam, the associated stresses should be analyzed subject to both the cracking and the creep processes; Step 3 - the present time mode of deformation of the dam is analyzed. The material properties of concrete are modified in accordance with the number of loading/unloading cycle’s n and time-period of dam operation t. If crack propagation is observed in the body of the dam during the analysis, the creep process may occur again due to stress redistribution (higher stresses) near the crack tip(s). For the present mode, the Boltzmann-Volterra theory of linear hereditary creep is employed to define the creep process in concrete.

1.3 Stage R-3 and Stage P-3 The parametric studies illustrated that the interface isoparametric quadratic singular finite elements can be effectively employed to model the problem of crack occurrence and propagation within the body of an elastic continuum. This elements can also be effectively employed to model the problem of crack occurrence and propagation at the base of an elastic continuum.

2 Effect of ageing of dam concrete and slow cyclic loading 2.1 Main principle of an approach The above discussed constitutive model (see 1.1 Stage R-1 and Stage P-1) for concrete in plane strain condition is modified in order to take into account the effect of ageing of concrete and the effect of material fatigue under slow static cyclic loading. Particular emphasis is placed on material response, simulating a realistic behaviour of a dam under actual conditions. This approach enables us to account for the effect of concrete ageing and also for the effect of concrete strength (fatigue) degradation under cyclic loading. To this end, the value of uniaxial compressive strength of concrete σ c in Equation 1 can be substituted by the value of the strength of concrete which

260


has been appropriately modified in accordance with the time-period of the operation of a dam t and the number of loading/unloading cycles n :

σ c = σ c (t , n ) (4)

The value of strain ε c associated with the maximum uniaxial compressive stress of concrete may also be modified in accordance with the number of loading/unloading cycles n and the time-period of the operation of a dam t:

ε c = ε c (t , n ) (5)

It is well known that the passage of time causes the concrete modulus of elasticity to increase. At the same time, cyclic loading causes the degradation of concrete stiffness. These two effects are accounted for in the present model by means of the time-period of operation of a dam t and modifying the value of the initial modulus of elasticity in accordance with the number of loading/unloading cycles n:

2.2

Effect of slow cyclic loading

E0 = E0 (t , n ) (6)

Cycling loading exhibits significant nonlinear behaviour and drastic changes in material properties of concrete. The result is a considerable degradation of material properties of concrete as the number of applied loading/unloading cycles increases. In the present work, the empirical relationships based on the experimental findings on cyclic behaviour [5] are adopted. Very briefly, following the tests carried out on the concrete specimens of Enguri arch dam, which were subjected to slow static cyclic compressive loading, the following relationships were established to define the degradation of the material properties of concrete in relation to loading/unloading cycles:

σ c (n ) = (1 − aσn lg n )σ c

( (n ) = (1 − a

) lg n )ε

E 0 (n ) = 1 − a En lg n E 0

εc n

n

n

ε

c

(7)

n

where parameters a σ , a E and a ε define the degradation of the material properties of concrete under slow static cyclic loading; and n is the number of loading/unloading cycles in accordance with the operation history of a gravity dam (it corresponds to the number of discharge/filling up cycles of the reservoir during the operation of the dam). The specific values of these parameters may only be determined by carrying out cyclic tests on concrete specimens. Nevertheless, a careful study of the results of the investigations suggests that the values of the above parameters vary within the following ranges: 0,05 ≤ α σn ≤ 0,25 ;

0,10 ≤ α En ≤ 0,30 ; 0,10 ≤ α zn ≤ 0,30 . It is also interesting to note that, in general, the rate of degradation of material parameters and the strength of concrete under static cyclic loading primarily depends on the level of stresses (i.e. the level of loading imposed on the specimen). For example, in tests,the value of modulus of elasticity for a concrete specimen was found to have been reduced by 51.5% (from 39780 MPa to 19300 MPa) after applying 150 loading/unloading cycles when the level of applied stresses was equal to 0,2σ c , with σ c = the uniaxial compressive strength of concrete. When subjected to the same number of loading/unloading cycles (i.e. 150), with the level of applied stresses set equal to 0,5σ c and 0,8σ c , the corresponding values of moduli of elasticity for the concrete specimen were found to have been reduced by 29,3% (from 33390 MPa to 23620 MPa), and 20,9% (from 28390 MPa to 22500 MPa), respectively [5]. In the present work, average values of the above-defined coefficients (which seem to provide reasonable approximation of an actual behavior of concrete under cyclic loading) were adopted, although, it should be noted that, in general, it is, indeed, possible to define the relationship between the parameters a σ n, a E n, a ε n and the level of stresses.

261


2.3

Effect of ageing of dam concrete

The same approach can be used to describe the change in material properties of concrete due to ageing of the material. Again, the logarithmic function seems to predict well the variations of the material properties of concrete with time. Namely, the relationships can be expressed as:

σ c (t ) = (1 + aσt lg t )σ c

( (t ) = (1 + a

) lg t )ε

E 0 (t ) = 1 + a Et lg t E 0 (8)

εc t

t

t

ε

c

t

where the parameters a σ , a E and a ε define the modification of the material parameters of concrete due to ageing of concrete; and t is the number of years in accordance with the operation history of a gravity dam. A careful study of the final results suggest that the values of the above parameters may lie within certain fairly well-defined ranges, depending on the type of concrete: obviously, specific values of these parameters are best found by in-situ measurements on concrete specimens. Tests reported by [5], suggest the above coefficients to lie within the following ranges: 0,05 ≤ α σt ≤ 0,15 ;

0,05 ≤ α Et ≤ 0,15 ; 0,05 ≤ α zt ≤ 0,10 . It can be seen that the values of the parameters a σ t, a E t and a ε t are generally lower than the values of the parameters a σ n, a E n and a ε n. This means that the effect of ageing of concrete on the stress-strain curve of concrete is less significant than that of static cyclic loading. However, it should be noted that there may be different combinations of the number of years and the number of loading/unloading cycles over a period of time. For example, if during the operation of a dam, there is only one discharge/filling cycle of the reservoir during a year: n = t, while for two discharge/filling up cycles of the reservoir during a year, the number of filling/unloading cycles will be twice as much as the number of years of operation of the reservoir- i.e. n = 2t. The above-defined values of the material properties of concrete were subsequently incorporated into the constitutive equations presented in Part 1 (i.e. the constitutive model for concrete under the plane strain condition) in order to account for the fatigue as well as aging of concrete under cyclic loading. The objective here was to find out as to how changes in material properties of concrete due to ageing and static cyclic loading affect the shape of the stress-strain curve for concrete under the plane stress and plane strain conditions, with the analyses performed only for the compression-compression and compressiontension modes of loading. As regards the tension-tension mode of loading, the corresponding stressstrain relationship was reasonably assumed to be linear with the tensile strength of concrete varying proportionally with changes in the value of the compressive strength of concrete. In total, about 200 computer runs were carried out, covering a wide range of different stress combinations α (with this parameter defining the ratio of the stress in the minor principal direction to the corresponding stress in the major principal direction), and for t = 0, 10, 25, 50, and 100, for the two cases of n = t and n = 2t. Some results for the tension - compression mode of loading are presented in Figure 1. Generally, the degree of the degradation of the material properties of concrete increase with increases in the number of loading/unloading cycles n. Ageing of concrete, on the other hand, does have certain favorable effects on the material properties of concrete, although the degradation of the material properties of concrete due to the combined (but contradictory) actions of aging and cyclic loading is still practically significant. Namely, when α = − 0,15; t = 100 and n = 200, the specimen was found to fail with a crack occurring in the plane strain condition when the compressive stress equals to only about 20% of the uniaxial compressive strength of concrete.

262


Figure 1 Comparison of uniaxial compression test, plane stress and plain strain conditions of concrete for tensioncompression state accounting for the number of loading-unloading cycles n and age of concrete t with α = −0,15

3 Effect of ageing of dam concrete and slow cyclic loading on interfaces The effect of cyclic loading on the shear stiffness k s 0 can be accounted for by means of modifying the value of the initial shear stiffness associated with zero normal stress

(k )

0 σ n =0 s .

263


In line with the approach adopted in Part 6, the following empirical relationship may be established to define the degradation of the magnitude of the initial shear stiffness associated with zero normal stress as a function of loading/unloading cycles n:

(k )

0 σ n =0 s

( )

= (1 − a kn lg n) k s0

σ n =0

(9)

where the parameter a k n defines the degradation of initial shear stiffness associated with zero normal stress under static cyclic loading; and n is the number of loading/unloading cycles in accordance with the operation history of a dam. The degradation of the shear strength at interfaces under static cyclic loading can be accounted for by modifying the value of cohesion c in Mohr-Coulomb criterion . The degradation of the magnitude of the shear strength corresponding to loading/unloading cycles n may reasonably be defined by the following empirical relationship:

τ c = c(1 − aτn lg n) + σ n tan φ (10)

where the parameter a τ n defines the degradation of the value of cohesion under static cyclic loading n with the values of the above coefficients lying within the following ranges: 0,10 ≤ α τ ≤ 0,25 ;

0,15 ≤ α kn ≤ 0,30

The effect of ageing of concrete is accounted for in a similar fashion. Namely, for the initial value of the shear stiffness associated with zero normal stress, one has:

(k )

0 σ n =0 s

( )

= (1 + a kt lg t ) k s0

σ n =0

(11)

The effect of ageing of concrete on the shear strength at interfaces may, on the other hand, be defined as:

τ c = c(1 + aτt lg t ) + σ n tan φ (12)

With the values of the relevant coefficients lying within the following ranges:

0,10 ≤ α kt ≤ 0,20

0,05 ≤ α τt ≤ 0,15 ;

Typical results, based on the above formulations, are presented in Figure 2, which depicts the effects of cyclic loading as well as ageing of concrete on the shear stress – relative displacement curves relating to an interface. It is clear that the cyclic loading and ageing of concrete cause a significant degradation of material behavior at the joints. Obviously, whenever possible, it is preferable to use input data, based on in-situ monitoring, for the analysis of a gravity dam, with the present results intended to demonstrate the practical importance of taking into account the effects of cyclic loading and ageing of the concrete in the body of a dam, when attempts are made to estimate the reserve strength(s) of old concrete gravity dam(s).

4 Conclusions Typical numerical studies, based on the proposed approach, suggest that: (1) the rate and extent of degradation of the concrete mechanical properties increases with the increasing numbers of discharge/filling up cycles of the reservoir; and (2) the degradation of the material properties of concrete due to the combined actions of loading/unloading cycles and ageing of concrete can still be practically significant. Namely, when α = − 0,15; t = 100 and n = 200, the specimen was found to fail with a crack occurring in the plane strain condition when the compressive stress equals to only about 20% of the uniaxial compressive strength of concrete.

264


Figure 2 Shear stress/relative displacement plots taking into account the number of loading/unloading cycles and the process of ageing of concrete t for an interface between concrete and rock, when σ n = 8,01 kg/cm2, c = 3,1 kg/cm2, tan φ = 1,1 k s0 = 230,0 kg/cm3, P a = 1,033 kg/cm2

5 Acknowledgement This research was supported by the Shota Rustaveli National Science Foundation grant. The authors would like to thank Professor Mohammed Raoof of Loughborough University and Dr. Vissarion Abuladze for their significant contributions to the development of this problem throughout the course of the study.

6 References [1] Hsieh SS., Ting EC., Chen WF. An elastic-fracture model for concrete. Proceedings of the 3d Eng. Mech. Div. Spec. Conf., ASCE. Austin, Texas 1979. pp. 437- 440 [2] Lin Z., Raoof M. (1993) A simple biaxial tangent constitutive model for concrete under static monotonic loading only. Proceedings of the Institution of Civil Engineers, Structures and Buildings, 95, pp. 49-54. [3] Lekhnitski SG (1963) Theory of elasticity of an anisotropic elastic body, Holden Day, San-Francisco [4] Motsonelidze A., Raoof M, Abuladze V. Prediction of concrete dam’s reserve strength based on its past loading history. Proceedings of the II International Congress on Dam Maintenance and Rehabilitation, Zaragoza, Spain, 23-25 November, 2010, pp. 777-786 [5] Osidze, V., Khoperia D. (1987) Deformation parameters of concrete of Enguri arch dam subject to static cyclic compressive loading. In: Construction of Hydro Power Stations in Mountainous Regions, Energoatomizdat, Moscow, pp 52-58

265


Delayed deformations of segmental pre-stressed concrete bridges: the case of the Savines Bridge J-P. Sellin1*, J-F. Barthélémy1, J-M. Torrenti2, G. Bondonet1 (1) Cerema/DTecITM (formerly Setra), Sourdun, France (2) Université Paris-Est IFSTTAR, Marne-La-Vallée, France Abstract: This paper is dedicated to the prediction of long term delayed deformations of prestressed concrete bridges with practical application to the Savines bridge. The reason why long term deflections increase with time is not fully understood and remains the purpose of many researches. The case of the Savines bridge has been chosen because of its age (60 years), the extensive available data and the problematic evolution of its deflection. A model is presented, based on the beam theory in which constitutive laws for materials are adjusted from current European standards. The measurements far exceed the calculated deflection. Empirical corrections of the material laws are then proposed in order to comply with data. Keywords: Bridge, Deflection, Serviceability, Creep, Standards

1 Introduction Since the first findings by Hyatt in 1907 (USA) and by Freyssinet in 1912 on the Veurdre bridge (France), it has become well known that concrete undergoes delayed deformations that may result in structural failures and serviceability problems. This phenomenon is highlighted in this paper with the representative example of the Savines Bridge. The Savines Bridge is one of the first generation of post-tensioned prestressed concrete bridges built in France. It has been constructed between 1958 and 1960 by the free cantilever method. It is a 77 m long box-girder bridge made of 13 spans, each of them consisting of 22 cast-in-place segments with depths varying from 4,15 m at main piers to 1,15 m at mid-span (see Figure 1). The concrete deck is post-tensioned with internal tendons. At the middle of each span, rotations and horizontal displacements are free of motion thanks to a horizontally sliding hinge which transmits only shearing forces. According to a symmetry argument, each cantilever can be considered in a first analysis as a statically determinate structure independent from the others. Of course this argument is questionable for a detailed analysis that must take into account defects from the articulations and deviations from the symmetry assumptions; nevertheless these discrepancies are disregarded in this study. One major consequence of this connection between adjacent cantilevers is that each mid-span hinge corresponds to a free displacement degree, leading to potential serviceability deformation problems.

Figure 1 Detailed view of a single span [1]

Only one year after the end of construction (summer 1961) excessive deformations were observed. For some spans, deflection at mid-span reached 80 mm in 1961 [1] and about 150 to 200 mm today. In terms of serviceability and user's comfort, these creep and shrinkage deformations may be considered not acceptable because of the non planar pavement. Hence, retrofitting consisted of *

Cerema/DTecITM, Technical Centre for Bridge Engineering, [email protected]

266


reducing the deflections with additional post-tension tendons. Besides, layers of asphalt were added onto the middle of the span. This kind of rehabilitation has been performed three times in 60 years, increasing stresses in the deck. In 2003, it was decided to eliminate this asphalt overload located at mid span and thus to accept a non plane deck in order to decrease the dead load effects. During its service life, the Savines Bridge has carefully been inspected (inside and outside the boxgirder and in the piers). These inspections have not highlighted structural cracks which could develop into failure or be the source of extra-deflection at mid-span.

1.1

Raw measurements

A continuous monitoring of the deflection evolution has been implemented on the structure since the first observation of large values. In the following analysis, the span named "f" is particularly considered because it shows the largest deflections and is well documented in terms of geometrical characteristics and design calculations. The analysis of the monitoring measurements has given rise to two problems: environmental conditions (relative humidity, temperature) vary from one dataset to another and the reference measurement baseline has changed several times without information cross-referencing. The deflections are therefore defined up to an unknown translation from one dataset period to another one. Consequently, for some dates, manual corrections are required in order to estimate likely values of absolute deflections. Raw values of the deflections at mid-span "f" are represented in Figure 2 for the four series of measurements.

1.2

Corrected deflections

Due to the nature of works carried out during maintenance operations and to possible variations of the bridge deflection with temperature, it is difficult to correct the raw measurements in order to obtain the real continuous deflection evolution. Indeed, the sensitivity to adjustments is rather high.

Figure 2 Gross deflection measurements for the span "f" for the four different campaigns performed

Figure 3 Evolutions of the deflections since the first measurement for two possible adjustments

The doubts about the measurement baselines prevent the long term behaviour from being easily identified. Finally, two possible trends are considered: the first approaches a horizontal asymptote and the second one continues to increase after 50 years (see Figure 3 representing the deflection variation between the current time t and the initial time t 0 chosen as the beginning of service life).

1.3

The origins of delayed deformations

The increasing evolution of the span deflection is caused by the delayed deformations of concrete combined with the relaxation of steel prestressing tendons. Many research results (see [2] for a review) have shown that delayed deformation of concrete can be split into two parts: shrinkage (independent of the loading) and creep (depending on the stress state). On the one hand, shrinkage results from the effect of capillary pressure due to self-desiccation (due to hydration) or drying. On the other hand, creep can be explained by two phenomena: basic creep related to viscous deformations at the scale of the hydrates and desiccation creep caused by drying i.e. water diffusion out of the concrete volume. As desiccation shrinkage and creep are generated by water diffusion, their evolution should be governed by the characteristic time of the latter process. Likewise self-desiccation

267


shrinkage should be controlled by the characteristic time of the hydration process. However the deep origins of basic creep and consequently of its kinetics remain a controversial issue in the scientific world: some authors argue that long term basic creep would be due to sliding between C-S-H sheets [3, 4] whereas others propose a mechanism based on dissolution and recrystallization of hydrates [5]. Anyway, for structures showing delayed deformations such as the Savines bridge, it remains difficult to identify the contribution of each phenomenon at the macroscopic scale.


Numerical model, input data

Numerical model The bridge is modelled with ST1 software [6] developed by the Cerema/DTecITM (formerly Setra). Only the half span "f" is modelled as a cantilever beam assuming fixed points between the deck and the pile. In the ST1 software, the creep calculation algorithm is based on the incremental theory [7]. The application of the superposition principle is consistent with the linearity assumption. According to Eurocodes standards [8, 9], the latter is satisfied if stresses do not exceed 0,45f ck , where f ck is the characteristic compressive strength value of concrete. Construction stages According to the initial drawings and the construction time schedule, the construction process for each segment is accurately modelled taking into account concrete casting, moving of the formworkcarrying traveller… The first measurement, which corresponds to the beginning of the bridge service life, has been performed on May 14th, 1960 (the day after the load test was completed on May 13th, 1960). Geometry The beam model required an average estimate of the notional size. The notional size as defined in NF EN1992-2:2006 [8] is h = 2 Ac / u , where A c is the cross-section and u is the perimeter of the member in contact with the atmosphere. The mean notional size of the box-girder is h = 45 cm.

2.2

Material properties

Concrete The mean compressive strength measured on site [10] is about 40 MPa after 28 days. According to the EN1992-1-1, the characteristic compressive strength value at 28 days is assumed to be 32 MPa as defined in Eurocodes [8, 9] f ck,28 = f cm,28 – 8 MPa. The concrete density based on its composition is evaluated to 2457 kg/m3. Considering bending, shearing and local reinforcement, the weight of the reinforced concrete is evaluated to 25,7 kN/m3. The ratio of bending steel reinforcement is set equal to ρ s = 0,5 %. Eurocodes 2 (EN1992-1-1 and EN1992-2 [8, 9]) enables a reduction of the Young modulus of concrete depending on the type of aggregates. The latter is not well documented: it is mentioned in design notes that aggregates are grid-rolled 10/20 and come from the local area of the Serre-Ponçon lake. Thereby, the worst case is chosen: the instantaneous modulus of elasticity value (at 28 days) is lowered by 30 %. Cement Available documentation and site test measurements state that the cement used conferred a high initial resistance. According to NF EN197-1:2001 [11] and to EN1992-1-1 [8], it is assumed that it is equivalent to a currently designated "R" cement. Post-tensioning cables Steel tendons characteristics such as section, anchorage slips, limit strength or coefficient of curvature friction are given in French technical agreements for S.E.E.E. F13 and for FU 4600 tendons [12]. Design justifications propose a 9% relaxation ρ 1000 after 1000 hours. Nonetheless, a study based on the crossbow method [13] suggests that relaxation has been underestimated for first generations of

268


prestressing tendons. According to [12], S.E.E.E. F13 tendons may be sorted in group 1 (1950-1958) or in group 2 (1959-1967) and they correspond therefore to the first generation of tendons. It reveals that relaxation of steel could be between ρ 1000 = 9 % and 12 %. Real values of stress applied on site for internal tendons gives 1250 MPa. Additional post-stressing In 1966, in order to limit deflection, additional post-stressing tendons were set into the structure. This maintenance is also modelled thanks to available drawings and design justifications [15].

2.3

Site conditions

Ambient relative humidity Despite this parameter is not accurately known, 20 dataset [16] over a three years period (1964 to 1966) are available. The chosen relative humidity used is the average relative humidity along these years and is RH = 75 %. This rather high value is explained by the fact that the bridge crosses the Serre-Ponçon Lake. Temperatures changes Effects of temperature changes are not taken into account in this study.

2.4

Comparison between measured and calculated values

Figure 4 Comparison of the predicted deflections with experimental measurements

Simulations are undertaken with two current standards: EN1992-1-1 [8] and EN1992-2.[9]. A first calculation involving shear force effects in the constitutive law highlights a negligible contribution of the latter (less than 1 %). At t 0 (date corresponding to the beginning of service life) prestressing induces an upward sag of 15.9 mm (for EN1992-2). The deformed structure at time t 0 represents the reference configuration for further calculations allowing comparisons with corrected measurements drawn in Figure 3. Maximum compressive stresses calculated along its service life are 5,3 MPa and 11,4 MPa respectively at the top and at the bottom flange. Thus, the assumption of linear visco-elastic behaviour is verified (see section 2.1). Results for each standard are compared to the in-situ measurements (see Figure 4). It can be seen that the deflections are underestimated by the two codes by a factor of 2 to 3.

2.5

Sensitivity of parameters

Sensitivity of input data The influence of the input data is observed comparing the deflection after 50 years after changing one parameter; the other ones corresponding to the mean value. Their influence are represented and compared to the mean calculation. Table 1 classifies main parameters according to their sensitivity for the EN1992-1-1 [8] model. It classifies parameters according to the variation of the relative deflection

269


Δf/f when the relative parameter Δp/p varies around the initial point. In order to classify these parameters, the mean slope s = (Δf/f ) / (Δp/p) is calculated. Table 1 Sensitivity of parameters for EN1992-1-1 Input data

Sensitivity

Slope s

Specific weight of the reinforced concrete

Very high

0,93

Relaxation of steel

Very high

0,85

Compressive strength

High

0,67

Ambient relative humidity

High

0,66

Weight of superstructures

Medium

0,46

Initial tension applied

Low

0,14

Notional size

Low

0,10

Sensitivity to shrinkage and creep models The effect of creep and shrinkage may be adapted by adding weighing coefficients k i in the decomposition of the delayed strains: ε = k as ε as + k ds ε ds + k bc ε bc + k dc ε dc where ε as , ε ds , ε bc and ε dc respectively denote autogenous shrinkage, desiccation shrinkage, basic creep and desiccation creep strains. Note that in the case of EN1992-1-1 there is no separation between basic and desiccation creep. In this case we consider only one coefficient k c for creep. EN1992-2 [9] also allows an optimization of the creep kinetics by means of another coefficient k kc . The creep kinetics is expressed t − t0 by: f(t − t 0 ) = where (t-t 0 ) is the loading age and β c = 0,40 ⋅ e 3.1 f cm f ck [8] t − t 0 + k kc ⋅ β c In order to know the major effects, the influence of these coefficients, each taken independently, is represented in Figure 5 (left) for EN1992-1-1 and Figure 5 (right) for EN1992-2. Variations of the deflections are compared to the initial calculation where all the coefficients are equal to 1.

Figure 5 Sensitivity to the deflection versus the weighting coefficients of shrinkage and creep for EN1992-1-1 model (left), for EN1992-2 model (right)

For the two models the most sensitive parameter is the one related to basic creep.

2.6

Model optimization

As mentioned above, standards underestimate the real deflections. Hence, weighing coefficients are adjusted in order to minimize the difference between in-situ measurements and numerical figures via an optimization algorithm based on the least square method.

270


The optimization process is only presented for EN1992-2 [9] because it allows adjustment more parameters related to the creep mechanisms and therefore gives the best results. Figure 5 shows that the deflection variation is linear with creep and shrinkage variations and it is expected that a loss of deflection due to a decrease of shrinkage strains could be compensated by an adequate increase of creep strains. Therefore, in order to simplify and identify the contribution of creep, shrinkage strains are not weighted. The chosen variable coefficients are in this case k bc , k dc and k kc (for creep), the relaxation of steel and the specific weight of the reinforced concrete because is has been shown that it is a relevant parameter (see Table 1). Experimental tests undertaken by Granger [17] have shown that the final creep after 3 years varies between ± 50%. fib bulletin 42 reports [18] that the interval of variation for the creep compliance function is about ± 60%. In accordance with these reports, weighing coefficients can move between 0,4 and 1,6. Table 2 summarizes the constraints and the optimized values for the two trends of the evolution of deflection and Figure 6 shows the ability of the optimized process to account for observed delayed deflections. Table 2 Results of optimization for EN1992-2 Optimized value Constraints

Trend 1

Trend 2

k bc

0,4 < k bc < 1,6

1,5

1,6

k dc

0,4< k dc < 1,6

1,5

1,6

k kc

0,1 < k kc < 7,5

1,4

1,4

ρ 1000

9 % < ρ 1000 < 12%

11,8

12

γc

25 kN/m3 < γ c < 27 kN/m3

26,7

27

Finally, from this optimization process the evolution of the deflections during the next 50 years can be predicted: it can be shown that the difference between the two sets of optimized values is limited to 5 cm (see Figure 6). With both sets the deflection is still increasing.

Figure 6 Prediction of the optimized computed deflection for the next 50 years

3 Discussion and conclusions Although measurements are precise enough, due to the change of baselines several times and the influence of external temperature, it is rather difficult to assess the real long term behaviour of the Savines Bridge. As shown in section 1.2, corrected deflection could either stagnate or increase by about 30 mm after a further 50 years of service life (as shown in Figure 3).

271


Despite these uncertainties, it is clear that in the case of the Savines bridge, all calculations based on standard parameters and laws systematically underestimate long term deformations of concrete. Further reasons may partly explain the discrepancy between calculations based on standards and experimental measurements: - Bažant et al. [19] present a numerous collection of deflection history of bridges. It shows that, in logarithmic time scale, deflections evolve systematically in a straight line. Current evolution of deflection suggests that deformation of the Savines bridge likely follows the trend 2 (see Figure 3) and evolves as a straight line in logarithmic time scale. In this case, the kinetics for creep could also be logarithmic as it is proposed for basic creep in Model Code 2010 [20]; - EN1992-1-1 suggests that the relaxation of steel tendons reaches an asymptotic value after 500,000 hours. If this phenomenon was considered as still active even after this period, the deflection could go on increasing. Besides, high temperature (for instance during summer under the asphalt layer) could also lower the stress within tendons [21]; - Shear lag effects, which are neglected in the present model based on beam theory, are known to be responsible of a significant part of the deflection of a cantilever beam (up to 25 % in [22, 23] or 20 % in the case of the Palau bridge [24]). Adaptation of creep laws with weighing coefficients, as allowed by EN1992-2 [9], enables compliance with in-situ measurements and makes it possible to obtain a trend for the future deflections. It might be concluded that current standards need to be adapted with new coefficients or reconsidered . But, the added coefficients do not have physical meanings. Indeed, their sensitivity (for EN1992-1-1 and for EN1992-2) [8, 9] shows that increasing one coefficient while decreasing adequately another one may lead to the same deflection evolution. In other words, the system has an infinite number of solutions. Calibration of code models is an interesting way of correction thanks to easy modifications, despite for the moment being purely empirical. At least, these coefficients should be justified scientifically by means of experiments and simulations. Furthermore, the uncertain values of some parameters like Young modulus, relaxation of steel, notional size or ambient relative humidity, have a high influence on the global behaviour of the bridge. Besides, it might also be worth investigating the implementation of new creep laws built upon consistency with measurements on different structures showing major troubles over long time periods.

4 Acknowledgements The authors wish to thanks Direction Interdépartementale des Routes de Méditerranée (DIR) for carrying out design justifications, drawings, national surveys measurements and maintenance reports. The authors wish also to thank François Toulemonde and Christian Cremona for their assistance and suggestions.

5 References [1] Patron-Solares A., Godart B. and Eymard R. (1996) Etude des déformations différées du pont de Savines (Hautes-Alpes), Long term deformations of the Savines Bridge (North Alps). Bulletin des laboratoires des Ponts et Chaussées, n°203, May-June 1996, pp 91-103. [2] Ollivier J.P., Vichot A. (2008) La durabilité des bétons: bases scientifiques pour la formulation de bétons durables dans leur environnement, Durability of concrete: basics for formulation of sustainable concretes in their environment, Presses de l'École Nationale des Ponts et Chaussées, in French, pp 167-216. [3] Acker P. (2001) Micromechanical analysis of creep and shrinkage mechanisms, Creep, Shrinkage, and Durability Mechanics of Concrete and Other Quasi-Brittle Materials, Proceedings of ConCreep-6 MIT, Elsevier, London, pp 15-25. [4] Sanahuja J, Dormieux L. (2010) Creep of a C-S-H gel: a micromechanical approach. Anais da Academia Brasileira de Ciências, n°82(1), pp 25-41. [5] Bažant Z. P., Prasannan, S. (1988) Solidification theory for aging creep, Cement and Concrete Research, n°18(6), pp 923-932. [6] DTecCITM (formerly Sétra), ST1 structural software: http://www.setra.fr/html/logicielsOA/logiciels_en.html, Structural Calculations and Softwares Division

272


[7] Acker P., Eymard, R. (1992) Fluage du béton: un modèle plus performant et plus simple à introduire dans les calculs numériques, Creep of concrete: A simplest and more efficient model for numerical calculations. In annales de l'institut technique du bâtiment et des travaux publics, n°507 (B295), in French. [8] NF EN1992-2:2006 Calcul des structures en béton, Partie 2: Ponts en béton - Calcul et dispositions constructives, Concrete structures, Part 2: Concrete bridges. [9] NF EN1992-1-1:2004 Calcul des structures en béton, Partie 1-1 : Règles générales et règles pour les bâtiments, Concrete structures, Part 1-1: Buildings. [10] Électricité De France (1960), Contrôle des bétons: Résultats des essais à la compression, site measurements: compressive strength of concrete, France, in French. [11] NF EN197-1:2001 Cement Part 1: composition, specifications, and conformity criteria for common cements. [12] Ministère de l'Equipement (1966) Agrément technique n° 82 et n°73-128: câble S.E.E.E. F13 et FU 4600: Agrément à divers procédés de précontrainte, Technical agreement n° 82 and n°73-128: S.E.E.E. F13 and FU 4600 tendons, http://www.piles.setra.developpement-durable.gouv.fr/documents-anciens-sur-lesouvrages-r331.html, Accessed 9 January 2014, in French. [13] IFSTTAR (2009) Guide technique: Mesure de la tension des armatures de précontrainte à l'aide de l'arbalète, Technical guide: measurements of tension in trends by means of cros-bow test, in French. [14] DTecITM (2014, to be published) Conception des réparations structurales et des renforcements d'ouvrages, Structural maintenances and retrofitting of bridges, in French. [15] CETE Méditerranée (1976) Reprise de la note de calcul, Design justification of additional post-tensioning tendons, Ministère de l’Équipement, Aix En Provence, in French. [16] Catinot L (1977) Rapport sur les résultats des mesures topographiques, Fieldworks surveying report, Électricité De France, in French. [17] Granger, L. (1995) Comportement différé du béton dans les enceintes de centrales nucléaires: analyse et modélisation, delayed deformation of concrete for nuclear power,(Doctoral dissertation, Ecole Nationale des Ponts et Chaussées), in French. [18] fib Bulletin 42 (2008) Constitutive modelling of high strength/high performance concrete, fédération internationale du béton (fib), Lausanne, Switzerland. [19] Bažant, Z. P., Hubler, M., & Yu, Q. (2011) Excessive creep deflections: An awakening. Concrete international, n°33(8), pp 44-46. [20] fib Bulletin 65 (2012) Model code 2010, final draft, fédération internationale du béton (fib), Lausanne, Switzerland. [21] Yu, Q., & Bažant, Z. P. (2013) Viscoplastic Constitutive Relation for Relaxation of Prestressing Steel at Varying Strain and Temperature. CONCREEP-9, MIT. [22] Krístek, V., & Bažant, Z. P. (1987) Shear lag effect and uncertainty in concrete box girder creep. Journal of Structural Engineering, n°113(3), pp 557-574. [23] Vítek, J., Křístek, V. (2000) Deflections and Strains of Prestressed Concrete Bridges, 16th congress of IABSE, Lucerne, 2000. [24] Bažant, Z. P., Yu, Q., & Li, G. H. (2012) Excessive Long-Time Deflections of Prestressed Box Girders. I: Record-Span Bridge in Palau and Other Paradigms. Journal of Structural Engineering, n°138(6), pp 676686.

273


Structural Longevity of FPSO Hulls Mark Tammer1*, Miroslaw Lech Kaminski2 (1) Institute for Engineering & Design, Faculty of Natural Sciences and Technology, HU University of Applied Sciences Utrecht, The Netherlands. (2) Ship Hydromechanics and Structures, Maritime and Transport Technology, Faculty of Mechanical, Maritime and Materials Engineering (3ME), Delft University of Technology (TU-Delft), The Netherlands. Abstract: The key challenge of managing Floating Production Storage and Offloading assets (FPSOs) for offshore hydrocarbon production lies in maximizing the economic value and productivity, while minimizing the Total Cost of Ownership and operational risk. This is a comprehensive task, considering the increasing demands of performance contracting, (down)time reduction, safety and sustainability while coping with high levels of phenomenological complexity and relatively low product maturity due to the limited amount of units deployed in varying operating conditions. Presently, design, construction and operational practices are largely influenced by high-cycle fatigue as a primary degradation parameter. Empirical (inspection) practices are deployed as the key instrument to identify and mitigate system anomalies and unanticipated defects, inherently a reactive measure. This paper describes a paradigm-shift from predominant singular methods into a more holistic and pro-active system approach to safeguard structural longevity. This is done through a short review of several synergetic Joint Industry Projects (JIP’s) from different angles of incidence on enhanced design and operations through coherent a-priori fatigue prediction and posteriori anomaly detection and -monitoring. Keywords: Floating Production Storage and Offloading assets (FPSOs); Structural Health Monitoring (SHM); Non-Destructive Evaluation/Testing (NDE/NDT); Risk Based Inspection (RBI); Condition Based Maintenance (CBM).

1 Introduction Firstly, this short paper will concisely outline current FPSO integrity management, after which the key paradigm of Structural Health Monitoring/Management is elucidated upon. Subsequently, the second chapter will briefly discuss the methodological constitutes, performed research and some key goals and outcomes of the current JIP’s. The focus of the third chapter will be on the distinct opportunities these collaborative parallel projects for safeguarding FPSO structural longevity offer in terms of synergetic effects and mutual strengthening to further operationalize Risk- and Probabilistic based approaches to design, construction and operations, including Inspection Repair and Maintenance (IMR) practices. Finally, this short paper will conclude with recommendations for future research.

1.1

FPSO Integrity Management

At the present time, the outcome of the periodical and event-driven asset inspections provide input for the determination of the components’ (compiled) Probability of Failure, which is combined with the Consequences of Failure to provide a risk profile and inspection scheme to prevent incidents, maintain a specific safety level and to enhance design and operational practices through feedback. In line with the aforementioned, in essence current Asset Integrity Management (AIM) models still consist of the a-priori determination of technical and organizational measures to ensure future economic system effectiveness and safety. Measure optimization is generally done by posteriori analysis on correlation and causality of usage, external influences and costs to improve the knowledge on physical system degradation, predict the future behaviour and further refine the measures accordingly.

*

[email protected]

274


Logically, the challenges stated above are further aggravated by the inspection optimum paradigm; the intersection between the economic principle of reasonableness and the fact that unnecessary, disruptive and costly inspection and maintenance could result in unintended and expensive downtime, subsequent damage and inherent Safety, Health and Environmental (SHE) risks. Hence, best-practices should be deployed to approximate the optimum of efforts to limit these risks and safeguard structural longevity. Nowadays, a multitude of research efforts and attention are directed towards more integrated forms of inspection management and (conditional) Risk Based Inspection (RBI) in particular [1-13]. The necessity of this paradigm-shift has been highlighted very appropriately in the paper of Fragola and Bedford [14] as reliability practices shift from the dominant common, singular failures to the dependable from both un- and anticipated interactions between (sub)systems and the internal and external environment. In addition, the dominant problematic details (hot-spots) are progressively conversed with measures due to recently gained experiences. This requires a focus-shift to assure structural longevity.

1.2

Structural Health Monitoring

Although asset complexity is further intensified by in-situ (embedded) systems, technological breakthroughs in Sensing- and Information Technology also pose a significant advantage in monitoring through analysing and discriminating the a-priori and posteriori structural- and functional health of assets. The acquired data can pro-actively control predictive models to determine the current state and optimal moment for restoring structural and functional integrity, which is respectively referred to as Structural Health Monitoring (SHM) and the arising IMR-actions as Condition Based Maintenance (CBM). The advantages of predictive, oncondition Asset Integrity Management are vast; the model of drivers for predictive IMR as constructed by Adams [15], graphically represented in figure 1, shows these perceived benefits [1]. In concreto, less downtime, minimal intrusion of the (sub)systems, the facilitation of a planned supply of maintenance resources and replacement before the actual failure, preventing subsequent damage [16], enhanced understanding of the design, modification of systems and equipment reliability [17] and less inspections and overall safer operations; lowering both the Capital Expenditures (CAPEX) and Operational Expenditures (OPEX).

Figure 1 - Potential Impact of SHM (Adams, 2007)

Notwithstanding, as outlined in the review 'Loads for use in the Design of Ships and Offshore Structures’ by Hidaris et al. [18] research efforts into the structural loading and longevity of ship and offshore structures are fragmented on the computation of wave induced loads (47%), specialist ship structure topics such as cargo sloshing, bow slamming, green water etc. (32%) and 7% on fatigue loading of ships and 10% for specialist offshore structures. Finally, a limited amount of 5% is directed on wave load uncertainty modelling and validation (of which 45% relates to fatigue load calculations). It is evident that with SHM-systems and the empirical data obtained, all research efforts as stated above can benefit tremendously as well as operations to determine IMR-regimes from the basis of economical and SHE-performance. Such a combined bottom-up (operational research and optimization) and top-down approach (fundamental scientific research) provides for maximum knowledge valorisation: the foundation on which the most successful JIP’s are built.

275


2 Methods and Joint Industry Projects The Ship and Offshore Structures’ section (S&OS) of Delft University of Technology manages and participates in several JIP’s from different angles of incidence, but all contribute to the common goal of linking research and education in the field of structural longevity of ship and offshore structures [19]. This is achieved through enhanced design (4D Fatigue), a-priori fatigue prediction (Monitas), posteriori anomaly detection and mitigation (HITS) and defect monitoring (GrackGuard). The dominant consideration comprehends the vision that phenomenological complexity requires a holistic view and an appositely constructed research portfolio.

2.1

4D Fatigue

To this very moment, the fundamental references in the design of ship and offshore structures and the inherent fatigue resistance are directed from uni-axial and constant-amplitude testing [20-22]. Nonetheless, during real-life conditions, structures are subjected to multi-axial, variable-amplitude loading including non-proportional characteristics for specific details. Unfortunately, the usability of current multi-axial practices are restrained due to limited validation efforts and finite academic scope in testing, which can be reverted to the general engineering perception that uni-axial loading is the predominant factor. Recent research has shown that conventional uni-axial methods significantly overestimate the fatigue lifetime, and lifetime predictions of multi-axial methods show significant differences [23]. The 4D-Fatigue JIP focuses on improving the current practices and estimations by performing both numerical and experimental research. This is achieved through a testing campaign on key details, which suffer from local multi-axial and non-proportional loading. The goal is to define a multi-axial criterion and design tool to assess the knowledge gap and provide for better estimations in fatigue lifetime of welded joints in ship- and offshore structures and constitute a precursor to modification of the prevailing rules and regulations. In short, realistic test-specimens will be FEM-analysed including systematic variation of control variables and limited geometrical variables. The focus will be on variation of ratios and phases of different load components. The load spectra will be defined from analysis of in-service measurements [24].

2.2

Monitas

The Monitas system is developed as an Advisory Hull Monitoring System (AHMS), which takes advantage of both emerging (Fatigue Damage Sensors) and conventional techniques (such as Environment Data Acquisition Systems and Cargo Loading Monitoring) and comparable methodologies from a multi-domain focus, to successfully facilitate the incorporation of sensing abilities. It combines and implements the methodologies as a more generic integrated monitoring entity to model the fatigue lifetime consumption of ship- and offshore structures based on comparison between the design and the actual fatigue lifetimes calculated by the fatigue design tool. The actual lifetime is based on measured data, which includes operational settings, environmental conditions, and hydro-structural response. Hence, the Monitas system presents, explains, and provides advice on the fatigue lifetime consumption of FPSO’s hulls [25-27]. Ultimately, apriory anomaly prediction allows for the reduction of operational costs and mitigation of both SHE-related and economical risks as deviations in fatigue lifetime consumptions are identified, anticipated and (re)acted upon through IMR-practices. Logically, the outcome will also trigger future design tool and -practise improvements. This will be further enhanced by a secondary project, as the system now predominantly focuses on current and historical environmental conditions, the impact of climate change - hence the discrepancy in design and future on-site sea-states - remains largely unexplored up to this moment. The university has commenced with future scenario-analysis to develop a methodology to evaluate the effect of climate change on sea-states to further enhance fatigue lifetime consumption estimations and design tools [28].

276


2.3

HITS

In addition to the aforementioned methods for assessing and predicting fatigue loading, the outlined knowledge gaps need to be accounted for as well as flaws in the design, construction and operational execution. Therefore, rules and regulations from authorities and stakeholders prescribe hull inspections and surveys to ensure the as-designed state and mitigate anomalies and risks through addon measures. These anomalies are often the result of differences between the design and on-site operating context, higher than anticipated residual stresses, fabrication issues (such as misalignments, inadequate welding etc.) and system effects. The latter consists of the inter-dependency of parts, failure modes and mutual dependencies [1]. Hence, one must anticipate for non-ideal design, construction and operations. Consequently, inspections still prevail as a necessity. The Hull Inspection Techniques and Strategies JIP (HITS) is initiated to provide for unambiguous industry guidelines on inspection procedures and -techniques. This is achieved by reviewing the multitude of existing guidelines from class societies, recommended practices and regulatory requirements to extract best-practices, identify and assess differences to derive robust inspection criteria, techniques and procedures for (e.g. regulatory, class society, company) criteria compliance. When more uniform practices become the industry standard, (censored) asset- and anomaly data can be shared, compared, interpreted and benchmarked more easily, providing empirical grounds for directed research efforts and enhanced design. Examples consist of the Bayesian updating of inspection findings and -schedules and the determination of the fatigue reliability of (un)inspectable joints c.q. details using structurally correlated inspection data. Both a valuable contribution, as disregard of structural findings and -correlation results in misunderstanding system reliability and inefficient use of beneficial information. Ergo, neglect of correlation on component level misjudges the reliability of (un)inspected components if system inspection information is available [29]. Hence, often, useful operational data is disregarded in operational decision-making and research.

2.4

CrackGuard

As outlined in the aforementioned JIP’s, when coping with fatigue as a primary degradation mechanism, engineers must avoid fatigue cracks by thorough design and fabrication processes, and operators c.q. authorized bodies must periodically inspect structures for the presence of cracks. Cracks exceeding thresholds in terms of size, location and/or propagation rate and risk are repaired or mitigated, e.g. through the application of additional strengthening and/or stopper holes. Cracks of an acceptable length or judged as non-effective repairs (e.g. non-critical design error) must be followedup during successive inspections. However, the complex nature of the phenomena makes it very difficult to estimate (near) future behaviour, which can be very capricious and - as stated - increasing inspection efforts and frequency pose both operational costs and risks. The CrackGuard JIP hinges on the principle of Quantitative Non-Destructive Evaluation (QNDE) for non-destructive/disruptive in-service inspection and monitoring. Current practices, as deployed in the HITS-JIP are basically limited to visual detection. Typically, after detection anomalies are assessed with strain monitoring, ultrasonic, magnetic and/or radiographic testing [30]. This project consist of precompetitive research and development of an affordable system for monitoring detected and allowable fatigue cracks based on the most recent achievements in crack propagation, sensing technology and wireless communication in order to reduce the extend and scope of successive inspections, while providing valuable information on the capricious nature of crack propagation [31]. Ergo, the final link in the chain of JIP’s - from design, conception and operation - for safeguarding the structural longevity of FPSO hulls. Note that this short paper delineates the research portfolio and JIP interdependency from a bird eyes view. Please refer to references for more detailed information on the performed research and results.

277


3 Synergetic effects Although the described JIP’s greatly differ in approach, from both a strategic and academic top-down vision, as well as an operational bottom-up mode and from procedures and regulations to fundamental research; the absolute strength lies in the synergetic effects and mutual strengthening of the holistic approach on mastering fatigue degradation. The interdependency greatly invigorates the efforts: Monitas and 4D Fatigue provide data for CrackGuard and HITS (and vice-versa through feedback), the latter steers the operational usage and requirements of AHMS-applications and design practices, tools and validation. Combined, (participation in) such a portfolio provides a much more solid basis for the implementation of Risk-Based Approaches as the combination of beneficial properties of the different methodologies is gained to provide data, to discriminate information and eliminate shortcomings from both perspectives. Hence, high levels of academic and professional participation between projects and direct valorisation into both operational and research practices.

3.1

Systems approach

The key methodological difficulty for safeguarding the structural longevity of FPSO hulls still lies in the collection of accurate data and the determination of the total accumulated fatigue damage for specific locations and (sub)systems [1]. The research efforts in the JIP’s have demonstrated that with monitoring systems the inspection schedule can be optimized in such a way that the annual reliability index of a structural detail will not drop below its allowable threshold value [1, 13, 29] while improving the performance of IMR-practices. The key overarching element consists of calibrating the probabilistic Fracture Mechanics model to the S-N approach [32] to keep the reliability model consistent with the conventional design method, to comply with rules and regulations and to provide for enhanced information on what, where and how to monitor and inspect. The methodology as proposed by the team consists of modifying two Fracture Mechanics parameters (primarily the geometrical faction, and secondarily the initial crack size) in such a way that the differences between the obtained reliability from both approaches are minimized [13]. This is an indispensable process for assuring the correct application, due to the inherent sensitivity of the reliability model. The combination of research and operational efforts and results from linking design, a-priori fatigue prediction and lifetime estimations, posteriori anomaly detection through empirical inspections and -monitoring with real-time monitoring of metocean and loading data closes the feedback-loop for continuous improvement of safeguarding structural longevity.

3.2

Recommendations

After this birds eye review of the conducted research and -portfolio of the S&OS section, this short paper concludes with recommendations for both future challenges, as well as the operationalization hereof: I. Research from the JIP-portfolio has indicated that the uncertainty (Standard Deviation, c.q. Coefficient of Variation) and hence the credibility of the RBI-methodology can be greatly improved by parameter-tuning of the long term stress range distribution. Additional research should focus on enhancing the estimations of these parameters; II. The assessment of the structural longevity of FPSO hulls should focus more on the combination of adverse effects due to system effects and compilations of failure modes, such as the incorporation of the effect of corrosion. A combination with the first recommendation and additional focus on both structural correlation and Bayesian updating of the findings are likely to further enhance longevity predictions; III. The combination of several JIP’s from very different angles of incidence, but on one overarching theme, has proven incredibly use- and powerful. To conclude, herewith a plea for strategic portfolio-management and research group composition with a clear spin-off and valorisation in both the industry as in the academic world.

278


Acknowledgements The authors would like to thank all JIP-participants, which primarily consist of offshore operators, oil and gas companies, shipyards, class societies and governmental institutions.

References [1] Tammer, M.D., and Kaminski, M.L. (2013). Fatigue Oriented Risk Based Inspection and Structural Health Monitoring of FPSOs. Proceedings of the 23rd International Offshore and Polar Engineering Conference (ISOPE) pp. 438-449. June 30–July 5, 2013. Anchorage, Alaska, USA. [2] American Bureau of Shipping (2003). ABS. Guide for surveys using Risk-Based Inspections for the offshore Industry. Houston, Texas, USA. [3] American Petroleum Institute (2002). API580:2002-09. Risk Based Inspection [API Recommended Practice]. First edition. Washington DC, USA. [4] Det Norske Veritas (2010). DNV-RP-G101:2010. Risk Based Inspection of Offshore Topsides Static Mechanical Equipment. Oslo, Norway: DNV. [5] Det Norske Veritas (2010). DNV-OSS-300-3:2010. Risk Based Verification of Offshore Structures. Oslo, Norway: DNV. [6] European Committee for Standardization (2008). CEN-CWA-15740:2008, Risk-Based Inspection and Maintenance Procedures for European Industry (RIMAP). Brussels, Belgium: CEN. [7] Goyet, J., Rouhan, A., L’Haridon, E. and Gomes L. (2011). Probabilistic System Approach for Risk Based Inspection of FPSOs. In: OTC 22684 Proceedings of the Offshore Technology Conference. 4–6 October, Rio de Janeiro, Brazil. [8] Health And Safety Executive (2001). HSE-363. Best practice for Risk Based Inspection as a part of plant integrity management. Norwich, UK: HSE. [9] Lee, A.K., Serratella, C., Wang, G. and Basu, R. (2006). Flexible approaches to Risk-Based Inspection of FPSOs In: OTC18364 - Proceedings of the 2006 Offshore Technology Conference. 1 - 4 May 2006. Houston, Texas, USA. [10] Norsok (2011). Z-008:2011 Risk Based Maintenance and Consequence Classification (third edition). Lysaker, Norway: Standards Norway. [11] Ship Structure Committee (2002). SSC-421:2002. Risk-Informed Inspection of Marine Vessels (chairman Radm Pluta, P.J.) Washington D.C., USA: SSC. [12] Straub, D. (2004). Generic Approaches to Risk Based Inspection Planning for Steel Structures. [online PhD dissertation]. Institute of Structural Engineering, Swiss Federal Institute of Technology, ETH Zürich. Available from: http://e-collection.library.ethz.ch/eserv/eth:1550/eth-1550-01.pdf [13] Tammer, M.D., Kaminski, M.L., Koopmans, M. and Tang, J.J. (2014). Current Performance and Future Practices in FPSO Hull Condition Assessments. ISOPE [IN PRESS]. [14] Fragola, J.R. and Bedford, T. (2005). Identifying emerging failure phenomena in complex systems through engineering data mapping. Reliability Engineering and System Safety [90] pp. 247–260. [15] Adams, D.E. (2007). Health Monitoring of Structural Materials and Components. Chichester, West Sussex, UK: John Wiley & Sons Ltd. ISBN: 978-0-470-3313-5. [16] Houtum, van, G.J.J.A.N. (2010). Maintenance of Capital Goods [Inaugural lecture]. Department of Industrial Engineering & Innovation Sciences, Eindhoven University of Technology, The Netherlands.

279


[17] Williams, J.H. (1994). Condition Based Maintenance and Machine Diagnostics. London, UK: Chapman & Hall. ISBN: 0-4124-6500-0. [18] Hidaris et el. (2013). Loads for use in the design of ships and offshore structures. Ocean Engineering. [IN PRESS] [19] Kaminski, M.L. (2011), Ingenious Ship and Offshore Structures [Inaugural Lecture]. Delft University of Technology (TU-Delft). Oct. 5, 2011. Delft, The Netherlands. [20] British Standards Institution (2005). BS7910:2005. Guide to methods for assessing the acceptability of flaws in metallic structures. London, UK: British Standards Institution. [21] Bai, Y., 2003. Marine Structural Design. Oxford, UK: Elsevier Science Ltd. [22] Horn et al. (2009). Report of Committee III.2 – Fatigue and Fracture. In: OTC17535 - Proceedings of the 17th International Ship and Offshore Structures Congress (ISSC). 16-21 August 2009. Seoul, Korea. [23] Horn et al. (2012). Report of Committee III.2 – Fatigue and Fracture. Proceedings of the 18th International Ship and Offshore Congress (ISSC), Volume 1, edited by W. Fricke and R. Bronsart, September 9-13, 2012, Rostock, Germany. [24] Besten, Den, J.H., Kaminski, M.L. and Huijsmans, R.H.M. (2013), Stress Intensity Factor Analysis Using Digital Image Correlation: A Post-Processing Approach Displacement Field Measurement for Crack Growth Parameters in an Aluminium MIG-welded T-joint. MARSTRUCT 2013, 4th International Conference on Marine Structures, March 25–27, 2013, Espoo, Finland. [25] Kaminski, M.L. and Aalberts, P. (2010), Implementation of the Monitas System for FPSO Units. Offshore Technology Conference, May 3–6, 2010, OTC-20871. Houston, Texas, USA. [26] L'Hostis, D., Kaminski, M.L. and Aalberts, P. (2010), Overview of the Monitas JIP. Offshore Technology Conference, May 3–6, 2010, OTC-20872. Houston, Texas, USA. [27] L'Hostis, D., Cammen, Van der, J., Hageman, R. and Aalberts, P. (2013), Overview of the Monitas II Project. . Proceedings of the 23rd International Offshore and Polar Engineering Conference (ISOPE) pp. 455-462. June 30–July 5, 2013. Anchorage, Alaska, USA. [28] Zou, T. and Kaminski, M.L. 2013. Possible Solutions for Climate Change Impact on Fatigue Assessment of Floating Structures . Proceedings of the 23rd International Offshore and Polar Engineering Conference (ISOPE) pp. 455-462. June 30–July 5, 2013. Anchorage, Alaska, USA. [29] Berg, van den, D., Tammer, M.D. and Kaminski, M.L. (2014). Updating Fatigue Reliability of Uninspectable Joints using Structurally Correlated Inspection Data. ISOPE [IN PRESS]. [30] Horst, van der, M.P., Kaminski, M.L. and Puik, E. (2013). Methods for Sensing and Monitoring Fatigue Cracks and Their Applicability for Marine Structures. Proceedings of the 23rd International Offshore and Polar Engineering Conference (ISOPE) pp. 455-462. June 30–July 5, 2013. Anchorage, Alaska, USA. [31] Horst, van der, M.P., Kaminski, M.L., Puik, E. and Lepelaars, E. (2014). Testing and Numerical Simulation of Magnetic Fields Affected by Presence of Fatigue Cracks. ISOPE [IN REVIEW]. [32] Paris, P. C., & Erdogan, F. (1963). A Critical Analysis of crack Propagation Laws. Journal of Basic Engineering 85(4) pp. 528-533.

280


On the very long term delayed behaviour of concrete J.M. Torrenti1, F. Benboudjema2, F. Barré3, E. Gallitre4 (1) Université Paris-Est, IFSTTAR, France- [email protected] (2) ENS Cachan, France (3) Géodynamique et structure, France (4) EDF Septen, France Abstract: Several nuclear power plant containment buildings are made with biaxially prestressed concrete. In order to maintain enough prestressing in the structure, the delayed behaviour of concrete on the very long term should be correctly assessed. In this paper two questions are considered: is the very long term creep of concrete asymptotic or not? And how to deal with biaxial creep? The observation of the very long term deformations of containments, the deflections of concrete bridges and laboratory results of creep tests are used to give answers to these questions. Keywords: Nuclear power plant, Prestressing, Creep, Concrete, Ageing

1 Introduction

The typical French reactor building consists of two concentric containments (fig. 1). The outer containment, designed to sustain external aggressions, is made with reinforced concrete. The inner containment is biaxially prestressed, from 80 to 120 cm thick, and designed to withstand an internal pressure of 0.5 MPa in case of an accident. In order to avoid tensile stresses in concrete, a prestress is applied, corresponding to compressive stresses in concrete of 8.5 MPa and 12 MPa along vertical (zz) and orthoradial (θθ), respectively. In the previous generation of French nuclear power plant (NPP), there was no metallic liner inside the internal containment. So the tightness of the containment in case of an accident is only assured by concrete. To limit leakage, cracking should be avoided and, so, tensile stresses should remain below the tensile strength of concrete. That is why the evolution of prestressing forces during time is critical for the NPP operation and why a good prediction of the evolution of delayed strains of the containment is needed.

Figure 1 Simplified diagram of the containment and its prestressing

281


To predict the evolution of prestressing forces with time, the delayed behaviour of the constitutive materials (relaxation for steel and creep for concrete) has to be modelled precisely on the long term. In this paper only creep of concrete is considered and two problems are discussed: is the very long term creep asymptotic or not and how to deal with biaxial stresses?

2 Is the very long term creep asymptotic? 2.1

Laboratory tests

Classical creep tests in laboratory last generally less 2 or 3 years (and sometimes less). With such duration it is possible using these tests to fit as well an asymptotic or a logarithmic evolution of strains. There is nevertheless one set of creep tests performed by Brooks [1] with a duration of 30 years. These tests performed on concretes with different water to cement ratio and with different aggregates types, show that creep is not asymptotic (figure 2). 80

specific creep (µm/m/MPa)

70 60 50 40 30 20 10 0 1

10

100

1000

10000

100000

time (log(days)) Figure 2 Creep of concrete – example of the results of Brook’s tests [1]

2.2

Long term deformations of prestressed structures.

The observation of the deflections of several bridges built of prestressed box girders indicates rather clearly that long term deflections are not asymptotic [2, 3, 4]. In the case of NPP, thanks to a large monitoring of the behaviour of the containment, global and local deformations are measured. Figure 3 presents the evolution of the deformations of the central part of the containment of a French NPP. Again, the long term behaviour is not asymptotic and seems in accordance with a logarithmic evolution with time. Note that delayed strains include here drying shrinkage and relaxation of the prestressing tendons.

282


Figure 3 Example of the evolution of the deformations in the central part of a containment of a NPP (V=vertical strains, T=tangential strains)

2.3

Consequences

With the measurements of the deformations of a sample during a creep test, it is possible to adjust an asymptotic hyperbolic function (like in EC2-1 - equation 1) or a combination of an asymptotic and a logarithmic functions (like in MC2010 – equation 2) – see figure 4. In the latter case the logarithmic function corresponds to basic creep while drying creep is asymptotic. 𝜀= 𝜀=

𝑐 𝜎 𝑡−𝑡0 𝑎 𝑓1 (ℎ) � (ℎ)+(𝑡−𝑡0)� 𝐸𝑐 𝑏 𝑔1 𝑐 𝜎 𝑡−𝑡0 𝑓2 (ℎ) � � �𝑎 𝐸𝑐 𝑏 𝑔2 +(𝑡−𝑡0)

+ 𝑑 ln(1 + 𝑒(𝑡 − 𝑡0))�

[eq. 1]

[eq.2]

where a, b, c, d and e are fitted constants. f 1 , f 2 , g 1 and g 2 are functions of the notional size h given by EC2-1 and MC2010. The notional size is here equal to 80mm.

The creep models being optimized on a sample, it is possible to predict the creep deformations of a larger structure. In our case a NPP has a notional size h=800mm loaded under a constant uniaxial stress of 12.5MPa and a RH of 50%. Figure 5 shows that in this case at very long term large differences occur despite a good calibration of the model on the same test. This is of course due to the differences in the nature of the equations (asymptotic or not) and it shows the importance of a physical manner to model the very long term behaviour of concrete structures. A bad prediction during design will underestimate the loss of prestressing and could affect the service life of the NPP. Note also that, in this case the contribution of basic creep is the most important in the long term despite the large notional size (figure 6).

283


1800,00 1500,00

creep strain 10-6

1200,00 exp. results

900,00

EC2-1 600,00

MC2010

300,00 0,00 0,0

0,0

0,0

0,1 1,0 Time (days)

10,0

100,0

1000,0

Figure 4 Comparison of experimental results [7] and models corresponding to equations 1 and 2. The applied stress is equal to 12.5 MPa and the relative humidity to 50%.

creep strain (10-6)

1600 1200 800 EC2-1

400

MC2010 0

0,01

0,1

1 time (years)

10

Figure 5 Comparison of the estimated creep strains using equations 1 and 2.

284

100


creep strain (10-6)

1400 1200 1000 800

basic creep

600

drying creep

400 200 0

0,00001

0,001

0,1 time (years)

10

Figure 6 Contribution of basic and drying creep in the deformation predicted by MC2010.

3 Biaxial creep 3.1

Laboratory tests

Few experimental results are available in the literature. By extending isotropic elasticity theory to the visco-elasticity, it is possible to define a secant creep Poisson ratio ν c : [eq.3] 𝛆(𝑡) = 𝐽(𝑡)[(1 + 𝜈𝑐 )𝛔 − 𝜈𝑐 tr(𝛔)𝟏] where ε is the creep strain tensor, σ the stress tensor and J is the uniaxial specific creep. If it is assumed that the material remains isotropic, creep Poisson ratio can be calculated using either equation 4 or 5 (after developing eq. 3): 𝜀11 (𝑡)⁄𝜀22 (𝑡) = [(1 + 𝜈𝑐 )𝜎11 − 𝜈𝑐 tr(𝛔)]⁄[(1 + 𝜈𝑐 )𝜎22 − 𝜈𝑐 tr(𝛔)] [eq.4] 𝜀11 (𝑡)⁄𝜀33 (𝑡) = [(1 + 𝜈𝑐 )𝜎11 − 𝜈𝑐 tr(𝛔)]⁄[(1 + 𝜈𝑐 )𝜎33 − 𝜈𝑐 tr(𝛔)] [eq.5] An analysis of several experimental results leads to the fact, in most of 1D (with the measure of strain in the non-loaded direction), 2D and 3D creep tests, the creep Poisson ratio is closed to 0.15. However, in some tests unphysical values (less than -1 or greater than 0.5) has been found [8], which may be due to uncertainty of sensors, tightness flaws in basic creep tests and friction between the concrete specimen and the loading frame. The analysis of available biaxial basic creep test is reported for biaxial creep tests in Figure 7. For experiments where strains have been measured in the 3 principal directions [9, 10, 11], the use of eq. 4 and 5 has leaded to similar results. For biaxial creep tests for which compressive stresses are different, it has been found very close results using either eq. 4 or eq. 5. These results show a significant discrepancy between results. After 100 days, the basic creep Poisson ratio is comprised between about 0.14 and 0.35. Even after about 4 years, its value seems to be continuously evolving. It is difficult to conclude at this time.

285


Mean creep Poisson ratio

0,5 Galenne (2013) Jordaan (1969) - test1 Gopalakrishnan (1969) Bergues (1972) Jordaan (1969) - test 2 Kennedy (1975)

0,4

0,3

0,2

0,1 1

100 Time [days] Figure 7 Experimental evolution of basic creep Poisson ratio.

3.2

10

1000

10000

Long term deformation of a NPP

In Figure 3, the evolution of the biaxial deformations during times in a NPP was presented. From these deformations one can express the evolutions of the sum and the difference of the vertical and tangential strains. These evolutions are presented figure 8 for three different NPPs [6]. It can be seen that the delayed strains in the containment wall follow two distinct kinetics: - the difference between tangential and vertical strains, which exhibits a fast kinetics and becomes almost constant after 500 days, - the average of tangential and vertical strains, which corresponds to a slow kinetics. If vertical and tangential strains are equal despite different values for the prestressing, this means that in the long term biaxial creep deformations are only sensible to the mean stress and independent of deviatoric ones. If biaxial creep is expressed using a tangent Poisson ratio, this is equivalent to a Poisson ratio equal to -1. If we assume that the kinetics proposed by MC2010 are correct, the long term slow kinetics corresponds to basic creep (figure 6). So there is a contradiction between the basic creep Poisson ratio obtained with experimental results and the one obtained from the global behaviour of the NPP. 1600 1400

strain (10-6)

1200

(V+T)/2 NPP1

1000

T-V NPP1

800

(V+T)/2 NPP2

600

T-V NPP2

400

(V+T)/2 NPP3

200

T-V NPP3

0 500

5000 time (days)

286

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands Figure 8 evolutions of the tangential (T) and vertical (V) deformations in the central part of the containment of three NPPs

4 Conclusion The observation of the global behaviour of containments of nuclear power plants confirms laboratory results and the observation of the long term behaviour of bridges: creep of concrete does not seem to have an asymptotic behaviour but an evolution with the logarithm of time. And it is very important to take into account this phenomenon in construction codes for the prediction of the very long term behaviour of structures. Concerning biaxial prestressed structures like the containment of a NPP, the global behaviour does not seem in accordance with laboratory tests. The analysis of these structures should be continued in order to understand the physical sources of this discrepancy.

5 References

[1] J. J. Brooks (2005) 30-Year creep and shrinkage of concrete.” Mag. Concr. Res., 57(9), 545–556. [2] Bažant, Z. P., Hubler, M., & Yu, Q. (2011) Excessive creep deflections: An awakening. Concrete international, n°33(8), pp 44-46. [3] Vítek, J., Křístek, V. (2000) Deflections and Strains of Prestressed Concrete Bridges, 16th congress of IABSE, Lucerne, 2000. [4] J.P. Sellin, J.B. Barthelemy, J.M. Torrenti, G. Bondonnet (2014) Delayed deformations of segmental prestressed concrete bridges: the case of the Savines Bridge, AMS conference, Delft. [5] Muller H., Anders I., Breiner R., Vogel M. (2013) Concrete: treatment of types and properties in MC 2010, Structural concrete, Volume 14, Issue 4, pages 320–334. [6] Chauvel D., Touret J.P., Barré F. (2006), Assessment of long term concrete deformations of nuclear structures based on in situ measurements, Advances on Geomaterials and Structures – Hammamet. [7] Granger L. (1995) Comportement différé du béton dans les enceintes de centrales nucléaires : analyse et modélisation, thèse de l’Ecole nationale des ponts et chaussées. [8] Benboudjema F. (2002) Modélisation des déformations différées du béton sous sollicitations biaxiales. Application aux enceintes de confinement de bâtiments réacteurs des centrales nucléaires, PhD Thesis, Université de Marne-La-Vallée. [9] Galenne E., Foucault A., Hamon F. (2013) Prediction of Delayed Strain of Nuclear Containment Building: from laboratory tests to an industrial mock-up, TINCE 2013, Paris. [10] Gopalakrishnan K. S., Neville A. M., Ghali A. (1969) Creep Poisson’s Ratio of Concrete Under Multiaxial Compression, ACI Journal, n° 66-90, p. 1008-1020 [11] Jordaan I.J., Illston J.M. (1969) The Creep of Sealed Concrete under Multiaxial Compressive Stresses, Magazine of Concrete Research, 21 (69), p. 195-204

287


Modeling Aging of Cementitious Pore Structure Neven Ukrainczyk1, 2*, Eduardus A.B. Koenders2 (1) Delft University of Technology, Delft, The Netherlands (2) University of Zagreb, Zagreb, Croatia Abstract: Coupled reactive-transport processes in cementitious materials play a crucial role in the aging process of building materials. Up to now, the effect of aging on microstructure and evolution of it’s properties, in three dimensions, was studied only by dissolution of certain phases at random locations. In this paper we present a new 3D reactive-transport model and results of leaching-induced aging simulations performed on virtual cementitious microstructures generated by Hymostruc model. The outputs are the morphology of the aged 3D microstructure together with aged properties characterised by fuzzy state of each voxel at different times. This allows to simulate the evolution of properties as a function of time as well as a function of the location within the microstructure.

Keywords: 3D modelling, reactive-transport, porous media, leaching.

1 Introduction

The knowledge of the relationship between microstructure (i.e. pore structure) and transport properties, e.g. effective diffusion coefficient, D ef of cement based materials is a crucial point to run properly the models coupling transport and chemistry. During transport mechanism a local microstructure of the cement based material is changing due to the dissolution effect and/or interaction of the diffusing ions with the solid phase. Thus, the coupled transport and chemical equilibrium effects result in a space and time dependency of the D ef . Both these dependencies {D ef = f(t, x)} present an important backbone of the coupled modeling concept and show significant impact on the final results of the predictions. Modeling of calcium leaching of cementitious materials is challenging [1-4] due to the multiscale porous and multi-phase nature of the cement matrix. The mechanism of calcium loss and the equilibrium calcium concentration (solubility) are different for each of these phases, depending mainly on their calcium to silicon ratio [4]. The most common approaches to overcome these difficulties consist on either limiting the number of phases considered or using a continuous equilibrium formulation based on Berner’s equation [5,6] for solid calcium concentration as a function of its equilibrium solution concentration. Rapid development of numerical models has provided novel methods to investigate the influence of microstructure on the evolution of the properties of cement based materials. A numerical scheme that reflects a representative porous network can be used to analyze the effective transport properties by means of either generating a porous network via advanced 3D numerical simulations (e.g. CEMHYD3D, μic or Hymostruc [7]); or by 3D sampling of a porous network using modern experimental imaging techniques such as X-ray computed tomography. A virtual 3D microstructure created with an available hydration model provides a basis for the analysis of the morphological influence onto the effective diffusion coefficient. Such an approach contributes to a better understanding of the phenomenology and thus improves the predicting reliability of the coupled models. The effect of aging on microstructure and evolution of it’s properties, in three dimensions, was studied only by dissolution of certain phases at random locations. In this paper we present a novel 3D reactive-transport model and results of leaching-induced aging simulations performed on virtual cementitious microstructures generated by Hymostruc model. This allows to simulate the evolution of properties as a function of time as well as a function of the location within the microstructure. *

[email protected]; [email protected], University of Zagreb, Zagreb, Croatia

288


1.1

1D homogenised model

The calcium leaching in cement-based materials is a coupled chemical equilibrium/diffusion phenomenon [2]. Kinetics of this transport-reaction processes are described by a simplified continuum (homogenized) model eq. 1 where the actual evolution of the pore morphology, chemical activity, convection, electrical coupling and precipitation effects are neglected. However, as an initial approach it makes a good prediction of the degraded depth and cumulative amount of leached calcium. The homogenized approach requires empirical relationships linking the diffusion coefficient to porosity.

P ( x, t= ) ∂u ∂t Deff ( x, t ) ∂ 2u ( x, t ) / ∂x 2 − ∂u solid ( x, t ) / ∂t

(1)

where u(x, t) is the Ca2+ concentration in the pore solution, usolid(x, t) is the concentration of Ca in solid phase, P(x, t) is the porosity and D eff (x, t) is the effective diffusion coefficient of Ca2+ ions (790 µm2 s-1 [9]). The first term on the right-hand side of eq. 1 stands for the diffusion process of the calcium in the liquid phase, which is assumed to be governed by Fick’s law. The second term accounts for the dissolution process, which leads to a source of calcium in the liquid phase. The calcium concentration in solid phase usolid(x,t) is calculated from its relationship with calcium concentration in solution. The existence of a solid phase assemblage with clearly defined dissolution fronts [1,2] is explained by instantaneous dissolution, i.e. establishment of the local liquid equilibrium concentrations. The kinetics of the dissolution reaction is governed by a diffusion process, because the diffusion rates are much slower than those of the chemical reactions. The evolution of Ca/Si ratio as a function of calcium concentration in solution corresponds to the degradation fronts from the non-degradated to the external zone of a cement material. The dissolution of Portlandite occurs suddenly for a threshold calcium concentration of 21 mol m-3 and explains the rapid drop at this value. The decalcification of CSH is gradual due to the various forms of CSH having a Ca/Si ratio ranging from 1.65 to 1. This explains the decreasing calcium concentration in pore solution between 21 and 2 mol m-3. A dissolution of ettringite and monosulphoaluminate also occurs in that zone. For calcium concentrations under 2 mol m-3, the solid phase corresponds to a silica gel.

2 3D implementation 2.1

3D virtual microstructure generation

The 3D virtual cementitious microstructures generated by Hymostruc can be simulated as a function of the random positioning of the cement particles inside a predefined REV, the initial particle size distribution (PSD), the degree of hydration, the chemical composition of cement, the morphological development of the hydration products, the water to cement ratio, and the temperature of the reaction process. The initial state of the microstructures is determined by stacking cement particles that follow a predefined particle size distribution (Figure 1a). For this, periodic boundary conditions were applied to minimize size effects and to comply with the volume balance induced by the water to cement ratio. Particles are stacked based on random selection of locations with equal probability of occurrence. Placing of the particles starts with the largest particles followed by smaller once, while obeying the particle size distribution. This process continues until all particles in the smallest fractions have been stacked. After having generated this initial particle structure, hydration algorithms are invoked that simulate the stepwise evolution of the particle hydration process and associated expansion of the outer shells of hydration products, whilst forming a 3D virtual microstructure (Figure 1b). During this hydration process, the solid volume fractions of the reactants, i.e. the anhydrous cement and free water, decrease, while, in return, the total fraction of formed hydration products increase. The outer expansion of the individual particles is calculated according to the so-called particle expansion mechanism [8], that accounts for the geometrical expansion of the expanding particles that overlap smaller particles and are located in the close vicinity. The positions of the 289


solid particles in space are described by means of a vector in a 3D Cartesian coordinate system. This includes the start (0, 0, 0) and end location (x, y, z), the diameter of anhydrous cement grain, and the thickness of inner and outer hydration layers that surround the shrinking core of the anhydrous cement particles.

a)

b)

a)

b)

Figure 1 a) 3D simulated microstructure (grey-cement, red-inner hydration product, yellow-outer hydration product): a) initial unhydrated cement; b) after hydration.

Figure 2 a) FD implementation, position and size of coordinates: width (x), height (y), and depth (z). Each sharing surfaces between neighboring voxels in x, y, and z directions has an assigned conductivity coefficient c x , c y , and c z , respectively. b) Steady state flux (J) across z-axis with (periodic or non-periodic) boundary conditions (b.c.) employed on 4 side faces parallel to the imposed flux.

2.2

Transient transport model

A finite difference (FD) based numerical scheme is derived for solving the transient transport problem (3D extension of eq. 1, only for transport). The algorithm starts with a discretization of the 3D virtual microstructure into a regular 3D mesh (e.g. Figure 2a), where each voxel in the mesh is assigned to be either a capillary pore phase or a solid phase, according to its actual position in the microstructure. In this algorithm, identification is done by the center point of a voxel. For each voxel plane that shares with a neighboring voxel in x, y, and z direction, a conductivity coefficient c x , c y , and c z , needs to be assigned, respectively (Figure 2 a). The 290


connectivity of all voxels are stored in three c vectors (whose lengths correspond to the number of voxels in the system, N). A six neighbor configuration was used, representing a situation where the voxel is connected to its neighbors by the six planes of a cubic voxel in x, y, and z direction. The conductivity coefficients of the surfaces that connect a central voxel to its neighbor voxel is calculated from a series connection approach using two conductors according to, Eq. 2.

= ci 1/ (0.5 / Di + 0.5 / Di + k )

(2)

where k = 1, w, or L. With this notation, w represents the number of voxels in a row and L the number of voxels in a layer. Ficks 2nd law is solved by a second order finite difference scheme provided in eq. 3. This equation shows a backward Euler fully implicit form of finite difference, for node i and its 6 neighbours.

cx ,i −1ui −1 + c y ,i − wui − w + cz ,i − wh ui − wh +

 ∆x 2  + cx ,i + cx ,i −1 + c y ,i + c y ,i − w + cz ,i + cz ,i − wh  ui + −  ∆t D  OLD 2 u ∆x + cx ,i ui +1 + c y ,i ui + w + cz ,i ui + wh = − ∆t D

(3)

Where u is the Ca concentration in solution, that needs to be calculated for all voxels and for each time increment, based on the known concentration, uOLD of the previous time step.

Assembling the equations for all (N) FD nodes forms a global system of equations, which can be represented in a matrix notation by Eq. 4.

Au = b

(4)

where u is the voltage vector (size of the total number of voxels in the system, N), A is a sparse and symmetric matrix with 7 diagonals (each voxel has 6 nearest neighbors) that contain information about conductivity coefficients of all the connections among the voxels, and b is the vector of knowns (i.e. boundary condition and previous time step concentrations, uOLD). The obtained system of equations (6, 7) is solved by a conjugate gradient algorithm with an optimized matrix-vector multiplication. This has been achieved by multiplying only those elements of the matrix that lie on the 7 diagonals while avoiding multiplication of a very large number of zero elements. Furthermore, since the size of the sparse matrix A is N times N, and may reach huge dimensions, the matrix is not stored explicitly but only by means of the vectors c x , c y , and c z which store the conductivity coefficients of the connections between voxel planes in the x, y, and z directions, respectively.

2.3

Boundary conditions

The main flow direction of the REV samples simulated in this study is considered to take place in z-direction. The 4 surfaces that are situated parallel to this main flow direction are subjected to two different types of boundary conditions, i.e. periodic and non-periodic. The non-periodic boundaries represent a full sealing off of the four boundary planes that run parallel to the dominant flow direction of the imposed flux (z-direction, Figure 2). This means that no transport can take place through these surfaces at all. On the contrary, periodic boundary conditions are also applied to the REV samples under consideration, and represent a full disclosure of the four parallel boundary planes, which are connected to the planes situated at the opposite side of the sample. 291


In the numerical algorithm, the boundary conditions are handled as two additional coefficient vectors (p x and p y ), each with a length equal to the number of boundary voxels that form one surface of the system: e.g. length=(h-2)*w, where h is the height and w the width of the surface. These two vectors store the conductivity coefficients for each element at the boundary-surface. If one of the voxels at a surface is a solid (zero conductivity) then the conductivity coefficient of the element it is linking to in the opposite surface plane is zero (no connectivity, and no flux). On the contrary, if a pore element in a surface is connected with a pore element in the opposite surface plane, then a flux is possible in that particular direction. After assembling of the main matrix according to eq. (6), the boundary conditions are applied

2.4

Degradation code

The numerical implementation is following the transient FD scheme for transient transport (Section Transport), including the following modifications to incorporate the degradation mechanism and source term for dissoulution of Ca. The main idea is to locate the solid voxels that are in contact with the pore solution, and impose their solution concentration from previous 0, corresponding to solid voxel, to a saturation value (e.g. 21 mol m-3 for portlandite). These phase boundary voxels are then dissolved by applying the mass balance calculation based on the total flux leaving out from the corresponding voxel to the nearest neighbors. The total amount of Ca diffused into solution (i.e. neighbors voxels) per time increment is equal to the amount of dissolved solid Ca, and is calculated by eq. 5. 6

∆n = D ∆t ∆x 2 ∑ ∆ui / ∆xi i =1

(5)

By knowing the rate of the dissolution, which was equalised to the diffusion rate, the change of porosity of the dissolving voxel with time can be calculated using the value of the bulk density (ρ = 2.27 g cm-3 for portlandite) and molar mass (M r ):

∆P = ∆n M r / ρ

(6)

The dissolution of the Ca(OH) 2 occures until the porosity reaches 0, and after that the voxel is treated as a (100%) pore voxel. Following the same approach, the change of Ca/Si ratio in the CSH phase can be calculated. For this phase the solubility, i.e. equilibrium concentration of the solution that represents the treshold value below which the dissolution occures, depends on the Ca/Si ratio. The phase boundary voxels are found according to criteria of having at least one ‘pore voxel’ neighbor. The FD equation for that voxel and the corresponding neighbours is updated. The pore concentration of the dissolving voxel is forced to saturation (equilibrium solubility) concentration. The connectivitiy coefficients c connecting the dissolving voxel to the nearest neighbours are changed from 0 (no connectivity with a solid voxel) to a relative value of 1, corresponding to a pore-pore connection). The FD eqs for the neighbours is also adjusted to consider the newly introduced saturated concentration on the dissolving voxels. This known concentration value goes to the right hand side, i.e. vector b of eq. 4.

3 Simulations

Two 3D virtual microstructures were generated to test the new degradation model: 1) simple sphere and 2) virtual cementitious microstructure. Specimens were cubic with a size of 50 µm and 1 µm voxel size. The cement pastes were prepared with an initial water to cement mass ratio of 0.4, hydrated for 28 days, resulting in 13% of capillary porosity. Model is based on the resolution of the mass balance equation for calcium concentration in solution and in solid. In this first presentation stage, the whole degradation process was demonstrated solely in terms of portlandite’s dissolution. In other words all of the solid phase 292


was treated as portlandite. Instantaneous dissolution is assumed, because the diffusion rates are much slower than those of the chemical reactions, so the kinetics of degradation is governed by a diffusion process. The dissolution of Portlandite occurs suddenly below a threshold calcium concentration of 21 mol m-3. Initial condition of the Ca concentration in the capillary pores was put to this saturation value. The aggressive solution is added by specifying Dirichlet boundary conditions (meaning fixing a constant concentration on the boundary to 0), simulating that one side of the model sample is placed in a reservoir of deionized water. This boundary condition is a very rigous, which, in reality, can be obtained by rapid (turbulent) flow of the water through the external reservoir.

1)

t=0s

10s

250s

500s

750s

1000s

2) Figure 3 A 2D slice of the 3D simulation results at degradation times: 0s, 10s, 250s, 500s and 1000s applied on two initial microstructures: 1) simple sphere (upper sequence) and 2) virtual cementitious microstructure (lower sequence). Grayscale map represents the concentration distribution (black is 0 and white is 21 mol m-3) while black also represents a solid.

4 Results and Discussion

Figure 3 shows 2D slices taken from the middle of the 3D results of the simulation obtained on two initial microstructures: 1) simple sphere (upper sequence in Fig 3) and 2) virtual cementitious microstructure (lower sequence). The Dirichlet boundary condition was applied on the left side of the slices. The presented degradation times are: 0s, 10s, 250s, 500s and 1000s. After 10s of simulation both initial microsturctures have increased in porosity, because the phase boundary solid voxels are transformed to a saturated concentration. With further degradation the results show a moving front of the concentration gradients that gradually dissolves the solid (Ca(OH) 2 ). The dissolution front is sharp, and is located always at the right end of the concentration gradient. Inside the porous matrix (more right from the dissolution front) there is no concentration gradient, and therefore no dissolution of the solid. The inclusion of CSH solubility in the model, that has variable solubility with Ca/Si ratio, is expected to create a more distributed dissolution in the matrix which is not only limited to the moving front. Next, the validity of instantaneous dissolution is discussed. Wang et al. [10] measured the dissolution rate of lime in solutions using a rotating disc of compressed lime to control the mechanisms involved in Ca(OH) 2 dissolution and the factors that affect the dissolution. The lime dissolution rate varied with the rotational velocity for values below 300 rpm, which is a threshold value for laminar and turbulent flow. When the disk rotational velocity was greater than 300 rpm, the dissolution rate was constant at 5.5 10-16 mol (µm2 s)-1 (at pH=7 and T=25oC). In Portland cement pastes, the pore solution typically has a pH from 12 to 13, the value depending in part on the concentration of soluble alkali salts in the cement. The dissolution rate decreased exponentially with increased values of pH over the pH range of 3-7, therefore, a value lower than 5.5 10-16 mol (µm2 s)-1 is expected in the cementitious materials. On the other hand, 293


the higher the surface area for a given portlandite mass, the higher the dissolution rate. Considering the surface area of 45 um2/um3 for CH [11] would yield 45 times higher dissolution rate, i.e. 25 10-15 mol (µm2 s)-1. This rate is still slightly higher than the maximal flux scenario occurring for the first time steps at the Diriclet boundary condition, where a 0 concentration is directly connected to a saturated concentration of the dissolving voxel, calculated as flux=D du/dx=790 um2 s-1 21 10-18 mol µm-4 = 16 10-15 mol (µm2 s)-1. Therefore care should be taken to check the validity of the assumption that the diffusion rate is the rate controlling mechanism. If the diffusion rate is higher than the dissolution rate the concentration of the dissolving voxel goes below the saturation value. In the proposed degradation algorithm this scenario is implemented by comparing the total flux from the dissolving voxel with the user prescribed dissolution rate (which depends on the temperature, pH and surface area), and if the flux is greater, the concentration of the voxel is lowered according to the amount of excess flux, porosity and volume of the voxel.

5 Conclusion A new 3D reactive-transport model and results of leaching-induced aging simulations were performed on virtual cementitious microstructures generated by Hymostruc model. The primarly simulation results were demonstrated by considering only the portlandite dissolution. The outputs of the new approach are the morphology of the aged 3D microstructure together with aged properties characterised by fuzzy state of each voxel at different times. The degradation simulation results show a sharp moving front followed by the concentration gradients that gradually dissolves the solid (Ca(OH) 2 ). A very rigous Dirichlet boundary condition (zero concentration) induces, at the first times of simulation, a high flux rates in contact to the nearest dissolving voxels. Therefore, for this case, care should be taken to check the validity of the assumption that the diffusion rate is the rate controlling mechanism. If the diffusion rate is higher than the dissolution rate, concentration of the dissolving voxel goes below the saturation value. The dissolution rate of Ca(OH) 2 depends on the temperature, pH and reactive surface area.

6 References

[1] Mainguy M Coussy O (2000) Propagation fronts during calcium leaching and chloride Penetration, J Eng Mech 126:250-257. [2] Mainguy M Tognazzi C Torrenti JM Adenot F (2000) Modelling of leaching in pure cement paste and mortar, Cem Concr Res 30-83-90. [3] Ulm FJ Torrenti JM Adenot F (1999) Chemoporoplasticity of calcium leaching in concrete, J Eng Mech 125:1200-1211. [4] Adenot F Buil M (1992) Modeling of the corrosion of cement paste by deionized water, Cem Concr Res 22: 489-496. [5] Berner UR (1998) Modeling the incongruent dissolution of hydrated cement materials, Radiochim Acta 44/45:387-393. [6] Buil M Revertegat E and Oliver J (1992) A model of the attack of pure water or under saturated lime solutions on cement, ASTM STP 1123:227-241. [7]Ukrainczyk N Koenders EAB (2014) Representative elementary volumes for 3D modeling of mass transport in cementitious materials, Modelling Simul Mater Sci Eng 22:035001. [8]Koenders E A B (1997) Simulation of volume changes in hardened cement-based materials, PhD Dissertation, Delft University of Technology, Delft, The Netherlands. [9] Atkins, P.W. (1994) Physical Chemistry. 5th Edition. W. H. Freeman, New York. [10] Wang J Keener TC Li G & Khang SJ (1998) The dissolution rate of ca(oh)2 in aqueous solutions, Chem Eng Comm 169:167-184. [11] Rodriguez-Navarro C Ruiz-Agudo E Ortega-Huertas M and Hansen E (2005) Nanostructure and Irreversible Colloidal Behavior of Ca(OH)2: Implications in Cultural Heritage Conservation, Langmuir 2005:10948-10957.

294


Protection of aged concrete structures: application of bio-based impregnation system Virginie Wiktor1*, Henk M. Jonkers1 (1) Delft University of Technology, Delft, The Netherlands

Abstract: This paper focuses particularly on the ageing of concrete due to micro-crack formation or freeze/thaw which results in an increased permeability of the concrete. The bacteria-based repair system presented in this paper aims at recovering the concrete permeability thanks to bacteria-induced calcium carbonate precipitation inside cracks/porosity. The performance of the bacteria-based repair system in laboratory and field application were very promising. The laboratory results showed that the crack sealing capacity of the repair system is very good as the cracks were completely sealed after impregnation. Results from field application are also very good as treated concrete had a significant higher resistance to freeze/thaw and only cracks that were not impregnated with the bacteria-based repair system were still heavily leaking. Keywords: Concrete, crack-repair, bio-based system, biodeposition

1 Introduction Concrete structures are, in the course of their lifetime, constantly ageing. They are exposed to a number of degradation processes such as carbonation, chloride ingress or freeze/thaw, which directly impact and degrade the properties of the material. Nevertheless, concrete structures can reach a service life of 50 years or longer as these ageing mechanisms are taken into account in the codes and predictive models used in the design phase of the structure. However, early-age deterioration is not easy to predict and the formation of micro-cracks for instance lead to a significant increase in the permeability of the concrete. The subsequent ingress of aggressive corroding substances can lead to the premature corrosion of the reinforcement and early failure of the structure. To maintain the integrity of the concrete structure measures for maintenance and repair have to be undertaken. Nowadays a wide range of repair products, such as for instance epoxybased fillers or silane-based water repellent, is available for concrete. However, besides their high cost, their short term efficiency and negative impact on the environment are an issue for the repair industry. Biodeposition, a method by which calcium carbonate (CaCO 3 ) precipitation is induced by bacteria, has been proposed as an interesting alternative approach to protect building materials against ageing. Among the various pathways involved in Microbial Induced Precipitation (MIP), enzymatic hydrolysis of urea in calcium rich environment is the most commonly used system. While it has been successfully applied as surface treatment in practice to limestone monuments, it has been considered only on a laboratory scale for cementitious material and crack repair. Also, besides cost issues, MIP using ureolytic bacteria might generate other problems, such as environmental nitrogen loading due to the production of ammonia during the hydrolysis of urea or negative effect to the material itself due chemical reactions with ammonium salt [1]. In addition, the time required for substantial amount of bacterially induced calcium carbonate may hold back the acceptance of MIP as efficient repair technique by the building industry. *

Corresponding author affiliation and e-mail address

295


Another interesting pathway for MIP is the metabolic conversion of organic salts through bacterial respiration. Successfully applied in self-healing concrete [2], this concept has been implemented by the authors for the development of bacteria-based repair system for concrete structures. In this way, a liquid ingress system for concrete transports a bio-based agent into aged porous/cracked concrete which results in reduced permeability and increased service life of constructions. The performance of the bio-based repair system in laboratory and in practice on 2 aged concrete parking garages are presented and discussed in this paper.


The bacteria-based repair system

The bacteria-based repair system combines advantages of both, traditional repair system for concrete (fast reacting, and short term efficiency), and bio-based methods (more sustainable, slow process, and long-term efficiency). The repair system consists in concrete compatible bacteria [3] and feed which produce calcite-based minerals decreasing concrete porosity. This system is composed of two solutions: (i) Solution A – sodium silicate (alkaline buffer, 4.8 g/L), sodium-gluconate (carbon source for bacteria growth, 125 g/L), yeast extract (vitamins for bacteria, 1g/L), alkaliphilic bacteria (1.6x108 spores/L). (ii) Solution B – calcium lactate (calcium for CaCO 3 precipitation) or calcium nitrate (nitrate source for denitrification when O2 is depleted and calcium for CaCO 3 precipitation, 500g/L), alkaliphilic bacteria (1.6x108 spores/L). The denitrification is the biological reduction of nitrogenous oxides to gaseous products during an aerobic (no oxygen) bacterial growth. This means that under the metabolic conversion of calcium nitrate, N 2 and CaCO 3 are produced. The sodium silicate in solution A ensures alkaline pH in the system and formation of a gel inside the crack. Although not very strong, this gel allows a rapid sealing of the crack (within few hours) and optimum environment for bacteria to precipitate calcium carbonate. By the time the gel becomes too weak, substantial amount of CaCO 3 has been precipitated to seal the crack.

2.2

Laboratory testing of the bio-based repair system

Mortar discs (∅ 17cm and h=2cm) were cast with ordinary Portland cement and aggregates (04mm) in plastic buckets as described by Wiktor & Jonkers[3]. The specimens were kept 28 days in sealed conditions at room temperature and then tested for 3 points bending test until failure. The bottom of the buckets is removed and the cracked mortar discs were glued in the buckets (Fig. 1). a. Casting of mortar discs

b. Breaking of specimen

c. The bottom of the bucket is removed

F bucket bottom removed

~ 2cm bucket ∅= 17 cm - h= 14 cm

d. Specimens are glued inside the buckets

Figure 1 Preparation of mortar specimens for water permeability test

Permeability test is performed before and 28 days after impregnation with the bacteriabased repair solution: the amount of water permeating through the crack in 1 hour is recorded. The permeability value before repair is noted P 1 and the one after P 2 . 296


The difference between these 2 values gives an estimation of the efficiency of the crack repair. To evaluate the resistance of the repair to ageing, the specimens are then subjected to five freeze/thaw cycles, and a third permeability test is performed afterwards (Fig. 2). Control specimens are impregnated with tap water. Calcium lactate is used as calcium source in solution B.

Figure 2 Impregnation of the mortar specimen with the bacteria-based repair system and evaluation of the crack-sealing efficiency by means of water permeability test

The decrease in permeability after impregnation and after 5 freeze/thaw cycles is given by equation 1 and 2 respectively: P decrease after impregnation =

P decrease after freeze/thaw =

2.3

( P2 − P1 ) x100 P1

( P3 − P1 ) x100 P1

(eq 1)

(eq 2)

Field testing of the bio-based repair system

In order to validate laboratory results, the bacteria-based repair system has also been applied in practice, on 2 parking garages named PG1 and PG2. In both cases, the concrete deck was suffering from cracking which resulted in a significant leaking of the structure. Also, in PG2, the concrete pavement on each side of the access ramp was aged and damage due to freeze/thaw.

Figure 4 Parking garage 1 (PG1) – (a) detail of a leaking crack before application of the bacteria-based repair system, (b) spraying of the bacteria-based repair system

In PG1, an area of 2x4m was sprayed with the bio-based repair system on both side of the concrete deck (Fig. 4). In PG2, part of the concrete pavement (area of 2x0.5m) and three cracks

297


(1-3mm wide) of the concrete deck were impregnated with the bacteria-based repair system (Fig. 5).

Figure 5 Parking garage 2 (PG2) – (a) spraying of the bacteria-based repair system on concrete pavement, (b) crack impregnated with the bacteria-based repair system

In both cases, solution A and solution B were each poured in a sprayer, and manually applied at the surface of the concrete in layers until saturation of the concrete treated area. Solution B contains calcium nitrate instead of calcium lactate as calcium source. In this way more calcium can be added and the system also has a nitrate source for bacterial growth in case the concentration in O 2 is limited. Two months after the application of the bacteria-based repair system the crack sealing performance was visually assessed in PG1. In PG2, the crack sealing efficiency was assessed by means of water permeability test performed on site. Rectangular wooden frames (1x0.5m) were placed on top of 3 treated- and 3 untreated cracks on the concrete deck. The wooden frames were sealed with silicon glue prior pouring 5L tap water. As the crack goes through the whole thickness of the deck, the sealing efficiency was assessed by monitoring visually, from the other side of the deck, how much water was dripping through the crack. Also, 6 cores were drilled from two different locations on the concrete pavement: 3 from the treated area and 3 from an untreated part on the same side of the access ramp as control specimens. The resistance to freeze/thaw and deicing salt was then evaluated in laboratory. The cores were tested according the NPR/TS 1239-9 and NEN-EN 13877-2. The test was performed by Cugla B.V. (Breda, the Netherlands).

3 Results and Discussion 3.1

Laboratory testing of the bacteria-based repair system

The results of the water permeability test are presented in figure 6. The cracks impregnated with the bacteria-based repair system are completely sealed as no water was leaking from the crack. For the control specimens on the other hand, the water permeability is even higher 28 days after the impregnation. It is also interesting to note that after 5 freeze/thaw cycles the permeability of control specimens tremendously increased while the specimens treated with the bacteria-based repair system still exhibit a lower permeability than the initial value. These results are very encouraging as only the specimens treated with the bacteria-based repair system showed improved properties. However, in order to validate these good results, the bacteria-based repair system should be tested on real concrete structure, outside in a noncontrolled environment.

298


Figure 6 Water permeability results of control and specimen treated with the bacteria-based repair system

3.2

Field performance of the bacteria-based repair system

- Parking garage 1 – PG1 The performance of the bacteria-based repair system for the first field application in PG1 was assessed qualitatively by visual observations. The pictures taken on PG1 two months after the impregnation with the bacteria-based repair system are shown on figure 7. It is worth noting that the pictures were taken only a few days after a raining episode. It can clearly be seen that the cracks that have not been treated with the bacteria-based repair system (Fig. 7a) seem wet and are lined with white precipitate formed due to ingress water through the cracks. On the other hand, all the treated area is dry meaning (Fig. 7b) that cracks treated with the bacteriabased repair system are not leaking anymore.

Figure 7 Visual inspection of cracks 2 months after application of the bacteria-based repair system in PG1 – (a) non treated area, the arrow points white precipitate still leaking from the crack, (b) treated area, no precipitate is observed

- Parking garage 2 – PG2 For the second field application in PG2, the performance of the repair system is assessed more quantitatively. The water permeability test to assess the crack sealing efficiency was 299


performed on site 2 months after the application of the repair system. The results were very encouraging as only cracks that were not treated with the bacteria-based repair system were still heavily leaking along their full length (Fig. 8a). Two cracks impregnated with the repair system had only few localized dripping spots (Fig. 8b), and the third one didn’t leak at all.

Figure 8 Observation of water leaking through the cracks during water permeability test – (a) control crack, (b) crack treated with the bacteria-based system

Relating to the freeze/thaw resistance of the concrete, the specimens treated with the bacteria-based repair system had a significant better resistance compared to the control as they exhibited 48% less mass loss (scaling) than the control. These results from field experiment confirm the good performance of the bacteria-based obtained in laboratory.

4 Conclusion

This paper presented the application of bacteria-based repair system as preventive solution against ageing of concrete. It focuses particularly on the ageing of concrete due to micro-crack formation or freeze/thaw which results in an increased permeability of the concrete. The bacteria-based repair system presented in this paper aims at recovering the concrete permeability thanks to bacteria-induced calcium carbonate precipitation inside cracks/porosity. The performance of the bacteria-based repair system in laboratory and field application were very promising. The laboratory results showed that the crack sealing capacity of the repair system is very good as the cracks were completely sealed after impregnation. Results from field application are also very good as treated concrete had a significant higher resistance to freeze/thaw and only cracks that were not impregnated with the bacteria-based repair system were still heavily leaking. This results are very encouraging for the application in practice of the bacteria-based repair as preventive measure against ageing of concrete. The bacteria-based repair is now being optimized in order to improve the freeze/thaw resistance and to raise the crack-sealing efficiency on site to 100%. The next step is to evaluate the long-term durability of the repair.

5 Acknowledgment The authors would like to thank Cugla B.V. (Breda, The Netherlands) for testing the resistance of the concrete cores to freeze/thaw. Also, the financial support from Agentschap NL (IOP grant SHM12020) for this work is gratefully acknowledged.

300


6 References [1] [2] [3]

Dhami NK, Reddy SM, Mukherjee A (2012) Biofilm and microbial applications in biomineralized concrete. In: Seto J (eds) Advanced topics in biomineralization. In Tech, Rijeka, pp 137-164. Wiktor V and Jonkers HM (2010) Quantification of crack-healing in novel bacteria-based self-healing concrete, Cement Concrete Comp. 33:763-770. Wiktor V and Jonkers HM (2012) Application of bacteria-based repair system to damaged concrete structures. Proceedings of the 2nd International Workshop on Structural Life Management of Underground Structures, Daejeon 19-20 October 2012 pp.31-34.

301


Estimation of Carbonation and Service Life of Box Culvert for Power Transmission Line Sang-Kyun Woo1, Yun Lee2*, Yong-Dong Choi3 (1) Korea Electric Power Corporation - Research Institute, Daejeon, Korea (2) Daejeon University, Daejeon, Korea (3) Korea Electric Power Corporation - Research Institute, Daejeon, Korea

Abstract: The construction of underground structures such as box culverts for electric power

transmission is increasing more and more, and the life extension of these structures is very important. Carbonation-induced corrosion in concrete may often occur in a high carbon dioxide environment. In this study, the risk of carbonation of two concrete box culverts in an urban area was evaluated by measuring the carbonation rate and concrete cover depth. Then, the carbonationfree service life at the depth of the steel was calculated, based on in situ information, by the Monte Carlo simulation. The service life of box culvert due to carbonation was estimated within 30~50 years via Monte Carlo simulation.

1 Introduction Box culverts for electrical power transmission began to be constructed in the mid-1970s in Korea and are increasing in numbers recently due to power demand growth and underground installation of transmission lines. The amount of related maintenance work also continues to increase because of the increasing usage life of the constructed box culverts. Corrosion of steel due to carbonation usually occurs particularly in an urban area which has a high level of carbon dioxide, emitted from vehicles and industrial factories. Carbonation can be defined as the chemical reaction between carbon dioxide present in the air and cement hydration products such as mainly calcium hydroxide and the CSH gel phase, which results in the formation of calcium carbonate. Thus, the risk of carbonation is more severe in urban or/and industrial area. Carbonation of concrete itself does not do harm in view of the performance of structure, adversely a marginal enhancement of the compressive strength was observed. However, when carbonation reaches at the depth of the steel, the high alkalinity of the concrete pore solution is neutralized and hydration products are dissolved then to lower the buffering capacity of hydrations against a pH fall. At this moment, the passivation layer on the steel surface, which otherwise would protect the steel embedment from a corrosive environment, is destroyed, and steel is directly exposed to oxygen and water, eventually to corrode.

2 METHODOLOGY. 2.1

Description of Carbonation Process

The limit state function is useful to render the risk/resistance of concrete carbonation, as being expressed into two parts: the resistance to carbonation and the carbonation risk, as given in Eq. (1). (1) g= (t ) R (t ) − S (t ) where R(t) and S(t) denote the carbonation resistance (concrete cover depth), and the carbonation risk (carbonation depth) with time. The carbonation depth is usually calculated as a function of time, as given in Eq. (2), with the carbonation coefficient encompassing the concrete quality and the environmental conditions. (2) x(t ) = K t *

Department of Civil Engineering, Daejeon University, 96-3, Yongun-dong, Dong-gu, Daejeon, 300-716, Republic of Korea, Email address: [email protected]

302


where x is the depth of carbonation, K for the carbonation rate coefficient and t for time of exposure to the atmosphere containing carbon dioxide. The limit state defined by Eq. (1), when subjected to a carbonation environment, is often regarded as the state that carbonation reaches the depth of the steel, in which steel starts to corrode then to reduce dramatically the performance of concrete structures..

2.2

Monte Carlo Simulation

The safety factor method and the Monte Carlo simulation are representative techniques in modelling the risk of carbonation. However, for the safety factor method in predicting the risk of carbonation, the safety factor was not rationally determined with the reliability index in previous studies. The safety factor has been usually derived from an engineering judgment rather than calculated from in situ information, in order to compensate for the variance of the parametric values such as carbonation rate and concrete cover depth, which might happen arising from the difference in construction environments and construction quality. The safety factor has been conventionally regarded as being 1.2 in some standards and guidelines. Furthermore, the distribution type for the parametric values, although the carbonation rate and concrete cover depth are obtained from the corresponding in situ examination, is always determined to the normal distribution only, possibly leading to an error in calculating the safety factor. Thus, it can be said that the carbonation-free service life predicted by the safety factor method might be less convincible. The Monte Carlo simulation requires a number of empirical measurements from a field in terms of the average value and the standard deviation of the parametric values, in order to validate its probabilistic prediction. Unlike the safety factor method, the carbonation resistance and risk factors are not required, when the probabilistic method is used, to consider the variance of the experimental information including the carbonation depth and concrete cover depth. Instead, the simulated deviation of the parametric values is taken into account in determining the carbonation resistance and risk. In the present study, the Monte Carlo simulation technique, as a probabilistic way, was used to assess the carbonation risk. For the simulation, mean value and standard derivation of the parametric values (i.e. carbonation depth and concrete cover depth) were obtained from underground field investigation. Then, 100,000 of random samples for the parametric value were generated by the Monte Carlo simulation. The probability for carbonation to reach the depth of the steel can be calculated from the 100,000 random trials and is defined as the ratio of the number of carbonation, calculated by Eq. (1), at the depth of the steel to the number of total trials, as given in Eq. (3).

= Pt n( g (t ) < 0) / N

(3)

where P t is the probability of carbonation at the depth of the steel and n(g(t) < 0) denotes the number of carbonation at the depth of the steel out of N trials (100,000 trials in this study).

3 FIELD AND INDOOR INVESTIGATON

Two concrete box culverts in Kayang and Kwacheon in Seoul, as typically shown in Fig. 1, were examined for carbonation after 18 and 16 years, respectively. The concentration of carbon dioxide in the box culvert reached 508 ppm in Kayang and 629 ppm in Kwacheon, respectively. Prior to measuring the carbonation depth, the concrete cover depth to the steel was measured using an ultrasonic cover meter in 84 points. After measuring the cover depth, 84 concrete core specimens were drilled out for durability test including carbonation depth investigation. In order to avoid further carbonation of the core specimen in later experiment, each specimen was sealed with plastic wrap up to indoor test. By splitting cylinder specimen in the laboratory, carbonation depth was measured in each specimen. 303


The measurement was done with a phenolphthalein pH indicator (i.e. 0, t0: time of loading, f (t, t0) is a probability density function. Knowing that F (t, t0) is the probability distribution function. It is obtained by direct integration with respect to time (t) of the probability density function f (t, t0) as follows:

(2)

F (t , t0 )   f (t , t0 )dt

The plot of F (t, t0) begins with an exponential form then inflected to reach in the end, an asymptotic threshold (figure 1).

1,0 0,8 0,6 0,4 0,2 0,0 0,01

0,1

1

10

100

1000

10000

Time(Days)

Figure 1: The distribution function of probability F (t, t0).

Resolution of this equation gives:

F (t , t0 )  e b (t t0 )

C

t 0

=

1  e b ( t  t 0 )

(3)

C

To take into account the evolution of the probability distribution function F (t, t0) to achieve asymptotic threshold, we multiply the equation (3) by a non-zero positive number (a) which gives the form final following: c

F (t , t0 )  a  (1  eb(t t0 ) ) 3. Estimation of model parameters: For the identification of the model parameters, we used the experimental results. The following relations give the expressions used:

322

(4)


V / S  2 a  1   2 .HR   3    HR  b   4   5 .V / S  

6

HR c   7   8 V / S  .HR





2

With: V / S: volume ratio surface (V / S) (mm) RH: relative humidity in percentage Tests and the corresponding constants in these expressions are summarized in Tables 1, 2 and 3. P (S) *: Probability of significance of each estimated coefficient. P (F) **: Probability of significance to the value Fisher test. Table1 : tests du paramètre « a »

Coefficient model

Standard Deviation

Student test

Fisher test

Correlation coefficient

Student

P(S)*

test

Fisher

P(F)**

R2

R

0,0000

0,9572

0,9558

test

β1

1,250042

0,0155

80,6407

0,0000

β2

-0,842352

0,0267

-30,8289

0,0000

β3

-0,000320

0,0001

9,6071

0,0000

682,517

2

Table 2 : tests du paramètre « b »

Coefficient model

Standard Deviation

Student test

Fisher test


Student test

P(S)*

Fisher

P(F)**

R2

R

0,0000

0,9723

0,9714

test β4

0,236297

0,0027

85,8150

0,0000

β5

-0,001000

8,7 E-05

-45,8909

0,0000

β6

0,002927

0,0004

6,0662

0,0000

323

1071,38

2


Table 3 : tests du paramètre « c »

Coefficient model

Standard Deviation

Student test

Fisher test


Student

P(S)*

Fisher test

P(F)**

R2

R

682,517

0,0000

0,7901

0,7202

test Β7

0,540869

0,0037

144,6832

0,0000

Β8

1,96E-06

5,5E-07

3,5082

0,0127

4. Validation of the model The model was validated by comparison with experimental results. The results are summarized in Figures 3a, 3b, 3c, 3d, 3e, and 3f.

Drying Shrinkage (mm/m)

We observe a perfect concordance between predictions of the model developed and the experimental results. Diameter of sample D = 76mm Relative humidity RH = 20%

1,2 1,0 0,8 0,6

Expérimental Conceived model

0,4 0,2 0,0

1

10

100

Time (Days)

1000

Drying Shrinkage (mm/m )

3a

Diameter of sample D = 102 mm) Relative humidity RH = 20% 1,0 1,2

0,8 0,6

Expérimental Conceived model

0,4 0,2 0,0

1

10

100

1000

Time (Days)

3b

324

2



Diameter of sample D = 76 mm) Relative humidity RH = 75%

0,8 0,6 0,4

Experimental Conceived model

0,2 1

10

100

1000

Time (Days)


3c


0,6 0,4


0,2 0,0 1

10

100

1000

Time (Days)


3d

0,8


0,6 0,4

Experimental Conceived model "B3" model

0,2 0,0

1

10

100

Time(Days)

3e

325

1000



Diameter of sample = 102 mm) Relative humidity = 50%

1,2 1,0 0,8 0,6


0,4 0,2 0,0

1

10

100

1000

Time (Days)

3f Figure 3: Comparison between the values of drying shrinkage predicted by the model and experimental results for different ratios V/S and for different relative humidity.

5. Application to the high performance concrete

Drying Shrinkage(mm/m)

Figures 4a, 4b, 4c, 4d, 3e and 3f gives the confrontation between values of drying shrinkage[1,2,3] gotten experimentally on the high performance concrete and these predicted by the developed model. It clearly appears that the developed model permits to predict correctly the drying shrinkage of high performance concrete. We observe a perfect concordance between predictions of the model developed and the experimental results.

0,3

High performance concrete Nuclear power station PALUEL

0,2

0,1

0,0

Experimental Conceived model 100

1000

10000

Time (Days)

4a

326



1,0 0,8

High performance concrete Nuclear power station PENLY

0,6 0,4


0,2 0,0

100

1000

10000 Time (days)


4b

0,5 0,4

High performance concrete nuclear power station CIVAUX - B11

0,3 0,2


0,1 0,0

100

1000

10000

Time (Days)


4c

0,4 0,3

High performance concrete Nuclear power station CIVAUX

0,2 0,1 0,0

Experimental Conceived model 100

1000

10000

Time (Days)

4d

327



600 500 400

High performance concrete Nuclear power station CHOOZ

300 200


100 0

100

1000

10000

Time (Days)


4e 0,6 High performance concrete 0,5 0,4

Nuclear power station FLAMMANVILLE

0,3 0,2


0,1 0,0

100

1000

10000

Time (Days)

4f

Conclusion The developed model is adapted well to describe the evolution of the drying shrinkage of hydraulic concrete. It has been justified by confrontation to the experimentally results. The developed model is simple to use and present the advantage to only contain a number limited of parameters. Results showed that the developed model gave good results with the high performance concrete. However, the developed model can be applied at any type of concrete.

References Journal article: [1] Bazant Z.P., Baweja S (2000) - Creep and Shrinkage Prediction Model for Analysis and Design of Concrete Structures: Model B3, ACI International SP-194-1 pp. 1-83.

328


[2] Bazant Z.P, Kim J.K (1987) - Wittmann F, and Alou F, - Statistical Extrapolation of Shrinkage Data-Part Il ; Bayesian Updating, AC1 Materials Journal, N° 84-Ml 0, pp. 83-91. Conference proceedings [3] Bazant Z.P (1996) - Fourth rilem international symposium on creep and shrinkage of

concrete: mathematical modeling", northwestern university, Evanston.

329


Bayesian modeling of bivariate extreme shocks Pasquale Cirillo(a) , (a)

DIAM - Delft University of Technology, Mekelweg 4 2628 CD Delft , The Netherlands

February 24, 2014

1

Introduction

Most shock models available in the literature are univariate, and this is particularly true when dealing with metal fatigue, see [12]. This means that we often consider a single-component system subject to random shocks of random magnitude. Most of the times, also shocks are simply assumed to be of one single type, even if some important exceptions are to be considered, as in the case of competing risk models (a review in [7]). The aim of this short paper is to introduce a first multivariate - actually bivariate, but the idea can be easily extended - shock model. In the literature, just a few pioneering works are present (e.g. [9], [10], [1] and [8]) on this topic, and they are all based on very strong, and sometimes unrealistic, assumptions. A basic multivariate shock model can be represented by a system made of different components. The components may have the same importance for the life of the system, but we can also consider the case in which a hierarchy is present. In this case, some parts are more important for the general survivorship of the system. As usual, the system is subject to random shocks of random magnitude and we can think of two types of shocks: • Component-wise shocks, i.e. shocks affecting a specific component; • Common shocks, i.e. shocks that inflict correlated damages on all the components of the system. Default can be caused by the failure of a certain amount of components, but also by the collapse of one vital component, if a hierarchy is present. Naturally damages can be cumulative or one-shot, as in the case of extreme shock models. It is also possible to consider different types of shock submodels for the different components of the system. An interesting feature of multivariate shock models is the possibility of assuming the replaceability 1

330


of non-vital components, introducing further flexibility and realism in the modeling. The importance of the multivariate extension is double: from the applied side, multivariate shock models are linked to several meaningful phenomena, when dealing with sets of defaults and interacting risks (in engineering, economics or medicine); from a theoretical point of view, multivariate models offer a lot of challenging problems to be solved. Some of these problems are in the wake of the many open questions of multivariate extreme value statistics, while some others are specifically related to shock models and the associated (multivariate) survival functions. Studying multivariate shock models also means deepening our knowledge of multivariate defaults and cascading failures, two topics that are particularly important in risk analysis at the moment. Until now, the majority of models in the literature have assumed shocks and risks to be independent, essentially because dependence makes everything more complex from a mathematical/probabilistic point of view. In reality, assuming independence corresponds to build up a model on some very weak foundations, because most of the actual phenomena we observe are clearly dependent: different risks often coexist and are strictly related.

The model in brief We take into account N defaultable systems, assuming them to be exchangeable. Every system is made of two elements subject to random shocks of random magnitude. No hierarchy is here assumed between components. A system may fail because of the collapse of any (or both) of the two components. The collapse is supposed to happen because of extreme shocks (see [5, 6]). The construction we propose is inspired by the model of [2]. Now, let Xn and Yn be the default times of the two parts constituting a system, for n = 1, ..., N . The couple (Xn , Yn ) thus identifies the n-th system. Let Zn , Vn and Wn be three independent reinforced urn processes (RUP) or, to be more exact, three urn-based extreme shock models, as the one defined in [3]. In order to simplify the notation, let αjZ and βjZ be the number of white and respectively black balls in urn U Z (j), j = 1, 2, ..., for the reinforced urn process {Zn }. As shown in [11, 3], the sequences {Zn }, {Vn } and {Wn } are exchangeable, and their de Finetti measure is a beta-Stacy process. Under a beta-Stacy prior, the results of [13] guarantee that the conditional survival function of {Zn } (and similarly of {Vn } and {Wn }) is SˆZ (z) = P [Zn+1 > z|Z∗n = zn ] =

z Y

(βjZ + rj (zn ))/(αjZ + βjZ + sj (zn )), (1)

j=0

2

331


where vector of observed systems up to n, mj (zn ) = Pthe Pn zn = (z1 , ..., zn ) is n 1 1 , r (z ) = j n k=1 [zk >j] and sj (zn ) = mj (zn ) + rj (zn ). k=1 [zk =j] Let now assume that Z1 , ..., Zn are independent and identically distributed random variables, subject to right-censoring. As known, censoring is a very common problem of survival studies. What we observe is (Z1∗ , δ1 ), ..., (Zn∗ , δn ), with Zi∗ = z, δi = 0 if a censoring took place, i.e. Zi > z, Zi∗ = z, δi = 1 if no censoring took place, i.e. Zi = z. With a quadratic loss function, the predictive distribution of Zn+1 given (Zn , δn ) is the Bayes estimator for the random distribution function. In particular, always from [13], we have " # z ∗ (z , d ) Y α + m j n n j SˆZ (t) = P [Zn+1 > z|Z∗n = zn , δn = dn ] = 1− Z (2) αj + βjZ + sj (zn ) j=0 P where m∗j (zn , dn ) = nk=1 1[tk =j,dk =1] and dn ∈ {0, 1}n . Notice that, for αjZ , βjZ → 0 ∀j, eq. 2 reduces to the standard Kaplan-Meier estimator. Now, for every n = 1, ..., N , set Xn = Zn + Vn , Yn = Zn + Wn .

(3)

Hence, by construction, we can notice the following: • for every system (Xj , Yj ), each component has a common element Zj and a specific one, Vj or Wj ; • the dependence structure is thus modeled without a parametric model, but with a simple, intuitive construction; • conditionally on Zj , Xj and Yj are independent; • σ(Zn , Vn , Wn ) = σ(Zn , Xn , Yn ); • the dependence is given by Cov(X1 , Y1 ) = V ar(Z1 ) ≥ 0, Cov(Xn+1 , Yn+1 |Zn , Vn , Wn ) = V ar(Zn+1 |Zn ), n ≥ 1.

(4) (5)

Main properties Let us briefly state some of the properties of the bivariate urn-based extreme shock model discussed in the previous section.

3

332


PROPOSITION 1.1. The systems {(Xn , Yn ), n ≥ 1} are exchangeable. PROOF. {Zn }, {Vn } and {Wn } are exchangeable by construction. Now observe that (Xn , Yn ) is a measurable function of (Zn , Vn , Wn ), which is an exchangeable sequence. Hence even {(Xn , Yn ), n ≥ 1} must be exchangeable.

Let FX be the marginal distribution of {Xn }, given the construction of the bivariate urn-based shock model, we have FX = FZ ∗ FV ,

(6)

FY = FZ ∗ FW .

(7)

Hence the marginal distributions of {Xn } and {Yn } are convolutions of betaStacy processes. As a consequence, if P is the probability function associated with F , PXY (x, y) =

x∧y X

PZ (z)PV (x − z)PW (y − z), ∀(x, y) ∈ N20 .

(8)

z=0

If σZ2 = V arFZ (Z), then CovFXY (X, Y ) = σZ2 .

(9)

From a Bayesian point of view, the use of a bivariate reinforced urn process to study systems made of coupled components subject to default is equivalent to the definition of a probability measure P2 on the space of the bivariate functions on N20 . More properties about P2 can be found in [4]. Let S(x, y) = P [X > x, Y > y] be the bivariate survival function, defined on N20 , describing the survivorship of a two-components system. Our interest is related to the predictive ˆ y) = P [Xn+1 > x, Yn+1 > y|Xn = xn , Yn = yn ]. S(x,

(10)

If the elements in (Xn , Yn ) are not subject to right-censoring, it is possible to compute eq. 10 explicitly, as stated in the following proposition. PROPOSITION 1.2. The predictive distribution P [Xn+1 > x, Yn+1 > y|Xn =

xn , Yn = yn ] is given by P [Xn+1 > x, Yn+1 > y|Xn = xn , Yn = yn ] = P [Xn+1 > x, Yn+1 > y, Xn = xn , Yn = yn ] . P [Xn = xn , Yn = yn ]

(11)

PROOF. Equation 11 is simply the result of the urn construction we have proposed. It can be computed by considering the numerator and the denominator

4

333


separately, and then calculating the ratio. The numerator is given by P [Xn+1 > x, Yn+1 > y, Xn = xn , Yn = yn ] = " min(x,y) min(xn ,yn ) min(xn−1 ,yn−1 ) min(x1 ,y1 ) zn+1 −1 x−zn+1 Y Y X X X X Z Z ··· fzn+1 (zn ) gi (zn ) giV (xn − zn ) vn+1 =0

vn =0

vn−1 =0

yn −zn+1

Y

giW (yn − zn )

i=0

v1 =0



n−1 Y

zj+1 −1

Y

fzZ +1 (zn ) j

j=1

xj+1 −zj+1 −1

giZ (zi )fxVj+1 −zj+1 (xj − zj )

i=0

Y

giW (yi − zi ) QZ z1

zY 1 −1

i=0 x1 −z Y1 −1

V QW y1 −z1 (1 − Qi )

giV (xi − zi )

QVx1 −z1 (1 − QZ i )

i=0

y1 −z Y1 −1

i=0

Y i=0



yj+1 −zj+1 −1

fyWj+1 −zj+1 (yj − zj )

i=0

i=0

# (1 − QW i ) ,

i=0

and the denominator by P [Xn = xn , Yn = yn ] = min(x,y) min(xn ,yn )

X vn+1 =0

X vn =0

min(x1 ,y1 )

···







fzZ (zn ) j+1

n−1 Y

X v1 =0

zj+1 −1

j=1

Y

giZ (zi )fxVj+1 −zj+1 (xj − zj )

i=0

xj+1 −zj+1 −1

yj+1 −zj+1 −1

Y

Y

giV (xi − zi ) fyWj+1 −zj+1 (yj − zj )

i=0 x1 −z Y1 −1

 giW (yi − zi ) QZ z1

i=0

i=0 V QW y1 −z1 (1 − Qi )

i=0

y1 −z Y1 −1

# (1 − QW i ) ,

i=0 αA b +mb (c) , gbA (c) A αA +β b b +sb (c) balls in urn U A (b).

where fbA (c) =

=

βbA +rb (c) , A αA b +βb +sb (c)

and QA b is the pro-

portion of white The possibility of fully observing all systems at a certain time point is however a strong assumption. In practical cases, many of the systems could be subject to right censoring. In such a case, equation 11 is no more valid, and it must be substituted with ˆ y) = P [Xn+1 > x, Yn+1 > y|X∗ = xn , δn = dn , Y∗ = yn , n = en ], S(x, n n (12) for which a closed form computation is not possible. Nevertheless, thanks to our urn construction, it is not difficult to develop a sort of MCMC algorithm to solve it. In particular, using eqs. 6 and 9, we can formulate the following procedure (see also [2]): 5

334

zY 1 −1

QVx1 −z1 (1 − QZ i )


1. Express an initial distribution FZ0 for Z. If a prior knowledge is available, use it! Otherwise, the choice is free. The only constraint being σZ2 = Cov(X, Y ). 2. Determine αjZ and βjZ . 3. Given FZ0 and the prior guesses FX0 and FY0 (these can be extrapolated 0 . from data), solve eq. 6 and get FV0 and FW 4. Compute αjV , βjV , αjW , βjW . Once we have set the parameters of the model, it is possible to compute ˆ y) as follows: S(x, 1. Given (X∗n , δn , Yn∗ , n ), the full conditional of Zn is PZn |Zn−1 ,X∗n ,δn ,Yn∗ ,n ∗ ∝ P [Vn∗ = xn − zn , δn = dn |Vn−1 = xn−1 − zn−1 , δn−1 = dn−1 ] ∗ × P [Wn∗ = yn − zn , n = en |Wn−1 = yn−1 − zn−1 , n−1 = en−1 ]

× P [Zn = zn |Zn−1 = zn−1 ] where, for example, ∗ P [Wn∗ = w, n = e|Wn−1 = wn−1 , n−1 = en−1 ] ( ∗ P [Wn∗ ≥ w|Wn−1 = wn−1 , n−1 = en−1 ] if e = 0, = ∗ ∗ P [Wn = w|Wn−1 = wn−1 , n−1 = en−1 ] if e = 1.

2. Since {Zn } is exchangeable, all the conditionals PZn |Z−j ,X∗n ,δn ,Yn∗ ,n , where Z−j = {Zi }ni=1 \{Zj }, have the same form. 3. Compute Vn∗ = X∗n − Zn and Wn∗ = Yn∗ − Zn . 4. Zn+1 , Vn+1 and Wn+1 are sampled according to PZn+1 |Zn , PVn+1 |Vn∗ ,δn and PWn+1 |Wn∗ ,n . 5. Naturally Xn+1 = Zn+1 + Vn+1 and Yn+1 = Zn+1 + Wn+1 .

References [1] Boyles, R.A., Samaniego, F.J., 1983. Maximum likelihood estimation for a discrete multivariate shock model. Journal of the American Statistical Association 78, 445-448. [2] Bulla P (2005) Application of Reinforced Urn Processes to Survival Analysis. PhD Thesis Bocconi University. [3] Cirillo P., H¨ usler J., 2010. Extreme shock models: an alternative perspective. Statistics and Probability Letters 81, 25-30.

6

335


[4] Cirillo P., 2011. A Bayesian model for bivariate extreme shocks. Submitted. [5] Gut, A., H¨ usler, J., 1999. Extreme shock models. Extremes 2, 293-305. [6] Gut, A., H¨ usler, J., 2005. Realistic variation of shock models. Statistics & Probability Letters 74, 187-204. [7] Kalbfleisch, J., 2002. The Statistical Analysis of Failure Rime Data. Wiley, New York. [8] Kulik, R., Szekli, R., 2005. Dependence orderings for some functional of multivariate point processes. Journal of Multivariate Analysis 92, 145-173. [9] Marshall, A. W., Olkin, I., 1993. Bivariate life distributions from Pólya’s urn model for contagion. Journal of Applied Probability 30, 497-508. [10] Marshall, A.W., Shaked, M., 1979. Multivariate shock models for distributions with increasing hazard rate average. Annals of Probability 7, 343-358. [11] Muliere. P., Secchi, P., Walker S. G., 2000. Urn schemes and reinforced random walks. Stochastic Processes and their Applications 88, 59-78. [12] Stephens, R.I., Fatemi, A., Fuchs, H.O., 2000. Metal Fatigue in Engineering. Wiley, New York. [13] Walker, S., Muliere, P., 1997. Beta-Stacy processes and a generalization of the Pólya urn scheme. Annals of Statistics 25, 1762-1780.

7

336


Residual Fatigue Life Evaluation of Rail at Squats seeds using 3D Explicit Finite Element Analysis Xiangyun Deng *, Maysam Naeimi, Zili Li, Zhiwei Qian, Rolf Dollevoet Delft University of Technology, Delft, The Netherlands Abstract: A modeling procedure to predict the residual fatigue life of rail at squats seeds is developed in this article. Two models are involved: a 3D explicit Finite Element (FE) model to compute the stress and strain at squats in rail, and the J-S fatigue damage model to determine the residual fatigue life on the basis of the computed stress and strain. In the FE model dynamic effects of wheel-rail system under rolling contact is taken into account. Bilinear isotropic elastic-plastic material properties are adopted to represent the hardening of wheel and rail. Squats are subject to multiple loading cycles. The geometry of the squat is varied in the simulation corresponding to a growing squat at different ages. It is found that small squats lead to fatigue failure while severe ones lead to ratcheting failure. Keywords: Fatigue Life, Finite Element Method, Rolling Contact Fatigue, Cracks Initiation, Rail Squats

1 Introduction

The dynamic behavior of wheel-rail interaction system has attracted considerable attention from the fatigue analysis and fracture mechanics viewpoints. Fatigue is known to be the principal cause of many mechanical failures in wheel-rail components. Due to this fact, fatigue analysis of wheel-rail elements has been an important topic of research in this field for many years. Among various failure mechanisms, Rolling Contact Fatigue (RCF) has been one of the major problems of wheel-rail material which not only increases the cost of railway operation and maintenance activities, but also influences the safety and availability of the system [1]. There is an extensive amount of research related to fatigue nature of rail material under dynamic wheel loading condition in the presence or absence of various defects in rail materials. In spite of many researches on RCF phenomenon in wheel-rail materials, the exact mechanism of fatigue initiation is still not fully understood. This is mostly due to the complex geometrical shapes and impact loading conditions of wheel-rail materials which result in multiaxial cyclic stress-strain states rather than uniaxial [2]. Squats are some kinds of singular rail surface defects which occur on the rail top and cause excitation in wheel–rail impact condition leading to large dynamic forces. Due to the large dynamic impact conditions, the fatigue life of material could be affected as a result of squat appearance in rail. Some examples of the recent studies about the influence of track geometry irregularities on rolling contact fatigue can be found in [3-6]. The main objective of the present investigation is to predict the residual life of rail after squats appear by applying suitable fatigue initiation models to the results of numerical simulations acquired from dynamic Finite Element Analysis (FEA). To this end, the values of stresses and strains in rails caused by impact loading condition of the wheels are considered as the input data to determine the residual life of rail fatigue (number of cycles to crack initiation) during multiaxial loading conditions. This was accomplished by further development and improvement of the FE models which were previously employed by Zhao et al. (2011, 2012)[4, 5]. The FE model of wheel–rail rolling contact problem is improved by applying some kinds of squat-shaped defects on the surface during transient simulation to compare the fatigue life of rail in the presence or absence of such RCF defects. *

PhD Candidate. Section of Railway Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology. Stevinweg 1, 2628 CN, Delft, The Netherlands. [email protected]

337


The methods of fatigue analysis in this study are not new and they are previously employed by other researchers for different structures such as pressure vessels, gas turbines, aircraft, vehicles and components such as rotating disks, axles and crankshafts and even railway track components. However, the application of these techniques together with the solution of frictional rolling contact with a 3D finite element method and dynamic simulation is the novel. Prediction of the fatigue life of rail material under the impact loading of the squats seeds in rail material is another pioneering adventure of this study. Based on numerical simulations and field observations, it is shown by Li et al. [7] that squat development is closely related to the dynamic contact force which is excited by the squat’s seeds and which is determined by the local eigen characteristics of the vehicle-track system. That work gives an appropriate understanding about the squat growth process and the wave pattern that often follows squats. The corresponding estimation of rail life time regarding various stage of squat growth has not yet been studied. This study is an attempt to acquire an estimative judgment about the effect of squats seeds on rail fatigue life behavior.

2 Fatigue analysis of wheel-rail material

In rolling contact fatigue, the causes of cracks initiation can be either ratcheting or low cycle fatigue[8]. To investigate the fatigue mechanism, it is necessary to recognize the loading path in rolling contact. In the wheel-rail rolling contact problem, the rail is subjected to a nonproportional multiaxial stress state, which results in the variation of the principal stress and maximum shear stress-strain directions during a passage of the wheel [9]. Therefore, to predict the life of rail, it calls for a multiaxial stress criterion including non-proportional loading. In this study, a model based on the energy density and a critical plane approach is used to predict the life of rail at squats. In addition, a well-recognized criterion for ratcheting failure in rail material proposed by Kapoor is used to predict the life of ratcheting. Jiang and Sehitoglu [10, 11] proposed a multiaxial low cycle fatigue criterion for RCF phenomenon based on critical plane approach. In this criterion, it is postulated that both normal and shear components of stress and strain, on the critical plane, contribute to the damage of the material. The model is expressed as following equation: ∆𝜀

𝐹𝑃 = 〈𝜎𝑚𝑎𝑥 〉 + 𝐽. ∆𝜏. ∆𝛾 (1) 2 where ∆𝜀 is the normal strain range, 𝜎𝑚𝑎𝑥 is the maximum normal stress, ∆𝛾 is the shear strain range, ∆𝜏 is the shear stress range, J is a material-dependent constant and 〈 〉 denotes the McCauley bracket 〈𝑥〉 = (|𝑥| + 𝑥)/2. All the stress and strain quantities in Eq. 1 are on the critical plane where the fatigue parameter FP is the maximum (FP max ). Through a tensor rotation for the stress and strain, the maximum FP and the critical plane are determined by surveying all the possible planes at a material point. The first term in Eq.1 considers the mean stress effect. The proposed multiaxial fatigue model has the correct form to capture the synergism between the shear and normal stress components. The relationship between fatigue parameter and crack initiation life is described by the following equation:

(2) (𝐹𝑃 − 𝐹𝑃0 )𝑚 . 𝑁𝑓 = 𝐶 where N f is the crack initiation life corresponding to fatigue parameter FP, FP 0 , m and C are material fatigue properties obtained by best fitting base line experimental data. Fatigue damage is assumed to accumulate linearly. When the fatigue parameter FP is equal to or smaller than FP 0 , no fatigue damage is predicted. According to Kapoor (1994) [12], if a material displays a constant ratcheting rate, the ratcheting rate parameter is considered constant in the present work as well. In [12], the equivalent ratcheting plastic strain per cycle can be used as the ratcheting strain, it is expressed as: ∆𝜀𝑟 = �(∆𝜀)2 + (∆𝛾/√3)2 338

(3)


where ∆𝜀 and ∆𝛾 are the average ratcheting normal and shear strain per cycle in three dimensional stress states. The equivalent ratcheting strain is calculated on the plane with the largest shear strain accumulation. The ratcheting life can be estimated by 𝜀

𝑁𝑟 = ∆𝜀𝑐

(4)

𝑟

where 𝜀𝑐 is the critical strain for failure by ratcheting. The above formulations allow identifying the parameters that control the failure rates. It is suggested that the failure mechanisms of fatigue and ratcheting were independent and competitive so that the life of the component was governed by whichever would be expected to cause failure in the shorter number of cycles. If a material has a low ductility, low cycle fatigue failure could occur in finite number of cycles in the form of crack initiation and propagation. For a ductile material, extensive ratcheting strain can cause the extrusion of thin slivers at the surface. In either case, the combination of plastic ratcheting and fatigue can be used as a measure of failure prediction under these conditions [10, 11]. The rail steel grade that is generally being used in the Netherlands is the typical pearlitic steel, normal grade 900A. The material constants which are used in the J-S failure model can be generated by normal fatigue tests like tension-compression and torsion fatigue tests. In the current study, however, these constants are obtained from [13] for the equivalent pearlitic steel as: J= 0.32, m=2.50, C=1500000, a constant value of 𝜀𝑐 = 7.1 is assumed for the critical strain of rail material as it is proposed in [14], the value of FP0 is assumed to be 0.2 .

3 Numerical simulations 3.1

FE modelling of vehicle-track system

A three-dimensional finite element model of vehicle-track system with the implementation of elastic-plastic material and transient dynamic simulation was used to study the states of stresses and strains in rail material as it is schematically shown in Figure 1. The wheel is at its starting position. The wheel rotates with an initial speed to reach a distance which is long enough to relax the dynamic effects that are induced by the sudden loading of the wheel. The current model of vehicle-track interaction system is previously employed by Zhao et al. (2012) [4] to study wheel-rail impact and the dynamic forces at discrete supports of rails. That numerical model is further developed in the current study to understand the detailed mechanism of rail fatigue and to obtain the RCF crack initiation life in rail material. The 3D FE model has been validated by Zhao and Li (2011) [5] in the normal and the tangential contact solutions against Hertz theory and Kalker’s program CONTACT. Applying some field observations with measurement of axle box acceleration, the model on the vehicle-track interaction has also been validated by field monitoring tests [7] when it was applied to squats. The relevant structures of the vehicle and the track are both considered in the model together with their actual sizes. The primary suspension of the vehicle is considered in the simulation. In order to simulate the high-frequency dynamic behavior of wheel-set and track, the detailed flexibilities of the vehicle system is disregarded and the sprung mass of the vehicle is lumped into M c that is connected to the wheel set through the primary suspension K c and C c (see Figure 1 and 2). The track system is a model of typical ballasted railway track, in which the supports of the rail are composed of the fastenings, the sleepers, and the ballast. Only a half wheel set and a half straight track are modeled in view of the symmetry of the system.

339


Mc

M Wheel

Kc , Cc

Solution zone, location of squat

Kc , Cc

End of simulation

450mm

1435/2 mm Wheelset

Rail

700mm Kp , Cp

Fastening Ms

Sleeper

Kb , Cb

Rail Kp , Cp

Ballast

Fastening Ms

Kb , Cb 600mm

Symmetrical boundary condition

Initial wheel location

Sleepe Ballast

300mm

Figure 1 Schematic geometry of the vehicle-track FE model in longitudinal (left) and lateral (right) directions.

Figure 2 Left: 3D representation of the vehicle –track FE model. Right: Finest mesh pattern in the solution zone.

The FE model (Fig.2) is analyzed using explicit time integration scheme. The wheel–rail interaction is calculated using a surface-to-surface contact algorithm. The parameters of the FE model are listed in Table 1, in which the variables for defining nonlinearity of the wheel-rail materials are also given. In this work, the bilinear isotropic hardening model is employed for the materials of wheel and the rail. The stiffness and damping parameters in this table are taken from [4]. A constant friction coefficient of 0.6 is assumed for the simulations. The tangential loading of the contact is modeled by applying a traction force corresponding to the traction coefficient. Typical velocity of the trains in this work is considered 140 km/h based on the maximum operating speed in the Dutch railway network. Two wheel passages are simulated in the present work. Table 1 The values of parameters used in the numerical simulations

Parameters (unit)

Static wheel load, M c (kN) Wheel weight (kg) Rail weight per length (Kg/m) Sleeper mass M s (kg) Friction coefficient Traction coefficient Rolling speed (km/h) Stiffness of ballast, K b (kN/m) Damping of ballast, C b (N.s/m) Stiffness of rail pad, K p (kN/m) Damping of rail pad, C p (N.s/m)

Values

116.8 900 54.42 280 0.35 0.15 140 45000 32000 1300000 45000

Parameters (unit)

Stiffness of primary suspension, K c (kN/m) Damping of primary suspension, C c (N.s/m) Young’s modulus of wheel-rail material, E r (GP) Poisson’s ratio of wheel-rail material, ν r Density of wheel-rail material, ρ r (kg/m3) Yield stress- work hardened rail (GP) Yield stress- work hardened welds (GP) Tangent modulus of elastic-plastic rail (GP) Young’s modulus of concrete material, E c (GP) Poisson’s ratio of concrete sleeper material, ν c Density of sleeper material, ρ c (kg/m3)

340

Values 880 4000 210 0.3 7800 1.12 0.99 21 38.4 0.2 2520


3.2

Modelling of squats on the rail surface

Squats are some kinds of singular rail surface defects which occur on the rail top and cause excitation in wheel–rail impact condition leading to large dynamic forces. Due to the large dynamic impact conditions, the fatigue life of material could be affected as a result of squat appearance in rail. Some examples of the recent studies about the influence of track geometry irregularities on rolling contact fatigue can be found in [4, 6, 15]. To understand the effects of dynamic forces excited by squats and find possible estimation of fatigue life of rail material, squats are modeled in the FE simulation as the geometric irregularities in the rail top. According to the observations and measurements in [7], a squat seed in its early stage is a small irregularity in the rail top surface without cracks, as shown in Figure 3(a). Such an irregularity due to, e.g. indentation, burn initially a V-shape in its longitudinal-vertical profile, and gradually develops into a W-shape. Cracks develop in later stages, usually first in the middle ridge of the W-shape, as shown in Figure 3(b). Based on these observations, four idealized defects are studied(scenarios1, 2, 3 and 4) in the present work as the geometric models of the squats either in V or W shapes at different stages, besides a smooth rail as the reference situation (scenario 0). Figure 4 shows the simulated defect profiles (including two V-shape and two W-shape defects) in the longitudinal-vertical plane and one 3D representation.

a

b

Figure 3 Squats in different stages, (a) a V shape in early stage, (b) a severe W shape in later stage. 0.05

Rolling direction from left to right

-0.05

-0.10

S0-Basic scenario(smooth rail) S1-Defect_V1 S2-Defect_V2 S3-Defect_W1 S4-Defect_W2 -40

-20

0

20

40

60

Longitudinal position (mm)

80

100

0.1588

0.1584

120

0.58

0.60 long itudin al, x 0.62 (m)

0.015 0.010 0.005 0.000 -0.005 -0.010

(m )

0.56

0.64

La te ra l, y

-0.15

-0.20 -60

Rolling direction

0.1592

Vertical, z (m)

Vertical height (mm)

0.00

Figure 4 All simulated scenarios of rail squats. (a) Longitudinal–vertical profile at the middle of the running band. (b) 3D representation of one W-shape defect sample.

3.3

Strategy of Calculation for the residual life

First of all, the material properties need to be appropriately considered to anticipate the fatigue life. The purpose of the present work is to predict the residual life of the rail under operation. Therefore, the object of investigation is a hardened rail material rather than a new rail. In the present work, hence, the value of 990MPa is used as the yield stress of the hardened rail material. This yield stress is induced from hardness of rail steel which was measured in rails under operation. The fatigue parameters for the new rail steel can be found in literature, ([13] for instance). Using these parameters, the whole life can be predicted. The residual life can be obtained by the whole life subtracting the current operation time which can be measured by

341


considering the relationship between yield stress and operation time in the field or the one obtained by plastic law. In the present work, however, only the first step is considered. Secondly, the state of stress and strain at squats is changed with the growth of squats, which can excite impact force. It is assumed in the present work that the stress and strain are constant in each cycle for each defect model. The variation of the stress and strain is achieved by modeling different geometry of defects as squats representative for different stages. Lastly, for the position of fatigue, according to the observations and detections, usually, squats cracks initiate in the middle ridge of the W-shape at later stages. The point of cracks initiation is called critical point in the present work. Since stress level influence the cracks initiation, the critical point usually has the maximum von Mises stress in one wheel passage at W-shape stages. In this work, therefore, the critical point in two W-shape defects is investigated to predict the rest life of rail. To predict the residual life of squats at different stages and the influence of growth of squats on the residual life, the same point in V shape defects and smooth rail are considered to predict the life. The material characteristics for all scenarios of defects are assumed to be the same.

4 Results of simulations and fatigue analysis 4.1

Stress-strain responses

Numerical analyses are performed for all the defects models (one case for smooth rail and four cases for models with squats seed) that are defined in the previous section. In Figure 5(a), it shows the maximum von Mises stress variation with time step in the second wheel passage for the same point which is the critical point determined in S3 and S4. The critical point is located at the middle ridge of the W- shape defect. This location is agreed with that of observations. It is probed then to see whether the plastic deformation occurs or not on the critical point at different stages of the squat’s growth process. In the present case, plastic deformation occurs in S3 and S4, while there is no plastic deformation in the rest three cases. The gradual evolution of different stress components at the critical points of the rail at different time steps of a wheel passage is depicted in Figure 5(b). Hereby only the result of S4 scenario is demonstrated as a sample. This shows that the stress distribution diagrams are getting higher values when the wheel is approaching to the critical point. Also the stress alteration diagrams of the critical point are non-proportional and the material is exposed to a complex stress states. b

S0 S1 S2 S3 S4

1000 800 600 400

-500

σyy σzz σxx τyz τzx τyx

-1000 -1500

200 0

500 0

Stresses (MPa)

Von Mises stress (MPa)

a

720

740

760

780

-2000

800

720

740

760

780

Time step

Time step

Figure 5 Variation of stress in the element with critical condition (second cycle), (a) Variation of Von Mises stress for all cases of defects, (b) Variation of stress components for scenario S4 as a sample.

4.2

Calculation of fatigue crack initiation life

Considering the stress and strain histories for the critical point, a tensor rotation technique is used to obtain the stresses and strains on an arbitrary material plane for the critical point in second wheel passage. According to the fatigue model, Eq. (1), the variations of fatigue parameter FP with respect to the angles 𝜃1 and 𝜃2 for scenario S1 and S4 as two samples are 342


calculated, as shown in Figure 6. 𝜃1 and 𝜃2 in this figure are the two characteristic angles of the normal vector of the material plane, while the third angle can be calculated by spherical law of cosines. b

a

0.6

FP

FP

0.6

0.3

0.3

0 0

0 0

45

45 90

θ2

0

45

90

90

θ2

θ1

0

45

90

θ1

Figure 6 An illustration of the search of the critical plane for scenario S3 (a) and S4 (b).

By applying J-S fatigue damage model (Eq. 2), the fatigue crack initiation life N f is predicted on the possible crack plane determined above. Figure 7 shows the results of so called fatigue curves for crack initiation life variations with the possible critical plane for S1 and S4 scenarios. It can be apparently seen that the fatigue life parameter progressively increases with the reduction of fatigue parameter (FP). All of the individual spots in the fatigue curve are corresponding to the all possible critical plane at the critical point. The minimum value of N f by the way is the most critical case of fatigue crack initiation in material since it is equivalent maximum value of FP. The smallest N f value is remarked in the abscissa to demonstrate the number of cycles to crack initiation in rail material for two scenarios. Employing the same procedure, the fatigue life of rail in other scenarios is determined. a

1

b

S1

0.8

Fatigue parameter, FP

Fatigue paramter, FP

0.6 0.4

0.2 Nf = 2796,028

106

107

108

109

1010

1

S4

0.8

Nf

1011

0.6 0.4

0.2 Nf =1737,732

106

Number of cycles to fatigue, Nf

Nf

107

108

109

1010

1011

Number of cycles to fatigue, Nf

Figure 7 Results of fatigue parameter and fatigue life of material for two samples of FE simulations, (a) S1 scenario, (b) S4 scenario.

4.3

Calculation of ratcheting failure life

In the current work only two sequential cycles of wheel passage are simulated for the ratcheting analysis since the realistic material model of hardened rail under operation is considered. The ratcheting rate parameter is considered constant in the present work. Figure 8 shows the shear stress-strain and the normal stress-strain on the plane with the largest shear strain at each time step in two cycles for scenarios S3 and S4 in which the plastic deformation occurs at the critical point.

343


b 0.003

a 0.003 Maximum shear strain First cycle Second cycle

0.002

0.002

Strain

Strain

0.001 0.000 -0.001

0.001 0.000 -0.001

Normal strain First cycle Second cycle

-0.002

Maximum shear strain: First cycle Second cycle

Normal strain: First cycle Second cycle

-0.002

-0.003 720

750

-0.003

780

Time step

720

750

780

Time step

Figure 8 Variation of shear strain and normal strain in the critical point for two scenarios, (a) scenario S3, (b) scenario S4.

From this figure, it can be found that the critical point experiences incremental shear strain and normal strain between first and second loading cycles. The accumulative shear strain and normal strain at the critical point is transferred to the equivalent plastic strain using Eq.3. These values are listed in Table.2. Using these incremental values and employing Eq. 4, the number of cycles to crack initiation by ratcheting (N r ) is estimated as shown in Table 2. Similar to the normal fatigue life, the minimum value of N r in the whole simulation stands for the ratcheting life of rail material. It is worth noting that the ratcheting life N r is unlimited for smooth rail and both S1 and S2 defect scenarios since no plastic deformation occurs in these cases. Table 2 Calculation of ratcheting failure life for all cases

4.4

Defect models S0- smooth rail S1- defect V1 S2- defect V2 S3- defect W1 S4- defect W2

Max. plastic normal strain 0 0 0 3.0E-05 5.9E-05

Discussion, dominant fatigue mechanism

Max. plastic shear strain 0 0 0 2.0E-05 2.2E-05

Nr ∞ ∞ ∞ 207,346 117,643

The values of N f and N r are calculated for all scenarios using the same procedure. The results are summarized in Figure 9 which shows the crack initiation life of material governed either by fatigue or ratcheting failure in all scenarios. For defect S0, S1 and S2, there are no plastic deformation occurring, therefore the ratcheting strain is mathematically zero and the ratcheting life is unlimited. In these three cases therefore, the crack initiation life of rail material is dominantly affected by fatigue. Comparing the required number of cycles for fatigue in all scenarios together with ratcheting life in S3 and S4, it can be found that the crack initiation life of material is dominantly affected by the presence of squat as the surface defects. N f is reduced once the larger size squat geometry is applied on rail surface. N r for S3 and S4 is moreover reduced with the growth of squat. The overall rate of reduction in the number of cycles has been more drastic for the fatigue life especially for transition between scenarios S3 and S4. In contrast with the ratcheting life, a gentle decrease in fatigue life of material is drawn for S3 and S4 scenarios. Comparing the results of N f and N r in the two scenarios, it is worth noticing that the value of N f has been relatively higher than N r . The amount of discrepancy between the fatigue life and ratcheting failure for these two cases is the relative life (N f /N r ) of 11 and 9.84, respectively. In S3 and S4 cases, therefore, the crack was initiated by ratcheting mechanism rather than fatigue one.

344


Number of cycles to fatigue

4.000.000

3.000.000

Nf Nr 2814660 2796030 2617010 2287000

2.000.000

1737730

1.000.000 207,346

0 S0

S1

S2

S3

176,759

S4

Defect cases

Figure 9 Results of N f and N r for all defect scenarios.

5 Concluding remarks A dynamic finite element model was employed in this study to simulate the dynamic interaction of vehicle-track system at squats. Based on field observation and measurement, a series defects with different geometry representing a squat at different stages are modeled to predict the rest life of rail at different stages of a squat. Jiang’s fatigue model and Kapoor’s ratcheting model are then employed for prediction of fatigue life N f and ratcheting failure life N r in material. The approach in this study therefore considered the initiation of cracks under combined ratcheting plastic strain and multiaxial fatigue. The results of N f and N r were summarized and compared for different cases. Based on these results, following conclusions can be made. - For smooth rail and small squats seeds, crack initiation is mainly caused by fatigue while ratcheting phenomenon does not exist. When squats seeds become typical W-shape, stress increases in their middle ridge and ratcheting strain is found. - Predicted lives of RCF crack initiation and ratcheting failure for the scenario S3 were observed around 5.71×105 and 2.07×105 cycles of wheel passage, respectively, followed by 4.34×105 and 2.07×105 for scenario S4. The ratcheting life for these two cases can be theretofore considered as the life to crack initiation. - The fatigue life of the rail material is significantly decreased by the presence of severe Wshape squats on rail surface. The main cause for this is ratcheting.

Acknowledgement This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organization for Scientific Research (NWO). ProRrail is kindly acknowledged for providing part of the funding and technical supports.

References [1]. [2]. [3].

A. Fatemi and D.F. Socie, A Critical Plane Approach to Multiaxial Fatigue Damage Including out‐of‐Phase Loading. Fatigue & Fracture of Engineering Materials & Structures, 1988. 11(3): p. 149-165. K. VAN and M. Maitournam, Rolling contact in railways: modelling, simulation and damage prediction. Fatigue & Fracture of Engineering Materials & Structures, 2003. 26(10): p. 939-948. K. Karttunen, E. Kabo, and A. Ekberg, A numerical study of the influence of lateral geometry irregularities on mechanical deterioration of freight tracks. Proceedings of the 345


[4]. [5].

[6]. [7].

[8]. [9].

[10].

[11]. [12]. [13]. [14]. [15].

Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 2012. 226(6): p. 575-586. X. Zhao, Z. Li, and J. Liu, Wheel–rail impact and the dynamic forces at discrete supports of rails in the presence of singular rail surface defects. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 2012. 226(2): p. 124139. X. Zhao and Z. Li, The solution of frictional wheel-rail rolling contact with a 3D transient finite element model: Validation and error analysis. Wear, 2011. 271(1-2): p. 444-452. X. Zhao, Z. Li, and R. Dollevoet, The vertical and the longitudinal dynamic responses of the vehicle–track system to squat-type short wavelength irregularity. Vehicle System Dynamics, 2013. 51(12): p. 1918-1937. Z. Li, R. Dollevoet, M. Molodova, and X. Zhao, Squat growth—Some observations and the validation of numerical predictions. Wear, 2011. 271(1): p. 148-157. Y. Jiang and H. Sehitoglu, Rolling contact stress analysis with the application of a new plasticity model. Wear, 1996. 191(1): p. 35-44. J.W. Ringsberg, Life prediction of rolling contact fatigue crack initiation. International Journal of fatigue, 2001. 23(7): p. 575-586. Y. Jiang, A fatigue criterion for general multiaxial loading. Fatigue and fracture of engineering materials and structures, 2000. 23(1): p. 19-32. Y. Jiang and H. Sehitoglu, A model for rolling contact failure. Wear, 1999. 224(1): p. 38-49. A. Kapoor, A re‐evaluation of the life to rupture of ductile metals by cyclic plastic strain. Fatigue & fracture of engineering materials & structures, 1994. 17(2): p. 201-219. O. Onal, D. Canadinc, H. Sehitoglu, et al., Investigation of rolling contact crack initiation in bainitic and pearlitic rail steels. Fatigue & Fracture of Engineering Materials & Structures, 2012. 35(11): p. 985-997. M. Akama, H. Matsuda, H. Doi, and M. Tsujie, Fatigue crack initiation life prediction of rails using theory of critical distance and critical plane approach. Journal of Computational Science and Technology, 2012. 6: p. 54-69. T.S. Deng, X. Zhao, B. Wu, et al., Prediction of crack initiation of rail rolling contact fatigue. Applied Mechanics and Materials, 2013. 344: p. 75-82.

346


Ageing effects of alkali-silica reaction in concrete structures Rita Esposito1*, Max A.N. Hendriks12 (1) Delft University of Technology, Delft, The Netherlands (2) Norwegian University of Science and Technology, Trondheim, Norway Abstract: The alkali-silica reaction (ASR) is a long-term deterioration process, which produces a hydrophilic and expansive gel causing damage. The ASR acts on concrete as an ageing phenomenon, modifying the material on the basis of its stress state. Focussing on the mechanical degradation of concrete, estimated trends for the mechanical properties in free expansive affected concrete are presented. These show the result of recent research which collected and statistically analysed laboratory tests on 54 concrete mixes performed by 11 authors. Comparable findings are also obtained through a multiscale material model, which aims to capture the micro and macro aspects of the problem. The ASR-affected concrete is seen as an evolving material, whose state should be followed over time taking into account chemomechanical coupling. Keywords: Alkali-Silica Reaction (ASR), degradation, mechanical properties, damage assessment, multiscale modelling

1 Introduction The Alkali-Silica Reaction (ASR) is a harmful deterioration process, which starts at aggregate level, with the combination of silica in the aggregates and alkali in the cement paste. Its product is a hydrophilic gel which swells and causes damage up to macro level, possibly influencing the integrity and capacity of the structure. The expansion process is directly related to the mix properties (e.g. aggregate and cement type, aggregate size, etc.) and to the environmental conditions. Moreover the stress state of the material has an influence on the redistribution of the gel in the concrete, thus on the swelling, and consequentially on the damage propagation. The reaction influences the performance of the material by leading to a relevant degradation of the mechanical properties. Recently the authors studied the influence of ASR on the mechanical degradation of concrete, by analysing available literature data regarding laboratory tests on free expansive ASR-affected concretes [1]. The collection of data included 11 authors, actually groups of co-authors, who tested 54 different concrete mixes. Considering the observed expansion and expansion rates as given, the specific aim is to find a trend between the deterioration of the mechanical properties and observed concrete swelling due to ASR, independent of the wide variety of concrete mixes used and experimental conditions applied. The research highlighted that the evolution of elastic modulus, both static and dynamic, is the best indicator for the identification and progression of ASR in concrete. Conversely, the evolution of compressive strength might veil ASR damage. The splitting test is to be preferred to capture the influence of ASR on the tensile behaviour of concrete. The research highlighted that the ASR-affected concrete appears as a substantially different material with respect to sound concrete and the known engineering strength-stiffness relationships, developed for the latter, cannot be adopted in the structural assessment procedures. Furthermore, considering that a proper material characterization is extremely relevant, a multiscale material model [2, 3] has been selected to perform structural analyses. The model accounts for the micromechanical changes provoked by the ASR and its swelling. By employing this approach a more fundamental model is adopted, which is able to capture the micro and macro aspects. The ASRaffected concrete is seen as an evolving material, the state of which should be followed over time taking into account chemical and mechanical loading conditions.

*

Rita Esposito, Delft University of Technology, Delft, The Netherlands e-mail: [email protected]

347


In this paper the attention is focussed on the mechanical degradation of concrete induced by ASR. First the recent findings regarding the statistical analysis of literature experimental date are presented. It includes different laboratory tests performed on free expansive ASR-affected concrete, to determine the evolution of concrete mechanical properties. The results are compared with a numerical simulation performed by adopting the proposed multiscale material model.

2 Mechanical degradation of concrete provoked by ASR The alkali-silica reaction in concrete is a long-term deterioration process, whose consequences are strongly related to the environmental and mechanical state of the material. The coupling between the chemical load, provoked by the ASR gel swelling, and the mechanical load is the key point which make the laboratory tests different from the real behaviour of a ASR-affected structures. By understanding first the mechanical effect of ASR and secondly the coupling phenomenon, it is possible to explain the behaviour of ASR-affected concrete structures. In order to quantify the mechanical degradation of concrete provoked by ASR, an extensive research was made [1]. Available literature experimental data [1, 4-12] was collected and statistically analysed to determine trends in degradation behaviour. The majority of the authors studied the degradation of the compressive strength (10 authors out of 11) and of the static elastic modulus (9 authors out of 11). The tensile behaviour was studied by 7 authors out of 11, who preferred the splitting tensile strength, above the modulus of rupture and the direct tensile strength. Non-destructive tests for the determination of the dynamic elastic modulus were chosen by 4 authors out of 11. The data were statistically analysed applying a normalization procedure: each property was normalized with respect to its reference value, which was calculated at an expansion equal to 0.05%. After the normalized property values were plotted versus the concrete expansion values, curve fitting procedure was applied. The fitting included two degradation laws: the S-shaped curve, which is a revised version of the law proposed by Saouma et al. [13], and the piecewise linear curve. In Figure 1a the best curve fitting results are presented together with the error band equal to 2σ. The piecewise linear curve is chosen for the description of the compressive strength behaviour, while the S-shaped curve was chosen for the other properties. The tensile strength behaviour is reported in terms of splitting test results. Both static and dynamic elastic modulus data are considered for the description of the stiffness degradation. The curve fitting defines the elastic modulus as the best indicator of ASR signs in concrete, in fact it presents a relevant deterioration already at early expansion; moreover its degradation rate is the fastest one. For high expansion values (ε > 2.00%) the residual stiffness is 20% of the reference value. Conversely, the compressive strength behaviour is described with an initial gain of 15% and a maximum reduction of 46%. However the estimation error is high, around 13%. The tensile behaviour appears to be well described by the splitting test results. Its deterioration starts at higher expansion values with respect to the elastic modulus. Its residual value is 64%. In Figure 1b the differences in degradation behaviour are shown in an alternative way. When the elastic modulus reaches 85% of its original value, both strengths reduce with a similar rate, but still slower than the degradation rate of the elastic modulus. At a normalized value of 0.50 for the elastic modulus, the normalized splitting strength reaches an asymptotic value of 0.60. The compressive strength is subjected to a drastic deterioration for a normalized value of the elastic modulus of 0.20. In engineering it is common practice to express the stiffness and the tensile strength of sound concrete as a function of its compressive strength. Using the strength-stiffness relationships proposed by Model Code 2010 [14], it was found that for sound concrete, the degradation rate of compressive and tensile strength is lower than the one for the elastic modulus (Figure 1b). This demonstrated that for ASR-affected concrete, it is not allowed to use the engineering strength-stiffness relationships to determine the elastic modulus and the tensile strength from the measured compressive strength.

348


(a)

(b)

Figure 1 Results of the statistical analysis: (a) Evolution law for elastic modulus, compressive strength and splitting tensile strength as a function of concrete expansion induced by ASR; (b) Relation between normalized elastic modulus and normalized strengths.

3 Multiscale modelling approach Since ASR was first discovered, many modelling strategies were developed to assess the behaviour of affected structures. Starting with rather straightforward engineering methods, the research moved to the material characterization adopting models on different scales. Micro-mechanical models focused on the reaction kinetics and mechanism, while macroscopic approaches tried to understand the damage effects observed in structures. Considering that a micro-mechanical material characterization is relevant for the description of the overall structural behaviour, a multiscale material model [2] is adopted (Figure 2), in order to account for the strong interaction between micro-mechanical aspects and the complex macro-mechanical state. The approach found its basis in the work of Charpin and Ehrlacher [15] and of Lemarchand et al. [16]

Figure 2 Modelling procedure for structural analyses.

The properties of concrete are evaluated considering a Representative Elementary Volume (REV), as reported in Figure 3a. The sound concrete is modelled as a heterogeneous material (Figure 3b) composed of aggregates and microcracks embedded in the cement paste. Each material is behaving elastically. The aggregates are modelled as spheres. The microcracks are modelled by three orthogonal families of penny-shaped inclusions. The alkali-silica reaction is simulated by changing the microstructure. The chemical process starts at aggregate level by consuming the silica available and forming the gel, which is modelled as spherical inclusions into the aggregates (Figure 3b). It is assumed that the gel has a volume bigger than the volume of the eroded aggregate, therefore a pressure is generated. The gel flows into the cement

349


paste by filling the microcracks around the aggregates. Eventually the pressure is high enough to generate damage in the system (Figure 3c). The chemical process is governed by the internal variable α which, for each family of aggregates (inclusion type 2), determine volume of aggregate material transformed in gel phase. It ranges between 0, not affected aggregates, and 1, completed eroded aggregates. The internal variable is also responsible for the link between time and expansion evolution. The reader can refer to [2] and [3] for further details. The damage evolution is formulated in the framework of linear fracture mechanics by employing an energy-based damage criterion. Both the external mechanical loads and the internal pressure contribute to the description of the damage, which leads to an increase of the crack radii. The effective properties of the medium are analytically evaluated by the Mori-Tanaka homogenization scheme.

(b)

(a)

(c) Figure 3 Micro-mechanical model of the REV: (a) 3D representation; (b) Sound and ASR-affected concrete; (c) Evolution of microstructures for ASR-affected concrete in free expansion conditions.

The model was adopted to simulate the degradation of mechanical properties in ASR-affected concrete samples in free expansion conditions. A sound concrete having an elastic modulus of 32188 MPa and a tensile strength of 3.20 MPa is considered. The example uses only one family of inclusions type 2 (aggregates), with a radius equal to the weighted average of the available aggregate sizes (d = 0 - 8 mm). Three families of inclusions type 3 (cracks), with the same initial crack radius and equally distributed in the three orthogonal crack planes are considered. The elastic moduli of aggregate and cement paste phases are assumed equal to 25000 and to 70000 MPa, respectively. The ASR gel bulk modulus is 1788 MPa. The compressive strength of the sound concrete is estimated, throughout a simulation of an uniaxial compressive test, equal to 30.43 MPa. In Figure 4 the numerical results are reported. With a sample that is free to expand and identical distribution of the three crack families, the macroscopic strains evolve isotropically. The characteristic S-shaped expansion curve is obtained (Figure 4a). The crack propagation follows a similar trend: in the first stage the crack radii are constants, because the internal pressure is still low and the gel is filling the available space in the concrete, afterwards the cracks isotropically propagate. In Figure 4b the degradation of the mechanical properties is reported as a function of the internal variable. They are evaluated at regular intervals of the internal variable, α, by simulating uniaxial

350


tensile and compressive tests. The properties are normalized with respect to the values of undamaged sound concrete properties (elastic modulus 32188 MPa, tensile strength 3.20 MPa, compressive strength 30.43 MPa). The properties start to degrade immediately, and present an asymptote for high values of the internal variable, α. The elastic modulus is the property which degrades fastest and presents the lowest residual value. The compressive strength follows a similar trend. The model predicts a lower degradation in terms of tensile strength.

(b)

(a)

Figure 4 Numerical results: (a) Evolution of macroscopic strain and crack radii; (b) Degradation of mechanical properties.

A through comparative study between the experimental and numerical results has still to be performed. However, preliminary comparison between Figure 1a and Figure 4b shows that the model is able to describe the degradation of the elastic modulus. Moreover the different degradation behaviour between elastic modulus and tensile strength is observed as well. However, the numerical results in term of compressive strength are not in agreement with experimental findings: the initial increase in compressive strength is not simulated and the predicted residual strength is too low. It is emphasized that the numerical prediction in Figure 4 is shown as a function of the internal variable and not in terms of concrete expansion. Moreover, the simulation is performed by assuming only one family of aggregates and reasonable (but disputable) input parameters. A further validation of the model is needed.

4 Conclusions The alkali-silica reaction is a long-term deterioration process which can slowly, but significatevly, influence the performance of concrete structures. The hydrophilic expansive gel, which is formed by the reaction between alkali in the cement and silica in the aggregates, can lead to concrete expansion with subsequent material deterioration. Recently the authors collected and analysed literature data regarding the degradation of mechanical properties in free expansive ASR-affected concrete. The data collation analysed 54 different concrete mixes studied by 11 authors. The majority of the research was focus on the estimation of compressive strength and static elastic modulus. The tensile behaviour of concrete was evaluated mainly through splitting tensile strength tests. Some authors adopted non-destructive test methods to estimate the dynamic elastic modulus. The data, expressed as normalized property values versus concrete expansion, were analysed in terms of curve fitting, adopting the S-shaped or the piecewise linear curve. The elastic modulus, both static and dynamic, results as the best indicator for the identification of ASR in concrete, showing relevant degradation at lower expansion values. Moreover its degradation rate is the fastest and it degrades by 90% of its original value. The behaviour of compressive strength, widely investigated probably because it is the principal test method adopted in the structural assessments, shows a non-monotonic trend. It displays an initial gain for expansion values lower than 0.15% and a subsequently reduction down to approximately 50%. The tensile strength, well described

351


by the splitting test results, shows a delay in the degradation with respect to the elastic modulus. However a similar deterioration rate for expansion values between 0.10% and 0.50% is observed. It reaches a maximum reduction of 64%. Comparing the degradation behaviour of compressive and splitting tensile strength with respect to the elastic modulus, a non-linear relation is observed. As a result, the ASR-affected concrete appears as a substantially different material and the known engineering strength-stiffness relationships, developed for sound concrete, cannot be adopted. In order to simulate the mechanical degradation of concrete due to ASR in free expansive concrete, a multiscale material model is adopted. The model aims to couple the chemical and mechanical effects in order to characterize the ASR-affected concrete in the structures, which is considered as an evolving material. The concrete is modelled at micro level as a multiphase material in which aggregates, cracks and gel formations are considered as embedded inclusions in the matrix, that is the cement paste. The development of the gel involves the aggregate’s erosion and swelling, together with a possible mechanical load, which can lead to crack propagation. The overall mechanical properties of concrete are analytically determined in agreement with the Mori-Tanaka homogenization method. This theory defines the average stress and strain state of the concrete as well as the effective stiffness tensor, which depends on the amount, the shape and the orientation of the inclusions. Therefore the model can be mechanically defined as a three-dimensional smeared approach. The damage state variables are directly linked to the microscopic crack families and their propagation is based on the principles of linear fracture mechanics. A preliminary comparison of the numerical results with the findings of the statistical analysis shows that the model is able to capture the degradation behaviour in terms of stiffness and tensile strength. The elastic modulus is confirmed as damage indicator, presenting the fastest degradation rate and the lowest residual value. Moreover the different degradation behaviour between elastic modulus and tensile strength is predicted as well. However, its performance in terms of compressive strength degradation should be improved. In conclusion, the ASR-affected concrete appears as an ageing of the material where the degradation is strongly related to the stress state. It does not follow the known engineering strengthstiffness relationship, derived for sound concrete. The multiscale material model, which aims to characterize the ASR-affected concrete in structures, appears as a complementary tool adoptable in combination with laboratory tests in the assessment procedure.

Acknowledgments This work is part of the project “Performance Assessment Tool for Alkali-Silica Reaction” (PAT-ASR, http://pat-asr.blogspot.nl/), which is developed in the context of the IS2C program (http://is2c.nl/). The authors wish to express their thanks to the Dutch National Foundation (STW), the Dutch Ministry of Infrastructures and the Environment (Rijkswaterstraat), SGS and TNO DIANA BV for their financial support.

References [1] [2]

[3]

[4] [5]

[6]

Esposito, R., et al. (2014) The Influence of Alkali-Silica Reaction on the Mechanical Degradation of Concrete (in preparation). Esposito, R. and M.A.N. Hendriks (2013) Multiscale Material Model for ASR-affected Concrete Structures. in XII International Conference on Computational Plasticity.Fundamentals and Applications (COMPLASXII). Barcelona, Sapin. Esposito, R. and M.A.N. Hendriks (2014) Modeling of Alkali-Silica Reaction in Concrete: a Multiscale Approach for Structural Analysis. in Computational Modeling of Concrete and Concrete Structures (EURO-C). St. Anton am Alberg, Austria: CRC Press Taylor & Francis Group. Swamy, R.N., ed. The Alkali-Silica Reaction in Concrete. 1992, Van Nostrand Reinhold, New York. Larive, C. (1998) Apports combinés de l’expérimentation et de la modélisation la comprehénsion de l’alcali-réaction et de ses effets mécaniques. Ph.D. Dissertation, Laboratoire Central des Ponts et Chaussées (LCPC). Monette, L.J., N.J. Gardner, and P.E. Grattan-Bellew (2002) Residual Strength of Reinforced Concrete Beams Damaged by Alkali-Silica Reaction- Examination of Damage Rating Index Method, ACI Mater. J., 99(1).

352


[7] [8]

[9] [10] [11] [12] [13] [14] [15] [16]

Ahmed, T., et al. (2003) The effect of alkali reactivity on the mechanical properties of concrete, Construction and Building Materials, 17(2): 123-144. Multon, S. (2004) Évaluation expérimental et théorique des effets mécaniques de l'alcali-réaction sur des structures modèles. Ph.D. Dissertation, Université de Marne-la-Vallée (in collaboration with LCPCEDF). Ben Haha, M. (2006) Mechanical Effects of ASR in Concrete Studied by SEM-Image Analysis École Polytechnique Fédérale de Lusanne (EPFL). Giaccio, G., et al. (2008) Mechanical behavior of concretes damaged by alkali-silica reaction, Cem. Concr. Res., 38(7): 993-1004. Sargolzahi, M., et al. (2010) Effectiveness of nondestructive testing for the evaluation of alkali–silica reaction in concrete, Construction and Building Materials, 24(8): 1398 - 1403. Lindgård, J. (2013) Alkali-silica reaction (ASR) - Performance testing. Ph.D. Dissertation, Norwegian University of Science and Technology. Saouma, V. and L. Perotti (2006) Constitutive model for alkali-aggregate reactions, ACI Mater. J., 103(3): 194-202. CEB-FIP (2012) Bulletin d'Information 65&66 - Model Code MC2010 Final Draft. Lausanne, Switzerland: International Federation for Structural Concrete (fib). Charpin, L. and A. Ehrlacher (2012) A computational linear elastic fracture mechanics-based model for alkali-silica reaction, Cem. Concr. Res., 42(4): 613-625. Lemarchand, E., L. Dormieux, and F.J. Ulm (2005) Micromechanics investigation of expansive reactions in chemoelastic concrete, Philosophical Transactions of the Royal Society a-Mathematical Physical and Engineering Sciences, 363(1836): 2581-2602.

353


A Poro-mechanical Approach for Assessing the Structural Impacts of Corrosion in Reinforced Concrete Members Esayas Gebreyouhannes1*, Takahashi Yuya2, Koichi Maekawa 3 (1) Addis Ababa University, AAiT, Addis Ababa, Ethiopia (2) Department Civil Engineering, The University of Tokyo, Tokyo, Japan (3) Professor, Department Civil Engineering, The University of Tokyo, Tokyo, Japan Abstract: Due to the larger volume of corrosion product than the parent steel, corrosion of steel bar in reinforced concrete members causes expansion pressure that induces tensile stresses. When the tensile stress exceeds the tensile strength, cracking occurs in concrete, thus jeopardizing the durability of the RC member. In this study, corrosion-induced crack initiation and propagation are investigated numerically using a poro-mechanical approach, by taking into account the formed corrosion gel and its interaction with concrete skeleton, pore systems and crack spaces. The effects of corrosion gel viscosity and its intrinsic permeability on crack initiation and propagation are studied using nonlinear FE sensitivity analysis and specific values are discussed by comparing with experimental results. Accordingly, a simple model has been proposed to consider the penetration of corrosion gel into capillary pore systems and migration to crack spaces. The model explicitly considers the dissipation of the expansion pressure caused by the corrosion product by considering the kinematics of corrosion-gel to the steel-concrete interfacial space, capillary pores and crack gaps. The reliability of the proposed computational platform is scrutinized in use of experimental results with varying water to cement ratio, cover depth and corrosion rate. It is found that poro-mechanics is a promising direction to realistically tackle the structural impacts of corrosion as functional aging. Keywords: Corrosion, poro-mechanics, Corrosion gel, crack spaces, pore

1 Introduction Corrosion of embedded reinforcing bars is one of the main causes for deterioration of structural concrete. The corrosion induces expansive pressure causing tensile stresses in surrounding concrete. Its effect on RC includes loss in load carrying capacity, reduction and eventual loss of bond between concrete and corroding steel, and cracking of concrete cover. These greatly affect the serviceability and strength of RC structures leading to accelerated ageing. In the maintenance scheme of structures, it is necessary to define a limit state, to rationally decide the time for repair. Currently, there are various definitions of limit state; as the initiation of corrosion [1]; as first concrete cover cracking [2, 3]; based on a certain limit of crack width [4,5] beyond which the cracks become unacceptable due to serviceability requirements. While each of these definitions has their own rational, the end results may significantly vary depending on the corrosion rate, cover depth, microstructure, concrete strength and nature of corrosion product. Thus, analysing the behaviour of RC structures after corrosion initiation i.e., predicting the time of first cracking, crack growth, the time when cracks become unacceptable, is of importance for selecting efficient maintenance and repair strategies [6]. To date, several analytical and experimental studies have been carried out on the corrosioninduced crack initiation and propagation in RC structures. One of the earliest analytical models, [7], considers concrete around a corroding reinforcing bar as a thick-walled cylinder. However, experiments and field observations indicated that the model significantly underestimates the time to the first cracking [2]. Without altering the conceptual approach, a number of modifications by introducing empirical coefficients which adjusted model predictions to *

Gebreyouhannes Esayas, Assistant Professor, Addis Ababa University, Ethiopia [email protected]

354


experimental factors have been proposed [2, 8, 9]. The main limitation of the models was, they did not account the interaction between concrete microstructure, cracks, and corrosion product. The process of corrosion involves soluble species that can dissolve in the concrete pore solution and subsequently migrate or diffuse through the cement paste matrix away from the corroding steel [10]. The pores near the corroding steel act as storehouses, hence delaying stress build-up that lead to damage of the concrete cover [11,12]. Indeed, recent theoretical models for predicting time to initial cover cracking have incorporated ‘free-expansion’ to account for the time needed for corrosion products to fill the porous cement paste around the rebar [9, 13, 14] Despite several attempts, the fundamental knowledge on the impact of corrosion on structural concrete is still lacking. One, the nature of the corrosion product influences the rate of ageing. Second, the studies consider a buffer zone where the initial corrosion products must fill before inducing expansive stress. The penetration of corrosion gel into capillary pores is a simultaneous process with the build-up of stress. Third, most of the studies to date are based on the accelerated test. However, the extent of penetration into the capillary pores is dependent on the corrosion rate. Importantly, existing analytical models are trying to tackle the problem using the single phase approach in use of the thick-walled cylinder analogy for a corroding reinforcing bar. The single-phase approach lacks flexibility in terms of addressing the realistic effects of the nature of the corrosion product, such as viscosity and intrinsic permeability. The multi-phase approach for concrete has been successfully applied for the assessment of kinematics of water in concrete subjected to high cycle loading [15]. In this study, the authors aim to propose and assess the versatility of poro-mechanical approach for the simulation of RC with corroded steel based on the multi-phase system with pore and skeleton. It is also the aim of the current study to assess, the effects major influencing factors that dictate the kinematics of corrosion gel such as corrosion-gel viscosity and its intrinsic permeability for the crack initiation and crack propagation of RC members.

2 Multi-phase Modelling of Corrosion 2.1

Pore-skeleton system

The process of corrosion involves soluble species that can dissolve in the concrete pore solution and subsequently migrate or diffuse through the cement paste matrix away from the corroding steel [10]. Basically, the kinematics of the corrosion gel plays a central role for crack initiation and crack propagation of a corroded RC system. This phenomenon is analogous to that of liquefaction, and the authors tackled the simulation of corrosion gel migration in capillary pores and crack gaps by applying the multi-phase model for the liquid-solid composites using Biot’s theory [16]. This new approach is integrated in the full 3D multi directional fixed crack model of RC with path and time dependent constitutive laws [17]. The governing equation for corroding RC in this study is based on Biot’s theory. RC with corroding steel is treated as an assembly of concrete skeleton, pore media (Capillary pores, air voids and cracks), corrosion product (solid and liquid component), water, and uncorroded steel. The liquid-solid interaction between the corrosion gel and concrete skeleton is considered, by defining the motion of corrosion gel with respect to the relative displacement from the concrete solid matrix. The movement of corrosion gel is denoted by U i is expressed by the average displacement of concrete skeleton u i and the substantial relative displacement of the corrosion gel phase w i as: 𝑈𝑖 = 𝑢𝑖 +

𝑤𝑖 𝑛

(1)

Biot’s original concept assumes isotropy of the soil-pore water system with particles surrounded by a liquid matrix. Thus the total stress (σ ij ) of concrete-liquid system as: 𝜎𝑖𝑗 = 𝜎𝑖𝑗 ∗ + 𝛿𝑖𝑗 𝑝

355

(2)


where, p is the isotropic pore liquid pressure and 𝜎𝑖𝑗 ∗ is the effective stress tensor defined on the concrete skeleton. This assumption of isotropy holds before crack initiation and be applied to address the migration of corrosion gel in to the capillary pores, interfacial zone and air-voids. The penetration of corrosion gel in to cracks can be formulated using the dynamic equilibrium equations with the term of dragging forces rooted in the permeability of corrosion gel in each direction –i through concrete pore and crack gaps as, 𝜎𝑖𝑗,𝑗 = 𝜌�𝑢̈ 𝑖𝑗 − 𝑔𝑖 � + 𝜌𝑓 𝑤̈𝑖 𝑝,𝑖 = 𝜌𝑓 �𝑢̈ 𝑖𝑗 − 𝑔𝑖 � + 𝜌𝑓

𝑤̈𝑖 𝑛

+

(3)

1 𝑤̇ 𝑘𝑖 𝑖

(4)

where, the last term of Eq. (4) defines the dragging action of corrosion gel and often referred as Darcy’s law of motion.

Figure 1 Schematic illustration for cracked constitutive model

The effective stress in the skeleton can be computed by using the cracked concrete constitutive model illustrated in Figure 1. The mechanical modelling of the liquid corrosion gel can also be expressed as; �𝑓 �𝑤𝑖,𝑗 + 𝜀𝑖𝑗 � 𝑝=𝐾 �𝑓 = �1−𝑛 + 𝐾 𝐾𝑠

𝑛 � 𝐾𝑤

(5)

(6)

where, 𝐾𝑠 and 𝐾𝑤 are bulk stiffness of concrete matrix and corrosion gel, and 𝑤𝑖,𝑗 + 𝜀𝑖𝑗 is the volumetric strain of condensed water inside cracked concrete. The overall framework of combined solid skeleton and condensed liquid water mechanics is summarized in [18].

2.2

Crack initiation due to corroding steel bar

After steel depassivation, corrosion products are formed at steel-concrete interface, having greater volume (two to six times) than that of the original steel. The formation of the corrosion product then induces an expansive pressure on surrounding concrete generating stresses in the concrete. When this stresses exceed the tensile strength of concrete crack will occur. The stress development due to evolution of corrosion is a complex phenomenon. This is because; the corrosion product is composed of liquid and solid components, the nature of corrosion product is dependent on the steel chemical composition and local environment, and migration of the liquid corrosion gel is strongly linked with the capillary pores of concrete. The existence of certain thickness (12.5µm [2]; 40µm [13]; 120µm [6]) ‘porous zone’ around the steel that must be filled with corrosion products prior to any development of expansive stress is proposed. 356


Experimentally up to 200µm thick penetration of corrosion products away from the steelconcrete interface is also observed [19]. For the purpose of the current study, a simple model is proposed to consider the penetration of corrosion products to the capillary pores, microcracks and air voids while stresses are evolving, based on the previous research results. Unlike the existing models, here a simultaneous penetration of corrosion products and build-up of stress is considered (see Figure 2). The mathematical formulation of the proposed model is expressed as: 0.2 𝐶𝑎𝑝_𝑔𝑒𝑙

Portion of corrosion product associated with stress build-up

∙ 𝑉𝑐𝑝 𝑓𝑜𝑟 𝑉𝑐𝑝 ≤ 𝐶𝑎𝑝_𝑔𝑒𝑙 𝑉𝑐𝑝_𝑠 = � 0.2 + 20 ∙ �𝑉𝑐𝑝 − 𝐶𝑎𝑝_𝑔𝑒𝑙� 𝑓𝑜𝑟 𝑉𝑐𝑝 > 𝐶𝑎𝑝_𝑔𝑒𝑙

(7)

1.0 0.8 = 0.2 + 20(V - Cap_gel) cp 0.6 Cap_gel = 0.03 + 0.02(-corrosion rate +1.0)

0.4 0.2

= 0.2 / Cap_gel * Vcp

0.0

Cap_gel

Cap_gel+0.04

Total volume of corrosion product Figure 2 Simultaneous penertation of corrosion gel with build-up stress model

where, Vcp is specific corrosion volume per unit volume of an element, Vcp_s is portion of the specific corrosion volume associated with stress, Cap_gel is the limit for the rapid penetration of corrosion product. Though there are no quantitative supporting experimental data, the penetration of corrosion products to concrete matrix must depend on the corrosion rate. As a matter of fact the experimental data in [20] indicate the presence of this corrosion rate dependency. The initiation of crack was significantly delayed for slower corrosion rate, indicating more damping effect of the expansive stresses due to corrosion. In the current study, with the intention of addressing the mechanism, the parameter Cap_gel is assumed to be dependent on the corrosion rate as:

2.3

𝐶𝑎𝑝_𝑔𝑒𝑙 = 0.03 + 0.02 ∙ (1.0 − 𝑐𝑜𝑟𝑟𝑜𝑠𝑖𝑜𝑛 𝑟𝑎𝑡𝑒)

Crack propagation due to corroding steel bar

(8)

Once crack is initiated, the expansion force is highly relaxed due to the penetration of corrosion gel in to cracks. In previous experimental studies, a linear relationship between the amount of corrosion and the crack width has been observed. Based on this observation a number of formulas relating the crack width and the amount of corrosion have been proposed. These relations provide valuable information for assessing corrosion associated risks in RC structures. From, mechanics point of view the crack propagation must depend on several parameters such as cover depth, concrete strength, nature of the corrosion product, distribution of cracks etc. Here, crack propagation is modeled by using the dynamic equilibrium equations with the term of dragging forces rooted in the permeability of corrosion gel in each direction. Thus the crack initiation and crack propagation of a corroding RC system can be studied by varying the major influencing factors such as viscosity for intrinsic gel migration and its intrinsic permeability. This sort of flexibility is only possible on a poro-mechanical based analytical platform. 357


After cracking of concrete the intrinsic permeability of corrosion gel increases rapidly. To account for such discontinuous behaviour due to cracking, an increase of 1 to 2 orders in permeability is considered. Tentatively, the model expressed in Eq. 9 is used for the intrinsic permeability of corrosion gel after cracking of concrete. 𝑍 𝑓𝑜𝑟 𝜀 ≤ 𝜀𝑐𝑟 𝑘= � 4 4 𝛽 ∙ 𝑍(1.0 + (𝜀 ∙ 10 ) ) 𝑓𝑜𝑟 𝜀 > 𝜀𝑐𝑟

(9)

where, Z is the intrinsic permeability of corrosion gel for uncracked concrete; 𝛽 is amplification factor for cracked concrete. The value of 𝛽 here is assumed to range from 10-100.

2.4

Experimental investigation (Bung Hwan Oh et al 2009)

Oh et al. experimentally investigated the critical corrosion amount to cause cracking of RC members. They carried out systematic experimental investigation by varying the concrete cover depth and concrete strength. In the experiment, the evolution of surface strain is monitored with the progress of corrosion using the accelerated test. The details of the experiment are indicated in Table 1 and Figure 3. More details can be referred from [21]. Table 1 experimental detail Oh et al (2009) f t (MPa)

E (MPa)

n

ε cr (x10-3)

0.55

fc (MPa) 27.5

3.10

24821

0.18

0.125

S2

0.45

40.3

3.93

30019

0.18

0.131

S3

0.35

44.3

4.12

31481

0.18

0.131

Test Series

w/c

S1

Figure 3 RC member with corroding steel at the middle portion Oh et al (2009)

For the purpose of parametric study, the experiments [21] are simulated using the poromechanical approach. Overlapping elements are used in the FE discretization, to represent the skeleton and pore-liquid system. The incremental corrosion product is given as forced increment. The nodal displacements of the two ends of the strain gage are computationally obtained and the corresponding evolution of strain is compared with that of the experimental results. A parametric study is conducted by varying the intrinsic permeability of corrosion gel (Z from 1.0*1E-12 to1.0*1E-12 for uncracked concrete) and the viscous parameter for migration of corrosion gel (T from 0.0001 to 0.01). The results are indicated in Figure 4. For all cases the volume expansion factor for the corrosion product is assumed to be 3.0. Figure 4 (a,c & e) indicate the effect of intrinsic permeability for a constant viscous parameter for migration of corrosion gel (T from 1.0*1E-02). The value 0.01 for the viscous parameter for migration of gel is consistent with the parameter for the migration of ASR gel, which will be discussed in accompanying paper by the authors. The sensitivity analysis results indicate that the evolution of concrete surface strain is remarkably influenced by the permeability of the corrosion gel. The experimental results for the strain evolution fall in between the corresponding values for the permeability values of 1.0*1E-10 and 5.0*1E-10. These values are 1 to 2 orders lower than the permeability of water. 358


Experiment S2-C5 S2-C5_Z=1.0E-10_T=1.0E-02 S2-C5_Z=1.0E-12_T=1.0E-02 S2-C5_Z=1.0E-09_T=1.0E-02 S2-C5_Z=5.0E-10_T=1.0E-02

700

500

600

Strain, micron

Strain, micron

600

400 300 200 100 0

0

2

400 300 200

4

6

8

10

12

0

14

2

4

6

8

10

12

14

Corrosion % b) Effect of gel migration S2-C5


600

Experiment S2-C4 S2-C4_Z=1.0E-09_T=1.0E-03 S2-C4_Z=1.0E-09_T=1.0E-04 S2-C4_Z=1.0E-09_T=1.0E-02 S2-C4_Z=1.0E-10_T=1.0E-02 S2-C4_Z=1.0E-10_T=1.0E-04

700 600

Strain, Micron

500 400 300 200 100 0

0

Corrosion % Effect of Intrinsic Permeability S2-C5

700

Strain, Micron

500

100

a)

500 400 300 200 100

0

1

c)

2

3

4

5

6

7

0

8

Corrosion % Effect of Intrinsic Permeability S2-C4

1

2

3

700 Experiment S2-C3 S2-C3_Z=1.0E-10_T=1.0E-02 S2-C3_Z=1.0E-12_T=1.0E-02 S2-C3_Z=1.0E-09_T=1.0E-02 S2-C3_Z=5.0E-10_T=1.0E-02

400

4

5

6

7

8

Experiment S2-C3 S2-C3_Z=1.0E-10_T=1.0E-02 S2-C3_Z=1.0E-10_T=1.0E-04 S2-C3_Z=1.0E-09_T=1.0E-02 S2-C3_Z=1.0E-09_T=1.0E-03 S2-C3_Z=1.0E-09_T=1.0E-04

600

Strain, micron

500

300 200 100 0

0

Corrosion % d) Effect of gel migration S2-C4

600

Strain, micron


700

500 400 300 200 100

0

1

2

3

4

5

6

7

8

9

0

10

0

1

2

3

4

5

6

7

8

9

10

Corrosion % Corrosion % e) Effect of Intrinsic Permeability S2-C5 f) Effect of gel migration S2-C5 Figure 4 Sensitivity analysis: effect of intrinsic permeability and viscous parameter for gel migration

Figure 4 (b,d & f) indicate the sensitivity of the viscous parameter for migration of gel under constant values of permeability. The results indicate a strong correlation of the viscous parameter with the build-up of stress. The higher the value for the viscous parameter, the easier will be for the corrosion products to migrate. However, for lower values of the viscous parameter the corrosion products tend to remain around the reinforcing bars resulting in a significant build-up of stress. Figure 5, indicate the crack pattern due to corroding steel for concrete covers of 5cm, 4cm and 3cm respectively.

359


a)

S2-C5 b) S2-C4 c) Figure 5 principal strain contour due to embedded corroding steel

S2-C3

Based on the parametric study, a value of 0.01 for the viscous parameter for migration of corrosion gel; 1.0*1E-10 to 5.0*1E-10 for intrinsic permeability of corrosion gel with 10 to 100 times for cracked concrete, are found to fairly capture the experimental results. Figure 6 shows the summary of computationally obtained results for the progress of surface concrete strain for concrete cover depth of 2cm to 5cm and the three different mixes indicated in Table 1. Most importantly, the evolution of surface strain due to corrosion is not a smoothly increasing curve rather it is serrated and shows subsequent rise and drop, both experimentally and computationally. This is hardly possible to capture using the existing single phase approaches. Experiment S2-C2 ; Experiment S2-C3 ; Experiment S2-C4 ; Experiment S2-C5 ;

Analysis S2-C2 Analysis S2-C3 Analysis S2-C4 Analysis S2-C5

500

500

400

400

300 200

Analysis S2-C4 Analysis S3-C4 Analysis S1-C4

300 200 100

100 0

Experiment S2-C4 ; Experiment S3-C4 ; Experiment S1-C4 ;

600

Strain, Micron

Strain, Micron

600

0

1

2

3

4

5

6

7

8

9

0

10

a)

0

1

2

3

4

5

6

7

8

9

10

Corrosion %

Corrosion %

Effect of concrete cover depth b) Effect of concrete strength Figure 6 Computational and experimental results

In use of the best fit parameters indicated above, the experiment by Yuxi Zhao [22] was analysed.. The analytically obtained crack pattern based on the poro-mechanical approach is indicated in Figure 7. In the experiment the first surface crack occurred at 28 days, while the corresponding analytical value is 36 days.

Figure 7 Cracking pattern due to corroding steel; Yuxi Zhao (2011)

3 Conclusion A new approach for tackling the crack initiation and crack propagation for corroding RC structures is proposed based on poro-mechanics. Accordingly, the overall mechanistic behaviour is found to be highly dependent on the kinematics of corrosion gel in concrete 360


capillary pores and crack gaps. Using sensitivity analysis, reasonable range of values for the major influencing parameters is determined. The effects of concrete cover depth and concrete strength are discussed and the poro-mechanical approach is found to be a promising and new direction of research for assessing the impact of corrosion in RC. It is also emphasized, the impact of corrosion on crack initiation and crack propagation is hardly understood and more studies in the future are required, experimental studies in particular.

4 Acknowledgment

This study was financially supported by JSPS KAKENHI Grant No. 23226011.

5 References

[1] Maage, M., Helland, S., Poulsen, E., Vennesland, Ø, and Carlsen, J. E.(1996). Service life prediction of existing concrete structures exposed to marine environment, ACI Mater. J., 93(6), 602–608. [2] Liu, Y., and Weyers, R. E. (1998) Modeling the time-to-corrosion cracking in chloride contaminated reinforced concrete structures, ACI Mater. J., 95(6), 675–681. [3] Torres-Acosta, A. A., and Martinez-Madrid, M. (2003) Residual life of corroding reinforced concrete structures in marine environment, J. Mater. Civ. Eng., 15(4), 344–353. [4] Andrade, C., Alonso, C., and Molina, F. J. (1993) Cover cracking as a function of bar corrosion: Part 1Experimental test, Mater. Struct., 26, 453–464. [5] Stewart, M. G. (2001) Spalling risks, durability and life-cycle costs for RC buildings, Proc., Int. Conf. on Safety, Risk and Reliability-Trends in Engineering, Malta, IABSE, Zurich, Switzerland, 537–542. [6] Dimitri V. Val, Leonid C., and Mark G.S. (2009) Experimental and Numerical Investigation of CorrosionInduced Cover Cracking in Reinforced Concrete Structures, J. Struct. Eng., 135: 376–385. [7] Bazant, Z. P. (1979) Physical model for steel corrosion in concrete sea structures-Application, J. Struct. Div., 105(6), 1155–1166. [8] Pantazopoulou, S. J., and Papoulia, K. D. (2001) Modeling cover cracking due to reinforcement corrosion in RC structures, J. Eng. Mech., 127(4), 342–351. [9] Bhargava K., Ghosh A.K., Mori Y., Ramanujam S., (2005) Modeling of time to corrosion induced cover cracking in reinforced concrete structures, Cem. Concr. Res. 35, 2203–2218. [10] Sagoe-Crentsil K.K. and Glasser F.P. (1989) Steel in concrete: part I. A review of the electrochemical and thermodynamic aspects, Mag. Concr. Res. 41 205–212. [11] Alonso, C., Andrade, C., Rodriguez, J., and Diez, J. M. (1998) Factors controlling cracking of concrete affected by reinforcement corrosion, Mater. Struct., 31 (211), 435–441. [12] Allan M.L., (1995) Probability of corrosion induced cracking in reinforced concrete, Cem. Concr.. Res. 25 1179-1190. [13] Petre-Lazar I., Gerard B., (2000) Mechanical behaviour of corrosion products formed at the steel–concrete interface, in: Testing and Modelling, EM2000 Proceedings of the 14th Engineering Mechanics Conference, ASCE, Austin, Texas,. [14] Chen D., Mahadevan S., (2008) Chloride-induced reinforcement corrosion and concrete cracking simulation, Cem. Concr. Comp. 30, 227–238. [16] Biot M.A.(1941) General Theory of Consolidation, Journal of Applied Physics, 12, 155-164 [17] Maekawa, K., Kishi, T. and Ishida, T. (2003) Multi-scale modelling of concrete performance: Integrated material and structural mechanics, J. of Adva. Conc. Tech., 1(2), 91-126. [18] Maekawa K. and Fujiyama C. (2013) Rate-Dependent Model of Structural Concrete Incorporating Kinematics of Ambient Water Subjected to High-Cycle Loads. [19] Wong H.S., Zhao Y.X., Karimi A.R., Buenfeld N.R. and Jin W.L. (2010) On the penetration of Corrosion products from reinforcing steel into concrete due to chloride-induced corrosion, Corrosion Science 52, 2469-2480 [20] Michel A., Pease B.J., Peterova A., Geiker M.R., Stang H., and Thybo A.E.A. (2013) Penetration of corrosion products and corrosion-induced cracking in reinforced cementitious materials: Experimental investigations and numerical simulation, Cement & Concrete Composites [21] Oh B.H., Kim K.H., and Jang B.S. (2009) Critical Corrosion Amount to Cause Cracking of Reinforced Concrete Structures, ACI Mater. J., 106(4), 333–339. [22] Zhao Y., Yu J., Wu Y., and Jin W. (2012) Critical thickness of rust layer at inner and out surface cracking of concrete cover in reinforced concrete structures, Corrosion science 59, 316-323.

361


The Influence of Drying Shrinkage on the Fatigue Life of RC Slabs Yoshisato Hiratsuka 1*, Koichi Maekawa 2 (1) SHO-BOND Corporation, Tokyo, Japan (2) The University of Tokyo, Tokyo, Japan Abstract: As moving load testing of RC slabs of highway bridges takes several months, drying shrinkage of concrete progresses concurrently with fatigue-induced damage during this time. Further, as actual bridges are placed in service after drying shrinkage has progressed to some extent, RC slab fatigue life test results do not correspond with the actual life of RC slabs in service. This study investigated the effect of the combined action of drying shrinkage and fatigue-induced damage by using multi-scale analysis based on thermodynamics. The increase in deflection under fatigue load was found to be the result of the combined action of shrinkage and load-induced fatigue damage, and the process by which damage progresses through their interaction was elucidated. Keywords: RC slab, fatigue life, drying shrinkage, wheel-type loading test, fatigue

1 Introduction

Owing to test period and space constraints, the wheel-type moving load tests used to assess highway bridge slabs are often conducted shortly after the fabrication of the specimens without allowing for the passage of time, and drying shrinkage progresses concurrently with moving load fatigue induced damage. On the other hand, a period of several months is expected to elapse between the completion of slab fabrication and the time when the actual bridge is put in service, which means that drying shrinkage develops to some extent before the bridge begins being subjected to moving traffics. This needs to be taken into account when estimating the fatigue life of actual bridges by using the moving load tests. This study seeks to assess the combined effect of damage caused by drying shrinkage and fatigue using a 3D integrated structural/material response analysis system (DuCOM-COM3) for long-term tracking of the changes in concrete material properties from casting to dismantling of concrete structures [1]. Its goal is to further the study of the fatigue life of bridge deck RC slabs through the verification of loading tests using these analysis results and through analysis with particular focus on the magnitude of drying shrinkage.

2 Reproduction of deflection through wheel load running test using coupled analysis 2.1

Test objects

The deck slab specimens selected for testing were haunched RC slabs designed according to the Specifications for Steel Highway Bridges (1964) having a span of 2,500 mm, and measuring 3500 mm in the bridge axis direction, 2800 mm in the transverse direction, and thickness of 160 mm. Figure 1 shows the shape and reinforcement arrangement of the RC slabs. Taking into account the rebar quality available at the time, SD295A was selected for D10 rebars, and SD345 for D13 and D16 rebars. The mix design of concrete used is shown in Table 1. Measurement of drying shrinkage was done using 100 × 100 × 400 mm unreinforced concrete specimens fabricated for this purpose and cured under the same conditions as the slabs. The specimens were kept in formwork until they developed sufficient strength to be demolded, approximately two days after casting, and following demolding, they were covered with a sheet and air cured. *

Email: [email protected]

362


Figure 1 Shape of RC slab and reinforcement pattern Table 1 Mix proportion of concrete used

2.2

W/C

S/a

C

W

S

G

AW

%

%

kg/m3

kg/m3

kg/m3

kg/m3

kg/m3

60

47.3

283

170

862

985

2.83

Reproduction of slab deflection by analysis

Measurement of the shrinkage strain of concrete was done by contact gauge. Owing to the unavailability of material test values for the aggregate used in the test, the maximum shrinkage of the aggregate was estimated from free shrinkage to be on the order of 300 μ. The shrinkage of the slab specimens was also measured at that time. The amount of shrinkage was obtained using as reference a section of specimen measuring 3.5 m in the longitudinal direction. Figure 2 shows the test results and the shrinkage strain from the analysis results [1][2]. The wheel load running tests were carried out at 47 days and 99 days after casting of the specimens.

Figure 2 Drying shrinkage strain and thermo-hygral simulation [2]

To determine the influence of drying shrinkage, the amount of shrinkage of a standard test specimen at the commencement of testing was converted to the amount of shrinkage of the slab specimen using the volume to surface ratio, and then structural analysis under the action of the fatigue load was conducted [3]. To further increase the investigatory depth of this study, reproduction of the simultaneous progression of the drying shrinkage process and fatigue damage process through numerical simulation was opted for. Taking into account the influence of drying shrinkage during testing and the acceleration of moisture migration and moisture loss associated with cracking, analysis of the high-cycle influence using DuCOM-COM3 was performed (hereafter referred to as link model). The influence of fatigue at the structural member level was expressed as residual deformation and cross sectional stiffness degradation. 363


The validity of this model has been established by previous studies showing this model to have sufficiently accurate tracking capability under various temperatures and loading conditions [4]. The elements used in the analysis are shown in Figure 3. A symmetrical model divided along the travel plane was adopted, and eight-node isoparametric elements using the enhanced strain model, which avoids shear lock, were used as the finite elements. The environmental conditions were held constant through the entire period at 20°C and 60% RH. With regard to the running of the wheel load, a stop time of 40 seconds was provided, after running it at 3 km/h during the actual test. Furthermore, deflection measurement was designed to run for approximately 10 minutes. In the analysis, the loading rate was set so as to reproduce a time period generally equivalent to the test period. A tire/ground friction area 300 mm wide × 450 mm long was used.

Figure 3 Analysis model

The analysis results are shown in Figure 4. The deflection value was evaluated by subtracting the first residual deflection from the measured value, as a result of which the slab deflection behaviour calculation accuracy can be seen to improve above that achievable by traditional structural analysis alone. In particular, this allowed reproduction of the increase in deflection from n = 10+3, when the interaction between drying shrinkage and fatigue appears.

Figure 4 Total deflection during load running

3 Influence of various factors on deck slab deflection 3.1

Aggregate shrinkage amount

Differences in the amount of drying shrinkage are due in large part to the effect of aggregate shrinkage [5]. Therefore, analysis was carried out by varying the maximum amount of aggregate shrinkage. In the link model, shrinkage is calculated from the specific surface area and maximum amount of shrinkage of aggregate, according to the internal humidity. Analyses were performed both for the case of no aggregate shrinkage and for the case of maximum aggregate shrinkage of 500 µ. The curing and other conditions, similarly to the test, were selected as initial curing at 20°C for 3 days, followed by sealing and air curing at 20°C and 60% RH. The moving load test was begun 45 days after casting. Figure 5 shows the slab deflection behaviour. The slab deflection can be seen to increase as shrinkage increases. By assuming deflection value Dn ≧ 12.1 mm as in the case of analytical slab failure testing, fatigue life becomes shorter as the amount of shrinkage increases. Numerical 364


analysis revealed that a more than two-fold difference in the fatigue life of a structural member exists depending on the amount of aggregate shrinkage.

Figure 5 Deflection behaviour of slab with varying amount of aggregate shrinkage

3.2

Curing conditions

Differences in curing conditions are assumed to result to result in different strength development and shrinkage amount. Thus analysis was performed by treating the slab curing conditions as parameters. The conditions used for the analysis were of three types, namely demolding at three days (3 Days) similarly to the test, demolding at 7 days (7 Days), and demolding at 28 days (28 Days). The maximum amount of aggregate shrinkage was assumed to be 300 µ, as the average value for Japanese aggregate. Figure 6 plots the 28-day strength by location in the transverse direction at mid-span under the different curing conditions. This being local strength, describing the relationship between the curing conditions and strength development based on these results is difficult. However, the analysis suggests that the shorter the curing period, the greater the loss of moisture from the surface of the slab, which delays strength development, tends to be.

Figure 6 4-week strength in transverse direction under different curing conditions

Figure 7 shows the slab deflection behaviour. There is no significant difference in initial deflection between the different curing lengths. This is due to the fact that in the case of large drying shrinkage, extensive cracking has already occurred, so that cracking under initial loading occurs to a lesser extent. On the other hand, when drying shrinkage is small, cracking is also limited, so that extensive crack development takes place under initial loading. In other words, because the difference with the initial residual deflection is used, in the case of 3 Days, the deflection due to shrinkage cracks is assessed as being large, while initial deflection is assessed as being small. In the case of 28 Days, initial deflection is estimated to be small and the effect of load-induced fatigue to be larger because little initial cracking takes place. As for fatigue life, not much difference between the various curing periods could be seen assuming deflection Dn ≧ 12.1 mm as with failure testing. Since the fracture morphology of the slabs was punching shear failure, the strength of the concrete around the centre of the structural element, which accounts for a large portion of the resistive cross-section, is considered to have a greater influence than that of the concrete near the surface. Moreover, it is surmised that if water were to be supplied to the slab surface, the fatigue life would differ owing to the effect of the water [6][7]. As this 365


effect is facilitated particularly when water penetrates through cracks, there is a strong possibility that the presence of initial cracks affects fatigue life. The coupled analysis in this study does not take into consideration the effect of water trapped inside cracks, the investigation of which is a subject for future study.

Figure 7 Deflection behaviour of slab under different curing conditions

3.3

Influence of number of days from casting until test start

Thus far, this paper has stated that the effect of shrinkage is factored into the calculation of the fatigue life of slabs because drying shrinkage progresses after the commencement of testing. On the other hand, the authors also assumed the case of loading when the effect of shrinkage has almost entirely disappeared. As shrinkage strain ceases almost entirely to increase after approximately one year, the analysis was performed for slabs cast one year before. For comparison purposes, an analysis of slabs cast six months before was also conducted. The maximum amount of aggregate shrinkage was set as 300 μ and the curing condition as 3 days of initial curing.

Figure 8 Deflection behaviour of slab loaded after a long time

Figure 8 shows the deflection behaviour of slabs subjected to loading tests after the lapse of differing long periods of time. Compared to the case where testing began on the 45th day, the longer it took for the testing to commence, the greater the initial deflection and the shorter the fatigue life. This is considered to be the effect of the progression of drying shrinkage over the lapse of a long time. On the other hand, the effect of shrinkage is offset to some degree by the development of concrete strength. The optimum number of days until testing is begun is considered to differ according to the specimen strength and the water-cement ratio. If one considers the moving load test as a test that simulates an actual bridge, a considerable length of time is required, from the fabrication of the slabs to the commencement of service, and it is estimated that performing testing after the lapse of one year yields results that approximate actual behaviour. In the case of specimens for which only a short time has elapsed since their fabrication, drying shrinkage has not had the chance to progress sufficiently, which opens the possibility that the fatigue life of said specimens will be overestimated.

366


3.4

Differences caused by humidity

As the development of drying shrinkage and concrete strength is influenced by humidity, an analysis was performed using humidity as a parameter. It was decided to cure the specimens for three days at 20°C in a sealed state, and then alter the environmental conditions. The maximum amount of aggregate shrinkage was set at 300 µ and the commencement of testing at the 45th day. Figure 9 shows the results of the analysis using the humidity parameter. As the humidity decreases, the effect of drying shrinkage appears earlier on, resulting in large initial deflection, whereas the opposite occurs when the humidity increases. Likewise, owing to the large drying shrinkage amount, the lower the temperature, the shorter the fatigue life, and the higher the temperature, the longer the fatigue life tends to be. In particular at 100% humidity, almost no drying shrinkage is observed, resulting in extremely small initial deflection and greatly increased fatigue life.

Figure 9 Deflection behaviour of slab at different humidity levels

4 S-N curves resulting from different shrinkage conditions To compare the effect of the various parameters on fatigue life, comparisons of extreme cases were made. The analysis conditions calculated in this paper are given in Table 2. Case 0 is the standard case used in this study. Case 1 is designed for a larger drying shrinkage effect, and Case 2 for a smaller drying shrinkage effect. Case 3 assumes the same conditions as those of Case 1, but with the difference that testing begins one year after casting. Table 2 Analysis conditions for fatigue life calculation

Shrinkage of aggregate

Sealed curing

Test start delay

Relative humidity

Micro

days

days

%

Case 0

300

3

45

60

Case 1

500

3

45

40

Case 2

300

28

45

80

Case 3

500

3

365

40

Figure 10 shows the slab deflection behaviour under different environmental conditions. Even when performing the comparison with the same load of 180 kN, drying shrinkage has a large influence on the defection behaviour, and in the case of the two extremes of Case 1 and Case 2, a difference of an order of magnitude or more can be seen. The fatigue life decreases as the number of days until the commencement of the moving load test increases. Particularly when the effect of drying shrinkage is large, the fatigue life declines by an order of magnitude or more, and as shown in Case 3, the decrease in fatigue life also becomes significant. The effect of changing the loading conditions in Cases 0, 1 and 2 on fatigue life was compared. Crack density and live-load deflection are frequently used to determine the serviceability limit during moving load testing [8]. However, as deflection at failure is element division dependent 367


during analysis, its use as is as a damage index is difficult. Therefore, in this paper, for the sake of convenience, the deflection value Dn ≧ 12.1 mm during failure at 180 kN during mobile load testing was used as reference, and failure deflection at other loads was defined as the product of that deflection value by the load ratio. The loads used to calculate the fatigue life (S/N) curve were, in addition to 180 kN, 220kN, 250kN, and 280kN.

Figure 10 Deflection behaviour of slab under different environmental conditions

Matsui proposes Psx as a failure parameter independent of slab span, slab thickness, and reinforcement ratio [9]. Figure 11 shows the S-N curves of each case using P/Psx comparing the load bearing capacity obtained with this equation and the applied load. Also shown is the moving load test data previously obtained by Yokoyama et al. [10]. Note that since f'c is often obtained through destructive testing of cylinders, compressive strength was obtained by analysing specimens as specimens for compression tests. The S-N curves traditionally used to calculate fatigue life were found to greatly differ owing to the influence of drying shrinkage.

Figure 11 S-N curves of slab under different environmental conditions

The fact that the plotted past destructive testing data is divided into two groups is considered to be simply due to differences in test equipment at the time. However, as the drying shrinkage measurement data of the time is not available, this is purely a matter of conjecture. Yet, if one assumes that differentiation to be caused by differences in the amount of drying shrinkage due to different aggregate shrinkage amounts, that differentiation may be explained rationally. It is not unreasonable to consider that differing testing institutions and equipment resulted in different aggregate shrinkage amounts and curing conditions, and that the two groupings of past data in Figure 11 reflect this. The influence of standing water on slab fatigue has been explained by Matsui [6], and measures to provide durability against this influence, such as providing a waterproof layer on actual structures, have been considered. However, S-N curves used to evaluate slab fatigue failure were found to greatly differ according to the amount of drying shrinkage. The effect of drying shrinkage on fatigue life used to be attributed to test error, but with regard to the definition of fatigue life, including the calculation of S-N curves, numerical analysis study suggests that drying shrinkage needs to be assessed properly.

368


5 Conclusions This paper has clarified the deflection behaviour of slabs under the combined action of drying shrinkage and fatigue. The findings are summarized below. (1) Fatigue tests that take a long time, include the effect of damage caused by progress of drying shrinkage in addition to the effect of fatigue, and therefore attention must be paid to the treatment of these respective effects. Through the use of link analysis, we were able to explain the increase in deflection by the progress of drying shrinkage. (2) Particularly in the case of specimens with a large amount of aggregate shrinkage or low environmental humidity, drying shrinkage has a larger influence because the increase in fatigueinduced deflection is compounded by the increase in deflection caused by drying shrinkage. (3) Fatigue tests have traditionally been performed soon after the fabrication of the test specimens, meaning that such tests were done while drying shrinkage was still in progress, raising the possibility of overestimating fatigue life. Commencement of testing after the lapse of about one year from test fabrication is desirable for testing that fits considers actual structures. (4) Generally, the S-N curves used for fatigue failure of slabs will significantly vary according to the amount of aggregate shrinkage and the environmental conditions. Such variation traditionally has been attributed to test error, but we demonstrated that an appropriate grasp of the amount of drying shrinkage makes it possible to explain such variation in terms of variations in the slope of S-N curves and their intercept values. The scope of this paper did not include a discussion of the influence of water on fatigue life, but as the presence of initial cracks is estimated to greatly affect fatigue life, this represents an area for future research.

6 Acknowledgements


7 References

[1] Maekawa K, Ishida T, Kishi T (2008) Multi-scale Modeling of Structural Concrete, Taylor and Francis [2] Maekawa K, Chijiwa N, Ishida T (2011) Long-term Deformational Simulation of PC Bridges based on the Thermo-hygro Model of Micro-pores in Cementitious Composites, Cement and Concrete Research [3] Hiratsuka Y, Senda M, Fujiyama C, Maekawa K (2013) Fatigue-Based Structural Behavior of RC Bridge Slabs with Different Loading Histories, EASEC-13, F-2-4 [4] Asamoto S, Ishida T, Maekawa K (2006) Time-Dependent Constitutive Model of Solidifying Concrete Based on Thermodynamic State of Moisture in Fine Pores, Journal of Advanced Concrete Technology, 4 (2) pp.301-323 [5] Asamoto S, Ishida T, Maekawa K (2007) Analysis of Concrete Shrinkage Coupling with Properties of Aggregate. Proceedings of JSCE, Vol. 63 No.2, 327-340 [6] Matsui S (1987) The Effect of Water on the Fatigue Strength of RC Slabs Under Moving Loads. Proceedings of the Japan Concrete Institute 9-2, pp 627-632 [7] Fujiyama C, Kobayashi K, Suzuki Y, Maekawa K (2012) Study of Factors Affecting the Fatigue Life of Deck Slabs Based on Experimental Data and Numerical Analysis. Proceedings of the Japan Concrete Institute, Vol. 34, No. 2, pp 667-672 [8] Matsui S, Maeda Y (1986) Proposal of Method to Assess the Degradation of RC slabs of Highway Bridges. Proceedings of JSCE, Vol. 347, I-6 [9] Matsui S (2007) Highway Bridge Slabs: Design, Construction and Maintenance. Morikita Pub. [10] Yokoyama H, Nagaya Y, Sekiguchi M, Horikawa T (2004) Assessment of Serviceability Limit of Highway Bridge Slabs Using Self-Propelled Testing Machine. Proceedings of Fourth Symposium on Decks of Highway Bridges, pp 49-54

369


Statistical models for interval censored time-to-event data Geurt Jongbloed Delft University of Technology, Delft, The Netherlands Abstract: Time-to-event data are collected and studied in many research fields. Examples include medical science and reliability theory. A complication often seen with these data is censoring. The time of the actual event is not observed, for each subject only an interval can be observed that contains this time. Parametric as well as nonparametric estimation procedures can be employed to estimate relevant quantities of interest. Also various models can be used to include explanatory variables in the model. In this paper the parametric and nonparametric approach to interval censored data are described. The aim is to show a glimpse of the possibilities of stochastic modelling and stimulate discussion on the development of models specifically in the context of ageing. Keywords: maximum likelihood, EM algorithm, accelerated life, Cox proportional hazard

1 Introduction Consider an experimental setting where a number of samples from a material is exposed to certain conditions. The question of interest is how long it takes until the quality of the material (can be related to strength, flexibility or other properties) drops below a certain level. Suppose the samples are ‘comparable’ and not obviously related to each other, say that these were obtained from one production line, but not too close together in time. Then, typically, not all samples break down at the same time. There will be variation in the observed failure times. A common approach in statistics is then to assume a stochastic model for the data. All survival times are modelled as random variables. Denote these by X 1 , X 2 ,..., X n . Comparability of the objects lead to an assumption that these random variables all have the same distribution, meaning that there exists one (distribution) function F such that for i = 1,2,..., n ,

Pr( X i ≤ x) = F ( x), x ≥ 0 . Lack of relation between the objects leads to the common assumption that the random variables are (stochastically) independent. This means that for all i ≠ j ,

Pr( X i ≤ x, X j ≤ y ) = Pr( X i ≤ x) × Pr( X j ≤ y ) = F ( x) × F ( y ), x, y ≥ 0 . The random variables are in this case called i.i.d., or independent and identically distributed. Now, given a data set of measurements, {x1 , x 2 ,..., x n } , the aim is to estimate the underlying distribution function F , so to approximate this distribution function only using the available data. In certain situations, one could impose specific assumptions on the distribution function F . For instance, that it is an exponential distribution function that can be written as

F ( x) = 1 − exp(−θx), x ≥ 0

for some parameter θ > 0 or a Weibull distribution function that can be written as

F ( x) = 1 − exp(−(αx ) ), x ≥ 0 for some parameter pair (α , β ) with α , β > 0 . See Figure 1 for a picture of some of these β

distribution functions (left panel) and corresponding probability density functions (right panel). A probability density function is the derivative of a distribution function.

370


1.0

2.0

0.8 1.5 0.6 1.0 0.4

0.5 0.2

0.0

0.0 0

1

2

3

4

0

1

2

3

4

Figure 1. The left panel shows four distribution functions, the right panel the corresponding probability density functions. Black: exponential (1), blue: exponential (2), red: Weibull (2;1), green: Weibull (5;0,5).

The class of exponential (or Weibull) distribution functions is called a statistical model for the data. Under such a parametric model, estimating the distribution function boils down to statistical parameter estimation, i.e. recovering the underlying parameter θ > 0 (or, in case of the Weibull model, the pair of parameters (α , β ) ) based on the available data. There are general methods to construct sensible parameter estimates within a parametric model. One of these is the method of Maximum Likelihood (ML). More on this method will be said in the next sections. Instead of assuming a well-known parametric model, one could also use a nonparametric estimator of the distribution function. The best known estimator is undoubtedly the empirical distribution function. This distribution function (depending on the data at hand) is defined by

Fn ( x) =

{i : xi

≤ x}

n

where ⋅ denotes the number of elements of a set. Figure 2 shows this piecewise constant empirical distribution function based on a data set of size n = 15 . The individual points can be identified as jump points of the empirical distribution function. The figure also shows two parametric ML estimates of the distribution function: one based on the exponential model, the other on the Weibull model. If a parametric model is adequate, the ML estimator corresponding to that parametric model will be close to the empirical distribution (for reasonable sample sizes).

371


1.0 0.6 0.8

0.5

0.4

0.6

0.3 0.4 0.2 0.2 0.1

0.0

0.0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

0

3.5

1

2

3

4

5

Figure 2. Empirical distribution function of a sample of size 15 together with the parametric ML estimates: the exponential (black) and Weibull (blue). The right panel shows the probability density functions corresponding to the parametric ML estimates.

In Section 2, this general statistical approach to estimating distributions will be described for the situation where there is so-called interval censoring. In that situation the event times are not observed precisely; for each subject only an interval is observed where the corresponding event time belongs to. A parametric and nonparametric approach to this problem is described. Section 3 deals with two particular types of regression model, the accelerated life model and Cox proportional hazard model.

2 Interval censoring Denote by T1 , T2 ,..., Tn a random sample from an unknown distribution function F . Instead of observing these random variables directly, n intervals are observed, (u1 , v1 ], (u 2 , v 2 ],..., (u n , v n ] , where the available information is that Ti ∈ (u i , vi ] . This censoring mechanism is called interval censoring. Now, based on the observed intervals, the distribution function of interest can be estimated. Again, a model is needed, meaning a set of potential distribution functions the true distribution function is assumed to belong to. This can be a parametric set, like the exponential distributions but also a nonparametric set. Examples of the latter are the class of all distribution functions or the class of concave distribution functions. Like in the estimation context with direct data, a natural estimation method is maximum likelihood. This approach can be taken in the parametric as well as in the nonparametric situation. Consider a data set consisting of the intervals. For a particular distribution function F , the log likelihood can then be defined as n

L( F ) = ∑ log(F (vi ) − F (u i ) ) . i =1

The ML estimator maximizes this function over the model at hand. For the exponential model, this function reduces to n

φ (θ ) = L( Fθ ) = ∑ log(exp(−θu i ) − exp(−θvi ) ) . i =1

372


For an artificial data set of size n = 20 visualised in the left panel of Figure 3, the function φ is given in the right panel of Figure 3. Its maximiser over θ , the ML estimate, is 0,79 in this case. It is easily found by a univariate optimisation procedure.

20 -15

15

-20

10

-25

5

-30

0

-35 0

1

2

3

4

5

0.5

1.0

1.5

2.0

Figure 3. Visualisation of the interval censored data. For 20 event times, intervals are shown known to contain the corresponding event time. The right panel shows the log likelihood function 𝝓.

Now, for the same data set, also a nonparametric maximum likelihood estimator can be defined and computed. The function L is then maximized over all (sub-)distribution functions, i.e. non-decreasing functions with range contained in [0,1] . This problem is well posed and well known algorithms like the Expectation Maximization (EM) algorithm (Dempster, Laird, & Rubin, 1977) and the Iterative Convex Minorant algorithm (Jongbloed, 1998). Figure 4 shows the exponential and nonparametric ML estimator obtained based on the data of Figure 3. 1.0

0.8

0.6

0.4

0.2

0.0 0

1

2

3

4

5

Figure 4. The nonparametric ML estimate for 𝑭 based on the interval data in the left panel of Figure 3 with the parametric ML estimate based on the exponential model, with estimate 𝜽 = 𝟎. 𝟕𝟗.

3 Regression models with interval censoring

Often, instead of only observing (intervals containing) event times, additional information of interest is available for all subjects in the study. There may be some subject specific material properties

373


available, some measurement of thickness etc. This extra information can be included in the model in various ways. The simplest model including covariates, is the accelerated life model. This model assumes the distribution functions of the n event times to be scaled versions of a single distribution function, where the scaling constant for the i -th subject is a fixed function of a linear combination of the covariate xi . More concretely, there is a function c on ℝ taking values in the positive numbers such that 𝑡 � Pr(𝑇𝑖 ≤ 𝑡|𝑥𝑖 ) = 𝐹 � 𝑐(𝛽 𝑇 𝑥𝑖 ) A popular choice for the function 𝑐 is 𝑐(𝑢) = 𝑒 𝑢 . In the simplest situation, the covariate vector is univariate, leading to Pr(𝑇𝑖 ≤ 𝑡|𝑥𝑖 ) = 𝐹�𝑡𝑒 −𝛽𝑥𝑖 � with the exponential choice for 𝑐. Using this expression for the (conditional) distribution function of the event times, the observed intervals can again be used to write down a log likelihood function: 𝑛 𝑛 𝑢𝑖 𝑣𝑖 �−𝐹� �� 𝐿(𝐹, 𝛽) = � log(Pr(𝑢𝑖 < 𝑇𝑖 ≤ 𝑣𝑖 )) = � log �𝐹 � 𝑇 𝑐(𝛽 𝑥𝑖 ) 𝑐(𝛽 𝑇 𝑥𝑖 ) 𝑖=1

𝑖=1

Another approach to incorporate background information in the analysis, uses the hazard rate of a distribution. Given a density with distribution function 𝐹 and density function 𝑓, this hazard rate (also known as failure rate) is defined by 𝑓(𝑡) 𝑑 𝜆(𝑡) = = − log(1 − 𝐹(𝑡)) = log 𝜖↓0 𝜖 −1 Pr(𝑇 ∈ (𝑡, 𝑡 + 𝜖]|𝑇 > 𝑡) 1 − 𝐹(𝑡) 𝑑𝑡 This function quantifies the instantaneous risk of breaking down at time 𝑡, given the item has functioned up till time 𝑡. Conversely, knowing the hazard rate 𝜆, the corresponding distribution function can be obtained via 𝑡

𝐹(𝑡) = 1 − exp �− � 𝜆(𝑠)𝑑𝑠� 0

The proportional hazard model assumes the hazard rates of all items to be proportional and a specific structure of the proportionality parameters as function of the covariate vector 𝑥: 𝜆𝑖 (𝑡) = 𝑐�𝑥𝑖𝑇 𝛽�𝜆(𝑡) where 𝑐: ℝ → (0, ∞) is a known function and 𝜆 is the so-called baseline hazard rate. The most popular model of this type is the Cox proportional hazard model, where 𝑐(𝑦) = 𝑒 𝑦 . The log likelihood of the observed intervals under this model is best written as function of 𝛽 and baseline hazard 𝜆, and given by 𝑛

𝐿(𝜆, 𝛽) = � log(Pr(𝑢𝑖 < 𝑇𝑖 ≤ 𝑣𝑖 )) 𝑖=1

𝑛

𝑢𝑖

𝑣𝑖

= � log �exp �−𝑐�𝑥𝑖𝑇 𝛽� � 𝜆(𝑠)𝑑𝑠� − exp �−𝑐�𝑥𝑖𝑇 𝛽� � 𝜆(𝑠)𝑑𝑠�� 𝑖=1

0

0

The models described briefly in this section, allow for great flexibility in modelling time-to-event data in the presence of interval censoring. The choice of function 𝑐, e.g., could be inspired by knowledge from the material and circumstances under which the model is used. The same holds for the (possibly parametric) statistical model for the ‘baseline distributions and hazards’ in the various models. From a statistical point of view, estimating parameters can be quite involved, but quite a lot is possible. Recent (and forthcoming) books dealing with modelling time-to-event data, also in a regression context, include (Klein & Moeschberger, 2003), (Sun, 2006), (Hosmer, Lemeshow, & May, 2011), (Kalbfleisch & Prentice, 2011) and (Groeneboom & Jongbloed, 2014). Papers that specifically deal with semiparametric regression models for interval censored event times include (Huang, 1996) and (Zhang, Lei, & Huang, 2010).

374


4 Challenges with statistics of ageing The aim of this paper is to show a broad audience that mathematical modelling of not precisely observed event data is quite flexible. The simple models presented in the previous section (and more of the kind) can be extended to become more realistic. A big challenge is to come up with tailor made extensions of the models, together with materials scientists and engineers and implement these for practical use in specific contexts of ageing. Experimental data on ageing of material is and can be obtained in laboratories. The circumstances can be chosen and measurements taken. Also real-life data are obtained from materials and infrastructures. Of course, there is a problem of scaling to a realistic time scale and also translating effects like ‘accelerated ageing’ or ‘multiplied risk’. It is a challenge to design useful experiments and corresponding models that can combine real-life data obtained from a realistic time scale with experimental data and come up with life time predictions for infrastructures and materials.

5 Bibliography Cox, D. (1972). Regression models and life-tables. J. Roy. Statist. Soc. Ser. B, 187-220. Dempster, A., Laird, N., & Rubin, D. (1977). Maximum likelihood from incomplete data via the EM algorithm. J. Roy. Statist. Soc. Ser. B, 1-38. Groeneboom, P., & Jongbloed, G. (2014). Nonparametric estimation under shape constraints. Cambridge University Press. Hosmer, D. W., Lemeshow, S., & May, S. (2011). Applied survival analysis: regression modeling of time to event data. Wiley. Huang, J. (1996). Efficient estimation ofor the proportional hazards model with interval censoring. Ann. Statist., 540-568. Jongbloed, G. (1998). The iterative convex minorant algorithm for nonparametric estimation. J. Comput. Graph. Statist., 310-321. Kalbfleisch, J. D., & Prentice, R. L. (2011). The statistical analysis of failure time data. Wiley. Klein, J., & Moeschberger, N. (2003). Survival Analysis: Techniques for Censored and Truncated Data. Springer. Sun, J. (2006). The statistical analysis of interval-censored failure time data. Springer. Zhang, Y., Lei, H., & Huang, J. (2010). A spline-based semiparametric maximum likelihood estimation method for the Cox model with interval censored data. Scandinavian Journal of Statistics, 338-354.

375


Long-Term Serviceability and Risk Assessment of Shallow Underground RC Culverts and Tunnels M. Kunieda1*, X. Zhu1, Y. Nakajima2, S. Tanabe2 and K. Maekawa1 (1) The University of Tokyo, Tokyo, Japan (2) Tokyo Electric Power Company, Tokyo, Japan Abstract: This paper aims to investigate the mechanism of the long-term excessive deformation of shallow RC box culverts, which have members of comparatively smaller thickness. The authors mainly focus on two points: internal moisture state of concrete members and increase in the vertical earth pressure acting on culverts caused by both uneven settlement of the foundation and RC structural deformation as a coupled action. To examine these effects, sensitivity analyses using multi-scale analysis and site investigation for deformational modes and steel strain are conducted. It is concluded that long-term excessive deformation may be attributed to these synergy effects accompanying risky delayed shear deformation owing to redistributed vertical soil pressure accelerated by the structural deformation associated with shrinkage and creep of concrete. Keywords: long-term excessive deformation, underground RC culverts, internal moisture state,

differential settlement, delayed shear deformation

1 Introduction 1.1

Research background

1.2

Scope

Since 1990s, the long-term monitoring of Tsukiyono Bridge’s deflection has been periodically reported by Hata et al. (1993) and the imperfection of design methods based upon the conventional linear creep law and shrinkage has been discussed (Watanabe et al. 2008). Recently, excessive deflections of cantilever PC viaducts have been reported worldwide by Bazant et al. (2011a, b) as well. Maekawa et al. (2010) point out two main causes of excessive deflection; one is the non-uniform thermo-dynamic state of moisture inside micro-pores and associated creep, and the other is the delayed average shrinkage of upper and lower flanges in time. (Maekawa et al. 2010, Ohno et al. 2011). The mechanism of the long-term deflections of PC viaducts has been made clear by considering these two factors. On the other hand, underground facilities that are essential parts of the urban infrastructure are used for a wide range of applications from small pipelines for lifelines to large underground structures including subway and highway tunnels. One of the maintenance problems of these underground structures on service is long-term excessive deflections of shallow RC box culverts, which have members of comparatively smaller thickness. After a few decades of service, the top slab deflection may exceed approximately 3 to 10 times the prediction by conventional design formulae. Large numbers of cracks at the inner surfaces of top slabs have been observed. This phenomenon of long-term excessive deformation of unknown cause is a great concern in terms of future risk and serviceability. This paper aims to investigate the mechanism of the long-term excessive deformation of underground box culverts coupled with soil foundation. In the scope of this study, two points that have been neglected for verification of the long-term serviceability limit state in design are focused on. One is the non-uniform internal moisture state of concrete members, which has been pointed out as one of the main causes of the long-term deflections of PC viaducts. The drying creep deflection derived from internal moisture loss of concrete members is assumed to be *

Department of Civil Engineering, The University of Tokyo, Japan , [email protected]

376


negligible in the Japanese Road Association Code (JRA 2002) because moisture containing soil consistently keeps concrete members wet. In consideration of the actual ambient environment of the targeted RC culverts, the relative humidity (RH) of the space inside the culvert is thought to be close to the average RH of outdoor air, but the outer surface exposed to the soil foundation is kept wet. This implies the non-uniform moisture state in concrete members. The other is the increase in the vertical earth pressure acting on culverts caused by both uneven settlement of the foundation and RC structural deformation as a coupled action. In the case of a shallow underground structure whose over-lay is not greater than 10m, a number of culverts carrying loads greater than the overburden soil have been reported (Abhijit 1991, Richard 2005). The increased vertical soil pressure on the top slab may accelerate long-term deformation as well. Displacement(mm)

0

3200

top slab

White line: chalked cracks

400

350

3900 350

Deflection shape of top slab (30years after completion)

Actual crack condition

※mm

-3

Expected in design -6 -9 -12

Measured

-15 -1500 -1000

-500

0

500

1000

1500

Distance from center(mm)

Figure 1 Observed long-term excessive deformation and crack condition 2. Increase in soil pressure by uneven settlement and structural deformation as a coupled action

1.Internal moisture state of concrete members

2010 Kuwano Ebizuka

Drying

RH： assumed to be nearly average of outdoor air

Moisture saturation： high

Figure 2 Two focus points for investigating long-term excessive deflection

2 Pre-analysis 2.1

Outline of multi-scale integrated model

To grasp the effects of two focused points on long-term deflection, the analytical system DuCOM-COM3 (Maekawa et al. 2008) is used. This is a multi-scale analysis code that links the thermo-chemo-physics platforms DuCOM (Maekawa et al. 1999, 2008) and COM3 (Maekawa et al. 2003) as shown in Figure 3. DuCOM is an integrated thermo-hygral analysis model that includes cement hydration in concrete mixture, micro-pore structure formation and mass transport models for concrete ranging from the 10-3 to 10-9 meter scales of micro-pores, while COM3 is a 3D finite element analysis platform for structural concrete with and without cracks. As a result, the linked system is capable of predicting changes in concrete material properties from casting to dismantling of entire structures and taking this material development into account for predicting the response of structural concrete. Through such integration, the long-term structural response under actual ambient conditions can be simulated in a more realistic manner. Figure 3 illustrates the code linkage for computing the nonlinear, time-dependent responses of reinforced concrete. A nonlinear path-dependent constitutive model for soil is also used to meet the challenge of discussing soil-structure interaction. In this study, soil is mechanically idealized as an assembly of a finite number of simple elasto-perfectly plastic elements connected in a parallel pattern as shown in Figure 3 (Maekawa et al. 2003. Towhata 2008). As each component is given different 377


yield strengths of plasticity, which may reflect the grading of the sand particle size, all components subsequently begin to yield at different total shear strains, which results in a gradual increase of entire nonlinearity. This analytical system can treat the time-dependent interaction behaviour of RC structures and soil in the same framework.

Figure 3 Outline of the multi-scale integrated analytical system and constitutive model for soil

2.2

Pre-analyses for evaluating increased soil pressure

Pre-analyses targeting the trap door test (Kuwano et al. 2010) were conducted for experimental verification of the vertical soil pressure owing to differential settlement. The overview of the test and the simplified equation to estimate the elevated earth pressure (Kuwano et al. 2010) is shown in Figure 4. The FEM mesh and the soil properties are defined in reference to the experiment. Joint interface elements are put on the boundary between soil elements and others, which allows free shear slip to reproduce the side layers of foundation. Increment of soil pressure

Pressure（kPa)

Initial soil weight

Shear plane

h:400mm Dr:95% (sand)

25 20 15 10 5 0

Analysis Experiment Initial pressure

-250 -200 -150 -100 -50

Joint Elements

Elastic element

50 100 150 200 250

Movable bed 200mm Downward forced displacement

Average pressure on stable door Pressure(kPa)

400mm

Dr:95% No cohesive stress

50mm

0

Distance from the center(mm)

Sand

Stable

Analysis： Principal strain contour

Soil pressure distribution (1.5mm downward displacement)

20 15 10 5 0

Stable door

Experimen t Analysi s 0

2

4

Exp：Shear plane

6

8

Downward displacement(mm) 100mm

Figure 5 Analytical and experimental results for change in Figure 4 Exprimental condition (Kuwano 2010) earth pressure with downward displacement and shear plane and half analytical Figure 5 showsmesh the comparisons of experimental and analysis results. In general, DuCOM-

COM3 is capable of reproducing the increased soil pressure on the stable door due to the downward displacement of the movable doors. Not only the rapid increase in soil pressure with 378


relatively small downward displacement (maximum soil pressure is observed at 0.5mm displacement), but also the shear plane, which has a strong relationship with the mechanism of increasing soil pressure, can be well simulated. In other words, this analytical system can be applied to simulate the phenomenon of the increase in vertical earth pressure acting on the culvert caused by uneven settlement when the amount of uneven settlement is approximately known.

3 Verification of main causes of excessive deflection 3.1

Details of four RC culverts

3.2

Analysis condition for sensitivity analyses

In order to examine the effects of two points (section 1.2), a series of sensitivity analyses were conducted. Furthermore, analytical results and site investigation of deformational modes and steel strains obtained by destructive testing were compared for discussion of the mechanism for generating long-term excessive deformation. As for analytical targets that have exhibited longterm excessive deflection, four RC culverts were modelled and analysed. The details are shown in Figure 6. The thickness of the concrete members are comparatively small (25cm or 35cm) and the height of the fill above the culvert is relatively low, ranging from 3.8m to 6.4m. The backfilled materials, which may affect the increase in earth pressure, differ for the four culverts. Culvert F, for which specific data about soil properties is available (typical sand used for backfilled material), is mainly focused on for sensitivity analyses.

6400

In reference to the structural details and material data, a half mesh domain was created. For all analyses, the culverts are exposed to different environmental conditions after 28 days of moisture curing. Construction process complexity was eliminated for the sake of clarifying the influencing factors. To figure out the effects of “increase in soil pressure due to uneven settlement” and “drying shrinkage, creep derived from the internal moisture state of concrete members” on the longterm deflection of the top slab, the following four cases are discussed. Case I: no internal driving forces provoked by concrete drying, no uneven settlement, Case II: consideration of uneven settlement, but no that internal driving force, Case III: consideration of internal driving force, but no uneven settlement, Case IV: consideration of all interactions. Thus, the difference between Case I and Case II, III, IV respectively represent each effect described above. Culvert M 1975 completion

350

350

150

250

Mix proportions (kg/m3)

1720

150

250

150

C

W

FA

CA

319

194

748

1121

W/C

Compressive strength(Mpa)

0.61

22.1

3600

1650

T(℃）

RH(%)

21.4-31.4

19.0-97.0

21.6-29.5

81.0-99.0

Internal frictional angle (°)

1-5

34.7

C

W

FA

CA

393

248

643

976

Loam

150

250

FA

CA

422

196

790

986

W/C


0.47

250

W

29.1

1800


C

RH(%)

T(℃）

RH(%)

15.0-95.0

11.2-27.4

95.0-99.0

250

T(℃） 10.2-28.1

1600

250


0.63

24.2

T(℃）

RH(%)

T(℃）

RH(%)

16.0-95.0

22.4-26.0

99.0

Culvert J 1968 completion Mix proportions (kg/m3)

150

3.2 W/C

17.8-28.9

3800

Culvert T 1982 completion

250

150

4200 1800

150

250

Inside air

250

Sand gravel

2400

250

Clay and silt

Cohesive stress(kN/m2)

Internal concrete near outside

Inside

400

250

RH(%)

N value


Inside air

T(℃）

Sand properties

150

250

Culvert F 1982 completion

350

3200

6100

Loam

Main target of sensitivity analyses

sand

C

W

FA

CA

287

175

734

1213

Inside air

W/C


0.61

16.2

Internal concrete near outside

T(℃）

RH(%)

T(℃）

RH(%)

16.9-27.7

15.0-95.0

16.9-26.6

84.0-93.0

The relative humidity (RH) of the inside and outside boundaries of the culvert was kept at Figure 6 Detailswhich of the targeted culverts view andofmaterial properties) 99.99% from casting, means RC that the (Cross-section internal moisture concrete members was consistently maintained. The ambient temperature and RH of Case III and Case IV were maintained at the annual averages (inside culvert: 25 oC RH 60%, outside: 20oC RH 99%).

379


Uneven settlement was reproduced by applying downward forced displacement at the bottom of surrounding soil. In Case II and Case IV, uneven settlement causing maximum increase in soil pressure was intentionally produced. That uneven settlement was applied over the 2 days immediately after completion in order to assume the severest condition, meaning the long drying creep effect of the top slab. The real effect of the increase in earth pressure due to uneven settlement was somewhere in between the two extreme cases (Case I and Case IV).

3.3

Analysis results

Inside RH(%)

Outside RH(%)

Uneven settlement

I≒design

99.99

99.99

Without

II

99.99

99.99

With

III

60

99

without

IV=real

60

99

with

Effect of uneven settlement Effect of internal moisture state Effect of both two effects

Case I

Deflection(mm)

uneven settlement contribution (Case II) Full analyses (Case IV)

-9

Case II and Case III added

0.01

0.1

1

10

100

1000

10000 100000

-5 -10

0

500

1

10

100

1000 10000

900

Measured

600 Case I

300

Soil settlement

0.1

1

uneven settlement contribution (Case II)

10

100

1000 10000

1000

Internal moisture contribution (Case III)

-200

Case I

-400 -600 -800

Measured -500

0.1

Full analyses (Case IV)

0

Strain（10-6）

Deformation (mm)

1.0

Figure 8 Tension steel strain at midspan with average vertical soil pressure on culvert

0

-1000

overburden soil weight(1.24kgf/cm2)

Elapsed time after curing(day)

Elapsed time after curing(day)

-15 -1500

1.2

0.01

Soil settlement

-15

1.4

0

Measured

-12

1.6

1200

-3 -6

1.8

0.01

Internal moisture contribution (Case III)

0

Strain（10-6）

Case

Average soil pressure(kgf/cm2)

Case II considering uneven settlement shows a small increase in deflection compared with Case I without uneven settlement. This is because the simulated steel strain grows with the increase in average vertical soil pressure due to the uneven settlement in the early age. The simulation considering uneven settlement is capable of reproducing the measured tension steel strains. This means the soil pressure acting on the culvert is also well simulated because the tension steel strains reflect the external force mainly governed by the soil pressure. The simulated increment of average soil pressure is up to about 1.37 times as the pressure of overlay soil mass. This value seems reasonable from both the calculated value (below 1.6 times soil pressure increment) based on the suggested equation (Kuwano 2010) and the other experiment using backfilled sand, which shows 1.32 times increase in soil pressure (Dasgupta 1991).

1500

-1000

Disance from center (mm)

0.01

Figure 7 Long-term deflection at midspan and deformational mode at current age

0.1 1 10 100 1000 10000 Elapsed time after curing(day)

Figure 9 Compressive steel strain at midspan internl contour figureand the As shown in Figure 9, the internal moisture is gradually lostwith from the RH inner surface,

drier the concrete members become, the faster the growth of compressive strain due to the drying creep. This effect on the long-term deflection seems to be limited because the greater part of the compression side is wet due to the moisture containing soil. The analysis of simply adding these individual effects leads to underestimated deflections, but the full-coupling analysis simultaneously considering the synergy effect of these two effects 380


shows good agreement with the measured results. The time-dependent long-term deflection from 15 years to 30 years of the full-coupling analysis also matches well the reported crack growth. The cause of this rapid increase in the long-term deflection after a few years is assumed to be the delayed shear failure around the haunch indicated by the principal strain contour (Figure 10). Full analysis

Bending strain

※20 times magnified deformation

Case II (with uneven settlement, no internal moisture effect)

Bending strain +

Bending strain

Shear strain!?

After settlement

300days

After settlement

10000days

300days

10000days

Figure 10 Principal strain contour of full analysis(left) and Case II (right) based on half mesh model

4 Mechanism of accompanying delayed shear and verification 4.1

Simulated deformational modes

To further investigate the delayed shear associated with excessive deflection, the other three culverts were also modelled and analysed. The simulated deformation of the other three culverts was more or less underestimated. Since soil properties are not known in these cases, the authors assume backfilled dry sand. But, according to the construction records, there is the possibility that cohesive soils were used. If so, the higher soil pressure induced by the foundation subsidence might be simulated. Another uncertainty is the seasonal change of culverts’ indoor relative humidity, which has never been recorded. Thus, the annual average was used for simulation. However, the principal strain contours of culverts J and T considering the uneven settlement and the internal moisture state show shear strain concentration near the haunch. In addition, the cracking actually seen close to the haunch is located in the flexural compression region. This fact also indicates the possibility of accompanying delayed shear.

4.2

Investigation into cause and mechanism of delayed shear

Delayed shear deformation is the new point of discussion in view of long-term serviceability. As a cause of accompanying delayed shear, the increase in the shear force acting close to the haunch should be first examined. Figure 12 shows the distribution of the vertical soil pressure acting on the culvert. The soil pressure at the middle is gradually released to about 60% of the overlay soil weight and the soil pressure near the haunch increases to approximately twice the initial soil weight. This redistribution of soil pressure is thought to be triggered by the deformation of the top slab, because large deflection at the middle and small deformation around the haunch take place in analysis. This redistribution of soil pressure is also indicated by measuring the actual soil pressure on the culvert (1991 Abhijit). As a result of this redistribution, the shear force acting on the cross-section showing the shear deformation is slightly increased after 100 days to 1000 days (Figure 13). This increase in the shear force triggered by the structural deformation could be one of the causes of accompanying delayed shear. The other cause of delayed shear is considered to be the reduction of the shear capacity by both the higher sustained load close to the static capacity and the long-term drying shrinkage. It is reported that the shear capacity of the simple beam which takes the influence of drying shrinkage is reduced by up to about 85% of the static capacity under the sealed condition (2011 Mitani). It is reasonably considered that the shear deformation occurs because of the sustained load close to the shear capacity for a few years under the constant drying condition. Experiments to determine shear capacity under sustained load near the capacity are required. 381


4.3

Site investigation into the existence of delayed shear crack

To verify the possibility that the actual RC culvert has shear deformation, a site investigation to carefully check for the existence of inner surface cracks near the haunch was conducted. Considering the bending moment acting on the culvert (Figure 14), the cracks should be concentrated on the tension side, which means the centre of the top slab. The observed cracks were not only on the inner surface of the centre, but also near the haunch, which is intrinsically the compression side in flexure. The largest crack width near the haunch is 0.9mm, which is more than three times the allowable value specified (Standard Specifications for Concrete Structures - 2007) even though the crack width at the centre is 0.3mm. Full analysis (sand)

0

clay

-6 -9

-12 season change of RH -15 -800 -400 0

Measured 400

-2

Displacement(mm)

-3

Displacement(mm)

Displacement(mm)

0

-4 -6 -8 -10 -12

800

-800


-400

0

400

800


2 1 0 -1 -2 -3 -4 -5 -6 -7

-800

-400

0

400

800


Figure 11 Simulated tension steel strain and deformational modes of current age with principal strain contour

After settlement

10000days

300days

Culvert

After settlement

300days

Reduced Increased

10000days

Uniform

Figure 12 Vertical direction contour (zz) showing distribution of soil pressure

Stress (kgf/cm2)

8

Shear capacity in design

Bending moment (Abhijit1991)

Full analysis

6

Tension area ＝crack

Crack site-investigation

B

A

Case II A

4

B

slight increase

check 2 0.01

0.1

1

10

100

1000

10000

Elapsed time after curing (day)

Figure 14 Assumed crack area and crack site-investigation

Figure 13 Average shear stress of one section accompanying delayed shear

It may be reasonable to understand that the large cracks near the haunch are caused not by bending moment, but by delayed shear cracks. Here, it must be noted that unlike a statically determinate structure in air, the investigated soil-structure system is computationally under stability even after diagonal crack propagation in the top slab. This is because the reduced member stiffness caused by shear cracks may lead to the re-distribution of soil pressure and a newly produced stress flow is formed to bear the load carrying mechanism as a result of the static soil-structure interaction.

382


5 Conclusions The following conclusions and the views were drawn from both thermo-hygral and mechanics analysis and site inspection: 1) Soil settlement brings about the imposition of excessive overlay loads upon culverts that exceed the design specification equivalent to the dead weight of soil above the top slabs. 2) The moisture gradient caused by the drying ambient states inside culverts accelerates longterm deformation over several decades. 3) Analytical results considering the interactive behaviour of structures and soil show that long-term excessive deformation accompanies delayed shear deformation owing to redistributed vertical soil pressures triggered by structural deformation. This scenario is partially supported by site inspection to observe the large cracks around the haunch. Further investigation is required to upgrade the serviceability limit state design of aging structures.

6 Acknowledments


7 References

[1] Hata Y, Oonishi N, Watanabe Y (1993) Creep behavior of prestressed concrete bridge over ten years, Proc. of FIP symposium, 305-310. [2] Bazant Z P, Yu Q, Li G H, Klein G J, Kristek V (2010b) Excessive deflections of record span prestressed box girder, Concrete International, ACI, 32(6). [3] Kuwano R, Ebizuka H (2010) Trapdoor tests for the evaluation of earth pressure acting on a buried structure in an embankment, Proc. 9th International symposium on new technologies for urban safety of mega cities in Asia, USMCA, Kobe, October 2010, CD-ROM. [4] Maekawa K, Chijiwa N, Ishida T (2011) Long-term deformational simulation of PC bridges based on the thermo-hygro model of micro-pores in cementitious composites. Cement and Concrete Research, 41(12), 1310-1319. [5] Maekawa K., Ishida T, Kishi T (2009) Multi-scale Modeling of Structural Concrete, Taylor & Francis. [6] Maekawa K, Chaube R P, Kishi T (1999) Modeling of Concrete Performance - Hydration, Microstructure Formation and Transport -, London, E & FN Spon. [7] Maekawa K, Pimanmas A, Okamura H (2003) Nonlinear Mechanics of Reinforced Concrete, SPON Press. [8] Ohno O, Chijiwa N, Suryanto B, Maekawa K (2012) An investigation into the long-term excessive deflection of PC viaducts by using 3D multi-scale integrated analysis, Journal of Advanced Concrete Technology, 10, 47-58. [9] Towhata I (2008) Geotechnical Earthquake Engineering, Springer. [10] Abhijit D, Bratish S (1991) Large-scale model test on square box culvert backfilled with sand, ASCE , Journal of Geotechnical Engineering, 117(1). [11] Bennett R, Wood S, Drumm E, Rainwater N (2005) Vertical loads on concrete box culverts under high embankments. J. Bridge Eng., 10(6), 643–649. [12] Mitani T, Hyodo H, Ota K, Sato R (2011) Discover and the evaluation of shear strength decrease of reinforced normal-strength concrete beams, Proceedings of JCI, 721-726, 33(2), 721-726.

383


A level set model for computational modeling of fatigue-driven delamination in laminated composites M.Latifi 1*, F.P. van der Meer 1, L.J. Sluys 1 (1) Delft University of Technology, Delft, The Netherlands Abstract: A new level set model for the simulation of delamination propagation in laminate composites under high cycle fatigue loading is proposed. For quasi-static loading conditions, interface elements with a cohesive law are widely used for the simulation of delamination. However, basic concepts from fatigue analysis such as the notion that the crack growth rate is a function of energy release rate, cannot be embedded in existing cohesive laws. Therefore, we propose a model in which the cohesive zone is eliminated from the computation. To demonstrate the accuracy of the model, several tests under different modes of fracture are conducted and the results are compared with experimental data, analytical solutions and results from a cohesive zone analysis. Keywords: Fatigue, Delamination, Fracture mechanics, Crack growth, Composites

1 Introduction

The increasing application of composite materials in engineering structures such as wind turbines and aircraft requires reliable computational tools to simulate their behaviour under different working conditions. Cyclic loading is often a critical aspect as it may give rise to fatigue failure. Crack growth in high-cycle fatigue is commonly described using the Paris law. The Paris law represents the load and material-dependent notion of crack growth refer as a function of strain energy release rate: G v = C   Gc 

m

(1)

Where Gc is the fracture energy and G is the energy release rate. The parameters C and m are material constants and must be determined experimentally. One of the key failure modes in composites is delamination which usually occurs due to high inter-laminar stresses acting between layers. For quasi-static loading conditions, interface elements with a cohesive law are widely used for the simulation of delamination. That is why, in recent years, researchers have tried to extend cohesive laws for use in high cycle fatigue analysis and in this context several models have been proposed in the literature [1-5]. In these models the cohesive law has been modified to incorporate the effect of cyclic loading. The core issue in these models is relating damage growth in the cohesive law to crack growth according to Paris law. However, in cohesive laws the energy release rate and crack growth rate are not explicitly defined and in order to estimate G and relate v to damage growth, inconsistent assumptions are typically needed. This problem motivates us to find an alternative for the cohesive zone approach. For delamination modeling a method without cohesive zone has been introduced in reference [6]. This method is based on the level set approach and has been developed for quasi-static analysis. In our work this model is extended to fatigue analysis.

*

PO Box 5048, 2600 GA

Delft, The Netherlands

384

[email protected]


2 Cracked laminate model The central idea in the approach developed in [6] is that the location of the crack front is described with the level set method. This means that there is a sharp front that does not have to be aligned with the finite element boundaries and the front can be located inside the finite elements. In order to model the kinematics of a partially cracked element (see Fig. 1), a weak discontinuity is inserted at the location of the front. As a consequence, there is can be a crack opening inside the partially cracked element. Furthermore, there appears a jump in stresses and strains around the front which can be used to compute the energy that will be released when the front moves.

Figure 1 Definition of crack front with level set field and schematic deformation of quadrilateral element containing the crack front [6]

3 Crack growth model The cracked laminate model described in the previous section allows for mechanical analysis of a partially delaminated structure. To define the crack growth rate, the level set function will be updated. In the quasi-static model [6] the crack growth is defined with a velocity that is computed along the front as a function of energy release rate. This definition is suitable for fatigue analysis . In the fatigue model the quasi-static definition of the front velocity is replaced with the Paris law :  G (s)  ν n ( s) = C ( β , R)    Gc ( β ) 

m( β ,R)

(2)

where the fracture energy Gc is a function of mode ratio and C , m are material constants that depend on the mode-ratio ( β ) and load ratio ( R ). The current energy release rate ( G ) can vary along the crack front and subsequently the velocity is a function of the location of the front ( s ) (see Fig.2).

Figure 2 Velocity is a function of the front location [6]

385


The dependence of fracture energy on the mode ratio is defined using an expression introduced by Benzeggagh and Kenane [7]: Gc =GIc + (GIIc − GIc )( β )η

(3)

where GIc and GIIc are fracture energy in mode I and mode II, η is a mode interaction parameter and β is the ratio between shear dissipation and total energy release rate, defined as:

β=

GII + GIII G

(4)

where GII and GIII are pure mode contributions to the energy release rate and G is defined in section 3.1. In this paper all tests are done at R = 0 . Therefore, the definitions of m and C under mixed mode condition are considered as a function of mode ratio following Blanco et al.[8]: log C = log CI + ( β ) log Cm + ( β ) log 2

CII Cm C I

m = mI + mm ( β ) + (mII − mI − mm ) ( β )

2

(5)

(6)

where CI and mI are crack growth rate parameters in mode I, CII and mII are crack growth rate parameters in mode II and Cm and mm are mode-ratio parameters obtained by curve-fitting mixed-mode experimental data.

3.1

Energy release

The energy release rate G used in the Paris law is computed from a modified virtual crack closure technique [9,10]. The energy release rate G is partitioned into three contributions which are related to pure modes of fracture: G =GI + GII + GIII

(8)

The energy release rate contributions are computed at the crack front location using the definitions [9]: GI = FZ u z , n  + Tn φn , n  + Ts φs , n  GII = Fn un , n 

(7)

(8)

GIII = Fs us , n 

(9)

Figure 3 Free body diagram of an infinitesimal extension of the bottom sublaminate around the front[9]

386


where Fz , Fs and Fn are the jumps in stress resultants which are equal to jump in generalized force in the top and bottom parts and Tn , Ts are distributed moments acting on the crack front(see Fig.3). Also, the differences in displacement gradients  u z,n  ,  u n,n  and  u s,n  are defined as: u  u = j , n=  n,n  n

3.2

− 0= , z 0+

− u j ,n

(10)

− n 0= , z 0− =

Discretization

For the level set update, the velocity needs to be known at the nodes. It is shown in figure (3.) that the distributed force vectors and the differences in displacement gradients are computed at the front. Consequently, G and the velocity field ( vn ) are defined on the front. To define the velocity degrees of freedom( Vn ) on the nodes, field Eq.(3) is discretized on the nodes. Satisfying Eq.(3) in an integral form and following Galerkin’s method using shape functions N results in : f [ M + K ]Vn = M=

(11)

∫ N ⊗Nd Γ

(12)

Γ

K κ h 2 ∫ (∇N  s ) ⊗ ( ∇N  s ) d Γ =

(13)

Γ

  G ( s) m   dΓ f ∫ N C  =   Gc   Γ  

(14)

where K is added to stabilize oscillations on the front, κ is a stabilization parameter and h is the typical element size.

3.3

Level set update

In order to update the level set field ( φ ) which is defined over the whole domain Ω , the velocity is extended. For this purpose, a fast marching method is used and a standard level set update is done to update the level set function [6]:

φ + υn ∆t → φ

(15)

Where ∆t is the time step size.

Table 1 Material properties for HTA/6376C carbon/epoxy[3,11] E11

E22=E33

G12=G13

G23

120.0 GPa

10.5 GPa

5.25 GPa

3.48GPa

GIc

GIIc

ν 23

ν 12 = ν 13

0.260

1.002kJ/m2

0. 51

0.3

387


4 Results and discussion Simulation of Mode I, Mode II and Mixed-Mode delamination tests were conducted to prove that the level set model can reproduce the response of the test specimens in fatigue loading condition accurately.

4.1

Simulation of a DCB specimen

Simulation of a double-cantilever beam under mode I loading is performed and the crack growth rate in the specimen is computed. The specimen was 150mm long, 20 mm wide, with two 1.55mm thick arms with an initial crack length of 35 mm. Material properties are shown in table 1. The laminate is modelled as an assembly of sublaminates in both delaminated and undelaminated regions which are governed by the shear-deformable laminate theory(see Fig.3). Triangle shell elements with 5 degrees of freedom for each node are used for discretization of laminate. One layer of elements is considered in each arm. In order to obtain a constant crack growth rate in each simulation, the specimen’s arms are loaded with two constant opposite moments(see Fig.4). Several tests with different values of applied moments have been conducted and the computed crack growth rate in each test is plotted versus energy release rate in Figure 4. This energy release rate is obtained from an analytical formula which relates energy release rate to applied moment: GI =

M2 bEI

(16)

where b is the specimen width, E is the longitudinal flexural Young’s modulus and I is the second moment of area of the specimen’s arm. A comparison between numerical results obtained from the level set model with experimental data [11] and cohesive model results [1] is presented in Figure 4. The comparison shows a good agreement between level set results and experimental data. The Paris relation that served as an input for the model is retrieved with high accuracy. This does not hold for the cohesive model due to aforementioned difficulties to implement the Paris law in the framework of the cohesive law.

Figure 4 Comparison of crack growth rate from level set model with experimental data and cohesive model results for mode II test

388


Figure 5 Comparison of crack growth rate from level set model with experimental data and cohesive model results for mode II test

4.2 Simulation of 4ENF The crack growth rate for different ranges of the energy release rate under mode II of fracture, is simulated. The dimension and material are the same as in the previous section. The specimen was loaded using the four point End Notched Flexure(4ENF) test. In mode II the energy release rate is related to the applied moment as:  cP  3  2 GII =   4bEI

2

(17)

where p is the applied load and parameter c is introduced in figure5. The finite element model used in these simulations is similar to that used for DCB tests, but the boundary conditions are different. The results are presented in figure 5. which demonstrate the ability of the model to reproduce the Paris curve.

4.3

Simulation of mixed mode loading case

The material and dimensions of the specimen loaded in mixed mode are the same as in the DCB tests. The specimen is loaded with two moments with the ratio between them (ρ), for a mode ratio of 50% given by [3]: 3 2 ρ= 3 1+ 2 1−

(18)

Table 2 Fatigue material properties extracted from curve fitting experimental data[1] CI(mm/cycle)

CII(mm/cycle)

Cm(mm/cycle)

0.0616

2.99

458087

mI

mII

mm

5.4

4.5

4.94

389

η 2.73


Figure 6 Comparison of crack growth rate from level set model with experimental data and cohesive model results for mixed mode test

The relationship between applied moment M and energy release rate is defined as [12]: G = G= I II

3

3 2 4(1 + ) 2

M2 bEI

(19)

The fatigue material properties used in the simulation are summarized in Table 2. The results from the simulations are shown in Figure 6. Once again, the results show a perfect match with the Paris law proving the suitability of the approach.

5

Conclusions

A level set model for simulating fatigue delamination propagation under fatigue loading condition is proposed. Unlike other fatigue models mentioned in the literature which follow damage mechanics and the cohesive zone approach, the level set model is developed based on fracture mechanics theory. The crack front location is described implicitly with a level set field and this field is updated according to the velocity computed from the Paris law. The model can predict the crack growth rate in different modes of fracture precisely. This is demonstrated by simulating the propagation rates of Mode I, Mode II and Mixed-Mode tests and comparing results with experimental measurements, the analytical Paris curve and results from a cohesive model. The analysis of results reveals the higher accuracy of the level set model compared with the cohesive zone approach.

6 References [1] Turon A, Costa J, Camanho PP and Dávila CG (2007) Simulation of delamination in composites under highcycle fatigue Compos, Part A Appl. Sci. Manuf., 38:2270–2282.

.

[2] Harper PW and Hallett SR (2011) A fatigue degradation law for cohesive interface elements – Development and application to composite materials, Int. J. Fatigue, 32:1774–1787. [3] Robinson P, Galvanetto U, Tumino D, Bellucci G and Violeau D (2005) Numerical simulation of fatiguedriven delamination using interface elements, Int. J. Numer. Methods Eng., 63:1824–1848.

390


[4] Naghipour P, Bartsch M, Voggenreiter H (2010) Simulation and experimental validation of mixed mode delamination in multidirectional CF/PEEK laminates under fatigue loading, Int. J. Solids Struct., 48:1070–1081. [5] Munoz J , Galvanetto U, and Robinson P (2006) On the numerical simulation of fatigue driven delamination with interface elements, Int. J. Fatigue, 28:1136–1146. [6] Van der Meer FP, Moës N and Sluys LJ (2012) A level set model for delamination – Modeling crack growth without cohesive zone or stress singularity, Eng. Fract. Mech., 79:191–212. [7] Benzeggagh ML and Kenane M (1996) Measurement of mixed-mode delamination fracture toughness of unidirectional glass/epoxy composites with mixed-mode bending apparatus, Compos. Sci. Technol., 56:439–449. [8] Blanco N, Gamstedt EK, Asp LE. and Costa J (2004) Mixed-mode delamination growth in carbon–fibre composite laminates under cyclic loading Int. J. Solids Struct., 41:4219–4235. [9] Van der Meer FP, Sluys LJ and Moës N (2012) Toward efficient and robust computation of energy release rate and mode mix for delamination Compos, Part A Appl. Sci. Manuf., 43:1101–1112. [10] Zou Z, Reid SR, Soden PD and Li S (2001) Mode separation of energy release rate for delamination in composite laminates using sublaminates, Int. J. Solids Struct., 38:2597–2613. [11] Asp LE, Sjögren A and Greenhalgh ES (2001) Delamination growth and thresholds in a carbon/epoxy composite under fatigue loading, J. Compos. Technol. Res., 23:55–68. [12] Williams JG, On the calculation of energy release rates for cracked laminates (1988) Int. J. Fract., 36:101– 119.

391


Introduction of structural ageing-specific functions for computational models based on synaptic networks Mavrikas G. 1*, Spitas V. 2, Spitas C. 3 (1) National Technical University of Athens, Athens, Greece (2) National Technical University of Athens, Athens, Greece (3) Delft University of Technology, Delft, The Netherlands Abstract: Individual computational models for various forms of ageing exist, but are neither sufficiently versatile through simple parameterisations, nor adaptable to different scenarios of ageing. Factors to be considered for ageing modelling include among others, material, geometry/ topology, load cases and environmental conditions, ontologies resulting in general to complex and cumbersome models. To address this problem, this work proposes a method to build up computationally complex models for structural ageing, based on a limited number of simple ageing-specific function primitives, which account collectively for a wide range of ageing phenomena. By implementing these functions in Synaptic Networks (Spitas 2013), the model is exposed to flexible changes, enrichment, computational effort savings and design optimisation potential. Keywords: synaptic networks, design for ageing, computational model, design optimisation

1 Introduction Due to the nature of all the materials tending to return to a thermodynamic equilibrium with the passing of time and the additional influence from the natural and human environment, structures tend to age with repercussions ranging from capital loss due to decreased performance, maintenance, failure and replacement to loss of human lives in the case of some catastrophic failures. Therefore, engineers undertake the task to interpret the behaviour of nature into complex empirical or analytical computational models, in order to use them for inspection and prediction of the lifetime and the performance of structures. The challenging task in terms of design is to manage the complexity of the ageing models in a versatile way; while competent models for many fundamental processes for ageing are known, the design models at hand mostly depend on structure-specific results drawn from specific applications. The result is that when modifying the design or operating conditions of a structure the models that describe its ageing behaviour do not automatically update/ adapt, even though the ageing mechanisms may change- thus in principle a new study is required with each design iteration, making design for ageing a cumbersome process at best. This is a typical problem in current multi-disciplinary modelling. The ageing mechanisms, from simple mass loss (e.g. by friction, corrosion, melting), to material and shape changes (e.g. phase transition, mechanical properties change, cracking and distortion), have been approached through multi-variable governing equations, which are, in a sense, already idealized and not suitable for general use; facts that force the engineer into producing mixed models which require significant computational resources. Moreover, the lack of versatility and adaptability of the existing modelling techniques pertinent to ageing requires additional resources to be committed. The work presented, aims into bridging this gap between the ageing laws and the actual process of designing for ageing, with no adverse effects in scientific reliability or validity of modelling. Initially, an approach on ageing-specific functions (for friction wear, mechanical properties downgrading and cracks propagation) and their translation into usable, mathematical models, has been made. After this introduction, the abstracted Synaptic Networks [1] [2] [3] [4] of the ageing-specific functions are built, in order to establish the basis for the attribution of design ideas and changes on the basic model of the selected case study. This is supported by an Object-Context-Goal (OCG) *

Corresponding author. Tel.: +30-6979629156 E-mail address: [email protected]

392


classification [5], which serves to streamline and keep the design options and process transparent. The usage of Synaptic Networks and OCG Analysis does not only pay off in terms of (visual) problem understanding and depiction of the natural flow of the phenomenon through the critical ageing parameters, but also provides the advantage of interaction with the problem at any time, in order to address different design goals and routes (e.g. different case studies, design outcomes, problemspecific constraints, detailed information), but also for further design optimization. In this paper, a primary approach to the aforementioned concept has been made as proof-of-concept.

2 Synaptic Networks and OCG Analysis Sourced from the need for a new way of ideas representation balanced between casual manner and the allowance for validations and implementation in computerized environments, ideas (anything conscious) and synapses (ideas acting as idea links/ operators) are the fundamental elements of this new approach. By expressing for any chosen case the pertinent ideas from the physical and the affective domain, we may implement them as an idea algebra [3] by recognising (modelling) meaningful links/ operations (synapses) among these ideas and quantifying them with specific functions; the result is a larger model in the form of a Synaptic Network (SN) with adjustable level of complexity and predictive ability. Through Object- Context- Goal (OCG) Analysis, ideas can be understood and classified as either objects or context (ideas affected/ adjusted by the SN, or not, respectively), or goals. Contrary to goals, which are independently set, objects and context come from characterising the system being studied and differ only in the sense that objects can be modified by design intervention, whereas context cannot. OCG ideas are by nature coupled and may be represented in a (multi-dimensional) OCG-space, allowing the relative qualification of design alternatives and Pareto optimization.

3 Ageing-specific functions For the purpose of the present study we limit the discussion to two major ageing failure mechanisms, as follows: • Friction Wear • Cracking Hereunder we revisit fundamental constitutive models for these mechanisms and express them functionally in terms of SNs.

3.1

Mass loss due to friction wear

Friction is found in countless structures where there is contact and relative motion at the interface of two or more bodies. The function which describes in a quantitative way the wear (Holm [6], Archard [7]), correlates the removed volume of the material (W), with the sliding distance (S), the normal pressure (p n ), the wear coefficient (k) and the surface hardness (H) of the worn material, as follows:

W = kpn S/ H (1) By differentiating the previous expression with respect to time (t). the following expression is obtained:

wH = pnu k

(2)

where (u) is the linear sliding velocity and (w) is the volume removal-rate of the worn material. As material is worn out the geometry and even the physical properties of the interface will change, resulting in changing distributions of the operating loads, which may aggravate or relive the phenomenon, depending on the system configuration. Other failure modes may be triggered in this

393


way. In most cases, when sufficient material has been removed by wear, the concerned (sub-) system may be considered to have failed. solid body

worn volume

state critical worn volume

pressure hardness (characteristic)

sliding distance

wear characteristic

worn volume

solid body

Figure 1 SN-expression of a constitutive law for wear (left) and wear failure criterion (right)

3.2

Fatigue crack growth and sudden failure

Crack initiation and propagation are modelled here based on fundamental laws expressed from Griffith (1921), Irwin (1957) and Paris (1963). Elaborating on a first model by Griffith [8], Irwin [9] described the stress state around a crack in terms of a stress intensity factor (K I ):

K I = σ πα

(3)

When the stress intensity exceeds a critical value (K IC ), which characterises the local geometry and material, then sudden cracking ensues resulting in total structural failure of the concerned (sub-) system. In terms of SN-expressions, the above are depicted in Fig. 2. state

solid body

solid body

stress

stress intensity stress intensity

critical stress intensity (characteristic)

crack size

Figure 2 SN-expression of a constitutive law for stress development at a crack tip (left) and fracture criterion (right)

The last critical input in the theory of crack propagation in terms of fatigue failure from cyclic loading, was Paris’ publication [10] about the correlation between the crack length (α) and the number of cycles (N):

dα / dN= C ∆ K m

(4)

Where C, m are material constants and ΔK expresses the variation of the stress value (Δσ) in the hysteresis-loop diagram. Eq. 4 constitutes the ageing-specific function which controls the mechanism of crack propagation, and thus forms part of the context in the SN.

394


4 Structure-specific functions The case study selected to illustrate the merits of using SNs to analyze ageing is the wear observed on a single-cutting disk surface of a tunnel boring machine (TBM) (Fig. 3).

4.1

The cutter definition

The tools usually employed in TBMs, are disc cutters (single or double) designed in a way to put through a double task; rolling of the cutterhead and cutting process. Depending on the kind of the rock which is about to be removed by the TBM, different design parameters are manipulated specifically. Critical parameters for the cutting process and conditions referring to the geometry of the cutter, are the outer radius of the cutter (r), the inner radius of the cutter (r in ) (i.e. the smallest radius where cutting process occurs), the position of the cutter in relation to the center of the cutterhead (R), the angle of the initial slope where the cutter has been positioned (β) and the function of the generatrix (f (x) ) which by revolving produces the cutting surface. In addition, the cutting disc may have an abrasive coating as depicted in Fig. 3.6. This coating has a variable thickness depending on the distance from the centre of revolution (x) and the distance from the centre of the cutterhead (R).

3.1 Cutting ring-area and cutting force on the cutter

3.2 Radial change of cutting force and velocity

3.3 TBM cutters

3.4 Active area of cutting A

3.5 Effect of the parameters x, R, β on wear

3.6 Coating Thickness

Figure 3 Geometrical layout definition of TBM

4.2

Kinematical and dynamical analysis of cutting

The orbit of the centre of the cutter is circular and if assumed that there is always only a line in contact with the material being removed (red lines in Fig. 3.1), then the swept area during one full revolution of the cutter is a ring (Fig. 3.1). The force (F PR ) at that area, is expressed through Eq. (5), while Eqs. (6), (7) show the expressions of the horizontal and vertical components of F PR .

FPR =

T i x + R + 2 xR sin( β ) 2

= FPRΗ F= PR cos( a )

2

(5)

T i (R + xsin( β )) x + R 2 + 2 xR sin( β ) 2

395

(6)


FPRV F= = PR sin( a )

xT i cos( β ) x + R 2 + 2 xR sin( β ) 2

(7)

where (i) is the number of cutters, and (T) is the torque of the cutterhead. Accordingly, for the linear velocity (u), the expressions are the following:

u= ω x 2 + R 2 + 2 xR sin( β ) (8)

= uH ω (R + xsin( β )) (9) uV = ω x cos( β )

(10)

The force and velocity which actually cause cutting are the vertical components (F V , u V ) while the horizontal ones cause the motion (slipping) of the cutter. Fig. 3.2 depicts F PRV and u V on an arbitrary radius of the cutter. Ageing of the TBM cutters, is a classic case of ageing from friction wear of the structure, therefore the ageing-specific function in this case study, is PV=constant (eq. (2)). As mentioned above, the important step is that the state variables of this kind of ageing are defined and in this case, these are the pressure (P C ) on the active area (A' =A·cοsβ) (Fig. 3.4) and the linear velocity (u V ).

PC uV ( x , R ) =

FPRV x 2T cos( β ) uV = = const. A cos( β ) Ai ( x 2 + R 2 + 2 xR sin( β ))

(11)

5 Goal functions Goals can be set arbitrarily. A typical basic goal is that no catastrophic failure will occur under given operating conditions, which are implemented by imposing the corresponding truth values to the pertinent ideas in the SN. Additional goals –usually addressing optimization- may be introduced by expressing them as SN-expressions and appending them to the SN. To demonstrate the latter concept, we replicate here a standard design approach, where we require uniform wear across the cutting area. solid body

position pressure

state

sliding distance Figure 4 SN-expression of the basic functionality goal (left) and the constant wear-rate goal (right)

6 The complete Synaptic Network 6.1

Joining the sub-networks; design and network expansion-contraction

Essentially the complete SN is a cognitive construct that describes our complete (current) understanding of the system and problem being solved, thus including the object, context and goals. As such, it is the union of all the pertinent sub-networks, including those constructed previously.

396


Joining two sub-networks requires that one or more of their ideas overlap and is visualized in Fig. 5, where the identical ideas are marked as think circles of same colour. Introducing new ideas and joining sub-networks expands the SN. While some of the ideas can be new, in the sense that they emerge from original analysis and modeling, typically most ‘building blocks’ for the SN are readily retrieved as existing knowledge, i.e. to link stress and normal pressure featured in the wear and cracking constitutive models to the force and the local geometry of the TBM, as shown in Fig. 6. Different configurations are potentially conceivable and in that sense it is useful to compare them by implementing them synchronously in the SN, i.e. the TBM cutting area may have a rectangular or cylindrical shape and different corresponding models for volume and surface area, as shown in Fig. 7. Clearly, an eventual embodiment can only implement one of them, as expressed by the network fragment shown in Fig. 7, which assures that ‘rectangular’ and ‘cylindrical’ have mutually exclusive truth values. Such features afford fail-safe conceptual automation.

Figure 5 Visualisation of joining SNs

stress

pressure

surface area

force

force

surface area

Figure 6 SN-expression for stress (left), pressure (right)

shape

worn volume

rectangular

cylindrical

volume

shape

rectangular

dimension

397

cylindrical


Figure 7 Composite (joined) SN-expression for volume (left) and shape classification compatibility (right)

6.2

Resolving the synaptic network

For analysis purposes, any ‘what if’ scenario can be represented by applying the values of known ideas and computing their effect on the goals or other ideas of interest within the SN. To use the SN for design, we may leave a number of values a-priori undefined and calculate them for chosen limit states of the goal criteria. We call these values design Degrees of Freedom (DOFs). The visual representation of a SN can help recognize the most suitable candidates to serve as DOFs: Ideally such candidates are closely linked to the goals and sufficiently decoupled from each other to not raise conflicts. The number of design DOFs admitting unique solution is in principle limited by the number (and format) of the goal criteria. It is possible that no solution exists for a given set of goals. If a single solution exists for a given problem formulation, then designation of more DOFs will typically result in an infinity of solutions. We demonstrate here one possible solution: With the goals of low wear and spatially uniform wear (with respect to the directions x and R), we chose to make no changes to either the geometry of the standard cutter head (standard angle β, cutters positions R, mask length L m ), or to the initial geometry of the cutter (r, r in , f (x) ). We influence the hardness and wear characteristic of the cutters by applying a hard coating to the surface of the cutter. For uniform wear, the goal to be fulfilled can be visualised as the flattening of the two left curves on Fig. 3.5. Observing the SN in the vicinity of the uniform wear goal, the wear constitutive model, and the local cutter force, it emerges that each cutter must retain a minimum abrasive layer wear (t m ), expressed through the control of the thickness (t) of the abrasive layer (coating) of the tool thus an appropriate function of thickness (t (x) ) needs to be applied together with an offset in thickness each time, due to the different R (t offset(R) ). The emerging solution is explicit in the form of rational polynomial functions as follows:

t( x )

k1 x 2 , (like PV (x) ) = k 2 x 2 + k3 x + k 4

toffset ( R ) =

k5 , (like PV (R) ) k6 R + k7 R + k8 2

= tm t ( x ) + toffset ( R ) , (like PV (R) )

(12)

(13) (14)

That the same network can be solved for the same goals in many different ways by selecting different ideas as design DOFs without essentially altering the network should give an idea as to the flexibility of this method. Said flexibility extends to considering any and all parts of the context, such as the operating conditions: i.e. the rotational speed, feed, rock toughness/ torque etc. In the same SN if the peripheral force should ever rise above a level that causes critical stress intensity in the cutters, then the dominant failure mode will automatically shift from wear to fracture, either sudden or preceded by a phase of crack growth, as per Eq. (3).

7 Conclusions Ageing phenomena in structures are complex and necessitate a combination of modelling flexibility, to account for different ageing mechanisms, and computational rigour. The application of synaptic networks to the sythesis of computational models based on ageing-specific functional primitives (context) was demonstrated, allowing the assessment of the effect of design solutions (object) on important aspects of ageing behaviour (goal). Specifically, through the case study of a TBM cutting head, SNs were used to provide: • Graphical-analytical representations for the modelling realisation of the problem, in terms of the ageing-specific functions which govern different ageing mechanisms • Emergence of a system-wide model helping in the identification of object-, context- and goal-related ideas, providing a clear overview of what an engineer is able to change/ design through his analysis and what are the possible and critical decisions towards this direction.

398


•

Capability of growing the SN by enriching it with more models and deriving more design solutions at any time. • Optimisation of the design parameters according to different goals (as i.e. the uniform wear criterion). • Capability for systematic reuse of the entire SN or fragments thereof, either by simply changing the values of key parameters, or by evolving the existing model. Further research is focusing on the implementation of this model in an automated computerised environment, providing a universal tool for buildinging, synthesising and resolving/ computing SN models of ageing and other engineering systems.

8 References [1] C Spitas, V Spitas, and M Rajabalinejad, "Case Studies in Advanced Engineering Design," in Proceedings of the 1st International Symposium on Case Studies in Advanced Engineering Design, 17-18 May, Athens, 2013. [2] C Spitas, "Beyond frames: A formal human-compatible representation of ideas in design using non-genetic ad-hod and volatile class memberships and corresponding architecture for idea operators," in International Conference on Engineering Design 2013 (ICED13), 19-22 August, Seoul, 2013a. [3] C Spitas, "Definition of a functional class of ideas for Integrated Product Development and supporting theory," in 9th International Workshop for Integrated Product Development, 5-7 Sept, Magdeburg, 2012. [4] C Spitas, "Analysis of systematic engineering design paradigms in industrial practice: Scaled experiments," Journal of Engineering Design, no. 22, 7, pp. 447-465, 2011. [5] C Spitas, "Object- Context- Goal Analysis," in SIG Decision Making workshop, Paris, 2013. [6] R Holm, "Electrical Contacts," Almqvist and Wilselles, 1946. [7] J F Archard, "Contact and rubbing of flat surfaces," Journal of Applied Physics, no. 24, 8, pp. 981-988, 1953. [8] A A Griffith, "The phenomena of rupture and flow in solids," Philosophical Transactions of the Royal Society of London, no. A221, pp. 163-198, 1921. [9] G Irwin, "Analysis of stresses and strains near the end of a crack traversing a plate," Journal of Applied Mechanics, no. 24, pp. 361-364, 1957. [10] P C Paris, M P Gomez, and W E Anderson, "A rational analytic theory of fatigue," The Trend in Engineering, no. 13, pp. 9-14, 1961.

399


Modelling of ion transport and Alkali-Silica-Reaction-induced damage in concrete using continuum micromechanics and phase field models Minh N. Nguyen, Jithender J. Timothy and Günther Meschke Institute for Structural Mechanics, Ruhr University Bochum, Universitätstraße 150, 44801 Bochum, Germany Abstract: This paper presents a multi-level model for the description of ion diffusion and ASR induced damage in concrete, considered as a fully saturated deformable porous medium with two interacting constituents: the solid skeleton and ions in the pore fluid. The multiple-ions diffusion is governed by the NERNST-PLANCK-POISSON system of equations. The influence of the pore space topology on ion diffusion is calculated through a novel continuum micromechanics based homogenization model. ASR occurs when diffusing alkalis and hydroxyl ions break the silanol and siloxane bonds in reactive aggregates, forming expansive gels. Once the gel volume exceeds the available pore space, swelling strain occurs, leading to nucleation and propagation of micro-cracks in and around the aggregates. Keywords: Multiple ion transport, tortuosity, alkali-silica reaction, fracture, phase field modeling

1 Introduction

Durability related material degradation of concrete structures exposed to ionic solutions and moisture is always initiated by the transport of ions into the interior of the structure, eventually leading to corrosive processes such as calcium dissolution, corrosion of the reinforcement or chemically expansive reactions such as the alkali silica reaction. Predictive computational models are needed for prognosis of the expected lifetime of concrete structures subjected to hazardous environmental conditions. Motivated by severe damage observed in concrete road pavements, this paper presents a multi-level model for the description of multiple ion species transport and ASR induced damage in concrete. During the last decades, a large number of numerical models related to the prediction of chloride ingress and calcium leaching in saturated and unsaturated concrete have been developed [18, 23, 27, 30, 32]. However, most of the above models focused on only one ion species, i.e. either chloride or calcium, without considering interaction between ions. Recently, a model for the diffusion of multiple ion species in porous materials based upon the NernstPlanck-Poisson equation [24-26] has been proposed, taking into account the nano-physics of ion-ion interactions at the level of the pore fluid. This model has been later used to numerically investigate the migration test experiment in fully saturated condition [7, 13, 25]. The up-scaling of ion diffusion from the level of the pore fluid to the level of the pore space of concrete requires the effective diffusivity of ions within the porous material which, in general is determined from experiments [7, 25]. Due to the complexity of the pore structure, the pore connectivity and the presence of dead-end pores, the diffusion of ions in porous media is slower than in pore fluid. These aspects are usually accounted by a single phenomenological parameter, named as tortuosity [8, 20] and calculated indirectly from the effective diffusivity measured in experiments. For a model-based determination of the tortuosity, the pore network model which attempts to numerically represent the pore structure [3, 4, 29] have been proposed. In this paper, instead, a continuum micromechanics approach previously proposed by the authors is used for calculating the tortuosity [16, 33]. In order to correctly predict the initiation phase of corrosive processes in concrete, the description of the alkali transport is prerequisite. Once reaching the aggregates, the penetrating alkalis and hydroxyl ions tend to break the silanol and siloxane bonds in reactive aggregates in the solid skeleton, provoking the so-called alkali-silica reaction (ASR), which forms an expansive gel exerting a high pressure on the matrix material, and eventually leads to the deterioration of 400


the material. Various models have been proposed to numerically simulate ASR and ASR-induced damage, including models at the micro-scale [9, 10, 22, 31] as well as macroscopic models [5, 17, 28]. Although there is still a debate, whether the ASR takes place inside the aggregates [9, 10] or outside the aggregates [22, 31], there is an agreement that swelling strain only occurs when the gel volume has become greater than the available porous volume and the swelling strain can be regarded on the macroscopic level as an imposed chemical strain in the solid skeleton [17, 21, 22, 28, 31]. However, most of the models available in the open literature concerned with ASR induced damage assume that the alkali ions involved in the reaction are already present in the concrete. Hence, they do not consider the initiation phase, i.e. the penetration of the alkali ions from the surfaces into the interior of the structure. The present work is concerned with a holistic modelling of the ASR process, including the coupling between the penetration of alkali ions into concrete and a chemo-mechanical model for ASR swelling.

2 Nernst-Planck-Poisson (NPP) equations for fully saturated diffusion of multiple ion species

The transport of multiple ion species dissolved in an electrolyte solution can be represented by the NERNST-PLANCK-POISSON model [12, 13, 25-26]. In this model, a set of n NERNST-PLANCK equations, one for each ion species involved, describing the diffusion process, is coupled with a POISSON equation, describing the variation of electric potential due to the interaction between various ionic fluxes. The diffusion flux of a single ion species is governed by the first FICK’S law ji = − Di ∇ci .

(1)

When ion-ion interactions are considered, the NERNST-PLANCK diffusion flux for the i-th ion species is given by [12, 24] F   ∇ψ  , (2) ji = − Di  ∇ci + z i ci RT  

where D i is the diffusivity of the i-th ion species, R is the ideal gas constant, T is the temperature measured in KELVIN scale, z i is the valence number, F is the FARADAY constant and ψ is the electrical potential. Within a porous medium, the diffusivity can be calculated by Di =

φ D0 , τ

(3)

where D 0 is the diffusivity in a fluid with a dilute concentration of ions. φ is the volume fraction of voids, namely the porosity of the porous medium, and is equal to the fractional volume of fluid phase in case of full saturation. τ is the tortuosity, characterizing the complexity of the pore structure, the pore connectivity, dead-end pores, pore path etc. The mass conservation law for each ion species gives

∂ (φci ) + divji = 0 , ∂t

(4)

which is actually the Fick’s second law of diffusion. The complete NERNST-PLANCK equation is then obtained by substituting equation (2) into equation (4) considering the relation (3) as ∂ (φci ) φ F   − div D0  ∇ci + z i ci ∇ψ   = 0 . RT ∂t   τ

(5)

Finally, the variation of the electric potential is governed by the POISSON equation 1  F (z i ci ) = 0 , (6) div ∇ψ  + τ   χ

∑

401


where χ is the dielectric constant of the medium.

3 Tortuosity in intact porous materials Using the cascade continuum micromechanics (CCM) model [33], the homogenized macroscopic diffusivity for intact porous materials is calculated as D intact = A f (φ , n )φD0 1 ,

(7)

in which 1 is the second order unit tensor, n ∈ [0, ∞ ) is the cascade/recursion variable and

A f (φ , n ) is the inverse of tortuosity parameter τ (compared with equation (3)) Af =

1 3 n (3φ − 1) , ∀n ∈ [0, ∞ ) and φ ∈ [0,1] . = 1− n τ φ − φ − n + 2φ ⋅ 3 n

(8)

The A f curves in term of n are illustrated in Figure 1. It is noted that τ = 1 when n = 0 .

Figure 1 Inverse of tortuosity parameter

A f plotted against porosity for various cascade levels n

4 Validation of the Nernst-Planck-Poisson model 4.1

Salt removal from a fired clay brick

In this example, the experiment of salt removal from fired clay brick described in [11] is numerically simulated. The fired clay brick sample with a length of 90 mm and tortuosity of 2 is initially saturated within 4000 mol/m3 NaCl solution. The initial and boundary conditions for the numerical simulation are summarized in Table 1. Table 1 Set up for the salt removal experiment

(x, t = 0) (x = 0, t) (x = L, t)

Na+ 4000 mol/m3 0 mol/m3 0 mol/m3

Cl4000 mol/m3 0 mol/m3 0 mol/m3

ψ 0V 70 V

In Figure 2a, the profiles of Sodium concentration at t = 2 and t = 16 hours obtained numerically are presented as solid lines, while the experimental data [11] as dots. Good agreement between the numerical model and the experiment can be observed. The experimental noise and a possible, unintended evaporation through the specimen surfaces may be the sources for the small differences between the numerical and experimental data. The porosity of the fired clay brick specimens is not reported in [11]. Therefore, the value φ = 0 ,2 is taken as a referenced porosity for fired clay bricks. With the cascade continuum model in Equation (8), the best fit 402


cascade number n = 1 is obtained, corresponding to a tortuosity of approximately τ = 2,33. The Sodium profiles obtained with this calculated tortuosity are depicted in Figure 2b.

a) b) Figure 2 Sodium profiles in the salt removal test obtained by experiment (Kamran et al. 2012) and by numerical simulation after 2 and 16 hours with the tortuosity τ obtained from a) Experiment: τ = 2, b) CCM model: τ = 2,33 with φ = 0 ,2 and n=1.

5 Chemo-mechanical modeling of Alkali Silica reaction (ASR) Considering ASR induced strain εASR as the imposed chemical strain and an elastic behaviour of the material, the constitutive law for chemo-elasticity reads [17, 28, 31] σ = C : ( ε − ε ASR ) = C : ε − 3 Kε ASR , (9)

where σ is the stress tensor, C is the elasticity tensor, ε is the strain tensor, K is the bulk modulus and εASR is the ASR expansion strain, usually considered as both volumetric and isotropic. It is assumed to be related to the reaction extent of the chemical reaction leading to swelling: ε ASR = β ξ − ξ 0 1 ,

(10)

where β is the asymptotic strain in free swelling and 1 is the second order unit tensor. ξ ∈ [0,1] is the reaction extent, in which ξ = 0 represents the beginning and ξ = 1 the end of the reaction.

The operator x returns x if x > 0 and returns 0 otherwise. The term ξ − ξ 0 is introduced to take into account the fact that the ASR-induced swelling occurs only when the volume of gel becomes greater than the available porous volume [22], i.e when ξ = ξ 0 . The reaction extent is calculated as the ratio of the concentration of ASR gel created, s gel , measured in moles per unit max volume of pore solution, and the maximum concentration of ASR gel that can be produced, s gel

ξ=

s gel

max s gel

,

(11)

max The maximum amount of gel, s gel , depends obviously on the amount of alkali ions and the

reactive silica involved in the reaction. Suppose the alkali ions are supplied continuously from the outside into the interior of the concrete structures, as is, for example, the case of concrete max depends only on the amount of reactive silica available in concrete. exposed to salt water, s gel

Assume that the ASR gel is produced with fixed composition: reactive silica, alkalis (e.g. Sodium) and water in which Na/Si = 2 [22], the amount of gel created is half of the amount of Na involved in the reaction: s gel = 0,5 s Na . The amount of Na participating in the reaction can be modelled as a sink term in the diffusion equation as follows

403


∂ (φc Na ) ∂ ((1 − φ ) s Na ) + + divj Na = 0 , ∂t ∂t

(13)

where c Na is the concentration of Sodium ions dissolved in pore fluid and φ the porosity of concrete. Note that the alkali-silica reaction takes place at reactive aggregates [5, 9, 10, 17, 22], i.e, the solid matrix and it can be inferred that the lost ions are related to the ion binding phenomenon [6, 14, 27, 30]. Thus the binding isotherms can be used to describe the relation between the alkali ions participating in the alkali-silica reaction and the free ones dissolved in fluid phase, such as Linear:

Langmuir:

Freundlich:

s Na = kc Na , k > 0 ,

(14)

β s Na = αc Na , α > 0, 0 < β < 1 .

(16)

s Na =

Ac Na , A > 0, B > 0 , 1 + Bc Na

(15)

For the sake of simplicity, the linear binding isotherm is assumed in this paper.

6 Numerical examples 6.1

Concrete beam subjected to ASR swelling

Figure 3 Concrete beam subjected to ASR swelling: Geometry and supports

In this example, the bending of a concrete beam due to ASR swelling is investigated. The geometry and the supports are illustrated in Figure 3. All surfaces, except the bottom surface, are assumed to be sealed. The bottom surface of the beam is exposed to 500 mol/m3 NaCl solution. The material parameters are as follows: Young’s modulus E = 25580 MPa, Poisson’s ration ν = 0,2 and porosity ϕ = 0,2. The tortuosity τ is calculated using Equation (8) with n = 5. The linear binding coefficient is assumed as k = 0,1. For the alkali-silica reaction, the asymptotic strain is adopted as β = 0,0035 [19]. ξ 0 and s gel max, which need to be determined from experiments, are assumed that ξ 0 = 0,1 and s gel max = 50 mol/m3. The simulation is performed using 40 x 10 finite elements with 365 time steps of size Δt = 1 days. Initially, the concrete beam is assumed to be free of ions. The ion concentration at the exposed surface is assumed to increase linearly to the concentration of the solution in 10 days. Although there is no external application of an electrical field, the electric potential on the exposed surface is set to zero, to serve as a reference potential. After 3 days, the beam starts to bend. Since there is no mechanical load applied on the beam, the bending purely caused by the ASR gel swelling. Increasing the time, the penetration depth of ions increases and thus, the area affected by ASR is larger, as visualized in Figure 4a by the Sodium profile and Figure 4b by the ASR strain calculated along the axis x = 200 mm. Due to the isotropy of the strain, only the first component is considered in Figure 4b. Figure 4b also shows that the ASR strain at points close to the exposed surface tends to be higher than points far away from it, due to the generally higher ion concentration at the exposed surface. However this observation is based on the assumption that the presence of reactive aggregates is uniform within the concrete. In reality, the reactive aggregates distribute randomly in terms of both the 404


aggregate size and the spatially position within concrete. These aspects could be phenomenologically simulated by randomly assigning each element e a value pe ϵ [0, 1] to indicate the influence of reactive aggregates within the element, in which pe = 1 means the full influence and pe = 0 means there are no reactive aggregate present in element e.

a) ASR strain b) Na concentration Figure 4 Concrete beam subjected to ASR swelling: profiles of a) ASR strain and b) Na concentration at points along the axis x = 200 mm

a) t = 360 days

b) t = 720 days

Figure 5 Concrete beam subjected to ASR swelling: topology of ASR-induced fracture

In order to predict the nucleation and propagation of ASR-induced cracks, a phase-field model [2, 15] is integrated into the present ion transport - ASR model. The elastic energy, which takes the ASR-strain into account, is given by

Φ=

1 ε el : C : ε el , 2

(17)

where ε el = ε – εASR is the elastic strain tensor. Further details of the variational phase-field approach for the description of evolving fracture are contained in [2, 15]. The previously described simulation of the concrete beam is performed with the coupling of the phase-field model to the transport model using 80 x 40 finite elements for a total time of 720 days of exposure to ionic solution at the bottom surface. The goal of this analysis is to investigate the resulting ASR-induced damage in the beam. The concentration of NaCl solution at the bottom surface is increased to 1000 mol/m3. The GRIFFITH’S fracture energy for concrete is G c = 0,095 N/mm. The length parameter controlling the width of failure zone is assumed to be l 0 = 0,5 mm. As a simplication, the influence of oriented micro-cracks on ion transport is not considered in this example. For consideration of diffuse microcracking in the context of the presented micromechanics model we refer to [33]. Figure 5 shows the crack pattern induced by the expansion of the ASR gel. It is characterized by a diffuse distribution of micro-cracks in the lower part of the beam. Due to the transport of alkali ions, the reaction starts in the vicinity of the exposed surface and gradually penetrates into

405


the interior of the concrete beam. Micro-cracks are nucleated at the reactive aggregates, then propagate and join together to form macro-cracks.

7 Conclusions

In the paper a NERNST-PLANCK-POISSON (NPP) model combined with a cascade continuum mechanics model was proposed for ion transport in concrete. The transport model has been coupled to a macroscopic chemo-elastic model for alkali-silica reaction and a phase-field model for the modelling of ASR-induced cracks. The NPP model for ion transport of multiple ion species has been validated with experimental solutions. Good agreement between model and experimental results was obtained. The cascade continuum model to calculate the tortuosity of intact porous materials using a multi-level micromechanics approach was described and validated, providing an analytical tool to evaluate the effect of pore space topology on the diffusion properties of cementitious materials. The cascade level n is characteristic for each type of material and can be obtained empirically. The proposed modelling approach of coupled chemoelasticity and ion transport enables the modelling of the evolution and the time-variant depth of the ASR-affected zone according to the penetration of alkali ions into the interior of concrete structures. Since the concentration of alkalis is higher near the exposed surface, it is expected that more ASR gel is created there, leading to deterioration near the exposed surface. The modelling of damage induced in concrete structures resulting from ASR swelling is based on the phase field description of fracture. However, the influence of oriented micro-cracks on the ion diffusion has not yet been considered. This will be the topic of further research.

8 References Journal article:

[1] Bangert F, Kuhl F and Meschke G (2004) Chemo-hygro-mechanical modeling and numerical simulation of concrete deterioation caused by alkali-silica reaction, International Jounal for Numerical and Analytical Methods in Geomechanics 28:689-714. [2] Borden MJ, Verhoosel CV, Scott MA, Hughes TJR and Landis CM (2012) A phase field description of dynamic brittle fracture, Computer Methods in Applied Mechanics and Engineering 217-220:77-95. [3] Bryntesson LM (2002) Pore network modeling of the behavior of a solute in chromatography media: transient and steady-state diffusion properties, Journal of Chromatography A 945:103-115. [4] Burganos VN and Sotirchos SV (1987). Diffusion in pore networks: Effective medium theory and smooth field approximation, AlChe Journal 33(10):1678-1689. [5] Charpin L and Ehrlacher A (2012) A computational linear elastic fracture mechanics-based model for alkalisilica reaction, Cement and Concrete Research 42:613-625. [6] Coussy O and Eymard R (2003) Non-linear binding and the Diffusion-Migration test, Transport in Porous Media 53:51-74. [7] Friedmann H, Amiri O and Ait-Mokhtar A (2004) A direct method for determining chloride diffusion coefficient by using migration test, Cement and Concrete Research 34:1967-1973. [8] Garboczi EJ (1990) Permeability, Diffusivity and microstructural parameters: A critical review, Cement and Concrete Research 20:591-601. [9 Ichikawa T (2009) Alkali-silica reaction, pessimum effects and pozzolanic effect, Cement and Concrete Research 39:716-726. [10] Ichikawa T and Miura M (2007) Modified model of alkali-silica reaction, Cement and Concrete Research 37: 1291-1297. [11] Kamran K, van Soestbergen M and Pel L (2012) Electrokinetic salt removal from porous building materials using ion exchange membranes, Transport in Porous Media 96:221-235. [12] Kato M (1995) Numerical analysis of the Nernst-Planck-Poisson system, Journal of Theoretical Biology 177:290-304. [13] Krabbenhøft K and Krabbenhøft J (2008) Application of the Poisson-Nernst-Planck equations to the migration test, Cement and Concrete Research 38:77-88. [14] Martin-Perez B, Zibara H, Hooton RD and Thomas MDA (2000) A study of effect of chloride on service life predictions, Cement and Concrete Research 30:1215-1223.

406


[15] Miehe C, Welschinger F and Hofacker M (2010) Thermodynamically consistent phase-field models of fracture: Variational principles and multi-field FE implementations, International Journal for Numerical methods in Engineering. [16] Meschke G, Leonhart D, Timothy JJ and Zhou M (2011) Computational mechanics of multiphase materialsmodeling strategies at different scales, Computer Assisted Mechanics and Engineering Sciences 18:73-79. [17] Multon S, Sellier A and Cyr M (2009) Chemo-mechanical modeling for prediction of alkali silica reaction (ASR) expansion, Cement and Concrete Research 39:490-500. [18] Oh BH and Jang SY (2007) Effects of material and environment parameters on chloride penetration profiles in concrete structures, Cement and Concrete Research 37:47-53. [19] Pesavento F, Gawin D, Wyrzrykowski M, Schrefler B and Simoni L (2012) Modeling alkali-silica reaction in non-isothermal, partially saturated cement based materials, Computer Methods in Applied Mechanics and Engineering 225-228:95-115. [20] Petersen EE (1958) Diffusion in a pore of varying cross section, American Institute of Chemical Engineers Journal 4:343-345. [21] Pignatelli R, Comi C and Monteiro PJM (2013) A coupled mechanical and chemical damage model for concrete affected by alkali-silica reaction, Cement and Concrete Research 53:196-120. [22] Poyet S, Sellier A, Capra B, Foray G, Torrenti JM, Cognon H and Bourdarot E (2007) Chemical modeling of Alkali Silica reaction: Influence of the reactive aggregate size distribution, Materials and Structures 40:229-239. [23] Saetta AV, Scotta RV and Vitaliani RV (1993) Analysis of chloride diffusion into partially saturated concrete, ACI Materials Journal 90(5): 441-451. [24] Samson E, Marchand J, Robert JL and Bournazel JP (1999) Modeling ion diffuion mechanism in porous media, International Journal for Numerical Methods in Engineering 46:2043-2060. [25] Samson E, Marchand J and Snyder KA (2003) Calculatoin of ionic diffuion coefficients on the basis of migration test results, Materials and Structures 36:156-165. [26] Samson E, Marchand J, Snyder KA and Beaudoin JJ (2005) Modeling ion and fluid transport in unsaturated cement systems in isothermal conditions, Cement and Concrete Research 35:141-153. [27] Sandberg P (1999) Studies of chloride binding in concrete exposed in marine environment, Cement and Concrete Research 29:473-477. [28] Ulm FJ, Coussy O, Kefei L and Larive C (2000) Thermo-Chemo-Mechanics of ASR expansion in concrete structures, Journal of Engineering Mechanics 126:233-242. [29] Wang L and Ueda T (2011) Mesoscale modeling of the chloride diffusion in cracks and cracked concrete, Journal of Advanced Concrete Technology 9(3):241-249. [30] Yuan Q, Shi C, de Schutter G, Audenaert K and Deng D (2009) Chloride binding of cement-based materials subjected to external chloride environment – A review, Construction and Building Materials 23:1-13.

Dissertation [31] Comby I (2006) Development and validation of a 3D computational tool to describe damage and fracture due to alkali silica reaction in concrete structures. PhD Dissertation, Ecole des Mines de Paris.

Paper in conference proceedings [32] Kuhl D and Meschke G (2003) Computational modeling of transport mechanism in reactive porous media – application to calcium leaching of concrete, in Bicanic N, de Borst R, Mang H and Meschke G (Eds). Computational Modeling of Concrete Structures, Swets & Zeitlinger, Lisse, pp. 473-482. [33] Timothy JJ and Meschke G (2011) Micromechanics model for tortuosity and homogenized diffusion properties of porous materials with distributed micro-cracks. In Proceedings of Applied Mathematics and Mechanics (PAMM) 11, pp.555-556.

407


Probabilistic corrosion forecasting of steel in concrete with potentialdependent chloride threshold Andrea N. Sánchez1*, Alberto A. Sagüés1 (1) University of South Florida, Tampa, Florida, USA Abstract: The concrete chloride corrosion threshold depends on the passive steel. Inclusion of potential dependence of threshold (PDT) requires adaptive modelling, where effects of prior activation events on corrosion and potential throughout the structure are modelled at each time step. Previous work showed that PDT leads to slower long term predicted damage development than with traditional potential-independent threshold (PIT) models. The paper expands the modelling approach to probabilistically distributed surface chloride content and concrete cover. Instances are revealed where PDT can also result in earlier onset of damage compared with PIT predictions, reverting to the previously established long term behaviour later on as the system ages. The results underscore the importance of considering PDT in modelling forecasts for ageing structures. Keywords: Corrosion, concrete, forecasting, chloride threshold, modelling

1 Introduction The corrosion-related service life of a reinforced concrete structural element exposed to a chloride rich environment is obtained as the sum of the durations of the initiation and propagation stages. The initiation stage starts when the structure is put in service and ends when the critical chloride corrosion threshold CT on the steel bar surface is exceeded. At this point, the propagation stage begins where expansive corrosion products accumulate on the steel surface. The propagation stage ends when surface cracking or other given limit state is reached. Often, the time-length of the initiation stage is longer than that of the propagation stage; hence the former receives more attention in the literature. Many of the durability models available assume a system with time-invariant CT typically within the range of 0.2 and 0.5% by weight of cement. Investigations to determine the value of CT have reported values that depend on type of cement, the area of steel, the concrete-steel interface, the pore network pH, and the steel potential while in the passive condition. [1-3] Despite the limited amount of work reported on the latter, findings suggest that for E < ~ -0.2 V (SCE)1 the CT value increases significantly. [3-7]. In that domain the relationship between CT and E value is reported to approximately follow: (

)

for ETa and RHs1) atmospheric relative humidity (RHa) saturated vapour pressure (pvo) partial pressure of vapour (pv)

(c) Wetting by surface condensation (Ts1)

atmospheric temperature (Ta) surface temperature (Ts)

Figure 2 Local hydrothermal condition near the concrete surface

When the surface temperature of concrete is higher than the atmospheric temperature as illustrated in Figure 2 (a), relative humidity near the concrete surface becomes lower than the atmospheric relative humidity and consequently drying of concrete is accelerated than estimated from the atmospheric relative humidity. On the contrary, when the surface temperature is lower than the atmospheric temperature as in Figure 2 (b), relative humidity near the concrete surface becomes higher than the atmospheric relative humidity and consequently drying of concrete is retarded or

418


adsorption is accelerated. When the evaluated relative humidity near the concrete surface is over 100% as in Figure 2 (c), condensation of water takes place on the concrete surface and concrete become wetting even when the atmospheric relative humidity is less than 100%. When water is condensed on the concrete surface, the condensed water is considered to be absorbed into concrete by capillary suction mechanisms.

3 Coupled heat and moisture transport analysis in concrete 3.1

Exposure test

As discussed in 2.2, surface temperature of concrete is important to calculate drying and wetting behaviour of concrete under natural environment because local hydrothermal condition near the concrete surface is dependent of the surface temperature. In fact, the authors confirmed that numerical analysis in which local relative humidity estimated from surface temperature was used could well simulate long term drying and wetting behaviour of concrete specimen exposed in the field. However, in case of real structure, surface temperature is generally unknown. Therefore, it is necessary to estimate surface temperature by numerical analysis from the measured hydrothermal condition, i.e., atmospheric temperature, relative humidity, hours of sunlight and precipitation. Numerical simulation of heat transfer in concrete subjected to natural environment was carried out and experimentally verified. Figure 3 shows concrete specimen used in the exposed test to investigate one-dimensional heat transfer. Water to cement ratio of concrete was 50%. The size of specimen was φ100x300 mm. Specimens were cured in the mould for 111 days. Thereafter, specimens were covered with insulators except the top surface as shown in Figure 3 to ensure one-dimensional heat transfer. Specimens were exposed in two environmental conditions as shown in Figure 4. In Case A, specimens were settled under the roof in the field and accordingly subjected to only temperature and humidity change. In Case B, specimens were fully exposed in the field and accordingly subjected to temperature and humidity change, rainfall and sunlight. Two specimens were prepared for each case, which were specimen for measurement of temperature and specimen for measurement of weight change. In the specimen for measurement of temperature, surface temperature and temperature profile with respect to depth were measured. Atmospheric temperature and relative humidity, precipitation and global solar radiation as a function of time are measured and recorded by a weather meter.

Figure 3 Concrete specimen to investigate one-dimensional heat transfer

419


Case A: specimens are subjected to temperature and Case B: specimens are subjected to temperature and humidity change humidity change, rainfall and sunlight Figure 4 Conditions of exposure test

3.2

Analytical method

Surface temperature of concrete is affected by heat transfer from atmosphere, heat transfer within concrete and, in case of exposed specimen, solar radiation. These factors were taken into account in the conducted heat transfer analysis. Finite difference method was employed as a computational method as shown in Figure 5. Loss of heat energy through the insulator was also considered to improve the accuracy. Atmospheric temperature and solar radiation were given based on the measurement results in the exposure test. In moisture transport analysis in concrete, relative humidity near the concrete surface was used as boundary condition. Relative humidity near the concrete surface was evaluated from the measured atmospheric humidity and calculated surface temperature as discussed in 2.2. Material parameters in the model were estimated from mix proportion of concrete. Atmospheric temperature

Heat flux through the boundary surface

q L = − m(U 0,n − U ext ) + Rn

Solar radiation Depth from surface

Figure 5 One-dimensional heat transfer analysis of the specimen by FDM

3.3

Experimental and Analytical results of surface temperature

Experimental and analytical results of surface temperature of the specimens in Case A and B as a function of time are shown in Figure 6. Measured atmospheric temperatures are also shown. Typical weekly results in summer season are presented for each case. Left figure shows results of Case A, in which specimens are put under the roof and subjected to time-dependent temperature and humidity change but not directly affected by rainfall nor sunlight. Surface temperature of concrete follows atmospheric temperature with slight delay. Surface temperature evaluated by heat transfer analysis can successfully simulate experimental result. Right figure shows results of Case B, in which specimens are subjected to time-dependent change of temperature and humidity, rainfall and sunlight. Surface temperature of concrete exceeds atmospheric temperature in the daytime because of directly heating by sunlight. Result of heat transfer analysis considering global solar radiation can simulate surface temperature of concrete precisely.

420


45

45

atmospheric temperature measured surface temperature calculated surface temperature

35 30

40

Temperature (deg)

Temperature (deg)

40

25 20 15 10

June 2013

5

35 30 25 20 15

atmospheric temperature measured surface temperature calculated surface temperature

10 5 0

0 1 1

1 2

1 3

1 4

1 5

1 6

1 1

1 7

1 2

1 3

1 4

1 5

June 2013 1 7

1 6

Time (day)

Time (day)

Case A: specimens are subjected to temperature and Case B: specimens are subjected to temperature and humidity change humidity change, rainfall and sunlight Figure 6 Experimental and analytical results of surface temperature of concrete specimen as a function of time

3.4

Experimental and Analytical results of drying and wetting behaviour of concrete

Experimental and analytical results of weight change of the specimens, i.e., change of total moisture content in the specimens, in Case A and B as a function of time are shown in Figure 7. Left figure shows results of Case A. In this condition, specimen is gradually dried. Right figure shows results of Case B. The weight of specimen in Case B increases during the exposure test due to adsorption and capillary suction of rain. This may be attributable to the imperfect sealing in curing process in the laboratory; the specimen in Case B might have been dried before the exposure test. Considering this, initial pore humidity in the numerical analysis was assumed 92% RH in Case A and 70% in Case B, both of which were determined based on the results of parametric analysis. Numerical analysis of moisture transport can simulate experimental results in both cases. 30

Change of moisture content (kg/m3)

Change of moisture content (kg/m3)

5 0 -5

-10 -15 -20

Analysis

-25

Experiment

-30 0

50

100

150

200

250

Elapsed time (day)

25 20 15 10 5

Analysis

0

Experiment

-5 0

50

100

150

200

250

Elapsed time (day)

Case A: specimens are subjected to temperature and Case B: specimens are subjected to temperature and humidity change humidity change, rainfall and sunlight Figure 7 Experimental and analytical results of moisture content of concrete specimen as a function of time

4 Simulation of long term drying and wetting behaviour of concrete based on automated meteorological data 4.1

Modeling of meteorological data

In order to predict long term drying and wetting behaviour of a concrete structure under natural environment, it is necessary to adequately evaluate environmental action on the objective structure as well as employing an accurate transport model. In this chapter, modelling of environmental action based on automated meteorological data is proposed.

421


Figure 8 Modelling of temperature, relative humidity, global solar radiation and actual sunshine duration based on automated meteorological data

Figure 9 Modelling of wetting action by rain based on automated meteorological data

In Japan, meteorological data in various locations is provided by AMeDAS (Automated Meteorological Data Acquisition System) from JMA (Japan Meteorological Agency). It is available through Internet. Making use of this, an environmental action model which is adopted as input data for moisture transport analysis in concrete is created. Figure 8 shows method of modelling of temperature, relative humidity, global solar radiation and actual sunshine duration. In this figure, procedure to take average hourly temperature in a day in January is presented. It is assumed that averaged hourly temperature is repeated every day in January. In the demonstrative analysis in the next section, averaged hourly temperature is calculated based on ten years’ data. Figure 9 shows method of modelling of wetting action by rain. Since the length of time of raining is sensitive on the wetting behaviour of concrete, three kinds of model are examined: monthly model, daily model and short cycle model. In the short cycle model, the length of one cycle is determined so that raining time in each cycle shall be one hour. For example, one month is divided into 124 cycles in Figure 9. This procedure is taken every month. In the demonstrative analysis in the next section, ten years’ data is made use of.

4.2

Demonstration of simulation results

Based on the proposed procedure, long term drying wetting behaviour of concrete exposed in Niigata, Japan is calculated. The results are shown in Figure 10. The blue curve shows calculated results based on the original hourly meteorological data. To investigate the influence of modelling method of raining, three kinds of raining model presented in Figure 9 are examined: monthly model, daily model and short cycle model.

422


Change of water content (kg/m3)

0 -5

-10 -15

-20 Original record Monthly model Daily model Short cycle model (each raining time is assumed one hour)

-25

-30 0

1

2

3

4

5

6

7

8

9

10

11

Elapsed time (year) Figure 10 Simulation of long term moisture content of concrete in Niigata, Japan

It is clear that modelling of wetting action by rain has great influence on the long term drying and wetting behaviour of concrete. If the length of raining time in each cycle is longer than reality, wetting action is overestimated. In the results in Figure 10, short cycle model, in which the length raining time in each cycle is set one hour, achieves good agreement with the calculation result based on the original meteorological data. Though the input data here was created manually, this procedure can be automated making use of the AMeDAS provided through Internet. Therefore, long term moisture content of concrete structure at anywhere in Japan can be precisely predicted, which will be available for prediction of durability, long term serviceability and, consequently, rational ageing management of concrete structures.

5 Conclusion In this paper, moisture transport model in concrete and model of environmental action available for prediction of long term drying-wetting behaviour of concrete are presented. (1) To simulate drying and wetting behaviour of concrete under natural environment precisely, it is necessary to take into account capillary suction of rain and local hydrothermal condition near the concrete surface as well as time-dependent change of atmospheric temperature and humidity. (2) Local hydrothermal condition near the concrete surface is affected by time-dependent change of atmospheric temperature and solar radiation. (3) The length of raining duration is sensitive in the prediction of long term drying-wetting behaviour of concrete. (4) Method of modelling environmental action on concrete structures based on automated meteorological data, which can be generally applied to predict long term drying and wetting behaviour of concrete structures, was proposed.

6 References [1] Shimomura T. and Ozawa K. (1992) Analysis of water movement in concrete based on micro pore structural model, Transactions of the Japan Concrete Institute, 14, 115-122. [2] Shimomura T. and Maekawa K. (1997) Analysis of the drying shrinkage behaviour of concrete using a micromechanical model based on the micropore structure of concrete, Magazine of Concrete Research, 49 (181), 303-322. [3] Thynn Thynn H. and Shimomura T. (2009) Hybrid computational method for capillary suction and nonsaturated diffusion in concrete, 4th International Conference on Construction Materials “ConMat'09”, Nagoya 24-26 August, 2009. [4] Thynn Thynn H. and Shimomura T. (2012) Modeling of environmental action for prediction of long-term water content in concrete, Proceedings of the Japan Concrete Institute, 34(1), 688-693.

423


Chemo-Hygral Modeling of Structural Concrete Damaged by Alkali Silica Reaction Yuya Takahashi1*, Koki Shibata1, Koichi Maekawa1 (1) The University of Tokyo, Tokyo, Japan Abstract: This study aims to develop an analytical tool for predicting the expansion due to the alkali silica reaction (ASR) and crack propagation in reinforced concrete (RC). Chemical transformations for the ASR and its rate are formulated considering multi-ion migration in the pore solution and integrated with the physical properties of solid concrete and liquefied gels in both micropores and crack spaces. The poromechanics-based models are verified with experimental facts; the results show that the proposed model may capture the ASR mechanics of damaged members under various confinement conditions and reactive aggregate contents. Finally, crack propagation in structural concrete is focused on with RC beams, and obvious cracks along steel bars are simulated as structural aging. Keywords: Alkali silica reaction, Poromechanics, Confinement, Expansion

1 Introduction

The alkali silica reaction (ASR) is characterized by its anisotropy of expansion under multidirectionally restrained conditions. Cracks due to ASR expansion in reinforced concrete (RC) structures usually occur along steel bars because of the stiffness anisotropy, and this is strongly related to the material properties of the ASR-gel. The ASR-gel is thought to be a semiliquid that can move along cracks or into pores, causing pressure release. Hence, stress fields generated due to the ASR in RC structures are highly complicated. In this sense, durability mechanics seems to be a key aspect in discussions on the long-term performance of concrete structures subjected to time-dependent expansion by the ASR. The pressure induced by the ASR is affected by not only material properties (chemical reactions, gel properties, etc.) but also mechanical conditions (confinement by steel, crack presence, etc.). Besides, material and mechanical processes strongly influence each other. Therefore, in order to understand and predict the post-ASR expansion behaviors of structural concrete, a coupled approach considering both material and mechanical phenomena is thought to be essential. Meanwhile, the authors are developing a multiscale chemo-hygral computational system, DuCOM-COM3 [1], which can conduct 3D multiscale analysis of structural concrete. Recently, Fujiyama et al. [2] studied the rate-dependent model in combination with the kinematics of condensed water and built a two-phase model [3] to consider water pressure and its intrusion into cracks and micropores under active load conditions. In this study, the two-phase poromechanics model is applied to ASR expansion and integrated with thermodynamic models to satisfy mass conservation. Then, the proposed models are verified with previous experimental facts, and the crack behaviors in RC structures are studied.

2 ASR-induced Expansion and Crack Propagation 2.1

Basic models for ASR expansion

The proposed ASR-induced expansion model is built on the basis of poromechanics, which has been used in geotechnical engineering applications such as consolidation and liquefaction of soil foundation. The ASR-gel is treated as the medium existing among crack spaces and microvoids. *

7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, JAPAN, [email protected]

424


Figure 1 Modeling of ASR expansion and pressure formation

Figure 2 Calculation scheme of ASR-expansion simulation

Figure 1 shows the modeling of the ASR expansion and stress distribution formation. First, based on the chemical equations for the ASR and expansion (Eqs. (a) & (b)), the ASR rate is formulated as a function of alkali concentration, free water amount, and reactive aggregate amount. Here, sodium and potassium are considered as alkalis reactive for the ASR, and for the first simple equation, a linear equation is applied (Eq. (c)). The coefficient of reaction rate, k, is set to 0.1×10-7 on the basis of the results of sensitivity analysis, and the effect of relative humidity (RH) is also considered in Eq. (d). The generated ASR-gel volume is calculated assuming X 2 Si 2 O 5 (H 2 O) 8.4 as the ASR gel molecular formula for each alkali, X ( X is Na or K)[4]; the consumed alkali amount and water amount are also calculated. It is important that the amounts of water and alkali considered when calculating the reaction rate as well as the amounts consumed be treated as global variables between the thermodynamic analytical system, DuCOM, and the 3D mesoscale structural analytical system, COM3. The overall calculation scheme is shown in Figure 2. In DuCOM, the multi-ionic approach developed by Elakneswaran et 425


al. [5] is used for Na+ and K+ mass balance computations. Through this scheme, strong coupling between the material properties and the mechanical phenomena can be achieved. For example, the water content calculated by the thermodynamic modeling strongly affects the ASR-gel production rate, while water consumption due to the ASR restricts further hydration. On the basis of the generated gel volume, stress formation is calculated. Here, we must take into account the physical properties of the ASR-gel. As mentioned before, the ASR-gel is supposed to be a semiliquid material. In order to express the solid-liquid coexisting states, parameter β is introduced, which is the ratio of the liquid phase to total ASR-gel. In the case of certain stress conditions, the solid phase part of the ASR-gel can expand isotropically under anisotropic pressure distribution (Eq. (j)), while liquid phase part expands while deforming anisotropically under isotropic pressure (Eq. (k)). The total pressure can be computed as the sum of those pressures (Eq. (i)). Parameter β is assumed to be time-and-pressure-dependent. First, β is set to 0.00, which implies that the gel is completely in the solid state, and then, according to the pressure gradient and time, β gradually increases up to 0.90. It physically expresses the intrinsic viscosity of the ASR-gel. When crack formation occurs, the ASR-gel migrates into the cracks and pressure can be released. However, because of the high viscosity of the gel, the migration is delayed. By considering the time dependency of β, the aforementioned delay can be simulated. In the modeling, we take into account not only migration into cracks but also absorption into large pores in cement pastes. The details are described in the next section.

2.2

Modelling of absorption into capillary pores

In previous studies, it has been proposed that some amount of the ASR-gel can penetrate the capillary pores in the surrounding cement paste and the absorbed gel does not contribute to pressure rise [6]. Laurent et al. [7] also suggest that the high-porosity zone around the reactive aggregate should be considered in calculating pressure with the ASR. In this study, such absorption into capillary pores is considered as per Wasuburn’s equation (Eq. (h) in Figure 1). The minimum radius to which the ASR-gel can penetrate under pressure can be specified by the equation, and pores whose sizes are larger than the minimum radius are regarded to contribute to the potential volume for gel absorption. If the existing ASR-gel volume is less than the potential volume, all of the gel is absorbed in the pores and no expansion may occur. Expansion occurs only after the gel volume exceeds the potential volume. Parameter Z in Eq. (h) is the coefficient for gel absorption, and it is determined based on the contact angle θ and viscosity σ of the substances. In the case of mercury, Z can be approximately 0.38, but there is no report on the θ and σ values for the ASR-gel. Hence, the value of Z for the ASR gel is determined by sensitivity analysis, using the free expansion test reported by Muranaka et al. [6]. The test and analytical conditions are explained in chapter 3. Figure 3 shows the results of the analyses for free expansion. We can see that the starting time as well as the total expansion change according to Z. By considering gel absorption into capillary pores, restriction of expansion for dozens of days in the early stage can be expressed. The appropriate value for Z was determined to be 0.5, by comparing the analysis results with the experimental facts.

Figure 3 Sensitivity analyses for coefficient of absorption into capillary pores

426


Thus, ASR expansion and the associated stresses are modeled and included in the existing multiscale chemophysical analysis system.

3 Verification: Uniaxial Confinement

To verify the proposed models, simulations of previous experiments are conducted. The target experiments are the uniaxial confinement experiments subjected to ASR expansion [6, 8-10]. Prism specimens are cast with various amounts of reactive aggregates, and steel bars are arranged in the longitudinal direction under specific steel ratios (Figure 4). Tables 1 and 2 show the mix proportions and experimental conditions used in targeted four previous studies [6, 810]. Tests A and B are performed with relatively high ratios of the reactive aggregate, while Tests C and D are performed with lower ratios. In Test D, a low steel ratio is considered. Meshes with sizes of 2–3 cm are used for the simulations, as shown in Figure 3, and length change developments in the longitudinal and transverse directions are observed.

Figure 4 Test image and mesh for uniaxial confinement tests Table 1 Mix proportions W/ C

W

C 3

S(non-

S

reactive)

(reactive)

3

3

3

G (non-

G

reactive)

(reactive)

3

Na 2 Oe 3

q

(%)

(kg/m )

(kg/m )

(kg/m )

(kg/m )

(kg/m )

(kg/m )

(kg/m3)

Test A (Muranaka et al.[6])

45

169

376

775

0

0

925

8.0

Test B (Tsukada et al.[8])

55

165

300

416

434

502

503

7.2

Test C (Yamura et al.[9])

45

203

450

660

0

912

101

9.0

Test D (Koyanagi et al.[10])

50

170

340

634

159

797

200

10.2

Table 2 Experimental conditions Specimen size

Curing conditions

Steel ratio (%)

Test A

100×100×400 mm

1 day sealing  40˚C RH100%

0.0, 0.7, 1.5

Test B

150×150×530 mm

Test C

100×100×400 mm

7 days sealing 40˚C RH100%

0.0, 0.71, 1.27, 1.99

Test D

100×100×400 mm

1 day sealing  40 ˚C RH ≥ 95%

0.0, 0.07, 0.14, 0.40, 0.90

28 days wet cloth curing  40 ˚C RH ≥ 95%

0.0, 0.3, 0.6, 1.3

The results for length changes in the longitudinal direction obtained with the proposed models and the experimental results are shown in Figure 5. In the analyses, the proposed model can express drastic decreases in the length change with increasing steel ratio. Thus, the proposed models can reproduce the overall trends concerning the effects of confinement by steel, yet there exists some discrepancy between the analyses and experimental results. 427


For lower reactive aggregate ratios, in the experiments, expansions were found to start at day around 10, while the analyses indicated a longer period without expansion. This is thought to be due to the difference in the reactivity of aggregates in the experiments. In Tests C and D, only 20% or 5% of the total aggregates are reactive for the ASR, but the total length changes of free expansion (steel ratio = 0.0%) are identical to or greater than the results of Tests A and B. The reaction rate in each test seems to be quite different. So far, a constant coefficient for reaction rate, 0.1×10-7, has been applied to all analyses. In the future, detailed analysis of this coefficient must be carried out as well as the formula itself for the ASR rate. In the analyses with lower steel ratios (the 0.07% case in Figure 5 (vii)), the total length changes are slightly larger than the free expansion. This is because of the existence of steel plates on both sides of the specimen. These steel plates confine expansion in the transverse direction at both ends of the specimen, and the confinement affects the increased expansion in the longitudinal direction. In the experiments, this effect seems to be negligible. Interaction between specimens and steel plates needs to be studied as well in the future.

Figure 5 Length changes in longitudinal direction

428


Figure 6 Length changes in the transverse direction

Figure 6 shows the analytical and experimental results for length changes in the transverse direction in Test B. It is well simulated that the steel ratio does not drastically affect the length changes in the transverse direction, in contrast to the large decreases in the longitudinal direction shown in Figures 5 (iii) and (iv). Such sophisticated deformation phenomena can be simulated with the proposed models considering the coupled chemical and physical phenomena.

4 Crack Propagation in RC

Based on the verified analytical system, crack propagation behaviors are studied via the analyses of RC beams. With various steel bar arrangements, ASR expansions are simulated and the locations and directions of the cracks formed are observed. Figure 7 and Table 3 show the analytical conditions for the simulations. RC beams with dimensions of 150 cm×10 cm×18 cm are simulated by referring to Koyanagi’s study [11], except for the steel bar arrangements. Taking advantage of symmetry, only half of the beam is modeled (Figure 7) and 3 different steel bar arrangements (Cases 1–3) are set, as shown in Figure 7. Case 1 is the standard case. Compared with this standard case, Case 2 has a larger cover at the bottom and Case 3 has a larger cover on the side of the beam. The total steel ratio and other conditions are the same among all the 3 cases. No web reinforcements are placed. The mix proportion is shown in Table 3. In the simulation, all the surfaces of the beams are kept wet for 14 days after casting and then exposed to the 40˚C RH99% condition.

Figure 7 Analytical meshes and steel bar arrangements Table 3 Mix proportion for analyses W/C

W

C

S (non-reactive)

S(reactive)

G(non-reactive)

G(reactive)

Na 2 Oeq

(%)

(kg/m3)

(kg/m3)

(kg/m3)

(kg/m3)

(kg/m3)

(kg/m3)

(kg/m3)

50

177

354

768

0

0

962

8.1

429


Figure 8 shows the distributions of normal strain on day 100 in the x, y, and z directions, which express the crack occurrence in each direction. We can see significant changes in the behavior of crack propagation with differences in the steel bar arrangement. First, focusing on the results for Case 1, the bottom part is restrained in the longitudinal direction (X-direction) with steel bars; hence, bending occurs and bending cracks appear on the upper face (Figure 8 (i)). For other surfaces, cracks along the steel bars can be observed at the bottom (Figure 8 (ii)) and on the side of the beam (Figure 8 (iii)). ASR-gels restrained in the longitudinal direction tend to move to other Y and Z directions, and that results in similar cracks. Besides, the corner of the beam is largely deformed according to the high confinement with the bended steel bars at the corner. Such a crack behavior can be seen in real structures as well. Figure 9 shows images of concrete structures subjected to serious ASR deterioration, obtained by the authors. Cracks along the main steel appear in the girder, and large cracks around the corner of the column can be seen. These are the typical crack situations in the concrete structures that suffer ASR expansion. Thus, we demonstrate that the analytical system proposed in this study can be used to simulate such a characteristic RC deterioration in the ASR. Comparing the results of other cases with Case 1, the crack behaviors are seen to be significantly different. In Case 2, wherein the bottom cover is thick, cracks on the bottom faces are reduced (Figure 8 (v)), and the crack position on the side face is shifted upward (Figure 8 (vi)). In Case 3, with a thick cover on the side surface, cracks on the side face are reduced. Such minute differences in the steel bar arrangement are seen to cause large differences in the crack behaviors. From these results, it is easily visualized that the crack behaviors of RC with multidirectional steel bar arrangements can be more complicated and difficult to predict. In this sense, the 3D analytical system with the material and mechanical models proposed in this study can be a powerful tool for predicting such complex behaviors.

Figure 8 Simulated normal strains in each direction

5 Conclusions In this study, a coupled calculation approach for the material and mechanical behaviors of ASR gel is adopted and applied to the existing analytical system for simulating the deterioration due to ASR expansion in structural concretes. Generated ASR gel volumes are calculated based on the mass conservation of water and alkalis. In association with the modified two-phase model 430


Figure 9 Crack situation in a real bridge

similar to the poromechanics of soil foundation, stress-strain relations are calculated by considering the gel pressure and its intrusion into both crack spaces and microcapillary pores. The proposed models are experimentally verified using plain and reinforced concrete. Uniaxial confinement tests under various steel ratios and amounts of reactive aggregates are simulated. The analytical results show significant reduction of expansion along steel bars, unlike the case of expansion in the orthogonal direction without the stiffness of steel bars. It is confirmed that overall behavioral trends can be reproduced with the proposed models, although differences in the reactivity of each aggregate needs to be studied in more detail. Crack behaviors of RC beams are also studied analytically. RC beams under different steel bar arrangements are targeted. The crack behavior observed in the study matches the typical crack patterns for the ASR deterioration modes of concrete members in real structures. It can be concluded that the proposed model may simulate anisotropic stress development accompanying ASR expansion in structural concrete.

Acknowledgments


6 References

[1] Maekawa K., Ishida T. and Kishi T. (2008). Multi-Scale Modeling of Structural Concrete, Taylor and Francis [2] Meakawa K. and Fujiyama C. (2013). Rate-dependent model of structural concrete incorporating kinematics of ambient water subjected to high-cycle loads, Engineering Computations, Vol. 30, Iss: 6: 825-841 [3] Biot, M.A. (1963). Theory of stability and consolidation of a porous media under initial stress, Journal of Mathematics and Mechanics, 12(2):521-541 [4] Ichikawa, T. and Miura, M. (2007). Modified model of alkali silica reaction, Cement and Concrete Research, Vol.37: 1291-1297 [5] Elakneswaran, Y. and Ishida T. (2013). Integrating physicochemical and geochemical aspects for development of a multi-scale modeling framework to performance assessment of cementitious materials, Multi-scale Modeling and Characterization of Infrastructure Materials, RILEM BOOKSERIES, Volume 8:63-78 [6] Muranaka M. and Tanaka Y. (2013). Development of physical and chemical model for concrete expansion due to ASR based on reaction mechanism, Journal of JSCE, E2/V-69, No.1: 1-15 [7] Laurent C. and Alain E. (2014). Simplified model for the transport of alkali-silica reaction gel in concrete porosity, Journal of Advanced Concrete Technology, vol.12:1-6 [8] Tsukada T., Koga H., Hayakawa T., Watanabe H. and Kimura Y. (2010). Basic study about expansion and restriction of structural concrete deteriorated with alkali silica reaction, Proceedings of JSCE, V-276: 551-552 (in Japanese) [9] Yamura K., Nishibayashi S. and Tanaka S. (1989). Study about influence of steel restriction on alkali aggregate reaction, Proceedings of JCI, 11(1): 135-140 (in Japanese) [10] Koyanagi W., Uchida Y., Iwanaga T. and Asano Y. (1998). Restraint effect of ASR expansion in RC members with small reinforcement ratios, Cement Science and Concrete Technology, No.52: 786-791 [11] Koyanagi W., Rokugo K. and Ishida, Y. (1985). Some properties of RC member with cracks due to alkaliaggregate reaction, Cement Science and Concrete Technology, No.39: 352-355 (in Japanese)

431


The influence of cement fineness on ageing of cementitious materials

Y. Zhang1 *，K. van Breugel1，Zhiwei Qian1 (1) Delft University of Technology, Delft, The Netherlands Abstract: In old days, many structures were built using coarse cement. The durability of these structures is better than that of the modern structures built with fine cement. Obviously, in those old structures ageing processes proceed slower. A probable reason is that there are many unhydrated cement particles in old structures and these particles start to hydrate many years after the structure is constructed. Hence it is an interesting topic to figure out the influence of fineness of cement on the durability of cementitious materials. In this study a modelling procedure is developed to predict the effect of cement fineness on the durability and ageing processes in cementitious materials. The HYMOSTRUC3D model is employed to simulate the development of the microstructure of cement paste. Then the lattice fracture model is used to capture the microcracks in a virtual microstructure due to tensile loads. After that the water permeability coefficient of the paste and the transport properties are predicted. Keywords: Aging, fineness of cement, further hydration, permeability, durability.

1 Introduction The infrastructures make out more than 50% of the nation’s assets. Concrete structures represent a large share of the infrastructures. For several reasons our infrastructure ages, for example due to interaction with the environment (temperature, moisture) and internal degradation process. The process goes very slowly and it is not easy to make an assessment of this process. The aging of infrastructure is a worldwide concern and huge amounts of money are spent in either upgrading of aging structures or new buildings. For example, in the U.S., more than 26 % of the bridges are rated either structurally deficient or dilapidated, requiring an estimated annual investment of about $17 billion for rehabilitation [1]. When exposed to corrosive environments like de-icer salts and seawater, serious durability problems have occurred in bridge decks, parking garages, undersea tunnels, and other marine structures less than 20 years old[2-4]. However, more than 2000 years old unreinforced concrete structures, such as the Pantheon in Rome and several aqueducts in Europe, made of slow-hardening, lime-pozzolan cements, are still exhibiting excellent performance, while many reinforced concrete structures built with fine Portland cement in the 20th century are quickly deteriorating. A probable reason is that there are many unhydrated cement particles in old structures. These particles start to hydrate many years after the structure is constructed, which makes these buildings less prone to ageing processes. It is an incontrovertible fact that the fineness of Portland cements has increased during the past decades and is continuing to increase. Figure 1[5] summarizes the mean values from three surveys as presented by Tennis and Bhatty[6] and also includes individual results from the Cement and Concrete Reference Laboratory (CCRL)[7]. Hence, it is worthwhile to figure out the relationship between the durability of cementitious materials and the fineness of cement.

*

Y Zhang Delft University of Technology, Faculty of Civil Engineering and GeoSciences / Microlab, Stevinweg 1, 2628 CN Delft, The Netherlands E-mail: [email protected]

432


Figure 1 Changes in the Blaine fineness of cements from the 1950s. Regression lines are provided for the Type I survey and CCRL Type I or I/II data sets only.

2 Problem Statement In terms of the influence of cement fineness on the aging of cementitious materials, two aspects should be taken into account: First, cementitious materials made using cement with different fineness cement have different microstructures. Different microstructures deliver different properties (both mechanical properties and transport properties). Eventually, materials exhibit different (micro)crack patterns, transport abilities, both of which are widely considered as key factors related to self-healing of unhydeated cement and further to aging problems. After construction, the anti-aging ability of concrete structures highly depends on the durability of cementitious materials and further on transport properties. Based on many experimental studies conducted by other researchers, it is believed that further hydration of the unhydrated cement contributes a lot to block (micro)cracks in cementitious materials. Compared to fine cement particles, coarse cement particles have more unhydrated cement. Therefore, coarse cement particles have a higher potential of self-healing. Hence different potential of self-healing ability caused by the size of cement particles should be considered.

3 Methodology

A modelling procedure illustrated in Figure 2 is developed to predict the proneness to ageing of cementitious materials, taking into account the influence of cement fineness. The HYMOSTRUC3D[8] model is employed to simulate the microstructure of cement paste for different finenesses of the cement. Then the Lattice Fracture model[9] is used to capture the microcracks due to tensile loads. Then the void elements, solid elements, cracked elements and self-healed elements are assigned different water permeability coefficients according to their transport abilities. The HYMOSTRUC3D model is able to take into account the influence of cement fineness on microstructure. Besides, this microstructure also shows the different potential of self-healing ability caused by the size of cement particles. In particular, Figure 3a is a microstructure by using coarse cement, while Figure 3b is a microstructure by using fine cement. After generating similar (micro)cracks, the amount of unhydrated cement in (micro)cracks in Figure 3a is more than that in Figure 4b, which indicates the two microstructures have different potential of selfhealing.

433


Figure 2 Structure of the modelling procedure

To simulate (micro)cracks, the 3D Lattice Fracture model is employed to generate (micro)cracks in cementitious materials. The combination of this model with HYMOSTRUC3D model makes it possible to consider the effect of cement fineness on microstructures and produce realistic multiple (micro)cracks.

a

b

Figure 3 Different self-healing potential caused by, a) coarse cement; b) fine cement

After that 3D Lattice Transport model is adopted to predict the transport properties of three networks, which are, initial lattice network, (micro)crack lattice network and self-healed lattice 434


network, as shown in Figure 4. The difference in transport properties of the initial lattice network and the (micro)cracked lattice network indicates the influence of the damage degree on the transport properties, while the difference between the (micro)cracked lattice work and selfhealed lattice network reveals the efficiency of the self-healing process.

Figure 4 modelling procedure to predict transport properties of three networks

4 Simulations In order to investigate the influence of cement fineness on water transport properties of the cement paste, two sets of cements with different Blaine values are used in the simulations. All parameters are the same except the fineness, as given in table 1. Cement Type OPC I OPC I

Table 1 Parameters of materials Blaine Values Degree of (m2/kg) Hydration 210 69% 600 69%

W/C Ratio 0.3 0.3

To limit the computation time, a cube of 20×20×20µm is cut form the original sample of 100×100×100 µm. Each direction is divided into 20 voxels, which means that each voxel is 1× 1× 1µm, as illustrated in Figure 5.

a

b

Figure 5 Microstructure of 20µm cube cut from 100µm cube : a) Blaine value 210 m2/kg; b) Blaine value 600 m2/kg;

From Figure 5; it can be seen that, compared with (b), (a) has coarser cement particles and, therefore, contains less particles. Although the original fineness of 100µm cube is 210 m2/kg and 600 m2/kg, the specific fineness of 20µm cube is not known. This paper focuses mainly on developing a procedure to investigate the influence of cement fineness on water permeability of 435


the paste. Nevertheless, the microstructures are apparently different, and there is more unhydrated cement in the sample with coarse cement (a). Hence, it has higher self-healing ability. The spherical particles in the microstructure are converted into a voxel-based digital image, based on which a lattice mesh and discretization process is carried out to generate initial lattice network, illustrated in Figure 6.

a

b

Figure 6 Initial lattice network of specimen cut from original sample with: a) Blaine value 210 m2/kg; b) Blaine value 600 m2/kg;

Each voxel randomly generates a node, and a triangular lattice system is preferred because it has a better connectivity compared with a quadrangular lattice system. The concave area in Figure 6 represents the void phase, as void contributes nothing to resisting mechanical loading, so no element is generated in void phase. After performing the tension test, a lattice network with 2000 broken elements is selected as a partially damaged system for the subsequent water transport simulation, shown in Figure 7.

a

b

Figure 7 Partially cracked lattice network of specimen cut from original sample with: a) Blaine value 210 m2/kg; b) Blaine value 600 m2/kg;

The connectivity of broken elements is also checked. All cracked elements, which are directly or indirectly interconnected via voids or broken elements to one of the six surfaces, are identified to be able to heal due to the self-healing ability. The reason is that water can reach the cracked elements and, therefore, unhydrated cement particles can hydrate further and block 436


microcracks. When calculating the water permeability, voids dominate the transport ability. Therefore, the lattice network should contain all voids, as illustrated in Figure 8.

Figure 8 Lattice network for water permeability calculation.

To predict the water permeability, different coefficients are assigned to different element types according to the transport properties of these elements, as shown in table 2. Table 2 Water permeability coefficient of different element types Cracked solid Element type Void element Solid element element Permeability coefficient (cm2/sec) 1.0E-8 1.0E-15 5.0E-9

Self-healed element 1.0E-10

The overall water permeability for different microstructures is listed in table 3. From this table; it can be seen that with different fineness of cement, the water permeability is obviously different. Different particle size distribution determines the geometry of the microstructure and hence exhibits different transport properties, and furthermore, possesses a different potential self-healing ability. Table 3 Overall water permeability

Blaine values(m2/kg)

Permeability of Original structure

Permeability of Cracked structure

Permeability of Self-healed structure

210

5.09E-10

9.01E-10

5.21E-10

600

7.15E-11

6.61E-10

3.06E-11

5 Results and discussions Although the selected size of the specimen is insufficient to demonstrate the exact relationship between cement fineness and water permeability of real concrete mixtures, the presented

437


approach can take into account the influence of cement fineness on water permeability. Several tentative conclusions are the following:

a) In this study several existing models are integrated and an approach is developed to take into account the influence of cement fineness on water permeability. A coarser cement results in a cement paste with a higher water permeability at the beginning, but exhibits better potential self-healing ability when compared with pastes made with a finer cement.

b) By combining HYMOSTRUC model and 3D Lattice Fracture model, the calculation procedure proposed in this paper can deal with the occurrence of multiple realistic (micro)cracks in the cement paste. The direction, the width, the length and the shape of (micro)cracks are arbitrary, therefore this model is suitable for various (micro)cracks. In further study, a representative element volume(REV) of the dimension 100×100×100µm will be used to quantify the influence of cement fineness on transport properties of the paste. The influence of other parameters also will be investigated, for instance, water cement ratio, degree of hydration, and cement type, etc. Several aspects need to be studied, though. For example, how to calculate the amount of further hydration products and their distribution in 3D space exactly and efficiently.

6 References

[1]. American Society of Civil Engineers (2009), “2009 Report Card for America’s Infrastructure”. [2]. Mehta, P. K. (1999) “Concrete Technology for Sustainable Development,” Concrete International, V. 21, No. 11, pp 47-53. [3]. Report of the National Materials Advisory Board (1987) “Concrete Durability — a Multimillion Dollar Opportunity,” NMAB-37, National Academy of Sciences, 94 pp. [4]. Litvan, G., and Bickley, J. (1987) “Durability of Parking Structures: Analysis of Field Survey,” Concrete Durability, Katharine and Bryant Mather International Conference, SP-100, J. M. Scanlon, ed., American Concrete Institute, Farmington Hills, Mich, 1987, pp. 1503-1526. [5]. Early-Age Properties of Cement-Based Materials. I: Influence of Cement Fineness [6]. Tennis, P. D., and Bhatty, J. I. (2005). “Portland cement characteristics—2004.” Concr. Tech. Today, 26_3_, 1–3. [7]. Cement and Concrete Reference Laboratory. (2007). http://www.ccrl.us. [8]. Klaas VAN Breugel (1997) Simulation of Hydration and Formation of Structure in Hardening Cement-based Materials. PhD Dissertation, Delft University of Technology. [9]. Zhiwei Qian (2012) Multiscale Modelling of Fracture Processes in Cementitious Materials. PhD Dissertation, Delft University of Technology .

438


CHARACTERIZATION AND MONITORING

439


Chloride Profiles in Concrete After 100 Year of Service in Panama Canal C. Andrade1*, N. Rebolledo1, A. Castillo1, R. Perez2, M. Baz2 (1) Institute of Construction Sciences (IETcc) - CSIC- Spain (2) Grupo Unidos por el Canal, GUPC Abstract Chloride ingress in concrete is one of the main causes of reinforcement corrosion and in consequence, its prediction in new or existing structures is of capital importance for the design or management of aging of reinforced concrete. Although several chloride penetration models have been developed in recent years, all of them lack of calibration at terms longer than around 30 years because the record of short-long term testing was not developed before. In present paper data of cores drilled from the 100 years old Panama Canal walls are studied to analyze the chloride profiles found and their interpretation on the light of concrete proportioning and microstructure. A “skin” thicker than 10 mm is found and total and water soluble chloride profiles indicate that they give similar apparent diffusion coefficients. The chloride concentration logically depends on the external salinity. Finally, a theoretical exercise is made in a back extrapolation into the possible initial concrete properties in order to deduce an “aging” factor which could be reasonable in terms of the observations made at present. Keywords: concrete, chloride, diffusion, long-term data.

1 Introduction Panama Canal has constituted a route for passenger and freight traffic of cardinal importance for world trade during the last 100 years. The works to build a two new set of locks were awarded to the “Grupo Unidos por el Canal” (Spanish initials, GUPC), whose engineering division is headed by a Spanish firm, Sacyr S.A. There are four levels of locks in the new facility between the Atlantic or Pacific Ocean and the (almost entirely freshwater) inland lakes. The Panama Canal Authority’s, ACP, specifications require a 100-year service life for the concrete in all members, defined to mean conformity with the 1000-coulomb for electrical charge set out in ASTM 1202 [1] and application of a reliable method for calculating service life. They also establish a series of requisites to minimise heat of hydration-induced cracking in the concrete and cover depths of around 10 cm. Concrete resistance to chloride ingress is based both on low porosity and the detention of chloride advance via absorption by the C-S-H in the cement paste or the reaction between chloride and the hydrated aluminium phases [2-4]. One parameter that can be used to measure such resistance is the chloride ion diffusion coefficient, or rather the “apparent” coefficient, which takes both transport across the pore network and the reaction with hydrated phases into consideration. It is normally modelled by solving for Fick’s second law of diffusion in a nonsteady state, assuming that the outside chloride concentration and its diffusion coefficient remain constant [5]. However, these assumptions do not reflect actual conditions because the coefficient has been shown to decline with time [6] and the chloride surface concentration to vary [6-7]. These observations cast doubt on the reliability of predictive models, whose empirical calibration has been limited to periods of not over 25 years. Present paper gives the chloride profiles that were obtained in cores drilled from the old Panama Canal and their analysis with regards to their features and back extrapolation to possible values of the diffusion coefficient at shorter terms. The results are compared with results obtained in the concrete mixes now being used in which aging factors are measured as previously published based in the monitoring of concrete resistivity evolution on time [8-9].

440


2 Experimental A total of 18 cores were drilled from different zones of the old Panama Canal. The zones were “Miraflores” Upper and Lower chambers and from “Pedro Miguel” East wall. The cores arrived to our laboratory shipped from Panama in a wooden box and wrapped with a wet black tissue and inside shielded plastic bags as is shown in Figure 1.

Figure 1 Aspect of the cores as received

The cores were immediately cut in slices and grinded the most superficial one in order to avoid any further change in the concentration profiles due to storage out of the structure. The procedure and techniques used for the characterization of the cores was the following: 1) Chloride profiles were made by grinding from core’s surface to approximately 300 mm in depth. Immediately after the powder samples were collected, they were individually sealed in properly identified plastic bags and send to the chemical laboratory for chloride analysis, 2) From the collected powders: chloride concentration, X-ray diffraction and Thermal differential analysis ATD/TG were performed, 3) In the inner part one slide of 40 mm was cut in order to measure the electrical resistivity and 4) From the most inner parts samples were taken for mercury porosimetry (MIP) and natural diffusion test (ASTM C1543). The chloride concentration was determined in accordance with ASTM C1152, Standard Test Method for Acid-Soluble Chloride in mortar and concrete. The samples obtained were analysed for the total chloride content as per Spanish standard UNE 83986. All profiles were measured starting from the core end corresponding to the concrete surface. The total chloride content was measured at various depth increments in multiple different cores from each chamber. Chloride penetration was also measured by bottom ponding (ASTM C1543). The test was conducted as laid down in ASTM standard C1543. A section of pipe was sealed to the top of the specimen with silicone and filled with a 30-g/L solution of NaCl. The entire piece was then covered with plastic film to prevent the solution from evaporating across the side walls of the specimen. The specimens were ponded for 90 days, during which time samples were tested to obtain the chloride profile by milling 2 mm off the surface of the specimen at a time. The non-steady state diffusion coefficient and the chloride concentration on the concrete surface were calculated by applying Fick’s second law of diffusion (Equation 1), where: C X is the chloride concentration at depth x and time t (dry wt% of the sample), C S is the chloride concentration in the concrete (dry wt% of the sample), Dns is the non-steady state diffusion coefficient (cm2/s), erf is the Gauss error function and t is test time, in seconds. 𝑥 𝐶𝑥 = 𝐶𝑆 ∙ �1 − 𝑒𝑟𝑓 � �� (1) 2�𝐷𝑛𝑠 ∙ 𝑡 Porosity (pores ranging from 400 to 0.01 µm) was measured on a Micromeretics PORESIZER mercury intrusion porosimeter following the procedure described in ASTM standard D4404. The 1-cm3 sample used in the test was extracted from the inside of a concrete wafer taken from the middle third of the specimen.

441


Concrete resistivity can be determined in a number of ways: here the Wenner four-point test, described in Spanish standard UNE 83988-2, was chosen (figure 2). The tips of four evenly spaced electrodes were moistened to ensure good contact and applied to the concrete surface. Alternating current (of up to 50 μA at a high frequency) was applied between the outer electrodes while measuring the change in the difference in potential between the inner electrodes. Resistivity depends on voltage, current and distance between electrodes.

Figure 2 Resistivity measured with a four-point resistivimeter

In a quasi-infinite medium, resistivity (ρ a ) can be found with Equation: ρ = 2π a R where R is electrical resistance, a is the distance between the electrodes and the factor 2πa is the co-called geometric proportionality factor. Since the specimens used were finite, a “form factor” had to be included in the calculations, whose value in the present case was 0.714. In XRD, a portion of the incoming X-ray beam is scattered if its wavelength is of similar dimension as inter-atomic distances present in the material under investigation. The photons are scattered in a number of Debye cones. Their opening angle is 2θ. The angular distribution of the scattered intensity can be used to identify the crystalline phases present in the sample. In this case, X- ray diffraction patterns were collected using D8 Advance de BRUKER AXS diffractmeter, operating in step scan mode, with Cu Kα radiation. Patterns were collected in the range 5-60° 2θ with a step size of 0.02° and a rate of 23´ 47´´per step. Differential thermogravimetric analysis was conducted on NETZSCH STA 449. Nitrogen was used as the purging gas. For each run, simple powders were loaded onto platinum simple pan and heated from 25°C to 1000°C at high-resolution heating rate of 10K/min, where the heating rate was dynamically and continuously modified in response to the changes in the rate of sample’s weight loss.

3

Results

Figure 3 to Figure 4 present the chloride profiles determined on the different cores. In the graphs, the ‘0’ coordinate on the X-axis correspond to the exposed surface the concrete core. The profiles, as expected due to the different salinity, are different in the Upper and in the Lower chambers. The profiles show an external part almost without chlorides, likely due to leaching of the cement paste and a maximum in the chloride concentration that is inside the concrete. This maximum results around 0.1% by dry sample mass in the Upper Chamber and around 0.3% for the Lower Chamber. In the case of Pedro Miguel Chamber the maximum is very low, lower than the theoretical chloride threshold for steel corrosion and the profile is almost flat. This is due to the low salinity level of the zones due they are near the interior lakes. The chloride content may be that of the cement or due to the low contamination in chlorides of the interior lakes with the passing of the ships. Only in the Lower chamber near the sea border the salinity is more significant. The chloride profiles obtained from the cores have been fitted to the Fick’s 2nd equation in order to obtain the apparent chloride diffusion coefficient, D ap , and the chloride concentration of the maximum C S (in % of dry sample mass). They are given in 1 and figures 3 and 4.

442


Figure 3 Chloride profiles measured on cores extracted from the Miraflores Upper and Lower chamber

Figure 4 Chloride profiles measured on cores extracted from the Pedro Miguel chamber

The cores P1 and P9 do not show a decreasing profile and then the fitting has not being made. In all the rest of cores the chloride concentration at the maximum, (named C s as it is taken as the surface concentration) is logically higher in the Lower Chamber than in the Upper one. The D ap values result a bit higher in the Upper Chamber (around 4E-12 m2/s) than in the Lower (around 2E-12 m2/s) which seems contradictory but it is a mathematical consequence of equation 1 in which the D depends on the ratio between the surface and the internal chloride concentrations. Table1 Non-stationary Chloride Diffusion coefficients C s * (% dry Coring place Core D ap (m2/s) mass concrete) Miraflores Upper chamber East 10 0,161 2,55E-12 Miraflores Upper chamber West 5 0,130 4,98E-12 Miraflores Upper chamber West 6 0,141 5,48E-12 Miraflores Upper chamber West 2 0,095 4,82E-12 Miraflores Lower chamber East 3 0,253 3,23E-12 Miraflores Lower chamber East 7 0,426 2,00E-12 Miraflores Lower chamber East 4 0,340 2,47E-12 Miraflores Lower chamber East 11 0,588 1,44E-12 Miraflores Lower chamber East 8 0,256 2,81E-12 Miraflores Lower chamber West 14 0.441 2.693E-12 Miraflores Lower chamber West 12 0.190 1.451E-12 Miraflores Lower chamber West 13 0.391 1.430E-12 Pedro Miguel 16 0.067 5.475E-12 * It is the chloride concentration of the maximum in the profile.

For Resistivity measurements, figure 5, one slice of 4 cm was cut from the more internal parts of each core, those with the smallest chloride concentration. All slices were saturated in a vacuum desiccator before the measurements were made through the “direct method” (UNE

443


83988-1). The values are above 450 Ωm which can be considered very high considering that the mixes have a relatively low content in aggregates (only 30-35%), that is, they have a higher volumetric fraction of the paste than modern concretes. 900

Resistivity (Ω.m)

800 700

849

800 728 641 626

600

602

655

681 543 538 543

486

500

665 519

558

450

400 300 200 100 0

Pedro Miguel P18

Pedro Miguel P17

Pedro Miguel P16

Pedro Miguel P15

MiraFlores Lower P13



MiraFlores Lower P8


MiraFlores Lower P4

MiraFlores Lower P7

MiraFlores Lower P3

MiraFlores Upper P2

MiraFlores Upper P6

MiraFlores Upper P5

MiraFlores Upper P9

MiraFlores Upper P1

35 30 25 20 15 10 5 0 MiraFlores Upper P10

Porosity (%)

Figure 5 Electrical resistivity

Figure 6 Total porosity (% in volume) by MIP

The MIP results have been determined in a slice of the concrete bulk at around or beyond 20 cm in depth from the concrete surface. 6 shows the total porosity (% in volume of sample). The values are the mean of three samples of the same core. The values are relatively high and compatible with the high content in paste of the concrete, but not in accordance with the high resistivity values.

444


Figure 7 DTG results: Loss of weight for the decarbonation reaction (calcium carbonate loss)and the water loss for the three concrete cores.

The XRD analysis was made in only three of the cores in the Chambers with different salinity: P5 from Miraflores Upper Chamber West, P3 from Miraflores Lower chamber and P16 from Pedro Miguel. The samples were the same used for Thermal Analysis. The compounds identified were Calcite, Ettringite, Quartz, Dolomite, An orthoclase Feldspat and Chloroaluminate. The three cores seem to have a similar composition. Not big differences are found. They have ettringite. Friedel’s salt is only identified in small amount, calcite that is mainly phase present in the external layers, and the Quartz and Feldespat are attributed to the aggregates. The DTG analysis was made in the same cores and samples than the XRD. The results of thermogravimetric tests obtained at a rate temperature of 10 ºC/min are summarized in figure 7. In the curves show two rapid weight losses. The first weight loss, located between 100 and 200°C, is the results of dehydration of the hydrates. The second major weight loss appears at 750°C and corresponds to the decarbonation of calcium carbonate. It has not appeared the weight loss due to the portlandite. They seem fully carbonated in spite of being in contact to water. For the rest of the core’ profile, it is not possible to identify the second weight loss, that is the calcite only is identified in the most superficial layers, with a decreasing trend towards the interior where neither portlandite nor calcite are detected with this technique.

4 Discussion

The concrete used was like a mortar with coarse aggregates with non-continuous grading. As mentioned, the chlorides have penetrated and show a profile that exhibits a maximum in the interior at around 15-20 mm from the surface. The concentrations of these maxima are taken as

445


“surface” concentration for the chloride fittings and the calculation of the diffusion coefficient. The maxima are logically higher in the Lower chamber (0.15%-0.42% by cement weight) than in the Upper one (0.06%-0.14%) and in Pedro Miguel which has the lowest ( concrete beam > concrete column. Furthermore, due to the poor construction technology, the concrete strengths of Chinese modern reinforced buildings are commonly low. With the long-term degradation of the performance, the concrete strength would decrease further. The concrete compressive strengths of these buildings are tested with the method of core drilling. The results are shown in Table 1, showing that most of these concrete compressive strengths are lower than 20MPa which is the minimum requirement in the present code of concrete structure [15]. Table 1 Test results of concrete compressive strength

Typical Case

Main hall

No. 1 of

of Tomb of

Zhongshan

Yu in

East Road

Shaoxing

in Nanjing

(1933)

(1935)

15.7MPa

15.4MPa

Huangpu

Dahua

Dacheng

Main hall

cinema in

factory in

of Nanjing

Nanjing

Changzhou

Museum

(1934)

(1935)

(1937)

12.4MPa

15.6MPa

17.0MPa

18.0MPa

14.7MPa

11.7MPa

17.8MPa

22.9MPa

23.2MPa

17.6MPa

hall of Jiangsu conference center(1931)

Average compressive strength of column Average compressive strength of beam

3 Analysis of carbonation life Modern reinforced concrete buildings are different from contemporary reinforced concrete buildings in rebar property, concrete property and building configurations. In general, techniques used are similar. Thus, the ageing mechanisms of the modern reinforced concrete and the contemporary reinforced concrete buildings are similar. As the water to cement ratio and the cement factor of modern reinforced concrete building are difficult to test and obtained on site, the carbonation life model considering the concrete compressive strength as the main parameter is recommended to be used in an adapted way. Consulting the relevant references [16] [17] and introducing the modified coefficient α which is related to material properties and building configurations, the calculation formulas of carbonation life of Chinese modern reinforced concrete buildings are given in Eq. (1) - Eq. (2).

t1 = ( c / k )

2

 58  = − 0.76  k 3α k j kco2 k p ksT 1/4 (1 − RH ) × RH 1.5  =  f cuk 

xc t0

(1)

(2) where t 1 is the carbonation life (a); c is the thickness of concrete cover(mm); k is the carbonation coefficient; k j is the position influence coefficient, when it is in corner, k j =1.4; when it is not in corner, k j =1.0; k co2 is the influence coefficient of CO 2 concentration, when it is in a very crowded place such as a teaching building or a cinema, k co2 ranges from 3.2 to 2.7; when it is in a crowded place such as an hospital building or a shopping building, k co2 ranges from 2.7 to 2.1; when it is in a common place such as a residence building or an office building, k co2 ranges from 2.1 to 1.6; when it is in a place with few people such as a garage or a underground park, k co2 ranges from 1.6 to 1.1; k p is the modified coefficient of casting surface, when it is in the casting surface, k p =1.3, when it is not in the casting surface, k p =1.0; k s is the stress influence coefficient, when it is in compression, k s =1.0, when it is in tension, k s =1.2; RH is the ambient humidity (%); T is the ambient temperature (℃); f cuk is the standard value of concrete 457


compressive strength (MPa); x c is the carbonation depth (mm); t 0 is the time from the construction beginning to the carbonation test (a). Through the regression analysis of the data of these modern reinforced concrete cases, α is presented to be 0.86. The calculation formula of carbonation life of Chinese modern reinforced concrete buildings could be obtained in Eq.(3).

    c   t1 =   1/4 1.5  58 − 0.76    3 × 0.86 × k j kco2 k p ksT (1 − RH ) × RH   f cuk  

4

2

Analysis of residual service life

(3)

Most of Chinese modern reinforced concrete buildings are the protected cultural relics or historical buildings with important historical and cultural values. The structural safeties of these buildings should be extraordinarily guaranteed. Thus it is suggested that the residual service lives of these buildings be calculated according to the steel bar corrosion and concrete cracking life method. In this paper, the corrosion degrees of the rebars of these modern cases are not tested. However, the steel bar corrosion and concrete cracking lives of these modern cases can be calculated according to the reference [17], as follows:

tcr= t1 +

δ cr λ

λ= 5.92kcl ( 0.75 + 0.0125T )( RH − 0.50 )

2/3

c

−0.675

f cuk

(4)

−1.8

(5)

δ cr =0.012c / d + 0.00084 f cuk + 0.018

(6) where δ cr is the critical corrosion depth (mm); λ is the corrosion rate of rebar(mm/a); k cl is the modified coefficient of rebar position, when the rebar is in corner, k cl =1.6, when the rebar is not in corner, k cl =1.0; c is the thickness of concrete cover; f cuk is the standard value of concrete compressive strength (MPa); T is the ambient temperature ( ℃); RH is the a t 1 is the carbonation life (a); t cr is the steel bar corrosion and concrete cracking life (a). According to the above calculation formulas of carbonation life and steel bar corrosion and concrete cracking life, the residual service lives of these modern reinforced concrete buildings are given in Table 2. The results show that the service lives of these cases have exceeded their carbonation lives. The actual states of these buildings conform to the calculation results. According to the calculated results, the residual service lives are less than 10 years. Table 2 Calculation results of residual service lives of these buildings Main hall of Typical Case

Tomb of Yu in Shaoxing (1933)

Carbonation life

No. 1 of Zhongshan East Road in Nanjing (1935)

Dahua

Dacheng

Main hall

cinema in

factory in

of Nanjing

Nanjing

Changzhou

Museum

(1934)

(1935)

(1937)

Huangpu hall of Jiangsu conference center(1931)

52 a

38 a

36 a

40 a

38 a

40 a

90 a

59 a

68 a

80 a

82 a

70 a

Service life

81 a

79 a

80 a

79 a

77 a

83 a

Residual life

9a

No

No

1a

5a

No

Corrosion cracking life

458


5 Conclusion The following conclusions can be drawn from the study: （1） The durability problems of Chinese modern reinforced concrete buildings are mainly large carbonation depth, rebar corrosion, low concrete strength and so on. The carbonation depths of columns and beams generally approach or exceed the thicknesses of concrete covers of columns and beams.The carbonation depths of slabs have generally exceeded the thicknesses of concrete covers of slabs. In general, the order of occurrence of the probability of rebar corrosion is: concrete slab > concrete beam > concrete column. （2） Testing some typical Chinese modern reinforced concrete buildings, a modified calculation formula of carbonation life is given which is adapted to Chinese modern reinforced concrete buildings. （3） The residual service lives of Chinese m odern reinforced c oncrete buildings have been calculated using the steel bar corrosion and concrete cracking life method. The service lives of these buildings have exceeded their carbonation lives, and the actual states of these buildings conform to the calculation results. According to the calculated results, their residual service lives are less than 10 years. So these buildings should be strengthened and repaired urgently.

Acknowledgements This paper is written with support of the fellowship provided by Civil Engineering Department at KUL ， National Natural Science Foundation of China (Grant No. 51138002) and the Fundamental Research Funds for the Central Universities(Grant No. 2242013R30001)

References [1] Huang Kexin, Wu Xingzu,etc. (1983) Reinforcement corrosion and protection in reinforcement concrete structure. Beijing: China Architecture & Building Press.(in Chinese) [2] PAPADAKIS V G,VAYENAS C G,FARDIS M N.(1991) Fundamental modeling and experimental investigation of concrete carbonation. ACI Material Journal， 88:363-373 [3] Zhu Anmin. (1992) Concrete carbonation and RC concrete durability.Concrete, 6:18-22(in Chinese) [4] Architecture Institute of Japan. (1984) Damage and durability countermeasure for the buildings. The Chinese Association of Metallurgical Construction Management. Beijing. [5] Lesage -de -Contenay C. (1995) Deterioration and repair， Bahrain Proc. 6:467-483. [6] Di Xiaotan, Zhou Yan. (1994) Concrete carbonation. China Academy of Building Research, Beijing.(in Chinese) [7] Zhang Yu, Jiang Lixue. (1998) A practical mathematical model of concrete carbonation depth based on the mechanism. Industrial Construction. 28(1)： 16-19(in Chinese) [8] Niu Ditao. (2003) Durability and life forecast of reinforced concrete structure. Science Press, Beijing. (in Chinese) [9] Bazant. (1979) Physical model for steel corrosion in concrete sea structures-Theory. Journal of Structural Division, 6: 1137-1153. [10] Liu Xila, Miao Shuke. (1990) Steel corrosion and the durability calculation of reinforced concrete structures. China Civil Engineering Journal, 23(4):69-78 (in Chinese) [11] Xiao Congzheng. (1995) Mechanism study and number-theoretic method for reinforcement corrosion in the RC Structures. Tsinghua University Press. Beijing. 1995(in Chinese) [12] Andres A. Torres-Acosta and Alberto A. Sagues. (2004) Concrete Cracking by Localized Steel Corrosion— Geometric Effects. ACI Materials Journal. 101(6):501-507 [13] Niu Ditao, Wang Qinglin, Wang Linke. (1996) Predeterminate model of steel corrosion extent in reinforced concrete structures before producing corrosion crack. Industrial Construction. 26(4): 8-10.(in Chinese) [14] Zhang Weiping. (1999) Damage prediction and durability estimation for corrosion of reinforcement in concrete structures. Tongji University Press, Shanghai. (in Chinese) [15] China Academy of Building Research.(2010) Code for design of concrete structures (GB50010-2010). Building Industry Press, Beijing.

459


[16] Xu Shanhua, Niu Ditao, Chen Xinxiao. (2002) The Life-Span Analysis of Reinforced Concrete Corrosion Cracking. Building Science. 18(5):32-35(in Chinese) [17] Xu Shanhua, Niu Ditao, Chen Xinxiao. (2002) The Life-Span Analysis of Reinforced Concrete Corrosion Cracking. Building Science. 18(5):32-35(in Chinese)

460


Laser Induced Breakdown Spectroscopy (LIBS) as a tool for the investigation of aging of infrastructure Thorsten Eichler, Gerd Wilsch, Steven Millar and Dieter Schaurich BAM Federal Institute for Materials Research and Testing, Berlin, Germany

Abstract: The laser induced breakdown spectroscopy (LIBS) is a combination of laser ablation and optical emission spectroscopy. Due to the possibility of direct measurements on the sample surface and a minimum of required sample preparation investigations of building materials can be conducted quite fast. In combination with a scanning technique (translation stage or scanning mirrors head) the element distributions are evaluated taking in to account the heterogeneity of the material. This is a significant advantage compared to standard procedures. LIBS measurements are also time and cost saving in comparison to standard methods. The automated measurement procedures minimise the liability for errors. All elements are detectable and after calibration results can be quantified. At BAM LIBS has been successfully applied for the investigation of distribution and transport of different ions in building materials. Quantitative measurements are performed for chlorine, sodium, potassium, sulphur, lithium and hydrogen. Beside this the identification of substances and the evaluation of quantitative ratios by means of an integrated marker are possible. An overview about the principle and the possibilities of LIBS investigations of building materials is presented and typical applications are shown. The LIBS technique is on the step from laboratory application to on-site analyses. Keywords: concrete deterioration, LIBS, element mapping, chemical analysis, heterogeneity, chlorine, sulphur, alkali silica reaction (ASR)

1 Introduction Reinforced concrete structures are generally dimensioned in consideration of a specific expected life time. For usual structures in civil engineering as e.g. multi storey car parks a minimum life time of approximately 50 years is expected. In some cases, the assessed life time is not attainable and ageing of specific structures happens far more quickly than expected. Environmental factors, influenced by weather, location or general exposition may cause the access of harmful species like SO4--, Na+, CO2 or Cl-, etc. to the structure and accelerate the ageing process due to deterioration of concrete and/or reinforcement. Laser Induced Breakdown Spectroscopy (LIBS) can be applied as a fast and reliable method in order to identify harmful species and accompanying damage processes. The LIBS method is a combination of material ablation, plasma formation and analysis of the emitted radiation by spectroscopic methods. For ablation and plasma formation usually a high energy pulsed NdYAG-laser is focused on the sample surface. Due to the high temperatures in the plasma all chemical bonds are broken and only elements may be detected. In a later stage of the plasma also some molecules may be observed. Due to the excitation process all elements are detectable. Also light elements like hydrogen or lithium. The plasma radiation is guided by an optical fibre to the spectrometer and then detected by a CCD-camera at the output slit of the spectrometer. The laser energy, pulse frequency, laser wavelength and the focus conditions are important for the sensitivity of the method. Also the spectrometer type, the spectrometer parameter, the fibre type and the CCD parameter have to be conformed to the elements under investigation. LIBS may be used on gases, liquids or solids. There are a lot of applications in different fields like environmental analysis, pharmaceutical investigation, biomedical investigation, forensic investigation or industrial applications like process control, recycling, sorting and quality control during manufacturing [1, 2].

461


2 Laser Induced Breakdown Spectroscopy (LIBS) technique A typical LIBS set-up used for the investigation of building materials is shown in figure 1. The radiation of a pulsed NdYAG-laser ( = 1064 nm, pulse rate 10 Hz, energy per pulse 100 mJ to 400 mJ, pulse duration = 8 ns) is focused to an area of one square millimetre to evaporate a small amount of the surface under investigation (some micrograms). Due to the high power density in the focus area plasma ignites and radiates for some microseconds. The plasma radiation is guided to a spectrometer system, which breaks the radiation into its spectral components. At the exit slit of the spectrometer the light intensities are detected by a CCD-camera. The focal length of the lens is 500 mm to minimize the influence of surface roughness. The volume around the plasma can be purged with air or helium to remove dust and enhance the results for the detection of some elements. The detection of chlorine and sulphur is more sensitive if helium is used as a process gas. The sample under investigation can be moved in a plane perpendicular to the laser beam to measure the spatial distribution of elements on the surface. The CCD detects light intensities in dependence of wavelength. Each element has typical lines in the spectrum caused by the inner structure of the atoms electron configuration. There are databases available which give the wavelength position of these lines [3]. Thus an assignment of peaks to elements is possible. The intensity of the radiation at a single peak contains information about the relative content of this element in the evaporated volume. There are different types of detectors from single peak detectors (photo diodes or channeltrons), different types of CCD cameras and complex echelle spectrometers. The choice of the detector is in close relation to the number of elements to be detected, the limit of detection and the financial budget.

Figure 1 LIBS setup (left) and translation stage with concrete core under investigation (right)

Figure 1 (on the right) shows a concrete specimen on the translation stage prepared for LIBS measurements. The specimen will be scanned in lines with a millimetre distance and a resolution of one measurement per millimetre for each line. The scanning of the 45 mm by 60 mm cross section of the core takes some minutes and the results are obtained directly after the measurement. The possible output of the measurement is the distribution of a specific element on the surface under investigation or, if the measurements are averaged per line, a gradient of element content. Due to the multi-element ability of the method the measurement of trace elements as well as the differentiation between cement and aggregates is possible. As a result the content of the trace element can be given in correlation to the cement content. The LIBS measurement displays statistical information about the specimen and brings out the heterogeneity of the building material. The LIBS set-up can be made portable. First experiences are given below.

462


The data evaluation needs some normalization procedures to eliminate interferences like fluctuations of pulse energy, surface properties or dust in the beam path. Software solutions are available which allow fast and automated data evaluation in only a few minutes. Nevertheless nowadays still a wellskilled operator is necessary to carry out LIBS measurements.

3 Measurements and Results 3.1

Chloride – deterioration of reinforced concrete

The durability of reinforced and pre-stressed concrete structures depends, amongst others, on the efforts taken to maintain passivity of the reinforcement or pre-stressing steel. The passivity can basically be lost by two different processes, carbonation of concrete and due to chlorides exceeding a specific concentration threshold [Cl-/OH-] at the steel / concrete interface. The chlorine concentration in the cement or concrete can be determined by evaluation of the emission line at 837.6 nm in the near infrared spectral range [4]. Other CI lines are available in theory but do not deliver the required sensitivity in practice. Alternatively adequate results can be achieved only with the VUV emission line at 134.7 nm under vacuum conditions. The chlorine line 837.6 nm needs to be classified as outside the sensitive range, unless the use of a gas purge with helium. Figure 2 shows the effect of helium purge on the chlorine signal at 837,6 nm.

Intensity [a.u.]

Helium Air

Wavelength [nm] Figure 2 Comparison of two spectra measured on the same sample. The bold line shows the measured intensities in air atmosphere. The dotted line represents the measurement in helium atmosphere.

463


Figure 3 Photo of cross section of a concrete core and the surface measured by LIBS (area 30 mm x 30 mm, red square, line spacing 1 mm). The direction of chloride ingress is indicated by an arrow.

In figure 3 the cross section of a concrete core taken from a parking deck and the surface measured with LIBS is shown. The direction of chloride ingress is indicated by an arrow. The lines produced by the measurement are clearly visible. The line spacing is 1 mm. The distance between successive measurements at a line was 0.5 mm. The spot size was about 0.2 mm. The time of measurement was less then 10 minutes. The evaluated chlorine signal is shown in figure 4. In the colour plot on the left hand side the chlorine distribution in the measured area is given. The colour’s intensity is correlated to the chlorine concentration. The graph in figure 4 (right) represents the averaged signal per measured line. A gradient of element content is obtained by quantification using a calibration curve.

Figure 4 Chlorine distribution in the measured area (left). Brighter red means higher chlorine concentration For each depth the mean value of the chlorine concentration in percent is shown (right)

A comparison of relative chlorine concentrations measured in correlation to the depth are displayed in figure 5. Again the measurements where performed on cross section of three different concrete cores. The cores where taken from a parking garage on different positions. The critical chlorine value at which corrosion may be initiated is marked by the blue line. The core indicated by the red marked values is in critical condition – repair work is urgently required. The core indicated by the green

464


marked values is close, but not yet critical – monitoring is necessary. The core indicated by the black marked values is in good condition.

Figure 5 Comparison of the relative chlorine concentration measured in correlation to the depth on the cross section of three different concrete cores. The critical chlorine value at which corrosion may be initiated is marked by the blue line. The cores where taken from a parking garage on different positions.

The evaluation of the three cores takes less than one hour. Thus a large number of cores can be evaluated which form a reliable decisional basis for civil engineer.

3.2

Sulphate ingress – concrete corrosion

The chemical action of sulphur in the form of sulphates or sulphides attacks concrete surfaces causing damage to, or breakdown of, the cement matrix. Considerable bulking is a potential problem with sulphates, e.g. due to the volume increase through ettringite formation, hence the destruction of the cement structure. One fact of particular note in the case of sulphides is the corrosive action of the sulphuric acid on sewage and waste water treatment buildings. The detection of sulphur by means of LIBS proves challenging insofar as, firstly, sulphur is already used in the bonding agent as a sulphate transporter and, secondly, there are precious few emission lines for sulphur identification in the visible range. One line which has proved suitable is the S line at 921.30 nm [5]. The zones where sulphuric acid penetrates a concrete test piece can be shown with LIBS imaging (see an example in figure 6). It is evident that the sulphur is mainly absorbed at the edges. Exact localisation of the areas contaminated by sulphur helps to assess the depth to which excavation of concrete is necessary.

Figure 6 Photo of cross section of concrete core (left) and colour coded sulphur content (right). Brighter yellow means higher concentration.

465


3.3

Alkali metals – alkali silica reaction

Figure 7 Photo of cross section of concrete core (left), measured sodium distribution (middle) and measured sodium distribution with excluded large aggregates (left). Brighter blue means higher concentration.

The term alkali silica reaction is used to describe the process whereby the siliceous minerals in some aggregates react with the soluble alkaline pore fluid elements (potassium, sodium) to form a alkali silicate gel which is eager to absorb water. In unpropitious conditions the volume of this gel increases to such an extent over time that the pressure gives rise to local swelling. This pressure can exceed the tensile strength of the concrete and damage the structure. Outward signs of this are cracking, efflorescence and pop-out. Added road salt with ingress of sodium ions can intensify the reaction. In figure 7 the results measured on a concrete core extracted from a highway are shown. On the left hand side a photo of the cross section of the concrete core is plotted. In the middle the measured sodium distribution is plotted. As easily can be seen there are high sodium concentrations in the large aggregates. Therefore on the right hand side the measured sodium distribution with excluded large aggregates is plotted. Brighter blue means higher concentration.

Figure 8 Comparison of sodium ingress profiles measured on the cross section of a concrete core. The blue curve is obtained by taking all points in to account. The red curve is obtained by excluding results measured on large aggregates.

466


In figure 8 the results are averaged per line. A gradient of element content in correlation to the depth is obtained. Blue curve obtained taking all points in to account. Red curve obtained by excluding results obtained on large aggregates. Only the red curve calculated by excluding the measurements on large aggregates represents the ingress profile of sodium.

Figure 9 Comparison of chlorine (red) and sodium (blue) ingress profile measured on the cross section of concrete core shown in figure 7.

In figure 9 a comparison of chlorine (red) and sodium (blue) ingress profile measured on the cross section of the concrete core is shown. The values are quantitative values after calibration. Both profiles are different. The results are showing evidently that the common practice using methods like µXRF, where the sodium concentration is educed from chloride profiles may lead to significantly erroneous results.

3.4

Mobile LIBS setup with scanner

A mobile LIBS set-up with fast on-line results is helpful for the engineer to estimate the condition of concrete structures and for quality assurance during concrete repair work. It ensures for example that all contaminated concrete is fully removed – as much as necessary and not more.

Figure 10 Left: Mobile LIBS system (www.secopta.com) and right: mobile scanner

Potential applications are parking garages or bridge decks with chloride contaminated concrete. During repair of sewage treatment plants the sulphur content of concrete structures may be evaluated. The set-up of the mobile system is shown in figure 10, left. The laser is located in the measuring head shown on the top of the main case. It is connected with the control unit via a cable. The cable length is

467


5 m. To maintain laser safety regulations, a cover in connection with a brush shields the laser beam path. For the investigation of heterogeneous materials like concrete a scanner unit moves the laser head over the surface (maximum area 17 cm x 17 cm, see figure 10, right). The plasma radiation is delivered to an optical fibre and analysed by the detection unit. The system is designed for measuring the chlorine, sulphur or alkaline elements. The content of chloride or sulphate is calculated from the measured chlorine or sulphur intensity. An automated data assessment program is integrated in the system. The system is designed for horizontal and vertical operation.

4 Summary It has been shown that using the LIBS-technique can be advantageous when evaluating structures with respect to harmful substances which can significantly contribute to faster ageing. It allows fast direct measurements in order to perform investigations on building materials. Only a minimum of sample preparation is required. In combination with a scanning technique (translation stage or scanner) the heterogeneity of the material can be taken into account when evaluating results. This is an advantage compared to the standard procedures of wet chemistry. LIBS measurements are also time- and cost efficient compared to the standard methods. The automated procedures minimize the occurrence of errors. LIBS has been successfully applied for the investigation of the distribution and transport of different ions in building materials. Quantitative results can be derived from intensity plots e.g. for chlorine, sodium, sulphur and lithium. It was shown that the ingress profiles of sodium and chlorine are not necessarily equal. Therefore the use of LIBS can be advantageous in comparison to other methods like micro-XRF, which is of limited suitability when sodium is an element of interest. The setup of rules and standards is the next step to establish LIBS as a standard procedure for chemical investigations of building materials.

5 Acknowledgements Financial support by German Federal Ministry for Economics and Energy (BMWi) is gratefully acknowledged.

6 References [1] Hahn, D. W. and N. Omenetto (2012). "Laser-Induced Breakdown Spectroscopy (LIBS), Part II: Review of Instrumental and Methodological Approaches to Material Analysis and Applications to Different Fields." Appl. Spectrosc. 66(4): 347-419. [2] Noll, R., Laser-induced Breakdown Spectroscopy - Fundamentals and applications, 2012, Springer Verlag, Berlin Heidelberg. [3] Ralchenko, Yu., Kramida, A.E., Reader, J. and NIST ASD Team (2008). NIST Atomic Spectra Database (version 3.1.5), [Online]. Available: http://physics.nist.gov/asd3 [2009, April 29]. National Institute of Standards and Technology, Gaithersburg, MD. [4] Wilsch G, Weritz F, Schaurich D, and Wiggenhauser H (2005) Determination of chloride content in concrete structureswith laser-induced breakdown spectroscopy, Constr Build Mater 19:724-730. [5] Weritz, F., S. Ryahi, et al. (2005). "Quantitative determination of sulfur content in concrete with laserinduced breakdown spectroscopy." Spectrochimica Acta Part B-Atomic Spectroscopy 60(7-8): 11211131. [6] Weritz F, Schaurich D, Taffe A, and Wilsch G (2006) Effect of heterogeneity on the quantitative determination of trace elements in concrete, Anal Bioanal Chem 385: 248-255. [7] Wilsch, G., D. Schaurich, et al. (2011). Imaging Laser Analysis of Building Materials - Practical Examples. Review of Progress in Quantitative Nondestructive Evaluation, Vols 30a and 30b 1335: 1315-1322.

468


Residual Strength of Aging Concrete T-Girders Strengthened with NearSurface-Mounted Composite System Tamer El-Maaddawy1*, Anes Bouchair1, Ashraf Biddah2, Amr El-Dieb1 (1) United Arab Emirates University, Al-Ain, United Arab Emirates (2) Ain Shams University, Cairo, Egypt Abstract: This paper aims at evaluating the residual strength of concrete T-girders strengthened with near-surface-mounted (NSM) composite strips then exposed to accelerated corrosion aging. Five beams were not corroded and five beams were subjected to 100 days of accelerated corrosion that corresponded to a tensile steel mass loss of 15%. Corrosion damage and cracking reduced the strength of the beams strengthened with NSM composite strips without U-wraps. The integration of U-shaped composite wraps in the strengthening regime prevented a premature peeling-off of the concrete cover and enclosed NSM composite strips. No strength reduction was recorded in the corroded beams strengthened with the NSM strips together with the U-wraps. The residual strengths of all corrodedstrengthened beams were substantially higher than that of the control uncorroded beam. Keywords: corrosion, concrete, NSM, composites, strengthening.

1 Introduction Flexural strengthening of deficient reinforced concrete (RC) infrastructure with advanced composites has gained a wide acceptance by the structural engineering community. The most popular composite system is the externally-bonded fiber-reinforced polymer (EB-FRP). Although the system is effective in increasing the flexural capacity, it is susceptible to acts of vandalism, fire, and mechanical damage. To protect the FRP reinforcement from weathering, it has been proposed to use a near-surfacemounted (NSM) FRP system, where FRP strips or rods are inserted into grooves pre-cut onto the concrete surface and held in place using an epoxy adhesive. The NSM composite system is a promising technique for strengthening of deficient RC structural components [1]. The NSM-FRP strengthening system can result in up to a two-fold increase in the flexural capacity [2-3]. Strengthening of RC T-beams with the NSM-FRP system provides higher flexural strength gain than that provided by the EB-FRP system with composite materials having the same axial stiffness [2]. Changing the type of the FRP reinforcement in the NSM strengthening system from carbon to glass, while keeping the axial stiffness of the FRP constant, has no significant effect on the flexural strength gain [2]. The gain in the flexural capacity is proportional to the amount of the NSM-FRP reinforcement and inversely proportional to the amount of the internal steel reinforcement [4-5]. The additional gain in the flexural capacity is not proportional to the added amount of NSM-FRP reinforcement [5]. The gain in the flexural capacity caused by the NSM-FRP strengthening is more pronounced for the high strength concrete girders than for the normal strength concrete girders [5]. The use of steel fiber reinforced concrete (SFRC) overlay in combination with the NSM-FRP strengthening has increased the service and ultimate loads of RC slabs by more than two and three folds, respectively [6]. The NSM-FRP strengthened beams and one-way slabs are vulnerable to fail by peeling-off of the concrete cover along with parts of concrete above the longitudinal tensile steel reinforcement. The premature detachment of the concrete cover and enclosed NSM-FRP reinforcement limits the flexural strength gain and reduces the beam ductility [1,3,5-7]. The NSM-FRP strengthening can result in up to a 34% reduction in the flexural ductility [3]. Only 30% gain in the flexural capacity in both the hogging and sagging regions of strengthened continuous one-way RC slabs has been recorded due to the premature peeling-off failure of the strengthening system [7]. * *

Associate Professor, United Arab Emirates University, UAE, E-mail: [email protected] Research Assistant Professor, Housing and Building National Research Center in Egypt (on leave)

469


To date, no data is available in the literature on the flexural performance of RC T-girders strengthened with the NSM-FRP system when exposed to corrosive environmental conditions. Since the FRP reinforcement in the NSM system are installed in the concrete cover, corrosion damage and associated concrete cover cracking could compromise the system performance. The effects of corrosion exposure on the flexural strength gain and ductility of RC beams strengthened with the NSM-FRP system need a thorough investigation before this system can be routinely employed in practical applications. This paper aims at evaluating the flexural response of RC T-beams strengthened with the NSM carbon fiber-reinforced polymer (CFRP) system then exposed to accelerated corrosion. The interaction between the corrosion damage, failure mode, amount of the NSM-CFRP reinforcement and presence U-wraps along the beam span has been elucidated and documented in this paper.

2 Experimental program 2.1

Test Matrix

The test matrix is summarized in Table 1. A total of ten RC T-beam specimens were constructed and tested. The specimens were divided into two groups, [A] and [B], five specimens each. In each group, one specimen was not strengthened to act as a benchmark. Two specimens were strengthened with two NSM-CFRP strips; one specimen without U-wraps along the beam span and one specimen with Uwraps. The remaining two specimens were strengthened with four NSM-CFRP strips with and without U-wraps. Following strengthening, specimens of group [A] were tested to failure whereas specimens of group [B] were subjected to 100 days of accelerated corrosion then tested to failure. In the specimen designation, LS refers to low strength concrete of 25 MPa; C2, and C0 refer to specimens with and without corrosion, respectively; 2F and 4F refer to flexural strengthening with two and four NSMCFRP strips, respectively; and U refers to U-shaped CFRP wraps. Table 1 Test matrix

Group

[A]

[B]

2.2

Time of corrosion exposure

No corrosion

100 days

Strengthening regime

Specimen designation

Longitudinal -

Transverse -

LS-C0

2 NSM-CFRP strips

-

LS-C0-2F

4 NSM-CFRP strips

-

LS-C0-4F

2 NSM-CFRP strips

U-CFRP wraps

LS-C0-2F-U

4 NSM-CFRP strips

U-CFRP wraps

LS-C0-4F-U

-

-

LS-C2

2 NSM-CFRP strips

-

LS-C2-2F

4 NSM-CFRP strips

-

LS-C2-4F

2 NSM-CFRP strips

U-CFRP wraps

LS-C2-2F-U

4 NSM-CFRP strips

U-CFRP wraps

LS-C2-4F-U

Test specimen

The specimens were designed in a way to ensure that a flexural mode of failure would dominate. A schematic of a typical test specimen is shown in Figure 1. The test specimen was 3200 mm long RC beam with a T-shaped cross section. The cross section had a web width of b w = 200 mm, flange width of b f = 340 mm, total depth of h = 260 mm, and an effective depth of d = 200 mm. The tensile and compressive steel reinforcement consisted of 3 No. 12 deformed steel bars and 4 No. 6 plain steel bars, respectively. Adequate shear reinforcement of 8 mm diameter deformed steel stirrups was provided at a spacing of 100 mm to prevent shear failure. The NSM-CFRP reinforcement used for strengthening

470


260

3 No.12 200

200

Section A-A

Section B-B No.8@100 mm

No. 8@100 mm

120 100

d = 200

260

No. 8@100 mm 3 No.12

B

340 4 No.6

50

340 4 No.6

d = 200

50

consisted of two or four strips, each having a cross section of 1.2 x 20 mm. The NSM-CFRP strips were embedded into slits, each having a width of 4 mm and a depth of 30 mm. The U-wraps consisted of U-shaped CFRP sheets, 100 mm wide each, placed at center to center spacing of 200 mm. The concrete mix used to cast the middle half of the beam up to a height of 120 mm, measured from the bottom soffit, had a 3% NaCl by weight of cement to depassify the tensile steel in this region and promote corrosion. The stirrups located within the corroded zone were epoxy-coated. As a further protection, the corners of these stirrups in contact with the tensile steel were wrapped with an insulation tape. The tensile steel rebars located outside the corroded zone were epoxy-coated to further restrict corrosion damage to the middle 1,500 mm of the tensile steel rebars.

400

A

Stainless steel tube

1500

100

Salted Zone

B

3000

A

100

Figure 1 Test specimen

2.3

Accelerated corrosion ageing

Accelerated corrosion by means of impressed current technique was adopted in this study. An internal stainless steel tube with an external diameter of 6 mm and a wall thickness of 1 mm was placed longitudinally at a distance 120 mm from the bottom face of the beam to act as a cathode during the accelerated corrosion process. The tensile steel rebars and the stainless steel tube were extended out of the beam to facilitate making electrical connections to the power supply. A constant current of 250 mA was impressed on the tensile steel reinforcement by means of external power supplies. The corroded specimens were connected in series to obtain a constant current through them. The accelerated corrosion process lasted for 100 days. The tensile steel rebars were connected to the positive terminal of the power supply to act as an anode while the stainless steel tube was connected to the negative terminal to act as a cathode. Potable water mist with a pH in the range of 7.5 to 8.3 was sprayed over the specimens during conditioning to facilitate corrosion reaction. Photos of the power supplies, electrical connections and test specimens under accelerated corrosion are shown in Figure 2.

(a)

(b)

(c) Figure 2 Accelerated corrosion; (a) power supplies, (b) electrical connections, (c) specimens under accelerated corrosion exposure

471


2.4

Materials

The cylinder concrete compressive strength at the time of structural testing was on average 25 MPa. The No. 12 and No. 8 deformed bars had nominal yield strength of 520 MPa whereas the No. 6 plain bars had nominal yield strength of 300 MPa. The NSM-CFRP strips had a tensile modulus of 165 GPa and tensile strength of 3100 MPa. The NSM-CFRP strips were bonded to the sides of the concrete grooves using an epoxy adhesive having a tensile modulus of 4.5 GPa, tensile strength of 24.8 MPa, and ultimate strain of 1%. The carbon fiber fabrics used in the U-wraps were unidirectional with a typical thickness of 0.17 mm, tensile modulus of 230 GPa, and ultimate strength of 3900 MPa. The carbon fiber fabrics were impregnated and bonded to the concrete using a compatible epoxy adhesive with a typical tensile strength of 30 MPa, E-modulus of 3.8 GPa, and elongation at break of 1.5%. Properties of the CFRP composite strips, carbon fiber fabrics, and epoxies used in strengthening were obtained from the manufacturer.

2.5

Test set-up

Following corrosion, the specimens were tested to failure under four-point bending with an effective span of 3000 mm and a shear span of 1300 mm. The load was applied incrementally by means of a hydraulic jack until failure. The mid-span deflection was monitored using a linear variable displacement transducer (LVDT). Electrical resistance strain gauges were bonded to the NSM-CFRP strips at several locations along half of the beam length to monitor the longitudinal strains in the NSMCFRP strips during loading.

3 Experimental results 3.1

Corrosion damage

The corroded specimens exhibited severe rust stains and longitudinal corrosion cracks in the concrete cover. The longitudinal corrosion cracks were parallel to the tensile steel reinforcing bars. At the end of the corrosion phase, the corrosion crack width was measured by means of a hand-held microscope. Corrosion exposure for 100 days resulted in a corrosion crack width in the range of 1.0 to 1.2 mm. Based on Faraday’s law, accelerated corrosion exposure of the present specimens for 100 days corresponded to a 15% tensile steel mass loss.

3.2

Failure mode

The control specimens LS-C0 and LS-C2 and also the uncorroded-strengthened specimens failed by tensile steel yielding followed by concrete crushing in the compression side (SY-CC) as shown in Figure 3a. The corroded specimens that were strengthened with the NSM-CFRP strips without the Uwraps failed by peeling-off of the concrete cover along with the enclosed NSM-CFRP strips without crushing of concrete (SY-PE) as shown in Figure 3b. The peeling-off failure of the strengthening system was preceded by tensile steel yielding. The integration of the U-wraps in the strengthening system prevented the premature peeling-off of the concrete cover and allowed the beam to develop its full flexural capacity where failure occurred by tensile steel yielding followed by concrete crushing.

(a)

(b)

Figure 3 Failure modes; (a) concrete crushing in compression, (b) peeling-off of the strengthening system

472


3.3

Flexural response

The main test results are summarized in Table 2. The load-deflection curves of specimens of groups [A] and [B] are depicted in Figures 4a and 4b, respectively. The response of the control uncorroded specimen LS-C0 is included in Figure 4b for the purpose of comparison. It is evident that the NSMCFRP strengthening significantly improved the flexural capacity and stiffness of the uncorroded specimens. Flexural strengthening of the undamaged specimens with two and four NSM-CFRP strips without the U-wraps increased the yield load by about 50% and 70%, and the ultimate load by 74% and 100%, respectively. The presence of the U-wraps had no effect on the gain in the yield load but slightly increased the gain in the ultimate load by about 6.5%, on average. The beams with the Uwraps were stiffer than the beams with the NSM-CFRP strips without the U-wraps. This is because the U-wraps improved the bond between the steel and concrete and also between the NSM-CFRP strips and concrete. The NSM-CFRP strengthening resulted, however, in a reduction in the ductility index relative to that of the control beam LS-C0. The ductility indices of specimens LS-C0-2F and LS-C02F-U with the lower amount of the NSM-CFRP strips were 74% and 80% that of the control specimen LS-C0, respectively. Specimens LS-C0-4F and LS-C0-4F-U with the higher amount of the NSMCFRP strips experienced a higher reduction in the ductility index relative to that of their counterparts with the lower amount of the NSM-CFRP strips. The ductility indices of specimens LS-C0-4F and LSC0-4F-U were 60% and 67% that of the control specimen LS-C0, respectively. The presence of the Uwraps increased the ductility index of the uncorroded-strengthened beams by 12%, on average. Table 2 Test results Yield Group

Specimen

Ultimate

Load

Deflection

Load

Deflection

P y (kN)

∆ y (mm)

Pu

∆ u (mm)

(kN)

[A]

[B]

Ductility index (∆ u /∆ y )

Failure mode

LS-C0

47

16.3

57

74.7

4.6

SY-CC

LS-C0-2F

70

19.8

99

67.0

3.4

SY-CC

LS-C0-4F

80

19.5

114

52.8

2.7

SY-CC

LS-C0-2F-U

70

15.9

105

58.7

3.7

SY-CC

LS-C0-4F-U

80

20.2

122

62.3

3.1

SY-CC

LS-C2

40

10.2

50

73.9

7.2

SY-CC

LS-C2-2F

60

13

90

49.5

3.8

SY-PE

LS-C2-4F

75

13.5

110

39.9

3.0

SY-PE

LS-C2-2F-U

60

12.8

105

74.2

5.8

SY-CC

LS-C2-4F-U

75

13.8

124

57.3

4.2

SY-CC

Results of the corroded-unstrengthened specimen LS-C2 indicate that a 15% tensile steel mass loss resulted in a comparable reduction in the yield load and a 12% reduction in the ultimate load. The effect of corrosion damage and cracking on the yield and ultimate loads of the strengthened specimens was dependent of the amount of the NSM-CFRP strips and the strengthening regime. The yield loads of specimens LS-C2-2F and LS-C2-4F that were strengthened with two and four NSM-CFRP strips without the U-wraps were 14% and 6% lower than those of their counterpart specimens LS-C0-2F and LS-C0-4F that were not corroded, respectively. The ultimate loads of the same specimens decreased by 9% and 3.5%, respectively due to corrosion. This revealed that corrosion of steel had a more pronounced effect on the flexural capacity of the specimens with the lower amount of the NSM-CFRP strips. For the specimens strengthened with the NSM strips together with the U-wraps, corrosion of

473


steel reduced the yield load but had no effect on the ultimate load. Despite corrosion damage and cracking, the residual flexural strengths of all strengthened beams were substantially higher than that of the control beam LS-C0. The ductility indices of the corroded specimens were typically higher than those of their counterparts that were not corroded. This is because the ductility of RC beams typically increases as the cross-sectional area of the internal steel reinforcement decreases. The ductility index of the strengthened beams reduced with an increase in the amount of the NSM-CFRP reinforcement. The ductility indices of the corroded specimens LS-C2-F2 and LS-C2-F4 were 83% and 65% that of the control beam LS-C0, respectively. The ductility index of the corroded beams strengthened with the NSM-CFRP strips together with the U-wraps was about 46% higher than that of their counterparts that were strengthened with the NSM-CFRP strips only. The ductility index of specimen LS-C2-4F-U was approximately 91% that of the control beam LS-C0. The specimen LS-C2-2F-U exhibited a ductility index higher than that of the control beam LS-C0.

(a)

(b)

Figure 4 Load-deflection curves; (a) group [A], (b) group [B]

3.4

CFRP strain at peak load

The effect of corrosion on the CFRP strain measured in the mid-span section at the peak load is depicted in Figure 5. The specimens with the higher amount of the NSM-CFRP strips exhibited lower CFRP strains at the peak load than those exhibited by their counterparts with the lower amount of the NSM-CFRP strips. The CFRP strains at the peak load measured in the uncorroded specimens LS-C02F and LS-C0-4F strengthened with two and four NSM-CFRP strips without the U-wraps were approximately 90% and 68% of the ultimate strain of the CFRP composite strip, respectively. The inclusion of the U-wraps in the strengthening system of the uncorroded specimens slightly increased the CFRP strains at the peak load by approximately 9%, on average, indicating better bond between the steel and concrete and also between the CFRP strips and concrete. Corrosion damage and cracking reduced the CFRP strain at the peak load for the specimens strengthened with the NSM-CFRP strips without the U-wraps by approximately 22%, but had no effect on the CFRP strains of the specimens with the U-wraps. The effect of including the U-wraps in the strengthening system was more significant for the corroded beams than for the uncorroded beams. The CFRP strains at the peak load measured in the corroded specimens strengthened with the NSM-CFRP strips together with the Uwraps were on average 50% higher than those of their counterparts that were strengthened with the NSM-CFRP strips only. This further demonstrates the importance of the inclusion of the U-wraps in the NSM strengthening system for better utilization of the tensile strength of the NSM-CFRP strips. The presence of the U-wraps prevented the premature peeling-off failure of the strengthening system, increased the strains in the NSM-CFRP strips and hence their contribution to the flexural capacity.

474


2.5

CFRP strain at peak load (%)

No corrosion 15% corrosion

2.0

1.5

1.0

0.5

0.0 2 CFRP

2 NSM CFRP + U wraps

4 CFRP

4 NSM CFRP + U wraps

Strengthening regime

Figure 5 Effect of corrosion on the CFRP strain at the peak load

4 Conclusions The effect of corrosion damage and associated concrete cover cracking on the flexural response of RC T-beams strengthened with the NSM-CFRP composite system was dependent of the amount of the NSM-CFRP strips and the presence of U-wraps along the beam span. Corrosion damage of a 15% tensile steel mass loss with a corrosion crack width in the range of 1.0 to 1.2 mm reduced the yield and ultimate loads of the specimens strengthened with two NSM-CFRP strips without the U-wraps by 14% and 9%, respectively. Corrosion of steel had a less pronounced effect on the yield and ultimate loads of the specimens strengthened with four NSM-CFRP strips without the U-wraps where only 6% and 3.5% reductions were recorded, respectively. For the specimens strengthened with NSM strips together with the U-wraps, corrosion of steel and cover cracking had no effect on the flexural capacity. Despite corrosion damage and cracking, the residual flexural strengths of all strengthened beams were substantially higher than that of the control uncorroded beam. The NSM-CFRP strengthening system typically reduced the beam ductility. The ductility index of the strengthened beams further reduced with an increase in the amount of the NSM-CFRP reinforcement. The ductility indices of the corroded specimens strengthened with two and four NSM-CFRP strips without the U-wraps were 83% and 65% that of the control uncorroded beam, respectively. The ductility index of the corroded beams strengthened with the NSM-CFRP strips together with the U-wraps was almost similar to that of the control uncorroded beam. The integration of the U-wraps in the NSM strengthening system was essential to prevent the premature peeling-off of the concrete cover and enclosed NSM composite reinforcement thus allowing the beam to develop its full flexural capacity and ductility.

5 Acknowledgment The authors wish to express their gratitude to the United Arab Emirates University for financing this research work under the NRF-UAEU research grant no. RSA-1108-00186. The authors would like to thank Mr. Faisel Abdelwahab, Eng. Abdelrahman Al-Sallamin and Eng. Tarek Salah for their assistance throughout testing.

6 References [1] De Lorenzis L and Teng J (2007) Near-surface mounted FRP reinforcement: An emerging technique for strengthening structures. Compos. B: Eng., 38(2): 119–143. [2] El-Hacha, R and Riskalla S (2004) Near-surface-mounted fiber-reinforced polymer reinforcements for flexural strengthening of concrete structures, ACI Structural Journal, 101 (5): 717-726.

475


[3] Barros J and Fortes A (2005) Flexural strengthening of concrete beams with CFRP laminates bonded into slits, Cement & Concrete Composites, 27 (4): 471-480. [4] Yost J, Gross S., Dinehart D, and Mildenberg J (2007) Flexural behavior of concrete beams strengthened with near-surface-mounted CFRP strips, ACI Structural Journal, 104 (4): 430-437. [5] El-Maaddawy T, Bouchair A, Biddah A, and El-Dieb, A (2012) Flexural behavior of concrete T-girders strengthened with NSM composite system, 7th Asian Symposium on Polymers in Concrete (ASPIC 2012), Istanbul 3-5 October, Istanbul, Turkey. [6] Bonaldo E, Barros J, and Lourenço P (2008) Efficient strengthening technique to increase the flexural resistance of existing RC slabs, ASCE Journal of Composites for Construction, 12(2): 149-159. [7] Dalfré G and Barros J (2011) Assessing the effectiveness of a NSM-CFRP flexural strengthening technique for continuous RC slabs by experimental research, Proc. of 1st Middle East Conference on Smart Monitoring, Assessment and Rehabilitation of Civil Structures, SMAR 2011, Dubai 8 – 10 February, Dubai, UAE.

476


Optical pH Imaging in Cementitious Materials Engui Liu1*, Masoud Ghandehai2, Weihua Jin2, Alexey Sidelev2, Christian Bruckner3, Gamal Khalil4

(1) New York University Abu Dhabi, Abu Dhabi, United Arab Emirates (2) New York University, New York, Unites States of America (3) University of Connecticut, Storrs, United States of America (4)University of Washington, Seattle, United States of America Abstract: We will report on new pH measurement methodology for cementitious materials. This method is based on an optical measurement that uses a ratio-metric Porphyrin-based sensing compound. The method has the potential to produce full-field 2D pH imaging of fractured concrete surface. The proposed sensor molecule incorporated into the sensing compound has two optical absorbance bands at 575nm, and 700nm. The two bands have opposite absorption responses to pH changes. The ratio of absorbance of the two bands is used to determine the pH value. We will present calibration and optical absorption performance of the sensor molecule in simulated concrete pore solution showing a dynamic pH range from pH 11 to pH 13.5. The digital imaging calibration of the sensing compound on fractured concrete surface presents the pH versus ratio curve for imaging application. Keywords: pH, sensors, degradation, imaging, concrete

1 Introduction

Deleterious chemical reactions lead to premature degradation of concrete and cause deterioration of our roads, bridges, dams, and levees. One of the major modes of distress in concrete is the high pH-dependent degradation. This includes Alkali-Silica Reaction (ASR), carbonation and rebar corrosion in concrete. In such cases, the pH gradient and fluxes of hydroxyl ion (OH-) play important roles in the initiation and propagation of micro-cracks in materials, highly potentially leading to structural damage. Over the past decades, ASR has been increasingly recognized as a major problem, as witnessed by the increasing number of research publications devoted to this topic [1-6]. At the same time, numerous laboratory studies have been carried out to evaluate the pH of carbonated cement paste and concrete, and to thereby address the chemically-induced corrosion of reinforced concrete [7-12]. Despite extensive worldwide research, large gaps remain in understanding the mechanism of high pH-dependent degradation. The knowledge gaps are due partially to lack of effective experimental methodologies to fully address and understand the extreme complexity of the chemical processes in situ. The determination of temporal and spatial fluctuations of internal pH levels and the transport of OH- in concrete in situ are perhaps the most important unresolved aspect of ASR and carbonation-induced distresses. Methods for the direct in situ measurement of high pH levels (around pH12- pH 13) have not yet been possible as the internal concrete pore solution is not readily accessible. Currently, chemical analysis for pH levels in concrete is carried out via pore solution extraction, developed by Longuet et al. in 1973 [13], and subsequently modified by Barneyback and Diamond in 1981 [14]. In this method, a small amount of pore fluid is extracted from hardened cement paste or mortar under very high pressures; the collected sample is then diluted for chemical analysis. Although the method provides an accurate measure of constituents of the extracted pore fluid, the overall process is tedious, destructive and costly. Extracting enough pore fluid for analysis can be extremely difficult, especially in the case of high performance concrete [15] where the *

Lecturer of Civil Engineering, New York University Abu Dhabi, [email protected],

477


water tocement ratio can be as low as 0.16 with crushing strength as high as 500 MPa, resulting in minimum amounts of extracted pore fluid. Finally, the measured chemical composition reflects only bulk information, with no information on spatial variations. It is therefore imperative to develop new, innovative, easy-to-use technologies that can indicate the chemical signature of cementitious materials on the temporal and spatial scales necessary for understanding the phenomena. The objective of this investigation is to make possible the direct measurement of pH profiles in Portland cement concrete with temporal and spatial resolution. For achieving that, this investigation seeks to develop a full-field optical imaging approach using a robust high pH sensing compound with a Porphyrin-based sensor molecule [16] incorporated for measurement of pH levels in cementitious materials.

2 Materials and Methods 2.1

Materials

(1) Simulated concrete pore solution The simulated concrete pore solution was prepared by dissolving Ca(OH) 2 , NaOH, and KOH in deionized water. The concentration levels of K+, Na+ and Ca2+ was kept at 0.4 M, 0.2 M and 0.001 M, respectively; thus the OH- concentration balances the sum of the alkali ions at about 0.6 M [15]. The solution was then diluted with deionized water to various pH levels.

(2) pH sensor molecule stock solution The sensor molecule was first dissolved into dimethyl sulfoxide (DMSO) with the concentration 0.4×10-3 M, and used as the sensor molecule stock solution.

(3) pH sensing compound The pH sensing compound was prepared by mixing the sensor molecule stock solution and polyvinyl alcohol solution (PVA solution, 30% weight in deionized water) together with a surfactant.

2.2

Methods

(1) UV-Vis Spectroscopy UV-Vis spectroscopic measurement was performed to record the absorbance of pH sensor molecule in simulated concrete pore solution at different pH levels from pH 10 to pH 13.5. The measurement was conducted in a cuvette system in which 1 mL simulated concrete pore solution and 0.1 mL pH sensor molecule stock solution was respectively added. The colour change of the solution was recorded using a digital camera, and the absorbance spectrum was measure using the set-up in Figure 1.

Figure 1 Experimental set-up for UV-Vis spectroscopic measurement

478


(2) Image Collection and Processing A Digital Singular Lens Reflex (DSLR) camera was used to record the images. Two Xenon lamps were employed as the illumination source, and two optical bandpass filters were used to capture image intensity at two wavelength ranges 575nm and 700nm (Figure 2). The wavelength range for filter 575 is from 550 nm to 600 nm with the maximum transmittance at 575 nm; and from 600 nm to 800 nm with the maximum transmittance at 700 nm for filter 700. The image processing generates a channel of ratio information which determines the pH levels based on the extracted intensities from images captured under filters 700 and 575.

Figure 2 Experimental set-up for images collection


Absorbance of sensor molecule in simulated concrete pore solution

The sensor molecule exhibits instant colour change after added to the simulated concrete pore solution. The colour changes from pink to bright yellow (Figure 3) against various pH levels in the range of pH 10-pH 13.5. The bright yellow presents at high pH values around pH 13; while the pink indicates lower pH levels near pH 11 and below; the intermediate colours between yellow and pink represent the pH levels about pH 12.

Figure 3 Colour change of pH sensor molecule versus pH in simulated concrete pore solution

The underlying mechanism for the observed colour change is the nucleophilic addition/attack of OH- onto the sensor molecule [16] in which the optical absorbance of the molecule changes after the addition/attack. Figure 4 shows the absorbance spectra of the sensor molecule in simulated concrete pore solution at various pH levels. The sensor molecule exhibits two characteristic absorbance peaks, one centred at 575nm and the other at 700nm. The 575nm peak represents the neutral/lower pH state of the sensor, e.g. pH 10, pH 11. The 700nm peak relates to the base/higher pH near pH 13. The two channels of absorbance change inversely

479


towards the change of pH levels, i.e., the absorbance intensity at 575nm decreases while that of 700nm increases, as the pH level increases.

Figure 4 Absorbance spectra of pH sensor molecule versus pH in simulated concrete pore solution

As stated earlier, the optical absorbance at two bands of 575nm and 700nm represent two forms of the sensor molecule versus pH levels. These two channels, as a result, were processed to derive the third channel of ratios between the two channels. In this way, the ratios determine the pH levels or the concentration of OH-.

3.2

Dynamic pH sensing range of the sensor molecule

Based on the absorption spectra (Figure 4), the absorbance intensities at band 575nm and 700nm were determined with intensity values at 800nm as the reference, in which, the absorbance value at 575nm and 700nm were obtained by subtracting the one at 800nm. The ratios between absorbance values at 575nm and 700nm were then calculated and expressed in Figure 5. Both of the ratios present the same pH dynamic sensing range from pH 11 to pH 13.5.

Figure 5 Ratio of absorbance at 575nm and 700nm versus pH based on absorbance spectra

480


3.3

Peak area integration from absorbance spectra of the sensor molecule

The objective of the method is to develop an imaging process to map the spatial variation of pH levels on fractured concrete surfaces. Unlike a spectrometer which records the full absorbance spectrum with intensity information for each wavelength, the digital camera captures intensity of all photons that arrive at the camera sensor matrix without the wavelength information. Referring to the spectra shown in Figure 4, the areas under the two peaks of 575nm and 700nm were quantified through integration (with y=0 as the base line), and the calculated intensities were used to calculate ratios which determine the pH values. This way of spectra processing is similar to the way the camera sensor acquires integraties of images. Therefore, during the images collection, two optical band filters were outfitted to camera lens for capturing intensities at the two peaks of 575nm and 700nm. This exercise was first carried out on each spectrum and the same baseline was chosen for both peaks. The integration range for the two peaks was determined by the specifications of the optical band pass filters shown as the highlighted yellow coverage in Figure 6.

Figure 6 Absorbance spectra showing areas covered by the band pass filters

Figure 7 Ratio of Integrated Areas under Peaks 575nm and 700nm versus pH

481


Figure 7 shows the ratios between integrated areas under the 575nm and 700nm peaks based on the spectra. Both ratios show good linearity from pH 11 to pH 13.5 which agree well with the peak value calibrations shown in Figure 5. This implies that optical band pass filtering of images taken by a digital camera has potential for assessing a wide range of pH levels on concrete surface.

3.4

pH sensing compound evaluation and Calibration

After preparation, the pH sensing compound was simply tested on a glass slide and on a piece of fractured undegraded cement paste. Figure 8 shows its color change on a glass slide after treatment with a drop of pH 13 concrete pore solution and on the undergraded cement paste surface, which has the same high pH range. In both cases the sensing compound changed color from pink to yellow over time. In the undegraded cement paste, a drop of acid at pH 1 made the already-yellow area return to pink, indicating the reversible performance of the sensing compound (bottom right).

Figure 8 The pH sensing compound on glass slide and treated with pH 13 concrete pore solution (top); process of sensing compound application on surface of undegraded cement paste (bottom)

The calbration of the sensing compound was to apply the sensing compound on undegraded cement paste with an known average pH level of pH 13, followed by imaging and processing to obtain an average intensity ratio which was assigned to pH 13. This ratio for pH 13 was then used to comapre with the one obtained from the spectra (Figure 7) and the further interpolation of the curve generates a new ratio-pH curve for imaging application. Figure 9 shows the sequence of sensing compound calibration process for imaging application. In Figure 10, an example of images and processed ratio of an undegraded cement paste sample was demonstated; and specially in part (c) and (d) of Figure 10, the processed ratio versus position on the undegraded cement paste surface was shown. The ratio values fluctuated around 1.5 and further processing of the curve generated an average ratio of 1.5, which was then assigned to pH 13. The final calibrated ratio-pH curve for imaging application (Figure 11), as earlier mentioned, was obtained through the interpolation process between ratios from processing spectra of the sensor molecule and images of sensing compound on undegraded cement paste samples.

482


Figure 9 Flow chart for pH sensing compound calibration process

Figure 10 Images and processed ratio for regular cement paste (undegraded, average pH 13) sample at pH sensing compound calibration process, (a) Colour image of Phenolphthalein applied on cement paste; (b) Colour image of pH sensing compound applied on the companion side of cement paste; (c) processed image showing ratio of intensities from images taken under filer 700 and 575; (d) processed curve showing average transverse ratio along the sample

483


Figure 11 Calibrated curve of pH versus ratio for imaging application

* calibrated by setting the average ratio of 1.5 (ratio of image: 700/ 575, Figure 10) to pH 13 as determined with undegraded cement pastes that are averagely pH 13

4 Conclusions The Porphyrin-based pH sensor molecule is capable of sensing high pH with the dynamic range from pH 11 to pH 13.5, suitable for applications to cement based materials, monitoring pH changes at early stages of degradations or hydration reactions. The pH sensing compound is capable of being applied onto fractured concrete surface for pH imaging application. The pH imaging protocol and calibrated pH versus ratio curve of sensing compound for application on fractured concrete surface has been developed.

5 References [1] Chatterji S. (2005) Chemistry of alkali–silica reaction and testing of aggregates, Cement and Concrete Composites, 27: 788-795. [2] Sargolzahi M., Kodjo S.A., Rivard P., Rhazi J. (2010) Effectiveness of nondestructive testing for the evaluation of alkali–silica reaction in concrete, Construction and Building Materials, 24: 1398-1403. [3] Bleszynski R.F., Thomas M. D.A. (1998) Microstructural studies of alkali-silica reaction in fly ash concrete immersed in alkaline solutions, Advanced Cement Based Materials, 7: 66-78. [4] Aquino W., Lange D.A., Olek J. (2001) The influence of metakaolin and silica fume on the chemistry of alkali–silica reaction products, Cement and Concrete Composites, 23: 485-493. [5] Garcia-Diaz E., Riche J., Bulteel D., Vernet C. (2006) Mechanism of damage for the alkali–silica reaction, Cement and Concrete Research, 36: 395-400. [6] Duchesne J., Bérubé M.-A. (2001) Long-term effectiveness of supplementary cementing materials against alkali–silica reaction, Cement and Concrete Research, 31: 1057-1063. [7] Alonso C., Andrade C., Gonzalez J.A. (1998) Relation between resistivity and corrosion rate of reinforcements in carbonated mortar made with several different cement types, Cement and Concrete Research, 18: 687-698. [8] Glass G.K., Page C.L., Short N.R. (1991) Factors affecting the corrosion rate of steel in carbonated mortars, Corrosion Science, 32: 1283-1294. [9] Gonzalez J.A., Algaba J.S., Andrade C. (1980) Corrosion of reinforcing bars in carbonated concrete. British Corrosion Journal, 15: 135-139.

484


[10] Suryavanshi A.K., Swamy R.N. (1996) Stability of Friedel’s salts in carbonated concrete structural elements. Cement and Concrete Research, 26: 729-741. [11] Yeih W., Chang J.J. (2005) A study on the efficiency of electrochemical re-alkalisation of carbonated concrete. Construction and Building Materials, 19: 516-524. [12] Huet B., L’Hostis V., Miserque F., Idrissi H. (2005) Electrochemical behavior of mild steel in concrete: influence of pH and carbonate content of concrete pore solution. Electrochimica Acta, 51: 172-180. [13] Longuet P., Burglen L., Zelwer A. (1973) The liquid phase of hydrated cement (in French) Rev. Mater. Constr., 676: 35-4. [14] Barneyback R. S., Diamond S. (1981) Expression and analysis of pore fluids from hardened cement pastes and mortars, Cement and Concrete Research, 11: 279-285. [15] Rasanen V., Penttala V. (2004) The pH measurement of concrete and smoothing mortar using a powder suspension, Cement and Concrete Research, 34: 813-820. [16] Khalil G. E., Daddario P., Lau K. S. F., Imtiaz S., et al. (2010) mero-Tetraarylporpholactones as high pH sensors, Analyst, 135: 2125-2131.

485


Effects of accelerated ageing on the adhesive bond between concrete specimens and external GFRP reinforcements Debayan Ghosh 1*, Showmen Saha 2, Somnath Karmakar 3 (1) Undergraduate Student, National Institute of Technology, Durgapur, India (2) Professor, National Institute of Technology, Durgapur, India (3) Assistant Professor, National Institute of Technology, Durgapur, India Abstract: This paper presents an experimental investigation on the durability of bond between glass fibre-reinforced polymer (GFRP) sheet and concrete, specifically as it relates to bonding of the GFRPsheet surface and behaviour of the sheet–concrete interface. The effect of temperature, moisture, pH level, and freezing-and-thawing cycles on the mechanical properties of GFRP-wrapped concrete was investigated. Aged GFRP concrete hybrid cylinders and prisms were tested in compression after 1000, 3000, and 8000 hours of environmental exposure. GFRP composite coupons were exposed to the same conditions and tested to identify the bonding mechanisms. The GFRP sheet was embedded in concrete and exposed to tap water at 23oC, 40oC, and 50oC to accelerate potential degradation. Keywords: Glass Fibre-Reinforced Polymer, Bond, Durability

1 Introduction

Infrastructure decay due to corrosion of embedded reinforcing steel stands out as a significant challenge worldwide. The main long-term deterioration mechanism involves moisture diffusion and the transport of dissolved harmful chemicals within the concrete, which can affect internal reinforcement. Glass-fiber-reinforced polymer (GFRP) materials have not been used in large-scale construction applications despite their numerous advantages over traditional materials such as steel. Yet the long-term performance of GFRP remains unresolved under some special conditions, such as in highly acidic environments. Nevertheless, GFRP’s low cost-to-performance advantage is driving its worldwide use and acceptance. GFRP materials are considered to provide high strength, while being lightweight, noncorrosive, and nonconductive. As a common feature of composites, prominent anisotropy in mechanical properties was observed, this has high fracture strength and stiffness along the fiber strengthening component. Yet its potentials are not fully realized due to moisture affecting the long-term life of composite properties. Preparing defect free composites are rather uncommon arising from their adopted processing techniques through which moisture penetration is quite common. Wide acceptance of FRP components in the construction industry requires comprehensive investigation of their structural and mechanical behavior to ensure their suitability for civilengineering applications. Bond development is a critical issue for their successful application as internal reinforcement in concrete structures. Several investigations have been carried out to determine GFRP durability under environmental conditions that could occur under actual service conditions. In addition, it is well-known that the coefficients of thermal expansion (CTE) of GFRP sheets are not the same in the longitudinal and transverse directions. The longitudinal CTE—depending on fibers—is lower than that of concrete, while the transverse CTE—depending on matrix—is about 2–4 times greater than that of concrete. Therefore, thermal gradients can lead to mismatching of transverse thermal expansion values between FRP and concrete, degrading the interface between FRP sheets and concrete and even resulting in concrete cracking. The effects of concrete environment on FRP and the mismatch of CTE between FRP sheets and concrete are major concerns affecting the long-term bond behavior of concrete structures with external GFRP reinforcement.

*

Undergraduate Student, National Institute of Technology, Durgapur E-mail: [email protected]

486


Designers and owners have concerns about the performance and long-term adhesion properties at the interface between concrete and GFRP sheets. This study sought responses to such issues through simulating field conditions: immersion of concrete wrapped GFRP sheets in tap water and characterizing the long-term performance of the concrete–sheet interface. In particular, the main objective of this study is to characterize the long-term durability of the interface between GFRP sheet and concrete with dynamic and static loading tests. This economical test is commonly used to assess bond behavior, even if it does not allow for direct measurement of design strength. The conditioning used in this study is more consistent with field conditions because the FRP material is attached to concrete.

2 Materials GFRP sheets manufactured by a company named ‘Vetrotex’ were used in this study (Fig 2). The GFRP sheets were made of continuous longitudinal E-glass-fiber strands and were bonded with the concrete coupons using polyether resin. The sheets of glass fibre were exposed to boiling water for 5minutes, 10minutes, 20minutes, 40minutes, 1hour and 2hours respectively to study its effects in terms of percentage weight loss and percentage weight gain (all are summarized in Table 1). The mechanical and physical properties-measured during preliminary tests-are summarized in Table 2. The concrete mixture consisted of 455kg of Type 10 cement (corresponding to ASTM I cement), 678kg of fine aggregate, 1171kg of crushed angular coarse aggregates and 203kg of water per cubic meter of concrete. The 28 days compressive strengths ranged from 22 to 26 MPa. These specimens were cylinders with diameter 150mm and height 300mm and prisms of cross section 100mm x 100mm and length 500mm. Fig 1 depicts such specimens. These specimens were singly wrapped using GFRP sheets. Table 1 Effects of boiling water on glass fiber Weight of fiber

Weight of fiber

Weight

Weight

before boiling

after boiling

loss

gain

Boiling Time (minutes)

(grams)

(grams)

(%)

(%)

5

17.5671

17.5534

0.078

--

10

15.6813

15.6604

0.133

--

20

16.0475

16.0330

0.090

--

40

15.9713

15.9490

0.140

--

60

16.2952

16.2980

--

0.017

120

15.8357

16.0060

--

1.075

Fig 1: Casted Concrete Specimens

Fig 2: Gfrp Sheet

487


Table 2 Properties of GFRP Sheet Property

Units

Value

Young’s Modulus

GPa

26

Poisson’s Ratio

0.26

Shear Modulus

GPa

10

Yield Stress

MPa

125

Breaking Strain

2%

Coefficient of Thermal Expansion

o

/C

19*10-6/oC

3 Testing Procedure This study involved accelerated aging of GFRP sheets wrapped to concrete specimens. Six number specimens were cast for each Cylinder and Prism. The said specimens were kept at saturated humidity for 3 days before initial conditioning, consisting of complete immersion in tap water at 25 oC for 28 days. The purpose of immersion was to stabilise the concrete’s mechanical properties. After completion of initial conditioning, specimen bond properties were measured and served as a baseline for bond properties. The other samples for testing were kept immersed in tap water for additional aging at different temperatures. Previous work at the University of Sherbrooke [6] has shown that long-term degradation was not significantly affected by the type of water used for accelerated aging (tap or deionized). The immersion receptacles were PVC containers specially manufactured for the study. The samples were separated from each other and the container bottom to allow the tap water to circulate freely between and around the GFRP wrapped samples. The water level was kept constant throughout the study to prevent the pH from increasing as a result of water evaporation (increased alkalineion content). The water temperatures were chosen to accelerate the aging effect, yet not high enough to trigger thermal degradation. The aging conditions in this study aimed at simulating actual application conditions. They were harsher than actual field conditions, however, since the specimens were continuously saturated with water. Following initial conditioning, specimens were fully immersed at three different temperatures (23 oC, 40 oC, and 50 oC) for three different lengths of time. The high temperatures accelerated degradation, as shown by the increased moisture diffusion rates in the concrete and sheets. The increased temperature during accelerated aging simulates the effect of time. Accelerated aging also provides information about the long-term behavior of the GRFP-sheet– concrete interface. At the end of each period, six specimens were removed from the water and subjected to static and dynamic flexure testing for prisms and split-tensile tests for cylinders to compare their average bond stress, mode of failure to those of the control specimens.

3.1

Tests

The GFRP wrapped prisms were subjected to static as well as dynamic testing using centrepoint-loading technique that conformed to ACI code. Each GFRP wrapped prism was instrumented with linear variable differential transformer (LVDT) at the point of loading to record deflection during testing. The test was carried out using a Hydraulic Actuator machine manufactured by HEICO, India. The static method was applied by applying an increasing load of 0.5kN/sec. For each test, the specimen was mounted on two roller supports and an iron pin of diameter 25mm was placed under the piston using an adjusting plate at the centre of the prism to give the centre point loading. The applied load, displacement, breaking load, load vs. displacement and load vs. time graph was recorded from the instrument 488


The GFRP wrapped cylinders were subjected to Indirect Tension Test by using Cylinder splitting tests. The test was carried out by placing the cylinder specimen horizontally between the loading surfaces of a compression testing machine and the load was applied until failure of the cylinder, along the vertical diameter. Fig 4 shows the test specimen. The test was carried out using a Compression Testing Machine (CTM) manufactured by HEICO, India. The load was applied by gradually increasing the load @ 4kN/min.

Fig3 : Testing of Specimens in Actuator

Fig4 : Testing of Specimens in CTM

Fig5 : Centre Point Loading

4 Results and Discussions The samples were examined after they were taken out of the environmental chamber. No initial cracks developed due to the difference in coefficients of thermal expansion (CTE) of the GFRPepoxy bond and concrete. The flexural strengths of the prisms subjected to various conditions are summarized in Table 3. Fig-6 shows the comparison of the Load vs Displacement curves of the specimens in terms of line chart. The values depict the average load at which the specimens fails. The flexural strength was only slightly affected after the samples were exposed to aggravated aging conditions. As depicted by Fig 6, the average load vs displacement curve also follows the same path for the unattacked free samples as well as the samples which were under attack and the initial displacement of the prism subjected to heat was slightly more than the normal one and this may be due to the reduction in stiffness of the GFRP-epoxy bond. But in the other two cases there was no significant effect.

489


Table 3 Average Flexural Strengths of prisms Failure Load Type of attack

kN

No Attack

13.6

Chemical Attack

13.1

Temperature Attack

13.4

Fig 6 The average failure loads of the cylinders subjected to Split-tensile tests are summarized in Table 4. Fig 7 shows the crack pattern developed in the specimen. Table 5 shows the percentage deviation in the strengths for cylinder and prisms under various conditions compared to the samples which were unattacked. Table 3 Average Failure loads of cylinders Failure Load Type of attack

kN

No Attack

299.667

Chemical Attack

295

Temperature Attack

299.333

Fig 7 Crack pattern in cylinder

Table 5 Percentage Deviation in the Failure Loads

490


Sample

Chemical Attack

Temperature Attack

Prisms

3.676%

1.47%

Cylinders

1.557%

0.111%

Now from Table 1 we see that effect of boiling water on GFRP sheets are negligible. Thus the above results depict that the effect of aggravated aging attacks on the epoxy bond between the GFRP sheet and the concrete surface is very negligible. From Table 5 it is clear that the effect on the epoxy bond was very negligible and the specimens took almost the same load till failure occured.

4.1

Modes of failure

All specimens tested, failed by crushing of concrete. After the tests the concrete samples were examined to check the bond failure mode. The GFRP sheet and fiber materials were still attached to the concrete, although the specimen had developed cracks. Failure occurred there because of the high tensile load developed in the fibre and the concrete sample. Fig 8 and 9 shows the crack pattern and the GFRP sheet after failure respectively.

Fig 8

4.2

Fig 9

Conclusions

Based on the above results, the following conclusions may be drawn: 1. The effect of aggressive environmental conditions on GFRP sheet and the epoxy bond was negligible even when subjected to very high temperatures. 2. The samples failed due to crushing of concrete and not due to degradation of the GFRP sheet or the epoxy bond 3. The concrete and the resin–fiber interfaces were not apparently affected by moisture absorption and high temperature.

5 Acknowledgements

The research was conducted with the help from Department of Civil Engineering, NIT, Durgapur. All equipments were provided by the Institute. The authors thank the Institute for all the cooperation and help provided in authoring this paper.

6 References

[1] American Concrete Institute (ACI). Guide test methods for fiber-reinforced polymers (FRPs) for reinforcing or strengthening concrete structures. ACI 440.3R-04, Mich: Farmington Hills; 2004. [2] American Society for Testing and Materials. Standard test methods for density and specific gravity (relative density) of plastics by displacement. ASTM D 792; 2000. [3] Karbhari VM, Stachowsky C, Wu L. Durability of pultruded E-glass/vinyl ester under combined hydrothermal exposure and sustained bending. J Compos Constr 2007; 19(8):665–73. [4] Porter ML, Barnes BA. Accelerated aging degradation of glass fiber composites. In: Proceedings of 2nd international conference on composite in infrastructure, Tucson, Arizona; 1998. p. 446–95.

491


[5] Riebel F, Keller T. Long-term compression performance of a pultruded GFRP element exposed to concrete pore water solution. J Compos Constr 2007; 11(4):437–47. [6] Wang P. Effect of moisture, temperature, and alkaline on durability of E-glass/ vinyl ester reinforcing bars. Ph.D. Thesis, University of Sherbrooke; 2005. p. 154. [7]Benzarti K, Quiertant M, Marty C, Chataigner S, Aubagnac C. Effects of accelerated aging on the adhesive bond between concrete specimens and external CFRP reinforcements. CICE 2010 - The 5th International Conference on FRP Composites in Civil Engineering September 27-29, 2010 Beijing, China. [8] Cognard, J. 2006. Some recent progress in adhesion technology and science. Comptes-Rendus de Chimie 9: 3-24.

492


Assessment of Mechanical and Chemical Deterioration of Artworks Roger M. Groves1*, Cristina Portalés2, Emilio Ribes-Gómez3 (1) Delft University of Technology, Delft, The Netherlands (2) Universidad de Valencia, Valencia, Spain (3) AIDO, Institute of Optics, Color and Image, Valencia, Spain Abstract: The preservation of cultural heritage is a common interest across Europe and the World. In the FP7 Syddarta project the project partners are developing a prototype instrument for the assessment of mechanical and chemical ageing and deterioration on paintings from the Baroque period. Mechanical deterioration is assessed by measuring shape using structured light projection and implementing morphological algorithms. Spectroscopic data is collected by hyperspectral imaging cameras recording in the visible and infrared. This hyperspectral data is processed using Multiplicative Scatter Correction and classified with Principal Components Analysis and Support Vector Machines. The processing algorithms for system calibration and assessment of deterioration are described in detail in this paper and examples are presented of the results from painted wooden panels. Keywords: mechanical deterioration, chemical deterioration, classification, spectral database, hyperspectral imaging

1 Introduction More than 21.000 institutions across Europe exhibit art in permanent or temporary collections, according to the EGMUS - European Group on Museum Statistics [1]. These museums and galleries play a key role in the sustainability of cultural heritage by combining tourism with the educational role of disseminating cultural heritage. An increasing interest in cultural heritage means that collections need to be actively managed to avoid deterioration of objects. Particular challenges are the transportation of movable cultural heritage and the possibility of damage during transit or different storage conditions, variations in museum climate (temperature, humidity, vibration) due increased numbers of visitors, ageing due to pollution, longer exposure times to natural and artificial light and the possibility of theft or damage. Mechanical deterioration can be addressed by measuring object shape or surface displacement under active or passive loading. Object shape can be measured using stereoscopic imaging, structured light projection [2] or low-coherence interferometry, depending on the scale of the object and accuracy required. Alternatively displacement may be measured with either shearography or holography [3] using active loading or by in-situ monitoring using passive loading [4]. Chemical deterioration assessment requires typically spectroscopic investigation. In Raman spectroscopy [5] microgram samples from the painting are analysed to provide detailed information on organic, inorganic and biological materials, phase transitions, corrosion and pollutants. Laser-induced breakdown spectroscopy (LIBS) is a useful method for in-situ determination of the elemental composition for both metal and light atomic elements nearly non-destructively (damage to the object is at the microscopic level). Fibre Optics Reflectance Spectroscopy (FORS) can be used for the characterisation of pigments by measuring point-bypoint reflectance spectra of small areas (diameter ~3mm). Hyperspectral imaging [6] uses filter wheels, electronically tuneable filters (ETF) or diffraction gratings to record spectral bands for each pixel in an image. In the visible and near infrared wavelengths (350-1000 nm) pigment identification has previously been performed using an 8-channel multi-filter system. From this the estimated spectral reflectance [7], as well as its first derivative, was compared to a database of 173 traditional Japanese pigments. Spectral imaging in the reflective IR region (700-2500 nm) was investigated in [8] for its potential to discriminate azurite, indigo, Prussian blue, lapis lazuli, *

Dr Roger M. Groves, Delft University of Technology, e-mail: [email protected]

493


cobalt blue, ultramarine and thalo blue pigments. Visible and infrared cameras with spectral band filters were used to record multispectral images of test panels and paintings. The FP7 Syddarta Project addresses the assessment of mechanical and chemical deterioration by the development of a portable instrument, which records shape data by structured light projection and spectroscopic data by infrared and visible hyperspectral imaging cameras. Based on the interest of museum partners in the consortium (IPCHS, Institute for the Protection of Cultural Heritage of Slovenia; RABASF, Real Academia de Bellas Artes de San Fernando, Spain) the Baroque period was selected for the case study and investigation of pigment properties. This paper will focus on the image processing routines developed to assess deterioration from shape and spectroscopic data. The prototype Syddarta instrument and control routines are introduced to describe the source of the data. Pre-processing is performed using system calibration algorithms to correct for perspective, chromatic and lens distortion. Then the techniques for mechanical and chemical deterioration will be described. Results from the measurement of paintings will be presented and discussed.

2 Hardware and Data Recording

This section describes the hardware used to record the shape and spectroscopic data. The individual measurement channels are described first, then the design and construction of the prototype instrument. Finally the instrument control software, operating procedure and data formats are presented. This section is intended to give a clear understanding of the type and quality of data that will be analysed using the algorithms described in Section 3. The system has been integrated in a portable prototype unit consisting of an optical head, power supply unit and laptop. The complete prototype system can be packed into a case for transportation by courier.

2.1

Structured Light Projection Channel

The light source for the Structured Light Projection (SLP) Channel is a CBT-90 WHITE LED (Luminus Devices Inc.) integrated with a V-9600 DLP Chip (Vialux GmbH), controlled by a V9600 DLP Controller Module (Vialux GmbH). This sub-system projects a series of phase-shifted sinusoidal vertical stripe patterns. Light scattered by the object is recorded by an acA204025gmNIR (Basler AG) with a C-mount camera lens. The camera views the object through a Varispec VIS Liquid Crystal Tuneable Filer (LCTF) (Perkin Elmer Inc.), also required by the Hyperspectral Visible Channel, described in Section 2.2.

2.2

Hyperspectral Visible Channel

The Hyperspectral Visible Channel (HV) uses the same hardware components as the SLP Channel, described in Section 2.1. The DLP Chip is set to project a constant illumination field on the object. The LCTF is controlled to scan through the visible wavelength range (400 to 720 nm), synchronised with the image capture by the acA2040-25gmNIR camera.

2.3

Hyperspectral Infrared Channel

The light source for the Hyperspectral Infrared (HI) Channel is a projection lamp type 7027 (Philips) in a custom optical assembly (Avantes BV). Light from the lamp is collimated and directed through a TF1700-1600-10-5-SD1 Dual-Transducer Acousto-Optic Tuneable Filter (AOTF) (Gooch & Housego PLC), controlled by two 16-channel AOTF drivers MSD038-11610UM-16x1-SR5 (Gooch & Housego PLC). Additional optics are used to direct the first-order beam to the object under testing and block the zero-order beam. Scattered light from the object is imaged by an OLES Macro infrared camera lens (Specim) onto a custom infrared camera (Xenics nv).

494


2.4

Control Software and Data Formats

The prototype instrument is controlled by Labview software using the ‘Queued State Machine – Producer Consumer’ (QSM-PC) architecture, with SubVIs for the different hardware devices. After system calibration, fringe patterns for 3D shape measurement are recorded, followed by the recording of hyperspectral data in the visible range, then hyperspectral data in the infrared region. All images are stored as PNG files, along with a datalog file giving all system parameters.

3 Algorithms for Digital Signal Processing

This section describes the algorithms developed for the analysis of mechanical and chemical deterioration. The main focus of the algorithms is to detect anomalies in the object shape and spectral signatures which are indicators of ageing of the structure and its components.

3.1

Structural Deterioration

Both canvas and panel supports were in use during the Baroque period. Canvas supports consist of a linen fabric tacked, or stapled on a wooden stretcher. The canvas will initially be taught and unwrinkled in the stretcher and may have been adjusted by keying the corners. Sizing, ground, paint and varnish layers are applied to this support. Structural deterioration of the canvas support and 3D deformation due to canvas tensions occurs due to the influence of temperature and humidity on the wooden stretcher and chemical aging of the canvas and applied layers. Wooden, or panel supports, deform mainly due to swelling and shrinkage of the wood due to variations in temperature and humidity. These changes may be reversible, i.e. elastic deformations, or irreversible, e.g. cracking, creep, compression set. The constraint of the panel, e.g. cross braces and cradles will have a significant effect on the deformation of the panel. Also of interest is the difference in behaviour due to the type of wood species, with poplar and oak being the most common materials in the Baroque period. For both canvas and wooden support deviation from a flat surface and the presence of holes/cracks are important. To analyse structural deterioration, the SLP Channel data is first processed to obtain the object shape, as described in [9]. The height profile of the object, g(x,y), is encoded as a function of the spatial phase of the fringe pattern [10], as expressed in equation (1):

g ( x, y ) = a ( x, y ) + b( x, y ) cos(2πf 0 x + ϕ ( x, y ))

(1) Where a(x,y) represents the background illumination, b(x,y) is the amplitude modulation of the fringes, f 0 is the spatial carrier frequency and φ(x,y) is the phase modulation of the fringes. The Fourier Transform (FT) method [10] in combination with phase-shifting of the projected patterns allows unwanted zero-order background illumination, a(x,y) and higher-order data to be removed by filtering, as described in [11]. This demodulation results in a wrapped phase map ψ(x,y), which after unwrapping [12] is transformed in a phase-to-height conversion, using the equations in [13] to give height values relative to a fictitious reference plane. From these triangulation is applied to reconstruct the object in the 3D reference frame. The geometric calibration of the camera and projector was performed following the procedure in [14]. A surface is then generated from the point cloud using the moving least-squares (MLS) technique [15]. From this surface profile, surface normals are determined using the eigenvector analysis based method [16], which can be compared within a specified radius, to determine shape imperfections. The algorithm uses r s (small radius, 100 points) and r l (large radius, 2500 points) to identify small scale and large scale surface imperfections.

3.2

Pre-Treatment of Spectroscopic Data

Spectral data contains non-linearities, due to scatter from particulates in the sample. There are several approaches to minimizing these non-linearities, so that the spectra can be more effectively modelled, such as Multiplicative Scatter Correction (MSC) [17] and Extended Multiplicative Scatter Correction (EMSC) [18], described below. The basic concept of MSC is to 495


remove non-linearities unrelated to the property of interest, due to scatter from particulates in the sample. First correction coefficients are estimated, as shown in equation (2) and corrected spectra are calculated using equation (3). 𝑥𝑜𝑟𝑔 = 𝑏0 + 𝑏𝑟𝑒𝑓,1 𝑥𝑟𝑒𝑓 + 𝜖 (2) xcorr =

xorg −b0 bref,1

(3)

Where b 0 , b 1 , etc. are correction coefficients, x org , x ref and x corr are the original, reference and corrected spectra respectively. The EMSC method allows a separation of physical light-scattering effects from chemical absorbance effects in spectra from powders or turbid solutions. This model-based method is particularly useful in minimizing wavelength-dependent light scattering variation. The mathematical description of EMSC is given equation (4), where the corrected spectrum is given in equation (5). 𝑥𝑜𝑟𝑔 = 𝑏0 + 𝑏𝑟𝑒𝑓,1 𝑥𝑟𝑒𝑓 + 𝜆𝑏𝜆,1 + 𝜆2 𝑏𝜆,2 + 𝜖 (4) 𝑥𝑐𝑜𝑟𝑟 =

3.3

𝑥𝑜𝑟𝑔 −𝑏0 −𝜆𝑏𝜆,1 −𝜆2 𝑏𝜆,2 𝑏𝑟𝑒𝑓,1

(5)

Spectra Analysis and Classification

Typical chemical reactions that cause damage are reactions with atmospheric contamination (pollution, acidification), photochemical reactions, oxidation and corrosion of metal components of the support, decomposition and hydrolysis of the linen, degradation of pigments (e.g. chalking, yellowing, oxidation, etc.), moisture and dirt penetrating the varnish layer and salt effloresce. Principal Component Analysis (PVA) for the pigment identification is described in [19-20] for Raman Spectroscopy spectra. The identification methodology is based on comparison between spectra, so all data should have the same dimensions and need to be normalized. Then polynomial fitting is applied in order to reduce the fluorescence (which is a radiated phenomenon, generated by many materials, which underlies the measured spectra.). As each spectrum is a vector of N (approximately 1000) points, the aim is to reduce the N-dimensional data to typically two or three dimensions, without loss of information. In [19] the correlation coefficient and a fuzzy algorithm are used, whereas in [20] Euclidean distance is preferred. The approach used here is described below in more detail. First PCA is used to reduce the data dimensionality and construct the feature space of the training data. Then classification is performed. Ten classification methods have been assessed, with 6 based on Euclidean distance and variations, and three on Support Vectors Machines (SVM) [21], see Table 1. # 1. 2. 3. 4. 5. 6. 7. 8. 9.

Table 1 Classification Methods used for Principal Component Analysis

Method description Minimum of the Euclidean distance to all training samples Minimum of the Mahalanobis distance to all training samples Minimum of the mean values of the Euclidean distances for each class Minimum of the mean values of the Mahalanobis distances for each class Minimum of the Euclidean distances to the mean training sample for each class Minimum of the Euclidean distances to the mean training sample for each class Support Vectors Machine, with linear kerned functions Support Vectors Machine, with polynomial kerned functions Support Vectors Machine, with Radial Basis Function (RBF).

Figure 1 shows an example of the classification for method #1 Euclidian Distance. The green, red and purple circles are reference samples in different classes.

496


Figure 1 Example of classification with method #1 Euclidian Distance. The test sample of orange colour would be classified in the ‘green’ class.

4 Results and Discussion This section describes the assessment of the algorithms developed for the analysis of mechanical and chemical deterioration. Results are presented for the eigenvector method for deviations in surface shape, scattering correction methods for spectral data and PCA classification methods.

4.1

Structural Analysis Results

Results are presented for the artwork ‘Sansón y Dalila’ (Sampson and Delilah), oil on canvas, painted by José Camarón y Miliá in the late 18th century. First the 3D surface is obtained from the point clouds using the MLS technique, as described in Section 3.1. Figure 2 shows a picture of the artwork and a close up view of the 3D surface of the man in the centre back of the picture.

Figure 2 shows a photograph of ‘Sansón y Dalila’ by José Camarón y Miliá (left) and a close up of the 3D mesh of face of the man in the centre (right).

Colour information can also be mapped on this surface, as shown in Figure 3. In this example RGB colour information from a colour camera is mapped onto the object surface. Shape imperfections from a flat surface as calculated using the large radius parameter, r l , of 2500 points and these regions are mapped in red on the surface. 497


4.2

Figure 3 shows RGB colour information and shape imperfections, identified by using the large radius parameter, r l , of 2500 points, mapped on the surface in red.

Pre-Treatment of Spectra Data

Figure 4 shows the original recorded spectra curves, Figure 5 shows the spectra curve after pretreatment with the MSC method and Figure 6 shows the spectral curves after pre-treatment with the EMSC method. For both tested methods, the EMSC was used to normalize data of the database, as it accounts for wavelength-dependent light scattering variation. After correction, as shown in Figures 5 and 6, the variance of the data is significantly reduced.

Figure 4 Originally recorded spectra curves after equalisation of the image

Figure 5 Spectra curves from Figure 4 after pre-treatment with the MSC method 498


4.3

Figure 6 Spectra curves from Figure 4 after pre-treatment with the EMSC method

Spectra Analysis and Classification Results

Table 2 shows a comparison of the success rates of the different classification methods. In order to quantify the reliability of the implemented classifiers, the procedure was tested with the data of the fourth sampled curve of each pigment, after EMSC correction. From an analysis of the results in Table 2, classification method #8 has the best performance, however methods #1 and #5 have good performance and are faster. When considering more than 30 classes, classification rates go below 95% and if more than 60 classes are considered, rates are under 90%. Table 2 Comparison of spectral classification methods by percent success rate. Results within 95-100% of the maximum classification are marked in grey. k-means classes 10 20 30 40 50 60 70

PCA Dimensions 10 30 10 30 10 30 10 30 10 30 10 30 10 30

#1

#2

#3

#4

#5

#6

#7

#8

#9

97,5 97,5 95,0 95,0 90,8 92,5 90,0 90,8 85,0 86,7 75,8 78,3 81,7 80,8

97,5 92,5 93,3 85,8 87,5 79,2 87,5 77,5 83,3 71,7 77,5 72,5 78,3 70,8

96,7 96,7 89,2 89,2 82,5 82,5 82,5 81,7 72,5 72,5 69,2 69,2 64,2 63,3

85,8 60,8 80,0 60,8 65,8 51,7 75,0 51,7 64,2 50,0 67,5 51,7 62,5 48,3

98,3 98,3 94,2 94,2 89,2 89,2 90,0 90,0 85,0 85,8 80,8 80,8 69,2 72,5

88,3 86,7 87,5 82,5 76,7 77,5 84,2 80,8 76,7 73,3 75,8 73,3 77,5 72,5

95,8 96,7 95,8 97,5 85,0 87,5 86,7 88,3 69,2 70,8 65,8 66,7 60,0 60,8

96,7 96,7 97,5 95,0 95,0 91,7 92,5 93,3 88,3 90,0 86,7 91,7 84,2 87,5

95,8 96,7 95,8 97,5 85,0 87,5 87,5 88,3 70,0 70,0 65,8 66,7 59,2 60,0

5 Summary This paper shows how structured light projection and hyperspectral imaging may be used to assess deterioration in artwork. Structured light projection in combination with the moving least squares and the eigenvector analysis based method are able to detect deformations in the painting surfaces. Spectral classification methods based on EMSC and SVM with polynomial kerned functions are able to classify pigments in a reliable way.

499


6 Acknowledgements This work was performed as part of the SYDDARTA project (www.syddarta.eu), funded by the European Commission under the 7th Framework Programme (project number 265151). The authors would like to thank Bart Slinger and Jelmer Rietsma, TU Delft for their support.

7 References

[1] EGMUS - European Group on Museum Statistics, http://www.egmus.eu/. Accessed 28 December 2013. [2] Sitnik R, Krzesłowski J and Mączkowski G (2011) Archiving shape and appearance of cultural heritage objects using structured light projection and multispectral imaging, Opt. Eng. 51:021115. [3] Tornari V, Bernikola E, Hatziyannakis K, Osten W, Groves RM, Georges M, Cedric T, Hustinx GM, Rochet J, Kouloumpi E, Doulgeridis M, Green T and Hackney S (2009) Multifunctional encoding system for the assessment of movable cultural heritage and resulting prototype device, in Osten, W and Kujawinska M (Eds.) Proc. of FRINGE 2009, Stuttgart, September 2009, pp. 680-687. [4] Uzielli L, Cocchi L, Mazzanti P, Togni M, Jullien D, Dionisi-Vici P (2012) The Deformometric Kit: A method and an apparatus for monitoring the deformation of wooden panels, J. Cultural Heritage 13: S94S101 [5] Ropret P, Miliani C, Centeno SA, Tavzes C, Manuali V (2010) Advances in Raman imaging of works of art, J. Raman Spectrosc. DOI 10.1002/jrs.2733. [6] Lewis EN, Schoppelrei J and Lee E (2004) Near-Infrared Chemical Imaging and the PAT Initiative, Spectroscopy 19:26-36. [7] Toque JA, Sakatoku Y and Ide-Ektessabi A (2011) Pigment identification by analytical imaging using multispectral images, in Karam L and Pappas T (Eds.) Proc. 16th IEEE International Conference on Image Processing (ICIP), Cairo, 7-12 November, pp. 2861-2864. [8] Delaney JK, Walmsley E, Berrie BH and Fletcher CF (2005). Multispectral Imaging of Paintings in the Infrared to Detect and Map Blue Pigments, in Hill Stoner J (Ed.) Scientific Examination of Art: Modern Techniques in Conservation and Analysis (Sackler NAS Colloquium), National Academy of Sciences. [9] Granero-Montagud L, et al. (2013) Deterioration estimation of paintings by means of combined 3D and hyperspectral data analysis, in Pezzati L and Targowski P (Eds.) Proc. SPIE 8790-38, Munich, pp. 879008. [10] Takeda M, Ina H and Kobayashi S (1982) Fourier-transform method of fringe-pattern analysis for computer-based topography and interferometry, J. Opt. Soc. Am. 72:156-160. [11] Fujigaki M and Morimoto Y (1997) Accurate shape measurement for cylindrical object by phase-shifting method using Fourier transform, in Chau FS and Lim CT (Eds.) Proc. SPIE 2921, International Conference on Experimental Mechanics: Advances and Applications, Singapore, pp. 557-562. [12] Ghiglia DC and Romero LA (1994) Robust two-dimensional weighted and unweighted phase unwrapping that uses fast transforms and iterative methods. J. Optical Society of America A 11(1):107-117. [13] Takeda M and Mutoh K (1983) Fourier transform profilometry for the automatic measurement of 3-D object shapes. Applied Optics 22(24): 3977-3982. [14] Zhang Z (2000). A flexible new technique for camera calibration. IEEE Transactions on Pattern Analysis and Machine Intelligence 22(11): 1330-1334. [15] Alexa M, Behr J, Cohen-Or D, Fleishman S, Levin D and Silva CT (2001) Point set surfaces, in Bailey M and Hansen C (Eds.) Proc. of the IEEE Conference on Visualization, San Diego 21-28. [16] Rusu RB (2009) Semantic 3D Object Maps for Everyday Manipulation in Human Living Environments, PhD Thesis, Technical University of Munich. [17] Martens H, Jensen SA, Geladi P (1983) Multivariate linearity transformations for near infrared reflectance spectroscopy, in Christie OHJ (Ed.) Proc. Nordic Symp. Applied Statistics, Stokkland Forlag, Stavanger, Norway, pp. 205–234. [18] Martens H and Stark E (1991) Extended multiplicative signal correction and spectral interference subtraction—new pre-processing methods for near-infrared spectroscopy. J. Pharm. Biomed. Anal. 9:625. [19] Castanys M, Soneira MJ and Perez-Pueyo R (2006) Automatic Identification of Artistic Pigments by Raman Spectroscopy Using Fuzzy Logic and Principal Component Analysis, Laser Chemistry 2006:1-8. [20] González Vida JJ (2011) Identificación automática de Espectros Raman de pigmentos mediante Análisis por Componentes Principales. Automatic Identification of Pigments by Principal Component Analysis of the Raman Spectra, MSc Thesis, Universitat Politècnica de Catalunya. [21] Cortes C, Vapnik V (1995) Support-vector networks, Machine Learning 20(3): 273.

500


Improving the durability of prestressed concrete structures by employing high strength duplex stainless steels H. Mahmoud1,2, M. Sánchez1, M. C. Alonso1*, L. Bertolini2 (1) Eduardo Torroja institute of construction science. C/ Serrano Galvache 4, 28033 Madrid. Spain. (2) Politecnico di Milano, Department of Chemistry, Materials and Chemical Engineering, via Mancinelli 7, 20131, Milano. Italy. Abstract: The corrosion behaviour of 2304 duplex high strength stainless steel has been analysed in the present work by determining the chloride threshold (CT) for corrosion initiation in mortar by means of accelerated tests. Effect of cold drawing has been studied by comparing parent low strength wires with cold-drawn high strength wires. Potentiostatic polarization tests in chloride-free and chloride mixed-in mortars have been carried out. The CT determined by using chloride penetration tests in chloride-free mortar varied from 3.06±0.5% total Cl¯ (with respect to cement weight) for cold drawn SS to 2.3% total Cl¯ for parent SS. On the other hand, the CT values determined by means of potentiostatic tests in chloride mixed-in mortar was 1.8±0.2 % Cl¯ for parent and 2.1±0.5% Cl¯ for cold drawn SS. Keywords: Prestressed concrete, Chloride threshold, Duplex stainless steels, High strength stainless steels.

1 Introduction In the last half century, the pre-stressing technology has been widely used for civil engineering applications, such as: water pipes, bridges, building etc. [1]. As well as conventional reinforced concrete structures, pre-stressed concrete structures (PCS) exposed to aggressive environments suffer of chloride-induced corrosion of steel. Indeed corrosion of prestressing tendons is one of the main problems that may threaten the PCS service life [2-4]. Nowadays, the application of high strength stainless steels (HSSS) in PCS stands out as a promising alternative to improve their durability [5-8]. HSSS wires are produced by a cold drawing deformation process of a parent stainless steel (P-SS) wire [7-9] which increases the tensile strength. This process reduces the cross section of the P-SS by stretching the steel wires. Cold drawn stainless steels (CD-SS) are usually obtained by applying a 5080% cold drawing degree, depending on the requirements for the specific use [9]. In the literature, available knowledge about the corrosion behaviours of HSSS wires in contact with concrete is scarce. Besides, most of the studies have been performed in alkaline solutions that simulate concrete pore solutions [5,10-12] and only few attempts have been performed in mortar [6-7]. Early researches to evaluate the corrosion behaviour of HSSS by determining the chloride threshold (CT) were focused on austenitic HSSS [5-7, 10-12]. Studies conducted by Recio [12] using potentiodynamic accelerated techniques in simulated alkaline concrete pore solutions (Sat. Ca(OH) 2 + 0.5 M KOH, pH 13.5) showed that the CT of austenitic 316 CD-SS is above 1.5 M. In addition, CT value of this stainless steel decreases with decreasing the pH of the alkaline solution. Alonso et al. [10] detected that pits were nucleated on the surface of austenitic 304 HSSS when the chloride content increased up to 0.75 M in saturated Ca(OH) 2 + 0.2 M KOH solutions (pH 13.2), and the application of external loads induced the nucleation of pits in the presence of relatively lower chloride concentrations (0.5 M). On the other side, Moser et al. [5] also used the potentiodynamic tests for the study of *

M.C. Alonso: [email protected]

501


chloride-induced corrosion of HSSS in alkaline solutions. The estimated values of CT for austenitic 316 and 304 SS in alkaline solutions of pH 12.5 were 0.25 M and 0.5 M respectively. Concerning carbonated solutions, austenitic 316 HSSS showed higher corrosion resistance in comparison with austenitic 304 HSSS [5,12]. Concerning the CT determination of in concrete, Nürnberger et al. [6-7] have concluded that the CT varies from 3 to 5% Cl¯ (with respect to the cement weight) for austenitic 304, 316 and 317 HSSS but adding chloride in the mix. As a consequence of the high cost of austenitic SS mainly induced by the high Ni contents, new search trends for the application of duplex CD-SS have been a subject of interest [5-8]. In this context, 2304 duplex stainless steel, which has already shown a good corrosion resistance in concrete, is a promising alternative [13]. However, the lack of knowledge about the corrosion behaviour of duplex HSSS when embedded in concrete is one of the main concerns that limit the employ of these materials in PCS. Consequently, the main objectives of the present work are to assess the corrosion behaviour of a duplex 2304 HSSS in mortar and the CT determination using accelerated potentiostatic polarization tests (in chloride-free and chloride mixed-in mortars). The analysis of influencing parameters such as the effect of cold drawing process on the CT of these duplex SS has been also considered.

2 Experimental 2.1

Materials

Duplex 2304 stainless steel (1.4362 according to EN 10088 standard) both in form of parent (P-SS) and cold drawn (CD-SS) wires of 9 and 4 mm in diameter respectively, was tested. The chemical composition and the mechanical properties of 2304 duplex SS are listed in Tables 1 and 2 respectively.

Table 1 Chemical composition (wt%) of 2304 duplex stainless steel. SS (ASTM)

%C

%Cr

%Ni

%Mo

%N

%S

%Mn

%Si

2304

0.03

22.9

4.29

0.1

0.1

0.01

1.79

0.6

Table 2 Mechanical properties of duplex 2304 stainless steel. Mechanical properties

Parent

Cold drawn

Maximum stress/ MPa

865

1629

Yield strength / MPa

569

1586

Young Modulus (E)/ GPa

171

178

In order to simulate the field condition, the steel wires were not submitted to any further mechanical nor chemical treatment and they were tested as-received from the production plant. Before testing, the samples were ultrasonically cleaned with water and acetone in order to remove any residue on the surface. The exposed area was delimited with an isolating tape. The CT of P-SS and CD-SS was determined by potentiostatic electrochemical tests in both chloride free and chloride mixed-in mortars. Two different experimental arrangements were used for the potentiostatic tests.

502


2.2

Tests with chloride penetration

The potentiostatic CT determination of P-SS and CD-SS was performed in this case using cylindrical chloride-free mortar specimens of 8 cm length and 2.3 cm in diameter, as shown in Figure 1-A. In this case, two series were tested. In the first series (S1), six specimens of both SS bars (two PSS and four CD-SS) were tested. For the second series (S2), only six specimens of CD-SS were analyzed. Ordinary Portland Cement (OPC) with low alkali and aluminate contents was used to reduce the bound chlorides. Mortar samples were prepared using cement/sand ratio 1/3 and w/c = 0.5. After casting, the mortar specimens were cured in a chamber at 95% RH and 20±2ºC for 15 days. After curing, the specimens were immersed in 1 M NaCl solution. Figure 1-B depicts the schematic shape of the three-electrode electrochemical cell employed. The working electrodes were the stainless steels wires, CD-SS and P-SS, connected in parallel to the potentiostat. The counter-electrode used was a stainless steel mesh located in the solution and surrounding the mortar samples. A Saturated Calomel Electrode (SCE) was employed as reference electrode. A constant potential of +0.25 V SCE was applied continuously until the onset of corrosion was detected in each SS bar. In this conditions the passive layer is formed on the surface of the SS at the specific polarization potential used previously the arrival of the chloride. The chloride ion is induced to penetrate the mortar barrier and consequently, the chloride concentration is slowly increased at the depth of the embedded stainless steels until reaching the chloride concentration (CT) at the bar level able to initiate an active pitting corrosion. The total cell current was periodically measured during the potentiostatic test. Moreover, the current passing through each immersed mortar sample was also measured periodically. The initiation of corrosion was detected by a sudden increase in the current of the system.

A)

B)

Figure 1 Scheme of (A) the single mortar cylindrical sample of 8 cm length and 2.3 cm in diameter and (B) the electrochemical cell used for the potentiostatic determination of the critical chloride threshold in chloride free mortar.

2.3

Tests in chloride mixed-in mortar

Regarding chloride mixed-in mortar, multi-specimen samples were tested. In this case, mortar was made with the same OPC of previous test and standard sand. Different concentrations of chloride (0, 2, 3 and 4 % Cl- with respect to cement weight) were added to the mixing water during casting as CaCl 2 . Cylindrical samples with 6 rebars (3 cold drawn + 3 parent SS) were prepared for each chloride content. The counter-electrode was an embedded MMO activated Ti-mesh as shown in Figure 2-A. As-received 10 cm long wires of P-SS and CD-SS were embedded in the 7 cm diameter multispecimen mortar samples. After casting, the chloride mixed-in mortar samples were cured at the same condition than chloride free mortar, and then immersed in saturated Ca(OH) 2 solutions, as shown in Figure 2-B. The potentiostatic test was carried out by progressive application of different constant anodic potentials

503


(+0.25, +0.40 and +0.55 V SCE ). An external SCE reference electrode was immersed in the saturated Ca(OH) 2 to control the applied potential (see Figure 2-B)

A)

B)

Figure 2 Multispecimen with mixed-in chloride samples (A) after casting and (B) during the potentiostatic electrochemical test.

2.1

Chloride analysis

At the end of each potentiostatic test, the chloride concentrations were analyzed in powdered mortar samples collected at the level of the rebar surface, 2 mm deep from the wire bed. The analysis of acid-soluble chloride (total chloride value) was performed by potentiometric titration [14-15], which involves (1) extraction of the chloride ions by digestion of a weighed portion of powdered mortar (2-5 g) in a boiling nitric acid solution (1:1) and separation of the resulting digested acid from the solid residue by filtration, followed by (2) titration of the extracted chloride ions dissolved in the filtrate with a standardized silver nitrate solution.


Chloride penetration potentiostatic tests

The depassivation process of 2304 duplex SS was first investigated by the evaluation of CT of PSS and CD-SS in chloride-free mortar using potentiostatic polarization tests (at +0.25 V SCE ). Figure 3 shows an example of the change in the current density with the immersion time in 1 M NaCl solution. In Figure 3-A, mortar specimens with P-SS and CD-SS of the first series (S1) are shown and in Figure 3-B the variation in the current density for the second series (S2) is displayed. Before chloride reaches the rebar surface, ageing in the alkaline mortar induces the passive film formation on the embedded stainless steel bars at the anodic potential of polarization (+0.25 V SCE ). During this period the chloride ions penetrate through the mortar barrier until reaching a critical concentration (CT) at SS surface, able to breakdown the passive film. In the S1 case (Figure 3-A), a sharp increase in the current was observed after certain period of time, indicating the initiation and propagation of active pits on the duplex SS surfaces (Figure 3-A). In the S2 case (Figure 3-B), certain fluctuations in the measured current densities were observed probably due to the formation of metastable pits able to be repassivated. However, at longer testing times, some of these metastable pits could propagate and induce an increase in the measured current density, as shown in Figure 3-B, in similar way as observed in the S1 case. However the onset of corrosion was considered to occur when the current maintain high values indicating that pits are active and able to propagate, further studies will be of interest to identify if metastable pits are already formed when fluctuation of current densities are observed.

504


10000

Cold drawn 2304 Parent - 2304

1000

10

(A) S1

i / µA.cm-2

i / µA cm-2

100

10 1

1

R1

R2

R3

R4

R5

R6

0,1

0,1

0,01

0,001

(B) S2

CD 2304

0

100

200 Time /days

300

0,01

0

100 200 300 400 500 600 700 Time / days

Figure 3. Current density evolution during potentiostatic test of 2304 duplex SS subjected to chloride penetration test. (A) First series (S1), and (B) second series (S2).

The chloride content analysis to determine the CT at the rebar level was performed immediately after the detection of the pitting corrosion onset. The CT values obtained for the tested stainless steels in mortar specimens, expressed as total chloride content in cement weigh, are listed in Table 3. The results show that the corrosion initiation was observed on the P-SS embedded in mortar when the chloride content at the steel surface was 2.3±0.2% by the weight of cement, while CD-SS showed higher corrosion resistance, i.e. 3.06±0.6% total chloride considering the two series. Table 3 CT values in % Clˉ total respect the cement weight of 2304 DSS, (A) first series S1, and (B) second series S2. (A)- S1 CT - % Clˉ

Parent R2 2.41

R3 2.43

R1 3.64

R2 3.71

Cold drawn R3 R4 3.49 3.52

(B)-S2 CT - % Clˉ

Cold drawn R4 R5 2.45 3.08

R1 2.14

R5 -

R6 2.78

R6 3.52

At the end of the test, the presence of pits on the SS surface was also confirmed by optical examination. Deep and widespread attacks were observed on all the specimens that have experienced the sharp increase in the current indicating the corrosion initiation.

3.2

Chloride mixed-in potentiostatic tests

Since penetration of chloride through the concrete cover requires time (e.g. when realistic values of concrete cover of the order of 30 mm or higher are considered), often the corrosion resistance of stainless steel is also studied with mixed-in chlorides; the testing technique, however, may affect the estimated value of CT [16]. In order to compare results obtained with previous tests of penetrated

505


chloride, the pitting corrosion risk of P-SS and CD-SS was also studied on wires embedded in chloride mixed-in mortars with different chloride concentrations up to 4% Cl¯ by applying a stepwise potentiostatic test. Figure 4 shows the variation of the residual current density of 2304 duplex P-SS and CD-SS embedded in mortar samples mixed-in with 2, 3 and 4% Cl¯. In each case, three stainless steel wires were tested and the current densities shown are the mean value of the three measurements. The chloride mixed-in mortars were immersed in saturated Ca(OH) 2 solutions during the potentiostatic test. Initially a potential value of +0.25 V SCE was applied, similarly to tests with penetrated chloride. According to the potentiostatic results depicted in Figure 4, the residual current density decreases by ageing until reaching a steady state current density. However, no sudden increase in the measured residual current, indicative of a clear corrosion onset, was observed after 7 days of polarization. Subsequently, the potential was increased to +0.4 V SCE and then to +0.55 V SCE (Figure 4). Only at the most positive polarization potential, near the oxygen evolution region (+0.55 V SCE ) [8], high current density values were measured.

Current density / µA.cm-2

10

P - 2% (A)

P - 3% P - 4% 1

0,1

0,01

+ 0.25 VSCE 0

10

2

+ 0.4 VSCE 4

6

8

Time / days

CD - 2%

10

+0.55 VSCE 12

14

16

(B)

Current density / µA.cm-2

CD - 3% CD - 4% 1

0,1

0,01

+ 0.25 VSCE 0

2

4

+ 0.4 VSCE 6

8

Time / days

10

+0.55 VSCE 12

14

16

Figure 4 Step potentiostatic tests of 2304 (A) parent and (B) cold drawn SS embedded in embedded in chloride mixed-in mortar with 2, 3 and 4% Cl¯.

Due to the low concrete cover of the wires (about 10 mm) the chloride content at the depth of the bars changed during tests due to leaching of chloride ions from the mortar to the external Ca(OH) 2 solution. The chloride analysis at the end of the test in chloride mixed-in mortar samples (initial 2, 3 and 4 % Cl⁻) at the rebar level were also performed to measure the chloride content in each sample at

506


the end of tests. The data obtained are listed in Table 4. The results of the chloride analysis of mortar samples show that the chloride contents measured at the end of tests were lower than the initially mixed-in. The results measured at the end of the tests, can be used as reference values for the estimation of the chloride threshold with this type of test. Table 4 Chloride analysis of chloride mixed-in mortar samples expressed by % total Cl⁻ by cement weight. Cl⁻ mix-in samples 2% 3% 4%

% Cl⁻ total

1.8 ± 0.2 2.1 ± 0.5 2.9 ± 0.1

Parent

Cold drawn

Pits Pits Pits

No pits Tiny pits Tiny pits

As no clear response of the current density was detected in Figure 4, the possible onset of pitting corrosion on the tested SS was evaluated by the optical examination of the surface at the end of the test. The presence of small pits could be observed at the end of the tests. So, in this case, corrosion onset was assumed to take place where such localized attacks could be observed (Table 4). In the P-SS case, the presence of pits at the surface indicated that the depassivation had occurred even with 1.8 ± 0.2 % Clˉ (by cement weight) at the steel-mortar interface. However, CD-SS showed higher corrosion resistance and the pitting corrosion initiation was only observed at 2.1 ± 0.5 % Clˉ. In this case, the determination of the corresponding potential to the pitting corrosion onset was difficult, as the optical examinations were performed after the stepwise potentiostatic test. However it has to point out that the addition of chloride from the initial stage can modify the passive layer formation at the SS surface that also affects the CT. Smaller CT values are obtained with chloride mixed-in tests than with the chloride penetration potentiostatic tests. However, it should be highlighted that the values obtained by both methods are not directly comparable as the two methodologies are different and the steel wire surface will be altered in different form under these different experimental conditions. The passive layer grown in mortar with chloride will differ of the passive layer formed in non-chloride mortar. In absence of chloride the passive film growth is favored at the earlier ages but when chloride ions are present at the interface from the beginning, certain changes are induced in the passive layer affecting its electrochemical properties and lowering its protective capacity [16,17]. The more extreme experimental conditions considered in chloride mixed-in tests, not only the chloride presence but also the higher testing potentials, must have contributed to the smaller values of CT determined. However, the testing time is considerably reduced what is an important advantage of this methodology. The above results clearly show that the estimation of CT depends on the testing procedure and it is remarkably affected by the: applied potential, the criterion used for corrosion detection, the measurements of the real chloride content at the steel surface, or the influence of the salt added to or penetrated in the mortar mix, NaCl vs CaCl 2 .

4 Conclusions The corrosion behavior of cold drawn wires of 2304 stainless steel in mortar was tested with different accelerated methodologies. The results showed that the estimated value of the chloride threshold CT was affected by the testing procedure. The estimated value of CT by the penetration potentiostatic tests in mortar vary from 3.06 ± 0.5 % Cl⁻ (by cement weight) for CD-SS to 2.3 ± 0.2 % Cl⁻ for P- SS. Lower critical chloride concentration for 2304 duplex SS (from 1.8 ± 0.2 % total Cl- for P-SS to 2.1 ± 0.2 % total Cl⁻ for CD-SS) have been estimated by potentiostatic tests with chloride mixed-in mortar. In spite of differences in the estimated CT values, the two experimental procedures, however, confirmed that 2304 duplex P-SS and CD-SS show high corrosion resistance against the chloride induced pitting corrosion and high CT and that the stainless steel microstructure influences the CT. In

507


fact, the CD-SS in both cases showed higher corrosion resistance than P-SS and both steels showed CT values higher than 2% by weight of cement, which are higher than the values expected at the depth of steel in prestressed and post-tensioned structures.

Acknowledgement Authors acknowledge to MINECO Spanish Ministry the financial support (project BIA201122760) and to INOXFIL for supplying the SS wires.

5 References

[1] Lin TY, Burns NED H Burns (1981) Design of prestressed concrete structures 3rd edn. John Wilky & Sons Inc., pp 1-38. [2] Nürnberger U (2002) Corrosion induced failure mechanisms of prestressing steel, Materials and Corrosion 53:591-601. [3] Nürnberger U (1997) Influence of material and processing on stress cracking of prestressing steel, Materials and Corrosion 48:602-612. [4] Alonso MC, Villegas MA, de las Heras E (2007) Developments in steel tendons. Design and materials, Polder R. B., Alonso M. C., Cleland D. J., Elsener B., Proverbio E., Vennesland Ø., Raharinnaivo A. (Eds.) New materials and systems for prestressed concrete structures. COST 534 Final report, European commission, pp. 6-34. [5] Moser RD, Singh PM, Kahn LF, Kurtis KE (2012) Chloride-induced corrosion resistance of high-strength stainless steels in simulated alkaline and carbonated concrete pore solutions, Corrosion Science 57:241253. [6] Nürnberger U, Wu Y (2008) Stainless steel in concrete structure and in fastening technique, Materials and Corrosion 59:144-158. [7] Nürnberger U, Wu Y (2009) Corrosion-technical properties of high-strength stainless steel for the application in prestressed concrete structures, Materials and Corrosion 60:771-780. [8] Sánchez M, Mahmoud H, Alonso MC (2012) Electrochemical response of natural and induced passivation of high strength duplex stainless steels in alkaline media, J Solid State Electrochemistry 16:1193-1202. [9] Mietz J, Rückert J (1997) Investigation of high strength stainless steels for rock and ground anchors, Materials and Corrosion 48:353-363. [10] Alonso M C, Sánchez M, Mazarío E, Recio F J, Mahmoud H, Hingorani R (2010) High strength stainless steels 14301 for prestressed concrete structure protection, 6th international conference on Concrete under severe conditions (CONSEC 10) Mexico, pp 1047-1054. [11] Alonso M C, Recio F J (2007) Corrosion performance of galvanized and high strength stainless steel tendons, stress corrosion test applicable to new prestressed steels. COST 534 Final report, European commission, pp. 24-39. [12] Recio F J (2010) Corrosión de aceros inoxidable y galvanizados de alta resistencia, como alternativa a los aceros convencionales de pretensados, PhD Thesis, Universidad Autónoma de Madrid, Spain. [13] Pastore T, Pedeferri P, Bertolini L, Bolzoni F (1991) Electrochemical Study on the Use of Duplex Stainless Steel in Concrete, Charles J, Bernhardsson S (Eds.) Duplex Stainless Steels. Les Editions Physique, pp 905-913. [14] Clemeña GG(2002) An alternative potentiometric method for determining chloride concrete content in concrete samples from reinforced concrete, PhD Thesis, Virginia Transportation Research Council. The University of Virginia. [15] ASTM (2008) Test Method for Water-Soluble chloride in mortar and concrete, C 1218/C 1218M – 99. [16] Bertolini L, Gastaldi M (2011) Corrosion resistance of low-nickel duplex stainless steel rebars, Materials and Corrosion 62:120-129. [17] Mahmoud H, (2013), High Strength Stainles steel protection capacity for prestressed concrete structures. PhD Thesis, Universidad Autónoma de Madrid, Spain.

508


Viscoelastic Stress Modeling in Cementitious Materials Using Constant Viscoelastic Hydration Modulus Will Hansen1*, Zhichao Liu1, Eduard A.B. Koenders2 (1) University of Michigan, Ann Arbor, USA (2) Delft University of Technology, The Netherlands Abstract: Viscoelastic stress modeling in ageing cementitious materials is of major importance in high performance concrete of low water cement ratio (e.g. w/c ~0.35) where crack resistance due to deformation restraint needs to be determined. Total stress analysis is complicated by the occurrence of internal stresses due to shrinkage, which requires estimating the stress relaxation effect from tensile creep. This study presents a new and direct methodology for viscoelastic stress analysis based on measurement of the viscoelastic hydration modulus. Autogenous shrinkage, if restrained, creates an internal tensile stress condition which is uniform within a cross section. Autogenous shrinkage stresses develop within the porous hydration products. They are compressive stresses and if restrained by reinforcement a net tensile stress develops. Results show that the viscoelastic hydration modulus is approximately 8000-9000 MPa and is a constant material property. Total stress analysis can now be separated into two components, an elastic stress based on the Young’s modulus (typically in the range of 28000-34000 MPa) and a viscoelastic (time-dependent) stress based on measurement of timedependent strains (creep and shrinkage). The importance of reducing paste content for shrinkage stress control is demonstrated using the Pickett’s model. Keywords: Autogenous shrinkage, High Performance Concrete (HPC), Shrinkage stresses, Modeling viscoelastic effects

1 Introduction Self-induced tensile stresses develop in concrete if the movements caused by cement hydration reactions are restrained [1]. During early age hydration two active mechanisms are involved in producing these movements, starting with thermal effects which dominate during the first 24-48 hours. Self-desiccation is another consequence of cement hydration as this process consumes water into solid hydration products [2]. As hydration proceeds internal pore drying develops with associated internal stress development from capillary tension in the pores [3]. These stresses transfer to the hydration products as compression and subsequent shrinkage. This type of shrinkage is known as autogenous shrinkage. It is characterized by a uniform volume reduction and at any time is a material property (that is, no moisture gradient), whereas drying shrinkage development is size-dependent and nonuniform. Thermal stresses are relatively short-term acting throughout the concrete composite, the timedependent shrinkage stresses are internal acting primarily on the porous hydration products. Autogenous shrinkage is intensified in high performance concrete of low water-cement (w/c) ratio (relative to conventional concrete) due to its generally higher cement content, reduced w/cm, and pozzolanic mineral admixtures. Prior results indicate that cementitious systems containing slag cement produces greater autogenous shrinkage at later ages [4-6]. The early age cracking problem in high performance concrete has become important due to the increased use of these materials [7-10]. The reasons were generally attributed to the higher chemical shrinkage, the finer pore structure, removal of calcium hydroxide as a shrinkage restraint, and a reduction in pore humidity associated with pozzolanic reactions. The fundamental basis for applying theory of viscoelasticty, developed for polymers, to concrete is the assumed analogy between creep compliance function and stress relaxation modulus [11-13]. The outcome in viscoelastic stress calculations is a reduction in concrete modulus by 50%-75% due to stress relaxation under full external deformation restraint [14-16]. *

Department of Civil and Environmental Engineering, University of Michigan ([email protected])

509


Recent numerical simulations, using the lattice model for predicting stress development from a mini temperature stress testing machine (TSTM) on small paste specimens, where thermal effects from heat of hydration are minimal, concluded that a poor agreement was obtained between predicted stresses by using three different models for obtaining stress relaxation modulus [17]. The first model was based on an assumption of instant relaxation that internal stresses will cause instant deformation of the micro-porous hydration products. In this model there is no relaxation over time. The second model assumes relaxation based on hydration and it uses an exponential relaxation factor approach [18]. In the last method, stress relaxation modulus is a reduced Young’s modulus, since the relaxation is the amount of stiffness lost over time. It was concluded that a better relaxation model is needed. An intriguing and novel test procedure was used by Bjontegaard [19] who measured the increasing autogenous deformation induced tensile stress development in a fully restrained test (TSTM) combined with parallel measurements of free (i.e. without external restraint) autogenous shrinkage. Although the focus of his study was the early age period (0-7 days) typical results show that once thermal contraction has ceased (typically within 24-48 hours), the two curves of stress and autogenous shrinkage development are parallel with a constant net viscoelastic modulus of about 11000 MPa.

2 Experimental Program Different mortar and concrete mixes of a 0.35 w/c ratio were prepared in the laboratory according to ASTM C192 and the mix design is listed in Table 1. Raw materials included Type I Portland cement, silica sand and limestone gravel. Cylindrical samples were cast and cured for one day before demoulding. Then they were sealed cured for different ages before the following test procedures were carried out. (1) Compressive strength and split tensile strength were tested on 100 mm × 200 mm cylinders according to ASTM C39 and C496, respectively. Three specimens were used for each age and both the average and individual results were reported. (2) Static modulus of elasticity was obtained from the stress-strain curve of 300 mm × 600 mm cylinders from the simultaneous measurement of uniaxial compressive load by a static hydraulic system, and linear deformation by a motion capture system (Figure 1). (3) Autogenous shrinkage was measured on duplicate mortar or paste specimens of 60 mm × 100 mm × 1000 mm where double polystyrene films were used to seal the specimen and an isothermal condition at 20±1 °C was achieved by circulating water through channels embedded into the sides and bottom of the rigs. In addition to the free shrinkage, the restrained shrinkage was measured by the embedment of four symmetrically located rebars in the specimen. (4) Sealed creep of the concrete mix in compression was measured according to ASTM C512 where the specimens were sealed to achieve a uniform moisture condition during the test. Table 1 Mix design (kg/m3)

Paste 20%agg. 40%agg. 60%agg. 70%agg.

cement 1497 1198 899 594 450

gravel 0 0 0 936 1093

510

sand 0 528 1055 646 753

water 524 419 314 208 157


Stress, MPa

60 50 40 30 20 10 0 0

(a)

500

1000 1500 2000 Strain, µm/m (b)

2500

Figure 1 (a) Static modulus measurement by a combination of a hydraulic test system and a motion capture system, (b) measured 7day stress-strain curve of concrete

3 Results and discussions 3.1

Basis for proposed viscoelastic stress-strain analysis of hydration products

Measurements and modeling of autogenous shrinkage of low w/c systems (0.35) for different internal restraint conditions (aggregate particle and steel reinforcement) forms the basis for the proposed activity. Autogenous shrinkage is a form of drying shrinkage, but without shrinkage gradients (Figure 2). A uniform internal stress develops due to the hydration process that proceeds without exchange of water (i.e. sealed curing). Internal stresses increase with increasing hydration resulting in more specimen shrinkage.

F

σ

Uniform moisture gradient due to self desiccation

Free autogenous shrinkage

Stress status if autogenous shrinkage is restrained

σ

Differential moisture gradient due to external drying

Free drying shrinkage

F

Stress status if drying shrinkage is restrained

Figure 2 Schematic of stress distribution from autogenous shrinkage and drying shrinkage [20]

The uniform stress condition allows for a straight forward tensile stress analysis when symmetrically placed reinforcement bars are used (Figure 3). Autogenous shrinkage results are shown in Figure 4 versus time. Force equilibrium analysis yields a linear correlation between free shrinkage and rebar restrained shrinkage (Eqs.1-2), from which a constant viscoelastic hydration modulus is

511


obtained based on minimizing error between predicted line (dashed) and measurements (Eq.3 and Figure 5). 𝜀𝑠 (𝑡)𝐸𝑠 𝐴𝑠 = [𝜀𝑐 (𝑡) − 𝜀𝑠 (𝑡)]𝐸𝑣 𝐴𝑐 Thus, A E 𝜀𝑐 (𝑡) = 1 + A s Es 𝜀𝑠 (𝑡) ≅ (1 + 𝑛𝑣 𝜌𝑠 )𝜀𝑠 (𝑡)

(1) (2)

p v

or

𝜀 (𝑡)

𝐸𝑣 = 𝐸𝑠 𝜌𝑠 /(𝜀𝑐(𝑡) − 1)

(3)

𝑠

Where 𝜀𝑐 (𝑡)= free shrinkage of plain mix, 𝜀𝑠 (𝑡)= deformation of steel in reinforced mix, 𝐴𝑐 = area of plain mix, 𝐴𝑠 = area of steel, 𝜌𝑠 = 𝐴𝑠 /𝐴𝑐 = steel ratio, 𝐸𝑠 = steel modulus, 𝐸𝑣 = viscoelastic hydration modulus, 𝑛𝑣 = 𝐸𝑠 /𝐸𝑣 .

Figure 3 Illustration of free shrinkage and restrained shrinkage by concentrically placed rebars

With rebar

Without rebar

-1200

-1200

-1000

-1000

Strain, ×106

Strain, ×106

With rebar

-800 -600 -400 -200

Without rebar

-800 -600 -400 -200

0

0 0,1

1 10 Time, day

100

0,1

(a)

1 10 Time, day

100

(b)

Figure 4 Autogenous shrinkage in (a) paste and (b) mortar containing 40% aggregate by volume (0.35 w/c)

512


1000

Ev = 8,500 MPa Ee = 32,000MPa

Absolute free strain, 10-6

800

600

400

200

035-paste

035-20%agg.

035-40%agg.

035-60%agg.

0 0

3.2

200

400 600 800 Absolute restrained strain, 10-6

1000

Figure 5 Effect of steel reinforcement on viscoelastic hydration modulus (𝑬𝒗 )

Pickett model for autogenous shrinkage prediction 7day 365day Pickett's model (90day)

28day Pickett's model (7day) Pickett's model (365day)

90day Pickett's model (28day)

Autogenous shrinkage, 10-6

1800 1600 1400 1200 1000 800 600 400 200 0 0

0,1

0,2

0,3 0,4 0,5 Aggregate content

0,6

0,7

0,8

Figure 6 Prediction of shrinkage strain by Pickett’s model

The Pickett’s shrinkage model is perfect for modeling autogenous shrinkage as it is developed for a uniform paste stress within a cross section [21]. 𝜀𝑐 = 𝜀𝑝 (1 − 𝑉𝑎 )𝑛 (4) Where 𝑉𝑎 is relative aggregate volume fraction and n is the shrinkage exponent, a measure of aggregate particle restraining effect. Free shrinkage results for different paste volume fractions can be fitted using the Pickett model with an exponent n ~ 1.5 in this case (Figure 6). This model is a powerful tool for evaluating effect of paste content (1-𝑉𝑎 ) on concrete shrinkage. 513


3.3

Total stress in HPC due to full shrinkage restraint

80 70 60 50 40 30 20 10 0

Measured static modulus Modeled static modulus 10

40000

8 6 4 2 0,1

10 Age, day

0 1000

Modulus, MPa

Compressive strength CEB-FIP model Split tensile strength S-curve model

Tensile strength, MPa

Compressvie strength, MPa

Based on the measurement and modelling of the compressive strength, split tensile strength (Figure 7(a)) and static modulus (Figure 7(b)) of concrete mix, the predicted elastic and viscoelastic shrinkage stress development from a full deformation restraint and sealed cured specimen is shown in Figure 8. The shrinkage stresses calculated with a constant viscoelastic hydration modulus are significant and increasing over time, thus reducing the crack resistance of HPC. Typically a failure limit of 1% is used. This corresponds to an allowable stress/strength ratio of 0.56 [22].

30000 20000 10000 0 1

10

100 Age, day

(a)

1000

(b)

Figure 7 Measurement and prediction on (a) compressive and split tensile strength and (b) static modulus

HPC (0.35 w/c ratio)

7

Stress,MPa

6 5 4 3 2 1 0 1

10 Time, day Elastic stress Relaxation modulus based stress Predicted shrinkage stress

100 Tensile strength 1% failure limit

Figure 8 Self-desiccation stress development in high performance concrete (w/c =0.35) subjected to full shrinkage restraint based on different stress prediction methods

This methodology replaces the need for relaxation modulus calculations using either the so-called analogy between creep compliance and relaxation modulus, or effective modulus method or viscoelastic modeling.

514


4 Conclusions Results from this study demonstrate that shrinkage stresses can be predicted using the hydration modulus which is obtained from autogenous shrinkage measurements. This methodology replaces the need for complicated creep and relaxation modulus analysis. The importance of reducing paste content for shrinkage stress control is demonstrated using the Pickett’s model.

5 References [1] Bosnjak D (2000) Self-induced cracking problems in hardening concrete structures. PhD Dissertation, The Norwegian University of Science and Technology, Trondheim [2] Copeland LE, Bragg RH (1955) Self-desiccation in Portland cement pastes, Research Department Bulletin RX052, Portland Cement Association, Skokie, IL [3] Koenders EAB (1997) Simulation of volume changes in hardening cement-based materials. PhD Dissertation, Delft University of Technology, Delft, the Netherlands [4] Hanehara S, Hirao H, Uchikawa H (1999) Relationship between autogenous shrinkage and the microstructure and humidity changes at inner part of hardened cement pastes at early ages, in Tazawa E-I. (Ed.) Proc. of the International Workshop Autoshrink’98, Hiroshima, Japan, E&FN Spon, London, UK, pp. 89-100 [5] Lura P (2003) Autogenous deformation and internal curing of concrete. PhD Dissertation, Delft University of Technology, Delft, the Netherlands [6] Lee KM, Lee HK, Lee SH, Kim GY (2006) Autogenous shrinkage of concrete containing granulated blastfurnace slag, Cem Concr Res 36: 1279-1285 [7] Bentz DP, Jensen OM, Hansen KK, Oleson JF, Stang H, Haecker CJ (2001) Influence of cement particle size distribution on early age autogenous strains and stresses in cement-based materials, J Am Ceram Soc 84 (1): 129-135 [8] Kovler K, Sikuler J, Bentur A (1993) Restrained shrinkage tests of fiber reinforced concrete ring specimens: Effect of core thermal expansion, Mater Struct 26: 231-237 [9] Toma G, Pigeon M, Marchand J, Bissonnette B, Barcelo L (1999) Early age restrained shrinkage: Stress build up and relaxation, in Persson B, Bentz DP, Nilsson L-O (Eds.) International Research Seminar: SelfDesiccation and Its Importance in Concrete Technology, Lund, Sweden, pp. 61-71 [10] Bentur A (2003) Early age shrinkage induced stresses and cracking in cementitious systems. RILEM Technical Committee 191-EAS, RILEM Publications [11] Bažant ZP (1977) Viscoelasticity of porous solidifying material–concrete, J Eng Mech 103 (6): 1049-1067 [12] Bažant ZP (Ed.) (1988) Mathematical Modeling Of Creep And Shrinkage Of Concrete, Chapter 2, John Wiley & Sons Ltd, pp.100-215 [13] Bažant ZP, Huet C (1999) Thermodynamic functions for ageing viscoelasticity: Integral form without internal variables, Int J Solids Struct 36 (26): 3993-4016 [14] Altoubat SA, Lange DA (2003) The Pickett effect in early age concrete under restrained conditions, in Kovler K, Bentur A (Eds.) RILIM International Conference on Early Age Cracking in Cementitious Systems (EAC'01), Haifa, pp. 133-143 [15] Grasley ZC (2006) Measuring and modeling the time-dependent response of cementitious materials to internal stresses. PhD Dissertation, University of Illinois at Urbana-Champaign [16] Pane I, Hansen W (2008) Investigation on key properties controlling early-age stress development of blended cement concrete, Cem Concr Res 38: 1325–1335 [17]Schlangen E, Leegwater G, Koenders, EAB (2006) Modelling of autogenous shrinkage of concrete based on paste measurements, in Marchand J, Bissonnette B, Gagne R, Jolin M, Paradis F (Eds.) Advances in concrete through science and engineering, Quebec city, Canada RILEM publications, pp. 1-13 [18] Breugel K van (1980) Relaxation of young concrete, TU Delft, Stevin report [19] Bjøntegaard Ø (1999) Thermal dilation and autogenous deformation as driving forces to self-induced stresses in high performance concrete. PhD Dissertation, Norwegian University of Science and Technology, Trondheim, Norway [20] Wei Y (2008) Modelling of autogenous deformation in cementitious materials, restraining effect from aggregate, and moisture warping in slabs on grade. PhD Dissertation, University of Michigan, Ann Arbor, USA [21] Pickett G (1956) Effect of aggregate on shrinkage of concrete and hypothesis concerning shrinkage, J Am Concr Inst 27: 581-590 [22] Lokhorst SJ (1999) Deformational behaviour of concrete influenced by hydration related changes of the microstructure. Report, Delft University of Technology, Delft, the Netherlands

515


Experimental study on the Corrosion Fatigue Performance of RC Beams Strengthened by Pre-stressed B/C HFRP Sheets Pan Jianwu1*, Chun Qing2, Zhou Xuehua1 (1) Civil Engineering Department, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China (2)Key Lab of Urban & Architectural Heritage Conservation, Ministry of Education, Southeast University, Nanjing 210096, China Abstract: Corrosive environment and fatigue load will greatly reduce the service life of reinforced concrete (RC) structure. In order to study the performance of RC beams strengthened with pre-stressed Basalt/Carbon Hybrid Fibre Reinforced Polymer (B/C HFRP) sheets, corrosion fatigue experiments on eight beams were carried out. Different load levels, different pre-stress levels and other factors were considered. The results showed that: (1) Corrosion of RC beams significantly reduced their fatigue lives; (2) After strengthened with B/C HFRP sheets, the fatigue lives under corrosive environment of these beams were significantly improved; (3) Conducting pre-stress on the B/C HFRP sheets reduced corrosion damage to the RC beams. With the improvement of pre-stress levels, the fatigue lives of these beams increased gradually. Finally, theoretical analysis of the fatigue bending stiffness was carried out. Keywords: corrosion fatigue, reinforced concrete (RC) beam, strengthen, pre-stress, basalt/carbon (B/C) hybrid fibre reinforced polymer (HFRP) sheet

1 Introduction Corrosion fatigue referred to the phenomenon of fatigue fracture under the combined effects of corrosive media and cyclic stress. It’s a special deterioration process [1]. The corrosive media could weaken the material properties and exacerbated the fatigue damage, and fatigue loads accelerated the corrosion environmental effects on the structures, but the coupling between the two effects was not a simple superposition. Corrosion fatigue involved in many civil engineering structures, essentially damp air, sea water and other corrosive media related. Reinforced concrete structures under interaction of cyclic loading and corrosive environment would significantly be reduced their service lives. Currently, large structures were always subjected to cyclic loading in different corrosive environments , including bridges across the ocean , coastal highway, railway bridges, and offshore oil drilling platform, etc.. Due to the presence of corrosive environments and cyclic loading, their durability problems were more and more prominent, and their service lives were also greatly reduced. [2] In China, the loss of reinforced concrete structures caused by corrosion fatigue problems was very serious. Especially in the coastal cities, the buildings were more susceptible to marine and other corrosive environments, so the durability of the buildings was greatly reduced. Around 1960, China's relevant departments investigated the operational status of reinforced concrete structures in 27 cities of southern China, and the results showed that: structures damage due to rebar corrosion was accounted for 74 % of the total survey. Harbor Engineering department also conducted a survey on 18 reinforced concrete piers in southern China, and the results showed that: 89 percent of the pier reinforcement corrosion was found. [ 3 ] Domestic and foreign researchers had carried out a series of fatigue tests on FRP sheets and reinforced concrete beams strengthened by FRP(Fiber Reinforced Polymer) after being corrosion , for example , Yi Jianwei et al [4] ,Deng Zongcai , et al [5] ,Sobhy Masoud , et al [6] EG Sherwood et al [7 ], Z.Wu, et al [8] . A common conclusion was that, ultimate bearing capacity of * Corresponding author affiliation: Civil Engineering Department, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China, e-mail address:[email protected]

516


strengthen beam was improved than non- strengthened beam with different degree.However, at present, their was still very few research on the coupling of corrosion fatigue of reinforced concrete beam strengthened by pre-stressed FRP.

2 Specimen In order to study the performance of RC beams strengthened with pre-stressed Basalt/Carbon Hybrid Fiber Reinforced Polymer (B/C HFRP) sheets, corrosion fatigue experiments on eight beams were carried out. Different load levels, different pre-stress levels and other factors were considered. The size of each specimen was 120mm × 200mm × 1500mm. The calculated span was 1200mm (Figure 1).

Figure 1 Specimen size

The standard value of actual compressive strength of test concrete cubes was 41.6MPa. Longitudinal reinforcement was HRB235, having a diameter of 10mm; stirrup was HPB235, having a diameter of 8mm. B/C HFRP(Basalt / Carbon hybrid fiber reinforced polymer) sheets were made by mixing basalt fiber and carbon fiber in the ratio of 2:1 in unidirectional layer[9,10]. Their thickness, width, elastic modulus, tensile strength were 0.167mm, 120mm, 1.3 × 105MPa, 2184MPa,respectively.In pre-stressed tension process, digital microscope was used to control the pre-stressed level. The real specimen was shown in Figure 2.

Figure 2 Real specimen

The corrosion solution was prepared more than 40 liters. Water, sodium chloride, sodium sulfate, magnesium sulfate, calcium sulfate, calcium bicarbonate, potassium were 40000g, 8359.2g, 1724g, 219.2g, 48.4g, 10g, 3.6g,respectively. Three-point loading was used. The loading equipment was self-made and had obtained China patent (Figure 3). During the test, several specimens were under fatigue loading simultaneously, when some specimens reached 2×104 cycles, 4×104 cycles, 8×104 cycles, 16×104 cycles, 32×104 cycles, 64×104 cycles, 128×104 cycles, the specimens were put down from the equipment and were placed in corrosion solution for corrosion, while other specimens were put in the equipment for fatigue loading. This process was repeated until 200×104 cycles were reached.

517


Figure 3 The loading equipment

3 Fatigue life During the test, the longitudinal steel bars ruptured generally prior to the fracture of HFRP sheets. When subjected to slight corrosion, the fatigue performance of the longitudinal steel bars would be greatly reduced. The cracks number of specimen under corrosion fatigue test was significantly reduced, but the crack width and crack spacing increased, and the height of the crack extension were relatively high. Corrosion made the flexural stiffness and fatigue life of the specimen greatly reduced. The main results were as shown in Table 1. Table 1 Main results

load

Pre-

fatigue life

stressed?

(104 cycles)

no

N/A

N/A

normal

---

yes

no

N/A

normal

natural

60%

no

N/A

40.12

L4

corrosive

60%

no

N/A

21.14

L5

natural

36%

yes

no

>200

No.

environment

L1

natural

N/A

L2

natural

L3

level

Strengthened?

failure mode

one longitudinal steel bar rupture crack width>1.5mm N/A crack

L6

corrosive

36%

yes

no

90.54

width>1.5mm,then one longitudinal steel bar rupture FRP debond,

L7

corrosive

50%

yes（low

yes

level）

46.55

crack width>1.5mm,then one longitudinal

518

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands steel bar rupture FRP debond, L8

corrosive

50%

yes（high

yes

level）

crack 50.23

width>1.5mm,then two longitudinal steel bar rupture

(a) L3（longitudinal steel bar rupture） (b) L4（crack width>1.5mm）

(c) L6（longitudinal steel bar rupture）(d) L7（longitudinal steel bar rupture）

(e) L8（longitudinal steel bar rupture） Figure 4 Failure modes

By comparison of the test data of fatigue life in Table 1, it could be seen: (1) The comparison between the test data of L3 and L4 showed that, due to the presence of corrosive environment, and at the same fatigue load, the fatigue life of reinforced concrete beam under corrosive environment was significantly reduced, and the reducing amplitude reached 47.3 %. Namely, the corrosive environment would significantly reduce the fatigue life of the specimen. (2) The comparison between the test data of L3 and L5 showed that, when strengthened with non-pre-stressed HFRP sheet, and at the same fatigue load, the fatigue life of strengthened beam(L5) was significantly improved, and was loaded to 200×104 cycles without being

519


destroyed. Namely, HFRP sheets strengthening would significantly improve fatigue life of the specimen. (3) The comparison between the test data of L5 and L6 showed that, when strengthened with non-pre-stressed HFRP sheet, and at the same fatigue load, the fatigue life of reinforced concrete beam under corrosive environment was significantly reduced, and the reducing amplitude reached 50%.But the fatigue life of L6 was far greater than that of L4. Namely, the improving effect of non-pre-stressed HFRP strengthening would be more obvious than the reducing effect of corrosive environment to the specimen. (4) The comparison between the test data of L6 and L7 showed that, although conducting pre-stress would reduce corrosion damage to the specimen and improve the fatigue life of the specimen, but load level had larger impact on fatigue life of the specimen. Fatigue life under high load level was significantly reduced, and the reducing amplitude reached 48.6%. (5) The comparison between the test data of L7 and L8 showed that, under different levels of pre-stressing, and at the same fatigue load, the fatigue life of specimen with higher prestressing level was slightly improved, and the improving amplitude reached 7.9 %. Namely, fatigue life of specimen would gradually increase with the improvement of pre-stressing level. (6) The comparison from the data of L6, L7 and L8 showed that, the impact of load level on fatigue life was more obvious than the impact of pre-stressing level on fatigue life of specimen.

4 Fatigue bending stiffness Under the three-point bending load, the relationship between bending stiffness (B) and mid-span deflection (f) can be expressed as [11]:

L3P B 48f

(1)

Where, B was the bending stiffness of the specimen; P was the applied load; L was the span. The fatigue bending stiffness of the specimens measured was as shown in Table 2: Table 2 The fatigue bending stiffness of the specimens

Specimen No.

L3

Specimen No.

L5

loading cycles 4

loading

bending 2

Specimen No.

cycles

bending stiffness（kNm2）

（10 ）

stiffness（kNm ）

0

845.12

0

696.02

4

739.13

2

339.09

8

567.72

4

310.81

16

568.26

8

280.29

32

480.88

16

231.34

loading cycles

bending stiffness（

4

（104）

L4

2

loading Specimen No.

cycles

bending stiffness（ kNm2）

（10 ）

kNm ）

0

959.42

0

767.61

4

823.29

2

415.60

8

719.39

4

390.17

16

694.64

32

349.31

32

654.55

64

281.62

64

608.93

（104）

L6

520


Specimen No.

128

594.72

loading cycles

bending stiffness（

4

L7

2

loading cycles

Specimen No.

bending stiffness（ kNm2）

（10 ）

kNm ）

0

901.32

0

967.40

2

474.82

2

622.96

4

474.14

4

595.68

8

386.14

8

566.28

16

371.52

16

524.14

32

338.61

32

571.09

（104）

L8

Kh

0

Meanwhile, theoretical analysis of he bending stiffness was carried out as follows. εc

(1-K)h

0

ηh

0

M

As

εs εcf

A cf

s SA S εE ( cf +εE cf Acf ε

Figure 5 bending stiffness calculation diagram

As shown in Fig.5, due to the pre-going strain caused by pre-stressing effect, so strain of HFRP was

 cf   

, from the balance of force and strain can be obtained: M  Es s As h0  Ecf ( cf   ) Acf ( h0   s ) (2)

s (1  K )h0   cf (1  K )h0   s

 sk   s Es  Acf  

BS 

(3)

M  Ecf ( cf   ) Acf ( h0   s )

(4)

( AS  Acf ) h0

Ecf  h0   s (1  K )h0   s Acf Es  h0 (1  K )h0

(5)

ES AS h02 ES AS h02 ES AS h02 (6)）   （4-13 6 E  0.65 ftk  E  6 E   1.15  0.2  1.15(1.1  )  0.2    1  3.5 f te sk 1  3.5 f

According to the test data in Table 2, we could get the fitting curve between the stiffness reduction factor (

 f  1.34779  0.12864lg N ( R2  0.97568) ) and the number of load cycles (N):

521


 f  1.34779  0.12864lg N ( R2  0.97568)

L5：

 f  1.01224  0.11514lg N ( R2  0.98862)

L6：

  1.00474  0.11255lg N ( R2  0.98908)

L7： f 2 L8： f  0.99605  0.08155lg N ( R  0.97433) So the fatigue stiffness after loading was：

B f   f BS

(7)

Where: Bs -short-term test beam flexural stiffness; AS -the total area of tension reinforcement   area (HFRP area could be translated by stiffness equivalence principle, namely, AS  AS  Acf ); h0 effective cross-section height;  - the cross-section arm coefficient, approximated 0.87; ES elastic modulus of steel bar;   -pre-going strain;  s - distance from the force point of longitudinal reinforcement to the edge of cross-section; K - the ratio of actual height of compression area of cross-section to the effective height of the cross-section, approximated 0.4;  - coefficient of strain uniformity of longitudinal reinforcement between the cracks; ftk - standard value of axial

tensile strength of concrete; te - Longitudinal reinforcement ratio calculated by effective crosssectional area of the concrete;  E - the ratio between the elastic modulus of steel bar and the elastic modulus of concrete;  - Longitudinal reinforcement ratio,   AS (bh0 ) ;   - the ratio between the compression flange area and the effective web area; B f -fatigue bending stiffness after a certain number of fatigue load cycles. In accordance with the above formula, the value of the fatigue bending stiffness of each specimen could be obtained, and then the calculated value was compared with the test value, and the error was also calculated, the final data could be obtained in Table 3 as follows: f

Table 3 Calculated values compared with experimental values under fatigue load cycles

No.

L5

L6

L7

L8

Value

Fatigue load cycles（104） 4

8

16

32

experiment

823.2996

719.3997

694.6474

654.5572

calculate

680.203

645.351

610.499

575.647

error

21%

11%

14%

14%

experiment

390.1792

370.468

364.124

349.3161

calculate

434.123

402.928

371.734

340.539

error

10%

8%

2%

3%

experiment

474.1459

386.149

371.5297

338.6105

calculate

483.850

450.18

416.51

382.824

error

2%

14%

11%

12%

experiment

595.6844

566.2897

524.1409

571.091

calculate

626.960

602.165

577.371

552.576

error

5%

6%

9%

3%

As shown in Table 3, the errors between the calculated values and experimental values were less than 21%, which indicates that, the discrete was small, but not small enough, because there were still some errors. The main reason of errors was that stiffness decline caused by corrosion was not considered in the formula. So we could introduce an empirical coefficient, namely, ‘a’, obtained by fitting experimental data. For specimen L5, L6, L7, L8, the values of ‘a’ were 1.15, 0.96, 0.90, 0.96, respecttively. The formula after correction was:

522


B f  a f BS

(8)

5 Conclusion The failure mode of RC beams strengthened by pre-stressed B/C HFRP sheets was mainly the longitudinal steel bar rupture. During the test, the longitudinal steel bars ruptured generally prior to the fracture of HFRP sheets. When subjected to slight corrosion, the fatigue performance of the longitudinal steel bars would be greatly reduced. The cracks number of specimen under corrosion fatigue test was significantly reduced, but the crack width and crack spacing increased, and the height of the crack extension were relatively high. The corrosive environment would significantly reduce the fatigue life of the specimen. The impact of load level on fatigue life was also very obvious. HFRP sheets strengthening would significantly improve the fatigue life of the specimen. The improving effect of non-pre-stressed HFRP strengthening to the specimen would be more obvious than the reducing effect of corrosive environment to the specimen. The fatigue life of specimen would gradually increase with the improvement of pre-stressing level. Finally, theoretical analysis of the fatigue bending stiffness was carried out, and the formula was presented, after taking into account a correction factor, the theoretical calculation method had certain reliability and accuracy.

Acknowledgements This paper is written with support of National Natural Science Foundation of China (Grant No. 51108238).

References [1] Xu Hao.Fatigue strength. Beijing: Higher Education Press,1988.(In Chinese) [2] Guo Zhenhai, Shi Xudong. Reinforced concrete principles and analysis. Beijing: Tsinghua University Press,2009．(In Chinese) [3] Luo Fuwu. Analysis and prevention of structural accident and defects. Beijing: Tsinghua University Press,2009.(In Chinese) [4] Yi Jianwei,Sun Xiaodong．Experimental study on fatigue performance of corroded reinforced concrete beams. Civil Engineering Journal，2007 40(3):7-10．(In Chinese) [5] Deng Zongcai．CFRP Strengthened RC Beams anti-fatigue and anti-corrosion effect. FRP / Composites， 2008 6:44-47．(In Chinese) [6] Sobhy Masoud ，Khaled Soudki ，Tim Topper．CFRP Strengthened and Corroded RC Beams under Monotonic and Fatigue Loads. Journal of Composites for Construction/November 2001:228-236． [7] E.G. Sherwood and K.A ． Soudki.Rehabilitation of corrosion damaged concrete beams with CFRP laminates----a pilot study．Composites，2000(31):453-459. [8] Z.Wu,KJ washita,et al.Fatigue performance of RC beams strengthened with externally prestressed PBO fiber sheets. FRPRCS-6,2003:885-894. [9] Pan Jianwu，Wu Gang，Yuan Xigui．Fatigue behavior of basalt-aramid and basalt-carbon hybrid fiber reinforced polymer sheets. Journal of Southeast University ( English Edition) ，2013,29 ( 1) : 84-87. [10] Yuan Xigui , Pan Jianwu , Wang Yan. Research on Preparation and Properties of a New Hybrid Sheet based on Basalt Fiber. New Technology and New Process,2011,7:105-108 (In Chinese) [11] Huang Peiyan, Zhao Chen, etc. Fatigue performance of RC members strengthened with FRP．Beijing: Science Press,2009 (In Chinese)

523


Investigation of time-dependent deformation of RC beam with flexural crack generated in early age considering shrinkage property of concrete Satoshi KOMATSU1*, Akira HOSODA1 (1) Yokohama National University, Yokohama, Kanagawa, Japan

Abstract: Time-dependent deformational behaviour of RC beams with flexural cracks generated in early age was investigated considering shrinkage property of concrete. Experimental results and numerical simulation proved that compressive deformation of concrete due to drying creep caused significant increment of tensile stress of reinforcement. Under high compressive stress/strength ratio, large creep deformation of concrete might be due to micro cracks generated at interfacial transition zone between cement paste and aggregates. Deterioration of bond between compressive reinforcement and surrounding concrete might be another reason of large creep deformation. A new methodology to calculate total flexural crack width using FEM simulation results was proposed and the behaviour of early age concrete in tension was discussed based on the analysis of the experimental results. Keywords: shrinkage property of concrete, cracking in early age, flexural crack, creep deformation, numerical simulation

1 Introduction Recently, in Japan, severe damages of concrete structures such as so large numbers of cracks due to excessive shrinkage of concrete have been reported [1]. However, the time-dependent behaviours of RC structures with cracks generated in early age have not been clarified sufficiently. Therefore, in this research, the time-dependent behaviour of RC beams with flexural cracks generated in early age is investigated considering the shrinkage property of concrete. The influences of the shrinkage property of concrete, the age of cracking, and environmental conditions on the time-dependent behaviour of RC beams are experimentally investigated. In addition, with a FEM numerical simulation, the effects of creep deformation of concrete on the time-dependent behaviour of RC beams are analysed. Furthermore, a new methodology to calculate total flexural crack width using FEM simulation results is proposed and the behaviour of early age concrete in tension is analysed.

2 FEM simulation system used in this research 2.1

Basic scheme [2]

In this research, a FEM simulation system integrating microscopic thermo-hydro physics and macroscopic nonlinear mechanics is utilized for investigating the time-dependent behaviour of RC beams. The responses are successively updated following time history in the software considering mutual linkage of microscopic characters of concrete composites and structural mechanics with damaging induced by loads and weather actions. With this simulation tool, the long-term deformation mechanism of PC bridges has been clarified. To consider the bond effect between concrete and reinforcement after cracking, the relationship between the average stress and the average strain of concrete is defined with stiffening parameter “c” (0.4 for deformed reinforcement). In the area where bond is not effective, crack is localized. Therefore, to express the effects of crack localization, “zoning *

Assistant professor, [email protected]

524


method” is applied and the fracture energy of concrete and the element size are considered in tension softening of the concrete in plain concrete zone. In this research, RC zone (bond effective area) was decided to be the area of double of cover thickness from the bottom of RC beams. To consider the shrinkage of cement paste, in case of high-middle humidity, condensed water in the pore forms meniscus at gas-liquid boundary, and negative pressure by the surface tension becomes driving force for shrinkage. In case of low humidity, tensile force is considered due to increment of surface energy in gel particles, and plastic deformation is modelled due to movement of the interlayer water. Additionally, ink bottle effect due to geometric structure in the pore is also considered. To consider the shrinkage of the aggregate, the stiffness of aggregate is defined as a function of the mean density of fine aggregates and coarse aggregates. The shrinkage of an aggregate is defined as a function of the maximum shrinkage strain when the aggregate is absolutely dried, and the degree of saturation of the aggregate.

2.2

Factors to be considered in investigating time-dependent behavior of RC beam since early age

It has been reported that creep deformation in concrete can be generated not only due to the water flow in the pore, but also due to the effect of micro defects between cement paste and aggregates [3, 4]. The model for concrete under sustained loading was proposed by El-Kashif [2]. However, the model has not been verified for early age concrete. There is a possibility that, in early age, creep strain can be developed due to the propagation of micro cracks at interfacial transition zone (ITZ). In addition, when the concrete in compression zone is under high stress/strength ratio in early age, there can be bond deterioration between concrete and reinforcement. In that case, substantial compressive stress of concrete becomes larger, leading to larger creep deformation.

2.3

Cracking standard strength

2.4

Caluculation methold of totall crack width

To avoid unnecessary cracking in numerical simulation, an imaginary tensile strength, “cracking standard strength” is set in this research. When the tensile strength calculated from compressive strength obtained in the numerical simulation is smaller than the cracking standard strength especially in very early age, the judgment of cracking is conducted by the cracking standard strength. In this study, the cracking standard strength was calculated from the experimental compressive strength at the age when continuous bending loading was applied to RC beams. The simuation tool used in this research is based on smeared crack model. Therefore, it is difficult to predict the generation and the propagateiton of individual macro cracks. In this research, as shown in Fig.1, by integrating the difference between the mean strain of reinforcement and the surface strain of concrete in the equivalent moment area, total crack

Fig.1 Calculation method of total crack width using the numerical simulation results

525


width is calculated. Comparing the calculated total crack width with the experimentally observed crack width, the time-dependent deformation of RC beams is analyzed. Here, in Fig.1, tensile strain is defined as positive, and compressive strain is defined as negative. In this calculation method, to obtain the strain of reinforcement, tension stiffening effect is considered. The following two factors are the driving force for the strain at the surface of concrete between flexural cracks. -

The difference of elastic strain between concrete and reinforcement due to shrinkage and creep of concrete - Strain caused by external force However, the second factor mentioned above is known as negligible [5]. In our calculation method, this factor is not considered. In order to calculate the elastic strain mentioned above, Ishibashi [6] proposed to assume an imaginary concrete member called “a small divided member (Mean crack spacing x Slab width x 1/5 of girder height)”. This small divided member was assumed to shrink freely whose six surfaces were exposed to the environment with 20°C and R.H. =70%. In this research, based on this idea, the strain at the surface of concrete is numerically simulated and used for predicting the crack width. The environment condition for the simulation was set as 20°C and R.H. =60% following the exposed condition. The length of a small divided member is decided as the observed mean crack spacing. The boundary condition of this member is set as sealed condition till the age of demolding of the RC beam, and after flexural cracks are generated, whole six surfaces are exposed to drying condition.

2.5

Size of element

In this simulation, average stress-strain relationship is applied. Therefore, it is recommended that an element size of concrete should be larger than the maximum particle size of coarse aggregates (20mm). On the other hand, the elements nearest to the surface should be small enough to prevent unnecessary numerically generated micro cracks. In this research, to prevent the propagation of those micro cracks for accurate simulation, the size of outermost layer was defined as 5mm. In the longitudinal direction of RC beam, the maximum size of element was set as 30mm.

3 Outline of experiment and properties of concrete 3.1

Details of RC beam and method of loading

In this experiment, the density of fine aggregate used is 2.53g/cm3. The density of coarse aggregates, limestone coarse aggregate and sandstone coarse aggregate are 2.69g/cm3 and 2.58g/cm3, respectively. The maximum size of coarse aggregate is 20mm. In Table1, the mix proportions of concrete are described. The fresh properties of those concretes were almost constant (Slump15±1.5cm, air content 4±1.5%). Table1 Mix proportions Unit content (%)

Type

Coarse aggregates

W/C (%)

s/a (%)

W

C

S

G

AE mixture (g/m3)

AE water reducing agent (g/m3)

Gs Gl

sandstone limestone

50

46 50

155 150

310 299

810 892

983 922

217 0

310 898

The parameters of RC beams are shown in Table2 and Fig.2. To compare the results in this research with the past research results, the dimension of RC beams was made same as in the experiment by Ishibashi [6]. To ensure the bond between concrete and reinforcement, a thin channel was made along the longitudinal rib of reinforcement. In the channel, strain gages were attached at each 30mm. The channel was filled with epoxy resin for waterproofing. The stress of reinforcement was calculated from the mean value of strains in reinforcement multiplied by the

526


stiffness of reinforcement. The distribution of the strain of reinforcement including yielding points should be expressed as a parabola using nearest three points [2]. However, in this research, the mean value was calculated by taking the mean value of nearest three points. In this experiment, two points loading was applied by controlling the strains of PC bars for loading. The mean tensile stress level of reinforcement was controlled as around 180N/mm2 which was almost the same stress level of an actual RC box girder (Higashi-sendai viaduct: Tohoku Shinkansen structure), which suffered severe flexural and shrinkage cracks [7]. In the equivalent moment area, strain gauges were attached at the compressive edge and at 50mm from the compressive edge. The deflection at the center of the span was also measured. Name of RC beam coarse aggregate demolding [day] loading [day] environment [ºC, RH]

Table2 Parameters of RC beams Gs(3) Gs(28) Gs(3)S sandstone 2.8 27.8 2.0 3.6 28.2 3.2 20ºC, 60% Sealed

Gl(3) Gl(28) limestone 2.8 27.8 3.6 28.2 20ºC, 60%

Fig.2 Dimension of RC beams

3.2

Shrinkage property of concrete

In the numerical simulation used in this research, the maximum shrinkage strain of aggregate in absolute dry condition is necessary to simulate the shrinkage of concrete. In this research, by the inverse analysis of the results of shrinkage tests of concrete by JIS method, the maximum shrinkage strain of aggregate was estimated (Gs = 2000×10-6, Gl = 500×10-6).

4 Test results of continuous loading of RC beams 4.1

Deformational behaviour of RC beams in static loading

The mean tensile stress of reinforcement during static loading is shown in Fig.3. The numerical simulation results agreed well with the experimental results, which meant that the mechanical properties of concrete were appropriately simulated based on the input data. The stress of reinforcement when the flexural cracks were generated at 3 days (205.2N/mm2) was larger than that at 28 days (190.2N/mm2).

Fig.3 The results of static loading

527


4.2 Time-dependent behaviour of RC beams 4.2.1 Influence of age of cracking The influence of the age of cracking on the increment of the mean tensile stress of reinforcement is shown in Fig.4. In the RC beam loaded at 3 days, significant increment of the mean tensile stress of reinforcement was observed. In the past research [6], it was reported that the increment of tensile stress of reinforcement after cracking was small (little influence on crack width propagation). Fig.4 shows that the increment of tensile stress of reinforcement after cracking is not negligible when continuous loading is started in very early age.

4.2.2 Influence of shrinkage property of concrete

The influence of the shrinkage property of concrete on the mean tensile stress of reinforcement is shown in Fig.5. In the RC beam whose concrete had larger shrinkage property (sandstone coarse aggregate), larger increment of the mean tensile stress of reinforcement was observed.

Fig.4 The effect of age of cracking on the mean stress development of reinforcement

Fig.5 The effect of environment condition on the mean stress development of reinforcement

4.2.3 Influence of environmental condition The influences of environmental conditions on the mean tensile stress of reinforcement and the strain of concrete at the compressive edge are shown in Figures 5 and 6. In the sealed RC beam, smaller increment of the reinforcement stress was observed. The strain of concrete at the compressive edge was also smaller in the sealed RC beam. In the sealed RC beam, decreasing of the mean tensile stress of reinforcement was observed, which may be due to the generation of a shear crack. The generation of a shear crack was also confirmed in the result of numerical simulation.

Fig.6 Strain developments of concrete at the compressive edge

528


In Fig.7, the distribution of the tensile strain of reinforcement is shown. In the RC beam exposed to the environment condition (20°C and R.H. =60%), reinforcement got yielded in some points. On the other hand, in the sealed RC beam, reinforcement remained in elastic range.

Fig.7 Strain distribution of reinfocement in equivalent moment zone

5 Analysis of time-dependent behaviour of RC beams with numerical simulation 5.1

Mechanism of increment of tensile stress of reinforcement

5.2

Mechanism of creep deformation at high compressive stress/strength ratio

The time-dependent behaviour of the RC beam with limestone aggregate loaded at 3 days is shown in Fig.8 (compressive stress/strength ratio =41%). In the simulation, stiffening factor “c” was changed in the range between 0.2 and 0.9. The influence of tension stiffening on the timedependent behaviour was small. Kakuta [5] pointed out the causes for the increment of reinforcement stress after cracking as below. - Due to compressive creep deformation of concrete, neutral axis gets closer to the compressive edge - Due to the deterioration of bond between tensile concrete and reinforcement The second cause above will not be the main mechanism in this research, because tension stiffening effect is originally small for a young age RC beam. The authors believe that the first cause above will be the main mechanism for the significant increment of reinforcement. The results of the numerical simulation of the increment of the mean tensile stress of reinforcement in the RC beam loaded at 3 days were much different from the experimental results. There must be some important factors unconsidered in the simulation system which largely affect the time-dependent behaviour of young age concrete. The authors’ hypotheses are as below: - Due to the deterioration of bond between concrete and reinforcement in compressive zone, the substantial compressive stress of concrete was increased. - Micro cracks were generated at ITZ between cement paste and aggregates in concrete, leading to large compressive creep deformation. These two hypotheses were simply expressed in numerical simulation as follows. First, the reinforcement in compressive zone was removed. Second, W/C of concrete in compressive zone was increased from 50% to 60%. Fig.8 shows the increment of the mean tensile stress of reinforcement in the RC beam with limestone aggregate loaded at 3 days. Considering both phenomena above, the simulation results (stress/strength ratio =98%) agree well with the observed results. Fig.9 shows the results of the RC beam with sandstone aggregate. Even though the both phenomena were considered, the simulation results (compressive stress/strength ratio =69%)

529


were still underestimating the observed result. In the RC beam with sandstone aggregate, the flexural cracking load was lower than that of the RC beam with limestone aggregate. This may be due to the difference of bond strength between coarse aggregate and mortar. In the past research, it has been reported that bond between sandstone aggregate and mortar paste was the weakest [8].The lower cracking load must have been due to the weaker bond at ITZ around aggregates. To investigate the effects of weaker bond at ITZ, a kind of cracking trigger elements of 10mm width were provided. In the equivalent moment zone, those cracking trigger elements were arranged so that the number of the trigger elements became equal to the number of observed flexural cracks. W/C of concrete in the trigger element was 10% larger than that in other elements. Consequently, as shown in Fig.9, significant increment of the mean tensile stress of reinforcement (compressive stress/strength ratio =98%) was observed. The reason why the rate of increasing was still smaller than the reality might be that the propagation of flexural cracks were not appropriately simulated.

Fig.8 Stress developments of reinforcement of RC beam wth limestone aggregate

5.3

Fig.9 Stress developments of reinforcement of RC beam with sandstone aggregate

Investigation of propagation of flexural crack width

Fig.10 shows the total crack width both in the experiment and the simulations. In the simulations, there are two lines. The upper green line shows the total crack width considering free shrinkage of the concrete between flexural cracks, and the lower blue line shows the total crack width considering no shrinkage of concrete. In the RC beam loaded at 28 days, the simulation results considering free shrinkage of the concrete between flexural cracks based on Ishibashi’s method was close to the experimental results. On the other hand, in the RC beam loaded at 3 days, the simulation results without considering the shrinkage of the concrete between flexural cracks were closer to the reality.

Fig.10 Developments of total crack width and schematic illustration of damage around flecural crack

530


The authors think that in the RC beam loaded at 3 days, the shrinkage of the concrete between flexural cracks due to drying might have been compensated by the tensile deformation of concrete following the deformation of reinforcement due to continuous loading. The tensile deformation of concrete was derived due to the less degradation of bond between reinforcement and surrounding concrete. On the other hand, in the RC beam loaded at 28 days, the bond between concrete and surrounding concrete was deteriorated accompanied with internal cracks [9]. To assume the free shrinkage of the concrete between flexural cracks, the calculation of total crack width agreed well with the observation. In the past research, it was pointed out that the larger shrinkage of concrete near flexural cracks might be restrained by reinforcement, which resulted in the agreement between the calculation assuming the free shrinkage of concrete and the observed crack width at the surface [10].

6 Conclusions -

-

-

The significant increment of the tensile stress of reinforcement after cracking was experimentally observed when RC beams were loaded at 3 days. The shrinkage property of concrete and the environmental conditions largely affected the increment of reinforcement stress. The significant increment of the tensile stress of reinforcement might be due to the compressive creep deformation of concrete. Under high stress/strength ratio of concrete, the increment of the tensile stress of reinforcement might be due to the deterioration of bond between concrete and reinforcement in compressive zone and due to micro defects at ITZ between aggregate and cement paste. Those effects were numerically simulated. In the RC beam loaded at 3 days, the calculation of total crack width without considering the shrinkage of the concrete between flexural cracks agreed well with the observation. This was due to the less bond degradation between tensile concrete and reinforcement. In the RC beam loaded at 28 days, the calculation of total crack width assuming the free shrinkage of the concrete between flexural cracks agreed well with the observation, which had been verified in the past research.

7 References

[1] JSCE, Final report of special committee for investigation of deterioration of Tarui viaduct, March 2008 (in Japanese) [2] K. Maekawa， T. Ishida， T. Kishi: “Multi-scale modeling of structural concrete”, Taylor and Francis, 2008 [3] S. Honda, J. Okunishi, K.Watanabe, Y. Tanimura: “Influence of micro defects on compressive creep”, Proceedings of the Japan Concrete Institute, Vol.35, No.1, pp.535-540, 2013 (in Japanese) [4] K. Watanabe, N. Sakakibara, W. Jason WEISS, J. Niwa: “Evaluation of relationship between micro defects and basic tensile creep with young age mortar by AE method”, Proceedings of the Japan Concrete Institute, Vol.33，No.1, pp.455-460, 2011 (in Japanese) [5] Y. Kakuta: “Maximum crack width in RC”, Concrete Journal, Vol.8, No.9, pp.1-10, September 1970 (in Japanese) [6] T. Ishibashi, T. Tsuyoshi: “Calculation method of surface flexural crack width in RC girder”, Journal of Materials, Concrete Structures and Pavements, No.484, V-22, pp.33-40, 1994.2 (in Japanese) [7] K. Matsuoka, S. Yoshino, “Application of expansive concrete to Girder –Construction method of Tohoku Shinkansen structure-”, Concrete journal, Vol.19, No.2, pp.30-36, February 1981 (in Japanese) [8] H. Kawakami: “Influence of type of coarse aggregate on mechanical behaviour of concrete”, Proceedings of the Japan Concrete Institute, Vol.13, No.1, 1991 (in Japanese) [9] Y. Goto: “Cracks Formed in Concrete Around Deformed Tension Bars, ACI Journal, No.68,-26, pp.244-251, April 1971 [10] T. Seki, T. Sakurai, T. Shimomura: “Influence of drying shrinkage on development of crack width in RC”, Proceedings of the Japan Concrete Institute, Vol.32, No.2, pp.211-216, 2010 (in Japanese)

531


The Influence of Fiber-Matrix Adhesion on the Linear Viscoelastic Creep Behavior of CF/PPS Composites M.H. Motta Dias1*, K.M.B. Jansen1, H. Luinge2, K. Nayak3 and H.E.N Bersee1 (1) Delft University of Technology, Delft, The Netherlands; (2) TenCate Advanced Composites BV, Nijverdal, The Netherlands; (3) Airborne Technology Centre BV, Den Haag, The Netherlands; Abstract: The influence of fiber-matrix adhesion on the linear viscoelastic creep behavior of AR-CF and SM-CF, reinforced PPS composite materials was investigated. Short-term tensile creep tests were performed on ±45° specimens under four different isothermal conditions. Physical aging effects were evaluated on both systems using the short-term test method established by Struik. The results showed that the additional surface treatment carried out on the SM-CF improved the fiber-matrix adhesion and thus enhanced the mechanical performance of CF/PPS composites but with minor effects on the creep response. Increasing retardation times with physical aging was observed in all test conditions. Compared to temperature effects, physical aging was shown to made a small contribution to the creep behavior of CF/PPS composites. Keywords: viscoelastic creep behavior; physical aging; CF/PPS composites; time aging-time superposition; time temperature superposition principle (TTSP)

1 Introduction Polymer matrix composites (PMCs) have long found utility in many engineering applications, where their specific characteristics such as, low cost, easy processing, corrosion resistance and enhanced strength-to-weight ratios, have enabled these materials to be used in place of traditional metals. In most cases, epoxy thermoset resins constitute the polymeric matrix used. Nevertheless, a growing trend has been the replacement of thermoset matrices with thermoplastics. One thermoplastic polymer that is a potential substitute for epoxy thermoset resins is polyphenylene sulfide (PPS), however, the use of this polymeric material for such purposes requires a greater understanding of its behavior under long-term stress/strain. It’s necessary for PMCs to retain their mechanical performance throughout their designed life time. Due to their polymeric matrix constituent, PMCs are viscoelastic in nature and, therefore, their time-dependent creep behavior can be significantly influenced by exposure to extreme-use environmental conditions, i.e., pressure, temperature, moisture and chemicals, mechanical loads, or a combination of both [1]. Thus, studying and understanding the effects of long-term exposure on the time-dependent viscoelastic behavior of PMCs is extremely important for their proper design and safe operation. As most PMCs are designed for long-term service lives, typically up to 20 years, measuring their viscoelastic responses over their entire lifetimes is fairly impractical. As a consequence, many studies have been conducted and published on accelerated test methods used to characterize the viscoelastic behavior of PMCs [2]-[5]. Accelerated test methods use short-term creep data and models to predict the long-term creep behavior of the material. A successful example of these models is the time temperature superposition principle (TTSP). Improving the fiber-matrix adhesion of high performance thermoplastic composite has been proven to be the key to increasing the performance to cost ratio of PMCs [6]-[8]. Therefore, the development of optimized pre-treatments and sizings has become a main focus of several fibers *

Faculty of Aerospace Engineering, Delft University of Technology, Kluyverweg 1, 2629 HS Delft, The Netherlands; [email protected]

532


manufactures around the world. Not much attention, however, has been given to the influence of the fiber-matrix adhesion on the viscoelastic creep behavior of composites. The goal of the research reported here was to investigate the linear viscoelastic creep behavior of ‘as received’ and ‘surface modified’ carbon fibers (AR-CF and SM-CF) reinforced PPS using elevated temperature to accelerate the viscoelastic response. Increased understanding of the relation between fiber-matrix adhesion and the creep response was gained by comparing creep compliance for AR-CF/PPS and SM-CF/PPS. Since creep is a phenomenon mainly dominated by the polymeric matrix, i.e., creep in fibers is considered to be negligible [2], [9], tensile tests of ±45° specimens, in-plane shear response, were chosen for this studies. As a preliminary step to obtain the time-temperature master curve, and to characterize the changes in matrix dominated properties over storage time, physical aging studies were also carried out using the short-term test method established by Struik [10].

2 Experimental Procedure 2.1 Test Materials and Specimen Configuration Two materials system were investigated in this study, AR-CF/PPS and SM-CF/PPS composites. The PPS thermoplastic resin was fabricated by Ticona and provided by TenCate Advanced Composites in the form of 80 µm thick film. The nominal PPS melt temperature, Tm, is 280°C. The CFs used were T300J 40B 3K standard modulus PAN CFs fabricated by Toray Soficar and provided by TenCate Advanced Composites in the form of 5-harness satin fabric with two different surface treatments: (1) As provided by Toray Soficar to TenCate Advanced Composites - referred to as 'as received’ (AR-CF) in this work. These fibers are surface treated and surface coated with 1% weight Bisphenol A diglycidylether type surface coating; (2) Woven AR-CF fabric which had received an additional industrial surface heat treatment carried out by TenCate Advanced Composites - referred to as 'surface modified' (SM-CF) in this work. No details about the heat treatment were disclosed to the author. The composite panels were produced by hand stacking of the CF and PPS resin film and subsequent thermopressing. The stacking sequences used were [0,90]4s, where [0,90] represents one layer of fabric: the layers were stacked so that they are symmetric with respect to the middle plane and that the warp side faces the outside. The Tg of the resultant material was found to be 102°C and defined as the temperature associated to the maximum of the tan delta peak, measured using a dynamic mechanical analyzer (DMA) at 1Hz and 2°C/min. The test specimens, measuring approximately 250mm long, in the load direction, by 25mm wide, were cut from 2,5mm thick laminated panels in an angle of ±45° to the fibers using a water cooled diamond saw. This specimen size is in accordance with the ASTM Specification D3518/D3518M. Paper tabs were glued to the specimen’s surface longitudinal ends to avoid slipping. Strain gages type KFG-5-120-C1-11, from Kyowa, were used for the strain measurements. In total, 3 strain gages were used for each tested specimen; 2 strains gages were mounted vertically aligned back-to-back in the center of the specimen, so that any possible buckling of the specimen would be detected; 1 strain gage was mounted longitudinally aligned in the center of the specimen. In order to separate load induced strain from thermal strain, temperature compensation gauges were mounted on another specimen placed close to the test specimen. This specimen remained unloaded throughout the test. The thermal apparent strain measured in this specimen was then used to correct the measured strain in the loaded specimen.

533


2.2 Test Equipment Testing in this study was performed using a 10kN Zwick 1445 tensile tester machine equipped with a convection oven. To obtain a better accuracy between the temperature displayed at the oven and the ‘real’ specimen temperature, a thermocouple was attached to the center region of the tested specimen surface. The temperature of the specimen during each set of aging and creep testing was maintained at a constant of 70, 80, 90 or 100°C (± 0.5°C) throughout the test. Prior to testing, the specimens were dried for at least 24 hours at 60°C in a vacuum oven. After each creep test the specimen were stored for 2 days in a desiccator to prevent moisture absorption, rejuvenated and then subjected to a new set of isothermal physical aging and creep testing at a different temperature.

2.3 Creep Testing The specimens were loaded at ±45° to the fiber orientation, in-plane shear response. All tests were carried out at sub-Tg temperatures. The test temperatures selected for this study were 70, 80, 90 and 100°C. These temperatures were selected to ensure that measurable changes in physical aging could be measured within a lab time frame. To ensure that all tests were performed within the linear viscoelastic range, a preliminary study was carried out to check that proportionality conditions and Boltzmann’s superposition were satisfied [10]. Specimens of AR-CF/PPS and SM-CF/PPS loaded at ±45° to the fiber orientation were repeatedly rejuvenated, quenched and subjected to sets of creep and recovery at tensile stresses ranging from 5 to 25MPa with increments of 2.5MPa at both the lowest and highest test temperatures. If the applied load places the material outside the linear viscoelastic range, the momentary compliance response will vary depending on stress level, indicating that the material is in the nonlinear range. Based on this methodology, a tensile stress level of 10MPa was chosen for this study. To promote comparability between AR-CF/PPS and SM-CF/PPS, all subsequent tests were performed at the same tensile stress level. The in-plane shear creep compliance response of the ±45° specimens is given by:

S

2 A(ex  ey ) P

(1)

where S is the in-plane shear creep compliance, P is the axial load applied on the specimen, A is the cross-sectional area of the specimen, ex is strain along the loading direction and ey is the strain along the transverse direction. Prior to test, the test specimen was heated to 120°C and kept at this temperature for 30 minutes, then high-pressured air was used to quench the specimen from 120 °C to the test temperature. The physical aging process was defined to start immediately after a specimen reached the desirable test temperature. The physical aging times selected for starting each creep segment were 4, 8, 24, 48 and 96hours and the duration of each creep tests, t, was chosen to be 0.125 te. These selections were based on lab opening hours, since the test machine used was not fully automated. After each creep segment, the specimen was unloaded and allowed to recover until the start of the next creep test. To account for any remaining residual strain due to the lack of full recovery, the strain measured in the creep segment was corrected by subtracting the extrapolated recovery strain from the prior creep curve. The procedures described above were repeated for all tested temperatures at both ARCF/PPS and SM-CF/PPS.

534


3 Results and Discussion 3.1 Physical Aging Effects on the Creep Behavior Momentary creep compliance versus time curves for AR-CF/PPS and SM-CF/PPS specimens at 70°C are shown in Figure 1. The lines connecting the measured compliance are just guides lines and similar curves were generate for 80, 90 and 100°C. AR-CF/PPS at 70°C and 10MPa -8

2,3x10

4hrs aging 8hrs aging 24hrs aging 48hrs aging 96hrs aging

-8

Compliance (1/Pa)

2,2x10

-8

2,1x10

-8

2,0x10

-8

1,9x10

-8

1,8x10

1

2

10

3

10

10

4

10

Time (s)

(a) SM-CF/PPS at 70°C and 10MPa

-8

1,7x10

4hrs aging 8hrs aging 24hrs aging 48hrs aging 96hrs aging

-8

Compliance (1/Pa)

1,6x10

-8

1,5x10

-8

1,4x10

-8

1,3x10

1

10

2

3

10

10

4

10

Time (s)

(b) Figure 1. Momentary creep compliance as a function of time (in semi-log scale) at 70°C, 10MPa and different aging times for (a) AR-CF/PPS and (b) SM-CF/PPS

Physical aging leads to a decrease in free volume of polymeric materials which, consequently, causes a reduction in segmental mobility, slowing down in creep response, and an increase in the degree of packing, stiffening. The curves shown in Figure 1 confirm these expectations. The creep curves shift towards longer times and the compliance levels decrease with increasing aging times, similar trends were obtained for all the other tested temperatures.

535


Time-aging time superposition was used to generate a single master curve for all of the data at a reference aging time of 96hours for both AR-CF/PPS and SM-CF/PPS. An algorithm was used to generate the aging shift factors, ae. These shift factors were chosen by minimizing the mismatching in compliance by a least square method, the resultant curve is presented in Figure 2. Similar curves were generate at temperatures of 80, 90 and 100°C. AR-CF/PPS and SM-CF/PPS at 70°C and 10MPa teref = 96hrs

-8

2,4x10

-8

Compliance (1/Pa)

2,2x10

-8

2,0x10

AR-CF/PPS

-8

1,8x10

-8

1,6x10

-8

1,4x10

T-CF/PPS

-8

1,2x10

1

10

2

3

10

10

4

10

Time (s)

Figure 2. Master curves obtained by time-aging time superposition using momentary compliance curves from Figure 3 for AR-CF/PPS and SM-CF/PPS. The reference physical aging time is 96hours

From Figure 2 it can be seen that the additional fiber surface treatment carried out in SM-CF improves the adhesion between the CF and the PPS matrix, resulting in a significant deduction of the measured compliance compared with AR-CF/PPS. Nevertheless, the fact that both curves appear to have an identical shape seems to imply that this improvement in the fiber-matrix adhesion has a minor effect on the creep response of these CF/PPS systems at 70°C. Similar trends were observed for the curves obtained at 80, 90 and 100°C, see Figure 6(a). The obtained ae versus the logarithmic te at 70°C values are shown in Figure 3. An small difference in ae for AR-CF/PPS and SM-CF/PPS can be observed at 4 and 24hours of physical aging, similar difference were also observed at 80°C. These difference can be explained by the uncertainty in the obtained ae due to the data reduction algorithm. Any data point with a certain inaccuracy has a bigger effect on the obtained ae using the reduction data than using raw data. If the curves in Figure 3 are fit using a linear regression, the slope of this straight line is the aging shift rate, μe. The shift rate values obtained by repeating all above mentioned procedures at all tested temperatures are shown in Figure 4. It can be seen that the aging shift rate decreases with increasing temperature. Note that above the Tg, the shift rate should vanish since the material is in thermodynamic equilibrium. Therefore, a sharp decrease of the aging rate towards zero can be expected above 102°C. The fact that the obtained shift rates for AR-CF/PPS and SM-CF/PPS at 70 and 80°C are not identical comes from the previously mentioned uncertainty in the obtained ae due to the data reduction algorithm.

3.2 Temperature Effects on the Creep Behavior The time-aging time master curves versus time for AR-CF/PPS and SM-CF/PPS specimens at all tested temperatures are shown in Figure 5. In this work, the 96hours reference curves were chosen to represent the time-aging time master curve. This choice was based on the fact that the 96hours curves are the ones with more data and perhaps the best in terms of the mechanical conditioning of the specimen.

536


AR-CF/PPS Linear fit AR-CF/PPS SM-CF/PPS Linear fit SM-CF/PPS

0,0

-logae

-0,2 -0,4 -0,6 -0,8 -1,0 -1,2 4,0

4,5

5,0

5,5

logte

Figure 3. Aging shift factor for time-aging time superposition at 70°C and 96 hours reference curve for ARCF/PPS and SM-CF/PPS

1,0

AR-CF/PPS SM-CF/PPS

Shift rate, µe

0,8

0,6

0,4

0,2

0,0 60

70

80

90

100

110

Temperature (°C)

Figure 4. Aging shift rate as a function of temperature for AR-CF/PPS and SM-CF/PPS

The well-known fact that higher temperatures accelerate molecular rearrangements in viscoelastic materials while lower temperatures slow it down [10], is shown in Figure 5. Furthermore, these curves shows that creep compliance was a function of test temperature, with an increase in temperature resulting in an increase in both compliance and related creep rate. The collection of time-aging time master curve for AR-CF/PPS and SM-CF/PPS at all tested temperatures was then collapsed into single material master curves using TTSP. The collapse was done successfully using the 70°C time-aging time master curve as the reference and horizontally shifting all other curves. The TTSP allowed the obtained master curves, along with the temperature shift factor versus temperature curves, to be used to predict momentary response at any temperature. The master curves and the temperature shift factors used to construct them are presented in Figure 6.

537


AR-CF/PPS te = 96hours and 10MPa -8

7x10

-8

Compliance (1/Pa)

6x10

96hrs aging 70°C 96hrs aging 80°C 96hrs aging 90°C 96hrs aging 100°C

-8

5x10

-8

4x10

-8

3x10

-8

2x10

-8

1x10

0 1

1x10

2

3

1x10

4

1x10

5

1x10

1x10

Time (s)

(a)

SM-CF/PPS te = 96hours and 10MPa -8

5x10

96hrs aging 70°C 96hrs aging 80°C 96hrs aging 90°C 96hrs aging 100°C

Compliance (1/Pa)

-8

4x10

-8

3x10

-8

2x10

-8

1x10

0 1

1x10

2

3

1x10

1x10

4

1x10

5

1x10

Time (s)

(b) Figure 5. Time-aging time master curves (96 hours curves) at different temperatures for (a) AR-CF/PPS and (b) SM-CF/PPS

Comparison of the two obtained TTSP master curves at the reference temperature of 70°C showed that AR-CF/PPS exhibited a higher compliance throughout the extrapolated time. The observed nearly constant difference in compliance between the two TTSP master curves clearly indicated that the additional fiber surface treatment (SM-CF) performed by the supplier (TenCate) improved the mechanical performance of CF/PPS composites with a minor effect on the time dependent increase in deformation, creep response. A series of tests to register improvements in the short and long-term mechanical performance of CF/PPS composites with CF with a surface treatment have been done at Delft University of Technology [11], [12]. Analyzing SEM images of fracture areas of AR-CF/PPS and SM-CF/PPS, it has been observed that, while the surface of the SM-CFs seems to be still covered by the resin, the one of AR-CFs is almost bare, showing the improved mechanical and chemical adhesion between the fibers and the matrix of the SM-CF/PPS specimens. The results presented in this paper seem to be in accordance with the findings discussed above.

538


1 year

AR-CF/PPS SM-CF/PPS

-8

6,0x10

Compliance (1/Pa)

10 years 20 years

-8

7,0x10

-8

5,0x10

-8

4,0x10

-8

3,0x10

-8

2,0x10

-8

1,0x10

0,0 0

10

2

10

4

6

10

8

10

10

10

10

12

10

Reduced time (s)

(a) AR-CF/PPS SM-CF/PPS

8

logaT

6

4

2

0 60

70

80

90

100

110

Temperature (°C)

(b) Figure 6. (a) Creep master curves of 96hrs aged specimens for AR-CF/PPS and SM-CF/PPS; (b) Temperature shift factors for AR-CF/PPS and SM-CF/PPS

3.3 Physical Aging Effects vs. Temperature Effects on the Creep Behavior In the previous sub-section it has been shown that both physical aging and temperature influence the linear viscoelastic creep behavior of CF/PPS composites. These influences, however, are characterized by changes in the viscoelastic creep response which seems to be different over time by several orders of magnitude. Furthermore, increasing both temperature and physical aging was shown to have opposite effects on the creep response. While higher temperatures accelerate the viscoelastic responses, longer physical aging times slow it down. A comparison between the ae used to shift the 4hours compliance curve at 80°C to collapse to the 96hours reference compliance curve at the same temperature, 80°C, and the aT used to shift the 96hours at 80°C compliance curve to the 96hours at 70°C reference compliance curve is shown in Figure 7.

539


SM-CF/PPS - 10MPa -8

2,4x10

96hours 70°C 4hours 80°C 96hours 80°C

-8

Compliance (1/Pa)

2,2x10

-8

2,0x10

-8

1,8x10

ae

-8

1,6x10

aT

-8

1,4x10

-8

1,2x10

0

10

1

10

2

3

10

10

4

10

5

10

time (s)

Figure 7. Creep compliance curves for SM-CF/PPS at 80°C, 4 and 96hours of physical aging and; at 70°C, 96hours of physical aging

The analyses presented in Figure 7 indicate that, compared to temperature effects, physical aging was shown to make a small contribution on the creep behavior of CF/PPS composites, thus a temperature change of 10°C, from 70 to 80°C, vice versa, generated a temperature shift factor, aT, that was approximately 20 times greater than the aging shift factor, ae.

4 Conclusions In this experimental study, the linear viscoelastic creep behavior of AR-CF/PPS and SM-CF/PPS using elevated temperature to accelerate the viscoelastic response was investigated. As a preliminary step to obtain the time-temperature master curve, and to characterize the changes in matrix dominated properties over the storage time, physical aging studies were carried out using the short-term test method established by Struik. Short-term compliance curves were obtained for different physical aging times at different temperatures. All tests were carried out at sub-Tg temperatures. As expected, the obtained curves showed an increase in retardation times, or stiffening, as the physical aging times became longer. Time-aging time superposition, was successfully applied to form aging master curves using the aging shift factors, ae, generated by an algorithm. The differences in aging shift rates, μe, for AR-CF/PPS and SM-CF/PPS at 70 and 80°C were probably due to uncertainty on the obtained ae. Using the 96hours compliance curves as representative of the time-aging time master curves, TTSP master curve were obtained by shifting these curves using a temperature shift factor, aT, and the 96hours at 70°C curve as the reference curve. The results indicated that the additional surface treatment carried out in the SM-CF improved fiber-matrix adhesion and thus enhanced the mechanical performance of CF/PPS composites by decreasing the creep compliance of CF/PPS composites throughout the extrapolated time range. This fiber-matrix adhesion improvement, however, was shown to have a minor effect on the linear viscoelastic creep response. In order to further improve the analyses of the result, an attempt for finding suitable prediction models has been started. Furthermore, additional tests to characterize the effects of the fiber-matrix adhesion on the creep behavior of other PMCs will be done to provide a useful foundation for new application areas for PMCs.

540


5 Acknowledgments The authors would like to thank TenCate Advanced Composites for supplying all the materials used in this work.

6 References [1]

D.W. Scott.; J.S. Lai; A.H.; Zureick. Creep Behavior of fiber-reinforced polymeric composites: A review of the technical literature. Journal of Reinforced Plastics and Composites 1995, 14, 588-617 [2] W.K. Goertzen,; M.R. Kessler. Creep behavior of carbon fiber/epoxy matrix composites. Materials Science and Engineering A 2006, 421, 217–225 [3] T.S. Gates.; D.R. Veazie.; L.C. Brinson. Creep and physical aging in a polymeric composite: comparison of tension and compression. Journal of Composite Materials 1997, 31, 2478-2505 [4] E.J. Barbero. Time–temperature–age superposition principle for predicting long-term response of linear viscoelastic materials. Creep and fatigue in polymer matrix composites, 1st ed.; R. M. Guedes, Eds.; Woodhead Publishing, 2011; pp. 48–69 [5] Chang, Y.S.; Lesko, J.J.; Case, S.W.; Dillard, D.A.; Reifsnider, K.L.; The Effect of Fiber-Matrix Interphase Properties on the Quasi-Static Performance of Thermoplastic Composites. Journal of Thermoplastic Composite Materials, 7(4), 1994, 311-324 [6] Madhukar, M.S.; Drzal L.T.; Fiber-Matrix Adhesion and Its Effect on Composite Mechanical Properties: I. Inplane and Interlaminar Shear Behavior of Graphite/Epoxy Composites. Journal of Composite Materials, 1991, 25(8), 932-957Chang, Y.S.; Lesko, J.J.; Case, S.W.; Dillard, D.A.; Reifsnider, K.L.; The Effect of Fiber-Matrix Interphase Properties on the Quasi-Static Performance of Thermoplastic Composites. Journal of Thermoplastic Composite Materials, 7(4), 1994, 311-324 [7] Madhukar, M.S.; Drzal L.T.; Fiber-Matrix Adhesion and Its Effect on Composite Mechanical Properties: II. Longitudinal (0°) and Transverse (90°) Tensile and Flexure Behavior of Graphite/Epoxy Composites. Journal of Composite Materials, 1991, 25(8), 958-991 [8] I.M. Ward, D.W. Hadley, Experimental studies of linear viscoelastic behavior as a function of frequency and temperature: Time-Temperature Equivalence. An introduction to the mechanical properties of solid polymers. John Wiley & Sons Ltd, 2004; pp. 95–120 [9] L.C.E. Struik.; Physical Aging in Amorphous Polymers and Other Materials. 1st ed.; Elsevier Science Ltd 1978 [10] G.C. Papanicolaou.; S.P. Zaoutsos. Viscoelastic constitutive modeling of creep and stress relaxation in polymers and polymer matrix composites. Creep and fatigue in polymer matrix composites, 1st ed.; R. M. Guedes, Eds.; Woodhead Publishing, 2011; pp. 1–47 [11] P. Carnevale.; A.A. van Geenen.; H.E.N. Bersee.; Study on fibre-matrix interfaces in carbon fibre reinforced PPS composites. SAMPE, Baltimore, USA, 2012. [12] S. Rasool.; P. Carnevale.; H.E.N. Bersee.; Effect of Fibre-Martrix Interfaces on Durability of Heavily Loaded Thermoplastic Composite Structures. ITHEC, Bremen, Germany 2012

541


The importance of chloride sensors stability in monitoring ageing phenomena in concrete structures: Ag/AgCl electrodes performance in simulated pore-water environment F. Pargar1*, D.A. Koleva1, E.A.B. Koenders1, K. van Breugel1 (1) Delft University of Technology, Delft, The Netherlands Abstract: One of the reported problems associated with the performance of Ag/AgCl electrodes in alkaline environment, as concrete pore water, has been their poor stability. Open circuit potentials are typically observed to be stable for a time period of a few minutes to some days depending on the thickness and microstructure of the sparingly soluble AgCl and the subsequently developed mixed potentials at the electrode/solution interface. In this paper the open circuit potential of the chloride sensors were monitored over time in simulated pore solution with and without chloride and their electrochemical response recorded via electrochemical impedance spectroscopy (EIS). The alterations in response of the sensors can be denoted to transformation of the AgCl layer to Ag2O in chloride-free high pH environment as in concrete pore water. Recovery of the AgCl layer takes place in chloridecontaining medium, where establishing of sensor stability is determined by the chloride concentration. Keywords: Chloride sensor, open circuit potential, electrochemical impedance spectroscopy

1 Introduction One of the most common issues within ageing infrastructure is steel corrosion in reinforced concrete. Among others, the level of (free) chloride concentration in the vicinity of the steel reinforcement is an important parameter as far as corrosion initiation and propagation are concerned. The traditional techniques for measuring the free chloride content in concrete require destructive sampling and thus cannot be used for continuous monitoring of the free chloride content. In addition, in these techniques the amount of chloride will be measured in the concrete volume under investigation instead of localized measurement, which leads to inaccurate results in the case of concentration gradients [1, 2]. The drawbacks of destructive techniques raised the attempts to measure the free chloride concentration in the pore solution using embedded ion selective electrodes [2, 3]. Silver/silver chloride (Ag/AgCl) electrodes are sensitive to chloride ions and ideally exhibit a certain electrochemical potential that depends on the chloride ion activity in the solution [1]. Although the principles of Ag/AgCl electrodes are well-established, the stability and durability of these electrodes in high alkaline environments such as those existing in concrete (pH~12–13) are of great importance and need to be fulfilled. Otherwise, the application in concrete environment is not feasible [4, 5]. As reported, the potential of Ag/AgCl electrodes (ion-selective (ISE), membrane ones) decreases with time upon immersion in solutions of high pH, in the absence of chloride. This can be attributed to the “damage” of the ISE membrane through the formation of silver oxides or silver hydroxide [5]. Previous studies showed that in high pH solutions with chloride, the potentials are almost stable over 6 months with only ±1.5 mV changes, whereas in the absence of chloride, the potential of the Ag/AgCl electrodes (chloride sensors) were less stable and decreased over time [7]. On the other hand, the relation between the chloride activity and electrode potential is logarithmic, i.e., the potential changes sharply at chloride concentrations close to zero even with a minimum variations in chloride concentration. As further discussed by Angst et al. [7], after 6 months immersion in chloride- free solution, the sensors were removed and placed in potassium hydroxide (KOH) solutions with the same alkalinity, but containing 0.5 M sodium chloride. In the case of 0.01M and 0.1M KOH they responded to the chloride concentration correctly and almost immediately. However, in the case of 1M KOH, the sensors *

E-mail address: [email protected]

542


approached the stable potential within a few hours. The authors concluded that the formation of silver oxide is fully reversible [7]. Similarly, other studies [6] report that Ag/AgCl electrodes demonstrate lower potentials when the pH is above 12 and approach the potential of silver/silver oxide at pH of 14. In this paper the variation in the potential of Ag/AgCl electrodes was monitored in simulated pore solutions (pH~13.6). Meanwhile, the electrochemical response of the sensors was monitored via electrochemical impedance spectroscopy (EIS) in an attempt to correlate electrochemical phenomena to possible changes in the microstructure of the chloride sensors within treatment.

2 Experiments Silver wires of 1 mm in diameter (99.99% purity, as received) were placed for 2 hours in concentrated ammonia and subsequently immersed in distilled water overnight. Then the wires were anodized for one hour in 0.1M HCl solution at a current density of 0.5 mA/cm2 (relevant procedure for the anodization regime as reported in [8]); platinum mesh served as a cathode. After anodization a brown-black AgCl layer was formed over the silver substrate. The anodized silver wire was welded to a copper wire and the non-anodized and welded zones were protected with an epoxy resin. Two series of tests were performed. In the first series, the sensors were immersed in simulated pore solution (SPS) (0.63M KOH+0.05 M NaOH+Sat. Ca(OH)2) with pH of 13.6, free of chloride, over 115 days. The open circuit potentials (OCP), versus saturated calcomel (SCE) reference electrode, were measured over time. In the second series the sensors were treated in SPS free of chloride for 4 days before immersion in SPS with 1000 mM and 125 mM chloride concentrations. OCP and EIS measurements were performed in these solutions. The EIS measurements were performed by superimposing an AC voltage of 10 mV amplitude (rms) in the frequency range of 50kHz (1kHz) to 0.01Hz, using Autolab, Metrohm, NL. Figure 1 shows the morphology of the AgCl layers, formed on the Ag substrate after anodization in 0.1M HCl (Fig. 1a) and the relevant EDS spectrum (Fig.1b), confirming the deposited layer as silver chloride.

Figure 1 SEM top view of AgCl layer deposited over the silver substrate (left); b) EDS of deposited AgCl (right)

3 Results 3.1. Time dependency of the potential of the sensor in alkaline solution Three replicate sensors were immersed in an open to air, chloride-free simulated pore solution (SPS) and their open circuit potentials (OCPs) recorded for 19 days. As shown in Fig. 2, the potential of these sensors decreased from an average value of 137 mV at the start of immersion to approx. 100 – 110 mV after 19 days. As aforementioned, in the absence of chloride and in high

543


pH environment (pH of 13.6 hereby employed), silver hydroxides form and further transform to silver oxides (Ag2O). The continuous formation of Ag2O shifts the potential of the sensors to more negative values, arriving at the recorded potentials of ~ 100 mV (Fig.2), which is according the Nernst equation and solubility constants for the Ag/Ag2O interface in conditions of close to room temperature and pH ~ 14.

Figure 2 Potential stability of the electrodes in simulated pore solution without chloride over time

3.2. The effect of chloride concentration on the response of the sensors In this test series, the effect of pre-conditioning of two replicate sensors in chloride-free SPS solution was monitored when chloride was added to the original solutions. When the sensors were exposed to the chloride-free SPS medium, silver hydroxide forms, transforms to silver oxide and OCP values shift catholically (Fig.3). The result is as previously commented and observed for the series of three replicate sensors (Fig.2), i.e. within immersion OCP values were in the range of 130 – 140 mV (start, Fig.3) and after 4 days of immersion established around 110 mV (day 4, Fig.3). The next step was “contamination” of the solutions with chloride: two chloride concentrations were employed – 1000 mM and 125 mM. The test was executed as follows: after 4 days of immersion in Cl-free SPS, the solution was adjusted to contain 1000mM chloride concentration and the response (OCP) of the sensors were recorded over time of 5 min to 1h (Fig.3). As expected, a cathodic shift of potential was immediately observed and OCP values stabilised around -6mV (which is the expected level of OCP relevant to the activity of Cl-ions in this solution, [8]).

Figure 3 Stability of the electrodes in simulated pore solutions: Cl-free and Cl-containing (1000 mM ,125 mM)

After potential stabilization, the sensors were re-immersed in SPS free of chloride ions (days 4 and 7 in Fig.3). The potential of the sensors shifted back anodically to ~ 120 mV (in contrast to

544


the previously observed 110 mV before immersion in SPS with 1000 mM chloride concentration). In other words, the sensors were partially recovered after immersion in the SPS with chloride ions. This change in the potential can be attributed to the formation of silver chloride (instead of silver oxide) on the surface of the sensors in the SPS with chloride, a predominant reaction when chloride ions are present. Between 4 and 7 days treatment in this “second” chloride-free treatment, the OCP shifted again toward the characteristic values for a Ag/Ag2O interface of ~ 110 mV (117 mV were observed, denoted to mixed potential from AgCl contribution at the stage of 7 days). Finally, the sensors were exposed to SPS with 125 mM chloride ion concentration. In this solution, the OCP of the sensors was established at the level of 46 mV after 24 hours of immersion (Fig. 3), 46 mV corresponding to the relevant chloride concentration in this case. It can be concluded that the “stabilization” time (or AgCl layer recovery) of the treated sensors differs depending on the chloride concentration in the solution. At higher chloride concentration, the sensors stabilize faster. This result is in line with the previous study [8] where it was shown that the intact sensors stabilize faster at higher chloride concentration.

3.3. EIS response of Ag/AgCl in SPS solutions The EIS response of the sensors in the previously discussed conditions and environment was recorded at open circuit potential, AC voltage of 10 mV amplitude (rms) in the frequency range of 1kHz to 10 mHz. The initial response of the Ag/AgCl sensors in Cl-free environment (day 1, Fig.4) presents two clearly distinct time constants, with possible contribution of an additional chemical /electrochemical process in the low frequency domain.

Figure 4 EIS response of Ag/AgCl sensors in Cl-free SPS solutions after 1,2 and 4 days of immersion Within immersion (day 2 and day 4, Fig.4), the magnitude of impedance and phase angle significantly increase, denoted to transformation of the AgCl layer into hydroxide/oxide layer in this environment. The charge transfer resistance at the Ag/AgCl interface is reduced and becomes dominated by diffusion limitation (blocking pores and voids) and/or mass transport phenomena within the formation of the already relevant Ag/Ag2O interface. Upon subsequent immersion of the sensors in Cl-containing environment, the EIS response immediately reflects the recovery of the Ag/AgCl interface – initial and subsequent drop in impedance and phase angle values are observed for the 125 mM and 1M chloride concentration (Fig.5, left). For comparative purposes and in order to clarify the contribution of previously or subsequently formed and/or modified AgCl/Ag2O layers, the EIS response of a Ag wire in 1M NaCl solution is presented in Fig.5 (right) as an overlay to the response of a Ag/AgCl sensor in SPS+1MNaCl. Fig. 6 presents the preliminary equivalent circuits, employed for fitting the EIS response, including an overlay of response (symbols) and fit (line) for the Ag/AgCl sensor after four days of immersion; Table 1 presents the preliminary fit parameters. As can be observed, the chosen

545


electrical circuits give a globally good fit result, however, adjustments are necessary in order to account for the exact surface phenomena. Although further investigation is needed in order to clarify the observed layers’ transformation through correlation of surface layer properties (via microscopy and X-ray analysis) and EIS response, based on the recorded behaviour at this stage, hypothetic behaviour is summarised in what follows. Within immersion of the sensors in chloride-free alkaline environment, the Ag/AgCl interface gradually transforms to Ag/Ag2O interface, reflected by the characteristic OCP changes (Figs. 2 and 3) and alterations in charge transfer/polarization resistance (Table 1).

Figure 5 EIS response of Ag/AgCl sensors in Cl-free SPS solutions (day 4) and in chloride-containing solutions (left) ; overlay of Ag rod and Ag/AgCl sensor response in 1MNaCl

Day1

Day2

1M,Ag

125mM

Figure 6 Equivalent electrical circuits (left) and overlay of EIS response and fit (right) Table 1 – EIS fit parameters – preliminary results Day1 Day2 Day4 125mM 1M 1M Ag

Rs Ω 4.1 4.1 3.9 3.6 4.3 5.3

R1 kΩ 3.5 1.4 2.5 0.8 0.05 0.03

Q1 Y o (μMho) 18.3 180 88 110 123 121

n 0.5 0.8 0.8 0.8 0.6 0.8

R2 kΩ 5 107 377 22 0.3 0.09

Q2 Y o (μMho) 133

n

W Y o (μMho)

0.6 134 92

240 7640 3100

0.5 0.4 0.6

χ² 0.03 0.002 0.001 0.04 0.04 0.05

The response for the Ag/AgCl sensors in chloride-free environment is initially modelled by resistors and pseudo-capacitances only, whereas for later immersion stages, a W impedance is introduced in order to account for diffusion/mass transport phenomena. Upon changing the environment to Cl-containing such, the electrochemical reaction of AgCl layer formation is the dominating one (Ag/AgCl in 125mM, Fig.5) i.e. W impedance is not observed, accompanied by drop in charge transfer/layer resistance and increase of pseudo capacitance values (Fig.5,6,

546


Table 1). Comparing the response of the Ag/AgCl sensors in SPS + 1M NaCl in the high frequency domain to the one of Ag in 1M NaCl (Fig.6), both with established OCP at -6 mV, the following is relevant: clearly the contribution of previously formed AgCl, subsequent transformation of AgCl to Ag2O and finally recovering of AgCl upon re-immersion in chloride-containing environment is reflected by more than 1 time constant in the response for Ag/AgCl sensors in the low frequency domain. However, the preliminary simulation/fit procedure, reveal similar capacitance and resistance values for the electrochemical reaction (at Ag/AgCl interface) and variation in these for the chemical phenomena (at AgCl/solution interface), which is consistent with the initiation of AgCl layer formation for the case of Ag rod only and transformation at interfaces for the case of Ag/AgCl sensors. The next step of this investigation, involving microscopic surface analysis and final EIS fit and simulation is expected to justify and visualise the observed phenomena on the Ag/AgCl sensors’ interface upon immersing in altered environment and thus justify the limitations of sensors’ stability with respect morphology and transformations in the AgCl layer.

4 Conclusion The following conclusions can be drawn from the present preliminary investigation: - In alkaline chloride-free solution, the potential of the silver/silver chloride sensors changes over time, denoted to the formation of silver oxide; - The more rapid change in sensors’ potential in SPS with 1000 mM chloride can be attributed to the enhanced formation of silver chloride (instead of silver oxide) on the surface of the sensors when chloride ions are present. In contrast, within lower chloride concentrations (as the 125 mM solution), longer stabilisation time is required (i.e. the reaction of AgCl formation is impeded compared to 1000 mM) - The impedance measurements show a pronounced difference in electrochemical response for the differently conditioned sensors, which is also with respect to the relevant environment. For chloride-free solutions, the magnitude of impedance increases with time of conditioning, denoted to the plausible formation of Ag2O within the AgCl layer, blocking pores and exerting microstructural changes in the AgCl layer itself. Additionally, in Cl-free environment, the Faradaic reaction associated with AgCl layer formation is not taking place – all these result in increased impedance values with time of conditioning. - In chloride-containing solutions, the EIS response shows significantly lower magnitude of impedance which corresponds well to the chloride concentration i.e. the lowest |Z| was recorded for the 1000 mM solution.

5 References [1] Angst U, Larsen CK., Vennesland O and Elsener B, (2009) Monitoring the Cl concentration in concrete pore solution by means of direct potentiometry, Proc. Intern.Conference on Concrete Solutions, Padua, Italy. [2] Elsener B, Zimmermann L and Bohni H, (2003) Non-destructive determination of the free chloride content in cement based materials, Materials and Corrosion, 54: 440–446. [3] Atkins CP, Carter MA and Scantlebury JD, (2001) Sources of error in using silver/silver chloride electrodes to monitor chloride activity in concrete, Cement and Concrete Research, 31:1207-1211. [4] Montemor MF, Alves JH, Simoes AM, Fernandes JCS, Lourenco Z, Costa AJS, Appleton AJ and Ferreira MGS, (2006) Multiprobe chloride sensor for in situ monitoring of reinforced concrete structures, Cement and Concrete Composites, 28:233-236. [5] Du RG, Hu RG, Huang RS and Lin CJ, (2006) In Situ Measurement of Cl- Concentrations and pH at the Reinforcing Steel/Concrete Interface by Combination Sensors, Analytical Chemistry, 78:3179-3185. [6] Svegl F, Kalcher K, Grosse-Eschedor YJ, Balonis M and Bobrowski A, (2006) Detection of Chlorides in Pore Water of Cement Based Materials by Potentiometric Sensors, journal of the Rare Metal Materials and Engineering, 35:232-237. [7] Angst U, Elsener B, Larsen CK and Vennesland O, (2010) Potentiometric determination of the chloride ion activity in cement based materials, Journal of Applied Electrochemistry, 40:561–573. [8] Pargar F., Koleva D.A., Copuroglu O, Koenders E.A.B. and van Breugel K, (2014) Evaluation of Ag/AgCl sensors for in situ monitoring of free chloride concentration in reinforced concrete structures. Young Researchers’ Forum II: Construction Materials, University College London, London.

547


Analysis and Simulation of the Performance of a Sensing Array for Gear Tooth Cracks Christos Spitas1*, Vasilios Spitas2, Andreas Nikolakakis2 (1) Delft University of Technology, Delft, The Netherlands (2) National Technical University of Athens, Athens, Greece Abstract: This work proposes a non-invasive measurement method and sensing array for detecting at high resolution developing cracks in gear teeth. The sensing array comprises a set of electrodes placed on the fatigue-crack prone areas of the gears, which by way of measuring electrical potential differences produced from the application of direct current, provide direct information on the length of a developing crack in real-time. The optimal configuration of the sensing electrodes on gear teeth depending on their geometry and developed stress field is determined and the suitability of the technique for real time crack monitoring of such systems is discussed. Keywords: gear tooth cracks, real-time crack monitoring, sensing array, electrical potential difference method, simulation

1 Introduction In-situ detection and monitoring of fatigue cracks in mission-critical gears is currently based on indirect calculations resulting from vibration spectral analysis of the power train. The assessment performed through this method yields highly inconclusive and ambiguous results mainly due to the complexity of the structure and the constantly changing dynamic performance of the gearbox due to ageing. Furthermore, the deconvolution and post-processing of the raw data usually require cumbersome algorithms and dedicated software, which, apart from their inherent inaccuracy, involve lengthy processing, therefore ruling out the possibility to use in real time. In this paper we propose to use the electrical potential difference method to address these shortcomings and non-invasively detect and monitor the propagation of cracks in gear teeth in real time. For this purpose we use an accurate model for the gear tooth geometry and mechanical load conditions, on the basis of which we predict the crack path, using the criterion that the crack follows the theoretical stress-gradient trajectories (i.e. vertical to the τ-max iso-lines), which is generally valid for short enough cracks. The electrical field resulting from applying a current via a pair of electrodes in a cracked specimen is then obtained by electrostatic FEA to determine sensitivity and optimal placement of a defined sensing array of electrodes. The FEA calculations consider single tooth models for efficiency, a practice that is widely used and accepted in the literature [1-2]. The electrical potential difference method has gained increasingly wide acceptance in fracture research as one of the most accurate and efficient methods for monitoring the initiation and propagation of cracks [3]. This technique relies on the fact that the propagation of a crack results in an increase of the electric resistance. It simply involves the application of a constant current through a cracked specimen or structure in such a way that a change in crack length alters the potential difference between suitably located electrodes in the vicinity of the crack. [4]. A method based on the potential difference technique, which allows the real-time measurement of fatigue crack growth on a shear specimen at high temperature, has already been investigated [5-6]. Also, a technique based on the electrical potential difference method for the real-time assessment of both the length and the direction of inclined cracks on a specially designed shear specimen propagating in Mode II and in mixed Mode I, II has been presented [7-8]. Nonetheless, this is the first study, in literature, where the potential difference method is applied on spur gear teeth.

*

Landbergstraat 15, 2628CE Delft, The Netherlands; [email protected]

548


2 Modelling of spur gear teeth A spur gear tooth geometry with the following characteristics was modelled: Profile shift (cm) = 0, Tooth thickness Coeff. (Cs)= 0.5, Pressure angle α = 20° and number of teeth z = 20. These parameters were fed into an in-house developed computer programme (MESHGEN.EXE), which produced the 2D geometry of each spur gear tooth. Following this step, the 2D geometry was imported as a sequence of curve coordinates in Solidworks and the 2D geometry of the teeth was converted to 3D by adding the same parameters for all cases: facewidth b = 1mm and module m = 1. The next step was to define the Highest Point of Single Tooth Contact (HPSTC) of each gear. It is known [7] that the position of the HPSTC of a gear depends only on its geometry and on the contact ratio of the pair and given by the formula: rHPSTC=

rk 12 + ( ε − 1 ) tg − 2 rk 12 − rg12

(1)

where: r k = tip radius , r g = involute base t g = base pitch, ε = contact ratio The designed spur gear tooth models were modelled as steel specimens with Young’s modulus E= 210 GPa, Poison’s ratio ν= 0.28. The contact ratio (ε) was chosen to be 1.2.

3 Location of crack initiation and predicted path

In order to assess the stress-state in the above spur gear tooth, Finite Element Analysis was carried out using Ansys Multiphysics. The mesh for the tooth model was created automatically with a fine density. A unitary load was applied on the working face of the gear tooth at the HPSTC, to represent the most severe load case. The Static Analysis results were used to calculate the crack path, assuming that crack initiation on the tooth surface happens at the point of the maximum principal stress (the second principal stress being 0 at the same point, this is also the point of maximum shear stress) and that the crack will propagate along the gradient of the maximum shear stress field. As such, nominally the crack will be perpendicular to the surface. The non-cracked stress field and the projected crack path are shown in Fig.1.

Figure 1 Calculated maximum shear stress distribution in non-cracked gear tooth model and prediction of the crack path

549


4 Potential difference sensing array layout and method On the surface of the gear tooth model four sensing arrays each comprising two current source-sink electrodes and two sensing electrodes, as seen in Fig. 2, were positioned corresponding to crack lengths of 10%, 20%, 30% and 40% respectively, where a length of 100% corresponds to a crack extending up to the centre line of the tooth. The electrical current was applied in turn between each pair of source-sink electrodes and respectively the potential difference was measured between the corresponding sensing electrodes. For the execution of measurements a constant current passes through the source-sink electrodes, creating a distribution of electrical potential over the surface of the gauge area of the model (Fig. 3). The potential difference between the sensing electrodes is constantly measured and changes as the crack propagates. Considering steel material, the specific resistivity was considered uniformly to be: ρ= 1.43×10−7 Ωm at 20 °C.

tooth working surface current source/ sink

predicted crack path

sensing array

y x

crack initiation

potential sensors

x field discontinuity (boundaries)

tooth root current flow line

real crack path

y

current sink/ source

Figure 2 Half-tooth model showing the placement of the sensing array in the predicted-crack-bound coordinate system (left) and explanation of working principle (right)

Ten different crack lengths ranging from 10% to 100% were investigated using FEA. The electrical current applied in the simulation was 5 mA. From the above results a relationship between crack length and normalised potential (i.e. V/Vo where Vo is the potential of the non-cracked specimen) was obtained and plotted in Fig. 4.

550


Figure 3 Equipotential contours (left) and equal current density vectors (right) for one location of the sensing array (10%) and crack length (70%).


Optimal longitudinal position of the sensing array

The graph of non-dimensional potential (Fig. 4) shows that as the electrodes’ position approaches the crack’s initiation point, the value of the non-dimensional potential - namely the sensitivity of the gear crack tooth sensor- increases. Therefore, it is initially concluded, that the optimum position of the electrodes is as close as possible to the crack’s initiation point.

Figure 4 Normalised potential versus crack length for sensing array positions between 10% and 40%

In order to be sure, that the optimum position of the electrodes is as close as possible to the crack’s initiation, electric field analysis was performed for six more sensing arrays located between 10% and 0% (exactly at the tooth surface). Particularly, the couples of electrodes were placed at 0.04, 1.5, 3, 6, 8 and 9% of the total crack. For each couple the analysis was performed for the ten designed crack lengths and the results are shown in Fig. 5.

551


In contrast to the trend in Fig. 4, Fig. 5 shows that between the position of 10% and 0%, as the sensing array approaches the position of 0%, the non-dimensional potential decreases significantly. The main reason for this is that the Potential Vo of the non-cracked specimens increases from the 10% to 0% exponentially, considering that when the electrical current is applied closer to the initiation of the crack, the current vectors tend to move only in the one side of the specimens’ surface. Conversely, the fluctuation of the electric potential Vo between positions 10% and 40% is almost imperceptible. Therefore, given that the optimum position for the electrodes is where the electric sensor perceives the highest value of non-dimensional potential, it is demonstrated that the position of 10% is the optimum.

Figure 5 Normalised potential versus crack length for sensing array positions between 0 and 10%

5.2

Optimal distance between the sensing electrodes

Initially, electrostatic FEA is performed on the non-cracked spur gear tooth and the results show that as the distance between sensing electrodes increases, the electric potential Vo is approximated by a linear increasing function, as shown in Fig. 6. This fact can be explained taking into account, that as the distance between electrodes increases, the imposed current must travel farther and hence the electrical potential difference Vo increases. Then, electrostatic analysis is performed on the cracked spur gear models for crack lengths ranging up to 100%, applying constant current on ten couples of electrodes, which have different distance between them and they are placed at the longitudinal position 10%. The results are shown in Fig. 7. It can be seen that the non-dimensional potential difference (and correspondingly the sensitivity of the measurement) decreases as the distance between the electrodes increases. At the same time, at large electrode distances their relationship is nearly linear, but at small distances it is clearly non-linear. The best positioning for the sensing electrodes is where the measurement sensitivity will be the highest. Thus, as per Fig. 7, their distance should be as small as possible. This is limited by the uncertainty in predicting the exact location of crack initiation, which gives rise to the concept of a ‘crack initiation zone’ around the theoretical position of maximum principal/ shear surface stress. The size of this zone will be dictated by variations in material properties and the presence of crackinitiating defects, variations in the tooth contact geometry that may shift the HPSTC etc Thus the sensing electrodes should ideally be placed at the limits of the crack initiation zone.

552


Figure 6 Reference potential difference of non-cracked specimen versus distance of the sensing electrodes

Figure 7 Non-dimensional potential difference versus crack length for different electrode distances

5.3

Real-time condition monitoring: Estimator function for crack length

From the simulation results in sections 5.1-5.2 it is possible to construct statistical regressions linking the measured non-dimensional potential difference to the non-dimensional crack length, allowing to infer the latter from the former reliably in real-time. The basic equation is: (2) = r αx + c The functions α ( d ) and c ( d ) can be found from the functions of linear interpolation: α ( d ) = 0.0411d −0.862

(3)

(4) c (d ) = −1.7455ln d + 2.165 where r is the ratio of the potential difference of cracked specimen to potential difference of the noncracked specimen, x is the non-dimensional crack length, and d is the non-dimensional distance between electrodes. Therefore by substituting Eqs. (3)-(4) into Eq. (2) we obtain:

553


r + 1.7455ln d − 2.165 (5) 0.0411d −0.862 The above mathematical formula illustrates the purpose of the designed electric sensor: the length (%) of a gear root crack can be assessed (real-time), given the sensor’s measurements of the nondimensional potential. x (r ,d ) =

6 Conclusions This paper presented a method and sensing array for measuring developing cracks in gear teeth in real time. By measuring the potential difference between two sensing electrodes it is possible to correlate it with the progress of a crack. Form the electrostatic analysis it is concluded that: - The non-dimensional potential increases as the crack propagates, due to the fact that the electrical resistance on the surface of the cracked gear tooth grows with the crack’s propagation. -

Between the position of 0 and 10% the Potential Vo of the uncracked specimens decreases exponentially. This occurs, because the current vectors tend to move only in the one side of the specimens’ surface when the electrical current is applied increasingly close to the initiation of the crack.

-

As the electrode couple’s position approaches the crack’s initiation point, the value of the nondimensional potential increases, namely the gear crack tooth sensor becomes more accurate. Nonetheless, this phenomenon appears until the position of 10%. Between positions 0 and 10% the non-dimensional potential decreases significantly, because of the exponential increase of the potential Vo.

-

As the distance between electrodes increases, the electric potential Vo is approximated by a linear increasing interpolation.

-

Considering that the optimum position for the electrodes is where the electric sensor perceives the highest value of non-dimensional potential, and given the aforementioned findings, it is concluded that the position of 10% is the optimum longitudinal position for the couple of electrodes and the distance between the electrodes must be the same size as the width of the crack initiation zone.

-

Furthermore, a mathematical formula was found, which allows the real-time assessment of the length of a crack at the gear root given the sensor’s measurements of the non-dimensional potential.

7 References [1] Pimsarn, M., Kazerounian K., Efficient evaluation of spur gear tooth mesh load using pseudointerference stiffness estimation method, Mech. Mach. Th. 37, 769–786, 2002 [2] Sirichai, S., Torsional properties of spur gears in mesh using nonlinear finite element analysis, PhD, Curtin: University of Technology, 1999 [3] Saxena, C.L. Muhlstein., Fatigue crack growth testing, in : H. Kuhn, D. Medlin (Eds.), Mechanical Testing and Evaluation, ASM Handbook, vol. 8, ASM International, Materials Park, Ohio, 2000. [4] R. Ghajarieh, M. Saka, H. Abe, I. Komura, H. Sakamoto, Simplified NDE of multiple cracks by means of the potential drop technique, NDT & E International, Vol. 28, No. 1, 23-28, 1995. [5] V. Spitas, C. Spitas, P. Michelis, Real-time measurement of shear fatigue crack propagation at high-temperature using the potential drop technique, Measurement 41, 424-432, 2008. [6] V. Spitas, C. Spitas, P. Michelis, A three-point electrical potential difference method for in situ monitoring of propagating mixed-mode cracks at high temperature, Measurement 43, 950-959, 2010.

554


[7] P. Michelis, Specimen Geometric Configuration for Uniform Shear Distribution During Testing, European Patent EP 0687899, 1997.

555


Fundamental Research on the Carbonation Control Effect by Coating Materials M. Sugiyama1* (1) Hokkai Gakuen University, Sapporo, Japan

Abstract: This research applied three kinds of coating materials to the concrete and examined the carbonation performance. Three kinds of coating materials are silicate system surface impregnation material, the silicon system surface impregnation material, and polymer cement. As a result, the covering material of polymer cement had the largest carbonation control performance. Keywords: Concrete, Carbonation, Coating

1 Introduction

When considering degradation of the reinforced concrete structure in the architecture field, neutralization is an important examination matter. Carbonation is a phenomenon in which the alkali performance of concrete is lost with carbon dioxide in the atmosphere. This research observed the performance of three kinds of surface protection materials. In this research, concrete imagined housing basic concrete. The purpose of this research is to find out the coat material with the largest durability for housing basic concretes. The experiment made the concrete test specimens which coated three kinds of surface protection materials. Three kinds of coating materials are silicate system surface impregnation material, the silicon system surface impregnation material, and polymer cement. This polymer is marketed in Japan and used for usual. The main ingredients of polymer are cement powder and an acrylic resin. Therefore, this polymer is not SBR. These test specimens compared the neutralization inhibition effect. This paper reported these test results.

2 Test plan and method 2.1

Experimental design

Table 1 shows the planning of experiment and Figure 1 shows the outline of experiments. The surface protection material used for the test is shown in Table 2. Acceleration neutralization was carried out about the test specimens which coated three kinds of surface protection material. A test plan is shown in Table 1. A test specimens are 100 x 100 x 400 mm. Silicate system surface penetration material elaborates a surface by an unreacted cement constituent, and a reaction and crystallization, and aims at the interception effect of degradation factors, such as water and carbon dioxide. The silicon system surface osmosis material forms a water absorption prevention layer in a surface part. And polymer cement system covering material aims at the interception effect of the degradation factor by covering the surface.

*

Prof. Faculty of Engineering,

[email protected]

556

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands Table 1 Planning of experiment

Mark.

The kind of surface protection material.

Acceleration neutralization test. Duration of test.

Shape of a test specimens.

1

With no treatment.

4 weeks,8 weeks, 12 weeks, 25 weeks.

Three 10*10*40-cm rectangular parallelepipeds.

2

Silicate system.



3

Silang system.



4

Polymer cement system.



Figure 1 Outline of experiments Table 2 Summary of surface protection materials Mark

2

3

4

Type

Silicate system.

Silicon system.

Polymer cement system.

Main components

Silicate sodium. Water glass system.

Silicon system compound.

The mixture of cement powder and a water-soluble acrylic resin.

Coating.

The first amount of recommendation is coated (0.15 kg/m2 of diluted solutions). + Curing by water spraying. + The second amount of recommendation is coated (0.15 kg/m2 of diluted solutions). + Curing by water spraying.

The first amount of recommendation is coated (0.3-0.4 kg/m2 of diluted solutions). + Dry curing. + The second amount of recommendation is coated (0.3-0.4 kg/m2 of diluted solutions.).

The 1st lower coating material (0.9 kg/m2) is coated.

557

+ Dry curing. + The 2nd glazing material (0.3 kg/m2) is coated.


2.2

Concrete test specimens

A test specimens is a prism with a form of 10x10x40 cm. The removal forme ware done on the day following placing. After doing underwater curing for six days, dryness among the air for three weeks was done. The conditions of air are the temperature of 20 ℃, and 60% of humidity. The concrete of the test specimens used W/C55%, the strength 24, the slump 18, and ordinary portland cement which are generally well used for housing basic concrete.

Figure 2 Accelerated carbonation test specimen, 10x10x40 cm

Figure 3 Location of carbonation depth Table 3 Mix proportion (kg/m3)

W/C

Cement

Water

Sand 1

Sand 2

Gravel

Admixture

55%

302

166

419

421

1009

0.755

Table 4 Quality of fresh concrete

558


Slump

Air

Temperature of placing concrete

Unit volume mass.

16.5 (cm)

3.6 (%)

26 (degree)

2317 (kg/m3)

2.3

Acceleration carbonation test method

The acceleration neutralization test method followed the measuring method of the neutralization depth of the concrete of JIS A 1153. The removal form of the concrete test specimens was carried out on material age the 1st, and it coated surface protection material to the test specimens which finished curing underwater (for six days), and in mind (three weeks). Acceleration neutralization was started after curing for two weeks (six weeks of material ages). Measurement of neutralization split the test specimens put in in the test tub, sprayed phenolphthalein liquid on the surface, and measured the neutralization depth. After this, the test specimens was split for every predetermined acceleration neutralization period, and it measured similarly. In addition, the inside of an acceleration neutralization test tub was set as the temperature of 20 degree , 60% of relative humidity, and 5% of carbon dioxide levels. The outline of a test specimens is shown in Fig. 2. The acceleration neutralization test tub used for the experiment is shown in the photograph 1.

3 Results and Considerations A proportioning table is shown in Table 3. The test result of fresh concrete is shown in Table 4. Under the condition of this time enforcement, the following things have been checked as a result of the test. On the basis of an unprocessed test specimens, the carbonation inhibition effect by each protective layer is considered.

Figure 3 Result of accelerated carbonation test, 25 weeks

559


3.1

Silicate system surface penetration material

The case of the rate of the carbonation depth to no treating where the period of acceleration carbonation was four weeks was 98%. Similarly, it was 96% when the period of acceleration carbonation was eight weeks. Similarly, the case where the period of acceleration carbonation was 12 weeks was 103%. Similarly, the case where the period of acceleration carbonation was 25 weeks was 101%.

3.2

Silicon system surface penetration material

The Silicon system surface penetration material was about 80% of neutralization compared with no treating. The case of the rate of the neutralization depth to no treating where the period of acceleration carbonation was four weeks was 70%. Similarly, the case where the period of acceleration carbonation was eight weeks was 85%. Similarly, the case where the period of acceleration carbonation was 12 weeks was 84%. Similarly, the case where the period of acceleration carbonation was 25 weeks was 86%.

560


3.3

Polymer cement system surface covering material

The protective layer with the highest (an effect is large) carbonation inhibition effect is polymer cement system surface covering material. The case of the rate of the carbonation depth to no treating where the period of acceleration carbonation was four weeks was 15%. Similarly, the case where the period of acceleration carbonation was eight weeks was 52%. Similarly, the case where the period of acceleration carbonation was 12 weeks was 43%. Similarly, the case where the period of acceleration carbonation was 25 weeks was 48%.

4 Conclusions A conclusion is the following as a result of this test. (1) The surface protection material with the highest (an effect is large) effect of carbonation inhibition is polymer cement system surface covering material. (2) The surface osmosis material with the next high carbonation control effect is a silicon system. (3) The surface penetration material with the smallest effect of carbonation inhibition is silicate system surface osmosis material.

5 References Conference proceedings [1] Masashi SUGIYAMA (2012) Fundamental research on the freezing thawing resistance of the concrete which delay-added the drying shrinkage reducing agent. Proceedings of the 2nd International Conference on Microstructure Related Durability of Cementitious Composites, Amsterdam 11-13 April 2012, TU-Delft [2] Masashi SUGIYAMA (2012) Fundamental research on the freezing thawing resistance of the concrete which delay-added the drying shrinkage reducing agent. Proceedings of the 10th International Conference on Superplasticizers and Other Chemical Admixtures, Prague 28-31 October 2012, CANMET/ACI, ISBN 978-0-9916737-2-8 [3] Masashi SUGIYAMA (2008) Freeze-thaw Test Results of Porous Concrete with Crushed Scallop Shell Material Added. Proceedings of the STRUCTURAL FAULTS + REPAIR-2008, Edinburgh 10-12 Juner 2008

561


DC current-induced curing and ageing phenomena in cement-based materials A. Susanto*, D.A. Koleva and K. van Breugel Faculty of Civil Engineering and Geosciences, Section of Materials and Environment, Delft University of Technology , Stevinweg 1, 2628 CN Delft, The Netherlands Abstract: This paper investigates DC current-induced “curing” and ageing phenomena in cement-based materials. Two current densities were used in a DC current regime i.e. mortar cubes were subjected to DC current flow of 1 A/m2 and 100 mA/m2; tap water and calcium hydroxide were external environment. Conditioning (“curing”) was performed for 112 days, during which period, compressive strength, porosity and electrical resistivity of the mortar specimens were monitored. Based on the experimental results, DC “curing” in the tap water tends to accelerate degradation processes in the mortar specimens i.e. results in increased ageing. Meanwhile, DC “curing” in calcium hydroxide solution has the potential to accelerate “curing” without negative side effects. Keywords: DC curing, ageing, water, calcium hydroxide, cement-based materials

1

Introduction

Ageing can be referred to as the time-sequential deterioration that occurs in all biological systems, materials and structures. In cement-based materials, ageing is understood to include reduction in mechanical properties (i.e. compressive strength, tensile strength, elastic modulus) and increased permeability. The ageing process in cement-based materials is not only affected by internal factors (i.e. degree of cement hydration and relevant kinetics) but also by external factors that contribute to the degradation process such as chemical attack, temperature/fire, aggressive environment. All these factors have the potential to accelerate ageing/degradation processes of cement-based materials. For instance, when cement-based materials are exposed to high temperature (>200°C), mechanical damage and chemical transformation occur simultaneously, which will change the materials’ microstructure. Chemical transformation can be attributed to dehydration of calcium silicate hydrate (CSH) and calcium hydroxide (Ca(OH) 2 ) that cause micro-cracking and irreversible changes in the microstructure [1]. Other factors that contribute to degradation and accelerated ageing of cement-based materials are chemical attacks such as carbonation, external sulphate attack, and chloride ions penetration. Various methods and techniques are used to reduce/prevent degradation of cement-based materials in order to extend service life, e.g. coatings, sealers, more durable and resistant cement-based mixtures, etc. Electrochemical methods (i.e. cathodic protection, realkalisation, electrochemical chloride extraction) are generally applied for steel corrosion protection in reinforced concrete structures, but these (although rarely investigated with this respect) would also affect microstructural properties as a result of the electrical current involved. Among all available methods, accelerated curing is also performed, since if properly executed accelerated curing is not only with beneficial effect on extended service life but also leads to reduced lifecycle cost [2]. ________________________________________ * Corresponding author: [email protected]

562


Several accelerated curing methods of concrete have been reported to achieve rapid gain of compressive strength at an early age due to temperature rise during the curing process [3,4]. Saul developed the maturity concept: “Concretes of the same mix at the same maturity have approximately the same strength whatever combination of temperature and age goes to make up that maturity” [5]. Various methods of accelerated curing of concrete have been used including steam curing, autoclaving (high pressure steam curing), microwave heating, hot water, hot air heating, infrared heating, and electrical curing [6-9]. This study is part of a larger investigation on the negative and possibly positive effects of electrical (stray) current on cement-based microstructure, where positive and/or negative effects would depend on current density and environment. In other words, possibility exists that DC (including stray current) can be beneficial at certain levels. This paper is focusing on the possible “curing” effects of electrical current flow i.e. electrical curing (e.g. direct current (DC) curing). In DC curing, electrical current is passed directly through the cement-based systems to produce joule heating effect, thus increasing the initial rate of cement hydration. Bredenkamp [6] stated that DC curing is one of the most energy efficient methods for accelerated curing of concrete and after the initial capital outlay for equipment, the running costs of direct electric curing are substantially lower than that of externally applied heat curing (steam, autoclave, etc). The energy cost of direct (DC) electrical curing with field strength between 300 and 500 V/m and rate of energy input 76kWh/m3 is about 5% of the total cost for a m3 of concrete. Other investigators mentioned that energy consumed per m3 of concrete are 30-40 (kWh/m3) [10] and 50-60 (kWh/m3) [11], respectively. As a comparison, this study investigated DC curing with regimes of 100mA/m2 and 1A/m2. The electric field strengths used were 4 V/m for the 100mA/m2 and 40 V/m for the 1A/m2 regimes respectively. The energy consumed per m3 of concrete specimen are thus 7.55 kWh/m3 and 75,5 kWh/m3, respectively during DC curing for 112 days. Although the benefits of DC curing in cement-based materials have been investigated, there is still limited information about the effect and influence of the external environment during the DC curing. This paper presents results for the effect of DC current flow as a possibility of DC curing for cement-based materials in water and calcium hydroxide solution as external environments, discussing mechanical and microstructural properties and thus differentiating “curing” phenomena from “ageing” phenomena when electrical current flow is involved.

2

Experimental materials and methods

2.1 Materials Mortar cubes of 40 mm×40 mm×40 mm (Fig.1) were cast, using OPC CEM I 42.5N with water-tocement ratio of 0.5 and cement-to-sand ratio of 1:3. The chemical composition (in wt. %) of CEM I42.5N (ENCI, NL) is as follows: 63.9% CaO; 20.6% SiO 2 ; 5.01% Al 2 O 3 ; 3.25% Fe 2 O 3 ; 2.68% SO 3 ; 0.65% K 2 O; 0.3% Na 2 O. After casting and prior to conditioning, the specimens were cured in a fog-room of 98% RH, 20°C for 24 hours; after de-moulding they were positioned in the containers. 2.2 Sample designation

Three groups of specimens were investigated: 1) group control - no DC current involved; 2) group 100 mA/m2 and 3) group 1A/m2, where DC current “curing” was relevant at the respective current levels (the mortar specimens were subjected to DC current flow, Fig.1). All groups were submerged in water or (Ca(OH) 2 ) solution as external environment. 2.3 Current regime

Figure 1 shows a schematic presentation of the experimental set up for DC curing using current density of 100 mA/m2 and 1 A/m2. The experimental set up is as previously used and reported [12]. The current density was adjusted by additional resistors (R 1 =2700 ohm and R 2 =270 ohm) 563


in the electrical circuit. The mortar cubes were completely submerged. Tap water and calcium hydroxide solution were used as external medium. The electrical current was applied from 24h hydration age (immediately after de-moulding of the specimens) and until 112 days of age.

a)

mortar

Tap water/Ca(OH)2

_

+ I (A)

27 cm

44 cm Electrodes and cables to apply current b)

c) R

Height of water solution=5 cm Height of mortar cubes=4 cm

12 V

A 27 cm

Current injected in the surface area A R1=2700 Ω

R2=270 Ω

44 cm

Figure 1 (a) Experimental set-up for DC curing (b&c) schematic experimental set up [11]

2.4 Methods 2.3.1. Standard compressive strength

Standard compressive strength tests were performed on 40×40×40 mm mortar cubes at the hydration ages of 3, 14, 28, 84 and 112 days. Three replicate mortar specimens were taken out from the conditioning set-up, cloth-dried and tested within a 30 min time interval.

2.3.2. Mercury intrusion porosimetry (MIP)

The sample preparation for MIP tests followed generally accepted procedures [14, 15]. The MIP tests were carried out by using Micrometritics Poresizer 9320 (with a maximum pressure of 207 MPa) to determine the porosity and the pore size distribution of the specimens.

2.3.3. Mortar electrical resistivity

Electrical resistivity of mortar was measured using an AC “2-pin method”, where the “pins” are metal plates with dimensions equal to the sides of the mortar cubes [12]. Resistance of the mortar specimens is measured by applying an alternating current of 1mA at a frequency of 1kHz. R-meter was used to record the resistance of the mortar. For the “under current” regime (groups 100 mA/m2 and 1A/m2 i.e. DC current “curing”), the resistance measurements were performed after current interruption of approx. 30 min and surface drying of the cubes. Electrical resistivity 564


was calculated using Ohm’s Law: ρ=R.A/l, where ρ is the resistivity in Ohm.m, R is the resistance in Ohm, A is the cross-section of the mortar cube in m2, and l is the length in m.

3

3.1

Results and discussions Compressive strength

Compressive strength is considered as a key property of concrete. It provides a general indication of concrete quality. The strength of a material is defined as the ability to resist stress without failure. Failure is identified with the appearance of cracks. There are several factors that affect strength of cement-based materials such as water/cement ratio, cement type, mixing water, admixtures and curing conditions including humidity, temperature and ageing (time) [16, 17]. The present study is focusing on the influence of DC curing and possible ageing on cementbased materials. DC current flow induces temperature elevation in the cement-based materials due to the Joule heating effect, which leads to the potential for altered mechanical properties. Figure 2 presents the evolution of compressive strength as a function of hydration age for mortar specimens submerged in water and calcium hydroxide solution in control and and DC curing-conditions. Generally, compressive strength increases with cement hydration and age.

Figure 2 Compressive strength as a function of hydration age for mortar submerged in (a) water and (b) calcium hydroxide solution

Fig.2a) depicts the results for water environment, showing that DC curing leads to compressive strength decrease for the mortar specimens after 112 days of cement hydration, compared to control specimens. It can be deduced that DC curing in water environment accelerates ion migration in the bulk matrix and hydration rate of the mortar specimens, subsequently leading to increased calcium leaching and decreased strength. The influence of current density at the level of 1A/m2 is more pronounced after 112 days, compared to the effects of current density at 100mA/m2. In other words, the ageing phenomena in terms of reduced mechanical properties in this case, depend on the level and intensity of applied current density. In contrast, Figure 2b) shows an opposite trend of compressive strength development when Ca(OH) 2 was employed as external environment. In this case, DC curing improves the mechanical properties of the mortar specimens, the most plausible mechanism being related to avoidance of calcium leaching in this environment, accompanied by enhanced ion/water migration and altered cement hydration. As a result, the compressive strength of mortar specimens under both current regimes of 1A/m2 and 100 mA/m2 end-up slightly higher than the control cases.

565


3.2

Microstructural properties

Figure 3 reveals porosity development as a function of hydration age for mortar specimens submerged in (a) water and (b) calcium hydroxide solution. As shown in Figure 3a, DC curing contributes to coarsening of the mortar specimens especially after 28 days of age. As expected, the total porosity for specimens conditioned in the regime of 1A/m2 maintained higher values compared to current density of 100mA/m2 and control specimens due to accelerated calcium ions leaching. In contrast, for DC curing in calcium hydroxide solution, the total porosity of mortar specimens under both DC current regimes remarkably decreases with conditioning. In this case, the total porosity decrease is larger when higher current density was applied. The pore structure development is in line with the compressive strength results and shows that in water environment the effect of DC current (DC curing respectively) can accelerate ageing phenomena, whereas in Ca(OH) 2 environment, the “curing” (i.e. positive) effects are prevailing. b) 15

a) 22

100mA/m2 1A/m2

porosity (%)

porosity (%)

Control

18

14

13

Control

11

100mA/m2 1A/m2

10

9

3d

14d

28d

14d

112d

Hydration ages

28d

84d Hydration ages

112d

Figure 3 Porosity as a function of hydration age for mortar submerged in (a) water and (b) Ca(OH) 2 solution

3.3

Electrical resistivity

Electrical resistivity measurements are generally used as a non-destructive technique to assess concrete durability [18]. The concrete pore water is essentially an electrolyte, containing mostly K+, Na+, Ca++, and OH– ions [19–21]. Additional ions, such as Cl-, can also be present as a result of exposure to various external sources (such as seawater and de-icing salts), increasing the concentration of ions in the pore water and, if minimal chloride binding is at hand, presumably reducing the electrolytic resistance of the pore water. There are several factors that affect resistivity such as paste volume, concrete composition, water–cement ratio, curing temperature, chlorides and moisture content [21, 22]. DC curing promotes temperature increase in the cement-based systems due to the Joule heating effect that contributes to accelerated cement hydration. Accelerating cement hydration influences resistivity development of cement-based materials. Figure 4 reveals the electrical resistivity development of mortar specimens, subjected to DC curing when submerged in water and calcium hydroxide solution. As shown in the Figure 4a, the electrical resistivity of mortar increases gradually with the progress of cement hydration. The electrical resistivity of mortar, subjected to current density of 1A/m2 is slightly higher compared with 100mA/m2 and control specimens until 21 days, denoted to enhanced cement hydration in this regime, compared to control and lower current density regimes. However, at later stages, the electrical resistivity of mortar under 100 mA/m2 and 1 A/m2 regimes maintained lower values compared to the control cases, attributed to the competing effects of enhanced ion/water transport (accelerated cement hydration) and calcium leaching. The former effects are initially pronounced, whereas the latter are relevant for later stages and determine the lowest electrical resistivity values, recorded in the 1 A/m2 regime. Calcium leaching contributes to microstructural alterations, including coarsening of the pore network, which 566


results in the observed variation of electrical properties of the bulk matrix (electrical resistivity of the mortar was measured in unsaturated condition).

Figure 4 Electrical resistivity evolution of mortar submerged in (a) water and (b)Ca(OH) 2

Figure 4b shows that the electrical resistivity of mortar specimens under DC curing in Ca(OH) 2 tends to increase gradually and maintained higher after 14 days, compared to the control specimens. This environment prevents or minimises calcium leaching, therefore a pronounced effect in this case is enhanced ion/water migration and accelerated cement hydration within DC current flow, resulting in increase in electrical resistivity of the bulk matrix. In contrast to accelerated aging phenomena in water, the DC “curing” effects, as positive such, are only observed for the specimens in Ca(OH) 2 environment.

4

Conclussions

This paper discussed DC current-induced “curing” and ageing phenomena in cement-based materials. Based on the experimental results, several conclusions can be drawn: a. DC current induced-curing in tap water tends to accelerate degradation process in the mortar specimens and is therefore responsible for enhanced ageing of the cement-based matrix in terms of reduced compressive strength, coarser pore structure and reduced electrical resistivity. b. DC current induced-curing (at the hereby investigated levels) in calcium hydroxide solution has the potential to accelerate curing and would result in a good quality concrete. c. Further investigation is needed to determine the threshold of positive and/or negative influence of DC current on the properties of cement-based materials. Curing (positive) or ageing (degradation) phenomena when DC current is involved need to be studied also with respect to the composition of the external environment and/or for sealed conditions.

5

Acknowledgements

The financial support from Directorate General of Higher Education Ministry of Education Republic of Indonesia is gratefully acknowledged. The authors would like to thank technicians of Microlab, Section of Material and Environment, Delft University of Technology for supporting an experimental set up.

6

References

[1] Zhang, Q., ‘Microstructure and deterioration mechanisms of portland cement paste at elevated temperature.’ Thesis (Ph.D). Delft: Delft University of Technology.

567

AMS '14 Proceedings of the Int. Conference on Ageing of Materials & Structures Delft 26 – 28 May 2014, The Netherlands [2] Cusson D, Lounis Z, Daigle L, Benefits of internal curing on service life and life-cycle cost of highperformance concrete bridge decks -A case study, Cement & Concrete Composites 32 (2010) 339– 350. [3] Federation Internationale de la Precontrainte, Acceleration of concrete hardening by thermal curing: guide to good practice, 1982, pp. 1–16. [4] J.G. Wilson, N.K. Gupta, An overview of the methods of accelerated curing of concrete by elevated temperatures with emphasis on electroheat techniques, in: Universities Power Engineering Conference Proceedings, University of Bath, UK, September 1992, pp. 107–110. [5] A.G.A. Saul, Principles underlying the steam curing of concrete at atmospheric pressure, Mag. Concr. Res. 2 (6) (1951) 127-140. [6] Bredenkamp S., Kruger D. and Bredenkamp G.L., 'Direct electric curing of concrete'. Magazine of Concrete Research, Vol. 45, No. 162, March 1993, pp. 71-74. [7] lan Heritage, Direct electric curing of mortar and concrete, Napier University, Edinburgh, UK, 2001. [8] John GW, Narendra KG (2004) Equipment for the investigation of the accelerated curing of concrete using direct electrical conduction, Measurement 35 (2004) 243–250. [9] J.G. Wilson, N.K. Gupta, Analysis of power distribution in reinforced concrete during accelerated curing using electroheat, IEE Proceedings–Electric Power Applications 143 (2) (1996) 172–176. [10] Kafry, I. D., 'Direct electrical curing of precast products', Precast Concrete, 1980, Vol. II, no. 7. [11] Wadhwa S.S., Srivastava L.K., Gautam D.K.,Chandra D., Direct electric curing of in situ concrete, Batiment International, Building Research and Practice, 1987, 15:1-6, 97-101 [12] Susanto A, Koleva DA, Copuroglu O, van Beek K, van Breugel K, Mechanical, electrical, and microstructural properties of cement-based materials in condition of stray current flow, Journal of Advanced Concrete Technology Vol. 11, 119-134, April 2013. [13] Susanto A, Koleva DA, van Beek K, van Breugel K, The effect of electrical stray current on material properties of mortar specimens, Proceeding the 6th Civil Engineering Conference in Asia Region: Embracing the Future through Sustainability, Jakarta 20-22 August 2013. [14] Sumanasooriya, M.S., Neithalath,N., Stereology- and morphology-based pore structure descriptors of enhanced porosity (previous) concrete, ACI Materials Journal. 106 (5) (2009) 429-438. [15] Hu J., Porosity of concrete, morphological study of model concrete, PhD thesis, Delft University of Technology, Delft 2004. [16] P. Kumar Mehta and Paulo J. M. Monteiro, Concrete Microstructure, Properties and Materials, October 20, 2001. [17] Nehdi and Soliman, Early-age properties of concrete: overview of fundamental concepts and state-ofthe-art research, Construction Materials 2011, 164: 57-77. [18] Millard SG, Gowers KR. Resistivity assessment of in-situ concrete: the influence of conductive and resistive surface layers. Proc Inst Civil Eng–Struct Build 1992;94(4):389–96. [19] Woefl GA, Lauer K. The electrical resistivity concrete with emphasis on the use of electrical resistance for measuring moisture content. J Cem, Concr Aggr 1979;1(2):64–7. [20] Andersson K, Allard B, Bengtsson M, Magnusson B. Chemical composition of cement pore solutions. Cem Concr Res 1989;19(3):327–32. [21] Elkey W, Sellevold EJ., Electrical resistivity of concrete. Norwegian Public Roads Administration, Publication No. 80; 1995. [22] Weydert Rand Gehlen C. Electrolytic resistivity of cover concrete: relevance, measurement and interpretation. In: Lacasse MA, Vanier DJ, editor. Durability of building materials and components 8, vol.1. Ottawa, Ontario, Canada; 1999. p. 409–419.

568


Novel Kelvin Probe electrode for non-intrusive corrosion rate evaluation of steel in aged concrete structures Michael T. Walsh *, Alberto A. Sagüés Department of Civil and Environmental Engineering, University of South Florida, Tampa, U.S.A.

Abstract: Dynamic polarization measurements of embedded reinforcing steel can be used to estimate corrosion rate in ageing structures. Potential in such tests is commonly measured using a conventional reference electrode but this can yield unstable readings, particularly on aged concrete with high nearsurface resistivity. Interaction between the concrete and the electrode introduces time-dependent liquid junction potentials requiring waiting for stabilization. The Kelvin Probe (KP) uses a vibratingplate capacitor principle to measure potential without contacting the concrete surface thus avoiding complications of electrolyte interaction. This paper extends application of a previously introduced KPbased system to include dynamic Electrochemical Impedance Spectroscopy (EIS) measurements of steel in concrete. Keywords: corrosion, concrete, reinforcement, contactless, polarization

1 Introduction

Corrosion rate evaluation of reinforcing steel in concrete structures is often conducted by means of steel polarization measurements. A small current density is impressed on the steel-concrete interface and the resulting small (e.g. 10 mV) potential change is measured either in the time domain or in the frequency domain, as in the case of electrochemical impedance spectroscopy (EIS) where impressed alternating currents over a range of frequencies range are used. In the simplest cases the corrosion rate is found to be proportional to the inverse of the polarization resistance which corresponds with the low-frequency limit of the ratio of the potential change to the impressed current density with appropriate correction for ohmic components. In more complicated cases the measurements require more sophisticated analysis. In both cases, obtaining accurate results requires that the potential, which is measured by means of a reference electrode normally placed on the concrete external surface, is stable in time despite any inherent instability of the reference electrode or changes occurring on the concrete surface during the test. An important source of potential drift stems from the nature of conventional reference electrodes (e.g. Copper/Copper Sulfate electrode), which require electrolytic contact between the internal electrode medium and the pore water near the surface of the concrete. The required contact normally involves a porous tip at the bottom of the electrode body and some intermediate moist body such as a small sponge in contact with the concrete surface. In most cases and especially in the case of dry or aged concrete surfaces, the contact with the sponge results in relatively large electrolyte intrusion to the initially nearly empty surface pore network. The intrusion results in the development of slowly evolving diffusional potential differences across the depth of the near-surface region. The consequent drift in the potential reading can be large (e.g. 200 mV) and continue over a long time (e.g. hours) thus delaying the start, or otherwise compromising the validity, of the corrosion rate derived from polarization measurements. The Kelvin Probe (KP) uses a vibrating-plate capacitor principle to measure potential without contacting the concrete surface thus avoiding complications of electrolyte interaction. The principle of operation of the KP is well documented in the literature [1-4] so only a brief review is presented here. The KP embodiment in these experiments used a 13 mm diameter austenitic *

University of South Florida, [email protected]

569


stainless steel (Type AISI 302) disk placed ~ 1mm over the concrete surface and made to vibrate perpendicular to the concrete surface at ~150 Hz with ~ 0.5 mm amplitude. An electromagnetic voice coil was used as a driver. The disk is connected through a current-sensing electronic circuit and a variable voltage source to the reinforcing steel that is embedded in the concrete directly beneath the disk. The natural Volta potential difference between the disk and the surface of the concrete, which together form a parallel-plate capacitor, results in a capacitive charge directly proportional to the value of the potential difference (nominally constant due to the interconnection) and the disk-concrete capacitance. Since the capacitance varies with the oscillating gap distance, the charge likewise varies resulting in an alternating current through the disk-steel interconnecting circuit. An automatic zeroing circuit changes the value of the voltage source until the current vanishes. The value of the source voltage at that point is equal but opposite to the natural steel-disk potential difference. That value is then recorded as the probe potential reading for that position on the surface of the concrete. Since the disk doesn’t contact the concrete and because its material properties are not subject to change, KP readings obtained by moving the disk over different parts of the concrete surface effectively constitute a survey of the potential gradients over the concrete surface. That information can be used to identify corroding regions of reinforcement within a concrete body, in the manner of the ASTM C-876 [5] potential mapping procedure but without disturbing the concrete surface. The authors have recently demonstrated the feasibility of using a novel customized KP as an alternative to a conventional electrode for potential mapping [6]. They also demonstrated the use of the probe for nearly non-intrusive polarization resistance measurements in the time domain that may serve for corrosion rate determination [7]. In this paper, recent advances are presented expanding the use of the KP to include frequency domain polarization measurements of steel in concrete, in the form of Electrochemical Impedance Spectroscopy (EIS) tests. This type of measurement can provide more accurate evaluation of polarization parameters with greater sensitivity, while avoiding any disturbance of the concrete surface at the location of measurement.


Kelvin Probe Embodiment for Polarization Measurements of Steel in Concrete

For steel polarization measurements the KP disk was placed at a single location on the concrete surface, with a relatively large (10 x 10 cm square with a 3 cm diameter hole in the centre) external counter electrode (CE) surrounding the sensing disk but not touching it. The CE and disk were separated by a grounded shield (Figure 1). The CE was made of a conductive elastomer (resistivity in the order of 1 ohm-cm) that does not moisten the concrete surface thus minimizing disturbance to the concrete footprint of the counter electrode and surroundings (including the place beneath the sensing disk) and was pressed against the concrete surface with a clamped steel plate. A polarizing current was introduced between the CE and the embedded steel, resulting in a change of the steel-concrete interfacial potential plus an ohmic potential drop due to the electric resistance of the intervening concrete between the surface of the concrete and that of the steel. Given the relatively large size of the counter electrode, the current flow toward the steel was nearly uniform and the KP sensed the combined ohmic and interfacial polarization response to the applied current. The values of the current and the combined potential change were recorded and the results were then, after making appropriate working assumptions of the effective steel area and other system parameters, processed to obtain a nominal polarization resistance and subsequently a nominal corrosion rate for the underlying steel. Calibration of the KP against a conventional electrode is unnecessary since only changes in potential matter. The feasibility of this approach was recently demonstrated by the authors for direct current galvanostatic pulse polarization resistance measurements in laboratory reinforced concrete specimens [7]. 570


C RW

Voice Coil

Shield

Disk

KP Control Unit Pressure Plate Conductive Elastomer CE

Steel Rebar Concrete Figure 1 Schematic diagram of combined KP reference (R) plus dry counter electrode (CE) system with the reinforcing steel bar (rebar) as the working electrode (W); red arrows show idealized excitation current path. Patent Pending

2.2

Electrochemical Impedance Spectroscopy

The extension to alternating current operation for EIS measurements, introduced in this paper, was implemented by connecting the system to the C, R, and W terminals of a Gamry 600 potentiostat and impedance analyzer unit. Because the vibrating disk operated at near 150 Hz and the electronic processing unit had typical settling times in the order of ½ second, EIS measurements were conducted only at frequencies 8)

(2)

Ca2+ + HCO 3 - → CaCO 3 + H+ (7.5 100 (mass)

31.8 + (l) => 48.9 (volume)

33.22 + (g) => 36.93 (volume)

MgO + H 2 O → Mg(OH) 2

Mg(OH) 2 + CO 2 + 2H 2 O → MgCO 3 ·3H 2 O

40.31 + 18 => 58.31 (mass)

58.31 + 44.01 + 2(18) => 138.37 (mass)

11.2 + (l) => 24.3 (volume)

24.3 + (g) + (l) => 74.77 (volume)

The final product nesquehonite resulting from MgO through hydration and carbonation was found to have volume expansion from 11.2 to 74.77, in percentage 568%. Furthermore, reactive MgO has great potential to decrease the pH level of the concrete [8]. In addition, concrete formed with stone aggregate has a void space for survival of bacterial spore; hence, lower pH may have benefits for bacterial growth and survival for a longer period. The hydration and carbonation reaction of MgO in the concrete demonstrated by [9] is: MgO +H2O →Mg(OH) 2

(4)

Mg(OH) 2 + CO 2 + 2H 2 O → MgCO 3 ·3H 2 O

(5)

One of the principal challenges in self-healing technology in concrete is to recover the mechanical strength after damage. However, two important mechanisms, pore blocking by impurities in water and debris produced due to crack sapling, are identified so far in self-crack healing of concrete [5]. However, the expansion products of MgO due to hydration and carbonation can potentially seal the crack and eventually heal the crack.

636


2 Materials and Method 2.1

Materials

Ordinary Portland cement (OPC)-CEM 1 (52.5N) was used as the basic cement material. Principal self-healing minerals was light burnt Magnesioum Oxide (MgO 92/200). Furthermore, local bentonite clay (supplied by Kentish minerals) was used for assisting MgO self-healing activity with OPC due to its swelling property. Furthermore, Superplasticiser (Sikament 700 satisfy BS EN 934-2, supplied by Sika Ltd), was used as a water reducing admixture. Different mixes prepared for experiment are illustrated in the Table 2 below. Table 2. Mix proportions of five different cement mixes.

Mix

2.2

MgO (- )

Bentonite -l ( )

Water

Superplasticiser

M1(PC100)-control

OPC (100 )

28.5

0.3

M2(PC95M5)

95

5

-

30

0.3

M3(PC90M10)

95

10

-

33

0.3

M4(PC90M5B5)

90

5

5

33.5

0.3

M5(PC85M10B5)

85

10

5

36.5

0.3

Methods

Mechanical and image analysis on cube (100X100X100mm) and prism (160X40X40mm) samples was conducted to investigate the self-healing performance. Compression test on cubes were performed in different period. Prism samples were used for forming cracks using threepoint bend test following (BSEN 12390-5) [10], and monitoring subsequent crack healing in different period. Flexural stress and strain was calculated using equation 6 and 7 below.

σf =

3PL 2bd 2

εf =

6 Dd L2

(6) (7)

In the equations, σ f = Stress in outer surface at midpoint (MPa), ɛ f = Strain in the outer surface (mm/mm), P= load (N), L= Support span (mm), b = Width (mm), d= Depth (mm), D= maximum deflection of the prism centre (mm). Three-point bend test setup is illustrated in Fig.1. P L/2

L/2

b d

L P/2

P/2

Figure 1 Three point bend test setup

637


Prism samples were cracked on day 1 after casting. Cracked prism samples were cured under water after cracking and they were placed in the tank vertically. No external compressive load was used in curing period. The healed samples were cracked again under three point bend test for measuring the flexural strength regain in 14, and 28 days. Further, 28 day healed samples were cured in the water tank after cracking and cracked again after 150 days for investigating cyclic self-healing. Strength recovery was compared using eq. 8. In this equation, R is the percentage of mechanical strength recovery; rec, σ f, max is the maximum flexural stress recovered with time (14/28/150 days); and day1, σ f, max is the maximum flexural stress in day one (first cracking day).

R=

rec, σ f ,max

day1, σ f ,max

(8)

X 100%

The healing was also monitored under digital microscope. Cracked prisms were marked in different places and there widths were measured after initial cracking. Images were taken in same crack position after 14 and 28 days for 1st cycle healing, and 150 days for 2nd cycle healing.

3 Results and Discussions 3.1

Compressive strength development

All mixes have showed steady development of compressive strength with time. Although, Mix 2 with 5% MgO substitution of OPC had slightly increased the compressive strength, 10% MgO in Mix3 slightly decreases it. On the other hand, bentonite clay reduced the compressive strength of mixes in Mix 4(PC90M5B5) and Mix 5(M85M10B5). Overall the compressive strength developments were similar and mix minerals had mix impact (Fig.2).

Compressive strength development with time

Stress (MPa)

120,00

M1(PC100)

100,00

M2(PC95M5)

80,00 60,00

M3(PC90M10)

40,00

M4(PC90M5B5)

20,00 0

10

20

30

40

Days

50

60

M5(PC85M10B5)

Figure 2 Compressive strength developemnt with time

3.2

Three-point bend test

MgO and bentonite addition had increased strength regaining capacity of OPC in Mix 2 to Mix4 (fig.3). Addition of up to 5% of both MgO and bentonite had steadily increased the recovery of strength in 14 days to 28 days. However addition of 10% MgO had lowered the strength recovery on 28 day compared to 14 day. This was due to the high expansion property of MgO in mixes. There was no reinforcement or fibre in prism samples. Hence the crack widths were random ranging between 0.1 to 0.5 mm and some samples were completely separated after cracking. Prisms samples that were completely separated after cracking have showed lower selfhealing (mechanical strength regain) performances compared to partially cracked samples. 638


Typical Recovery on Day 14 M1

1

M1

2

M2

0,8

M2

M3

1,5

M4

1

M5

0,5

Stress (MPa)

Stress (MPa)

Typical Day 1 2,5

0

M3

0,6

M4

0,4

M5

0,2 0

0

0,001

0,002

0,003

0

Strain (mm/mm)

0,002

0,003

Strain (mm/mm)

2nd Cycle Recovery on Day 150

Typical Recovery on Day 28

1,5

M1

0,5

M1

M2

0,4

M2

M3

1

M4 M5

0,5

Stress (MPa)

Stress (MPa)

2

0,001

0

M3

0,3

M4

0,2

M5

0,1 0

0

0,001

0,002

0

0,003

Strain (mm/mm)

0,001

0,002

0,003

Strain (mm/mm)

Figure 3 Three-point bend test on day 1 and recovery in different period

Percentage Recovery

Mechanical Strength Recovery 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% M1 (PC100)

M2 (PC95M5)

M3 (PC90M10)

Mixes

Avg. recovery in 14 days Avg. recovery in 28 days 2nd Avg. recovery in 150 days

M4 (PC90M5B5) M5 (PC85M10B5)

Figure 4 Percentage of mechanical strength recovery of mixes in 14, 28, and 2nd cycle 150 days.

Strength recovery in 2nd cycle of healing after cracking 28 day self-healed samples was also promising (Fig.3). M4 and M5 mixes have showed better average strength recovery in 2nd cycle after 150 days (Fig.4). Additions of 10% MgO had made the samples more expansive. Thus, without reinforcement, external loading [11], or end restraining, there were limited recoveries.

639


3.3 Mix

Self-healing image investigation Recovery in Day 14

Recovery in Day 28

2nd Cycle Recovery in Day 150

Mix1

1mm

1mm

1mm

1mm

1mm

1mm

Mix2

Mix3

1mm

1mm

1mm

Mix4

1mm

1mm

1mm

1mm

1mm

1mm

Mix5

Figure 5 Crack healing images in 14, 28 days and 2nd cycle in 150 days

640


Digital microscopic image analysis have confirms that all the mixes healed better compared to control OPC mix (Fig. 5). M2, M4, and M5 mixes have showed effective crack sealing potential within 14 days. M4 with both 5% MgO and bentonite have showed most efficient crack sealing and healing potential (Fig. 5). Investigations have also revels that cracks up to 0.15mm were healed most efficiently and cracks greater than 0.35 mm were healed partially. Crack healing and crystallisation had increased with curing time. Crystallisation was initiated from both side of the crack surface and heals the crack gradually with time. All mixes showed better crack sealing after 150 days of curing in case of 2nd cycle self-healing, whereas M4 and M5 with both MgO and bentonite showed complete closer of cracks with efficient strength recovery.

Hence, combination of both MgO and bentonite considerably had increased the autogeneous self-healing capacity of OPC. The expansion of light burned MgO was triggered by the swelling property of bentonite in contact with water. Their mixtures not only had sealed the crack faster but also healed it in 28 days, and also had considerable second cyclic recovery in 150 days.

4 Conclusions

Our experiments have revels that little substitution of OPC with optimum combination of MgO and bentonite has improved the autogenous self-healing capacity of cement. Property of minerals, therefore MgO expansion and bentonite swelling have showed better crystallisation in the crack surface. Addition of MgO up to 5% in M2 have positively impact the compressive strength development, although 10% MgO reduces it. Bentonite always had reduced the compressive strength. Hence, optimum combination of minerals for self-healing would be 5% MgO with little amount of bentonite in PC. Crack width had also played an important rule in self healing, where smaller cracks up to 0.15 mm were healed faster. MgO minerals have not only improved the autogeneous self-healing of the cement, but also have opened great prospects for any autonomous self-healing techniques to be incorporated with for efficient self-healing. Preliminary SEM and XRD investigation on self-healing materials have confirmed the presence of CaCO 3 , MgCO 3 , and combined MgCaCO 3 products. The presence of the MgO also reduced the pH in the cement paste which would encourage any biological healing process incorporated within such PC-based mixes. Further investigation of self-healing materials formed and combination of autonomic bacterial self-healing process with optimum minerals mix proportion is underway.

References

[1] Ahn and Kishi (2010) Crack Self-healing Behavior of Cementitious Composites Incorporating Various Mineral Admixtures. Journal of Advanced Concrete Technology, 8(2), pp.171–186. Available at: http://joi.jlc.jst.go.jp/JST.JSTAGE/jact/8.171?from=CrossRef. [Accessed October 12, 2012] [2] Wagner E.F. (1974) Autogenous Healing of Cracks in Cement-Mortar Linings for Gray-Iron and Ductile-iron Water Pipe. Journal of the American Water Works Association, Vol. 66, pp358-360 [3] Westerbeek (2005) Self-healing materials. Radio Netherlands. cited on Joseph, C. et al., 2010. Experimental investigation of adhesive-based self-healing of cementitious materials. Magazine of Concrete Research [4] Ramm, W. &Biscoping, M. (1998) Autogenous healing and reinforcement corrosion of water-penetrated separation cracks in reinforced concrete. Nuclear Engineering and Design, 179(2), pp.191–200. Available at: http://linkinghub.elsevier.com /retrieve/pii/S0029549397002665. [Accessed February 1, 2013] [5] Edvardsen (1999) Water permeability and autogenous healing of cracks in concrete. ACI Material Jounral, 96(4), pp.448-454. Web: http://www.unitedstatesconcrete. com/AutogenousHealingofCracks.pdf [Accessed February 1, 2013] [6] Homma et al. (2009) Self-Healing Capability of Fibre Reinforced Cementitious Composites. Journal of Advanced Concrete Technology, 7(2), pp.217–228. Available at: http://joi.jlc.jst.go.jp/JST.JSTAGE/jact/7.217?from=CrossRef [Accessed January 29, 2013] [7] Unluer (2012) Enhancing the carbonation of reactive magnesia cement –based porous blocks, PhD thesis, Department of Engineering, University of Cambridge [8] Zhang, T., Cheeseman, C.R., Vandeperre, L.J. (2009) Development of novel low pH cement systems. DIAMOND'09 Conference, York, UK.pp.1–4

641


[9] Vandeperre, L., Liska, M., and Al-Tabbaa, A. (2007) Reactive MgO cements: Properties and Applications. International conference on sustainable construction materials and technologies. Coventry. [10] British Standards Institution, (2009) Testing Hardened Concrete Part 5: Flexural Strength Test Specimens. BS EN 12390-5:2009 [11] ter Heide, N and Schlangen, E. (2007) Selfhealing of early age cracks in concrete, Proceedings of the 1st International Conference on Self Healing Materials, Noordwijk aan Zee, The Netherlands

642


Colloidal nanosilica healing ability for reinforced concrete repair M. Sánchez1*, M.C. Alonso1, I. Díaz1,2, R. González1 (1) Institute of Construction Sciences “Eduardo Torroja” (IETcc – CSIC), Madrid, Spain (2) Autonomous University of Nuevo León, Monterrey, Mexico

Abstract: The migration of colloidal nanosilica is proposed as non-invasive technology to promote the sealing of hardened concrete. The nanosilica sealing ability and its interaction with the mortar solid phases were assessed by different characterizing techniques on mortar samples 28 days after the treatment finished. The mortar consolidation after the treatment was deduced from the decrease of the smallest pores (1 m), evolving with mortar ageing. The chemical interaction was observed between the migrated nanosilica and the cementitious matrix. The pozzolanic activity of silica nanoparticles was confirmed. Also the formation of new gels C-S-H enriched in silicon (C/S ratio between 0.75 – 1.1) was observed by BSE-EDS and thus, a healing character of the treatment can be expected. A penetration of 1 mm was measured by BSE-EDS. Keywords: non-invasive repair techniques, electromigration, colloidal nanosilica, healing properties

1 Introduction Nowadays, the development towards more sustainable solutions is becoming a priority in developed countries, and the most competitive industries need to incorporate these aspects into their targets. Also in the construction field innovative concepts such as nanotechnology or self‐ healing materials become of great interest to look forward a more competitive industry. Different attempts have been reported concerning the application of nanotechnology in concrete [1‐2], and also, several approaches can be found about self‐healing of cementitious materials [3]. The application of such novel solutions has been mainly reported for new structures, by adding nanoparticles or self‐healing materials to fresh concrete. However, the huge amount of existing concrete structures involves a great economic effort on repair, demolishment or refurbishment projects around the world and thus, developing non‐invasive technologies for reinforced concrete repair appears as one of the greatest challenges in construction industry. Not only the repair but also the performance improvement may be an interesting point to be incorporated in the repair / maintenance projects of existing concrete structures. The addition of nanosilica to fresh concrete has been reported as an innovative solution to improve the performance of new concrete structures [4‐5]. The pozzolanic activity of nanosilica promotes its reaction with portlandite decreasing the concrete pores by forming new gels C‐S‐H [6] and thus, improving the durability of the concrete structure. If the nanosilica reactivity would remain even when applied in hardened concrete, a healing character could be expected from the incorporation of these new gels C‐S‐H into the solid phases of the hardened cementitious matrix. The application of electrochemical methods has been proposed as accelerated technique to successfully enhance the penetration of charged particles through the concrete pores. That is the case of the introduction of corrosion inhibitors from the concrete surface during an electrochemical treatment of chloride removal (ECE) [7‐8]. Also the electromigration of particles after an ECE treatment has been reported to stimulate the electrodepositing of different compounds for crack sealing in concrete beams [9‐10]. *

M. Sánchez, IETcc-CSIC, [email protected]

643


The migration of colloidal silica nanoparticles covered with alumina from the hardened concrete surface has also been proposed as a final stage of electrochemical repair methods, taking advantage of the electric field connected for the removal of the aggressive agents during the treatment. The migrated nanoparticles promoted the sealing of the concrete surface hindering the further penetration of aggressive agents [11‐12]. Although the efficiency of nanosilica on decreasing concrete porosity and improving rebar corrosion behaviour after the treatment was reported [13‐14], no discussion was found concerning the nanosilica / cementitious matrix interaction mechanisms. Author’s approach is based on the migration of colloidal nanosilica under the action of an electric field to promote the sealing of the concrete surface by filling the pores with the migrated nanoparticles [15‐16]. In the present study, the mortar consolidation associated to the nanosilica migration treatment was assessed by non‐destructive measurements, such as the mortar electrical resistivity, and by destructive techniques, such as mercury intrusion porosimetry (MIP). The reactivity of the migrated nanosilica with the cementitious matrix was analysed by differential thermal analysis (DTA) and thermogravimetric analysis (TGA), back‐ scattering microscopy with X‐Ray microanalysis (BSE/EDS). The last one also allowed estimating the nanosilica penetrability through the concrete cover. The reactivity of the silica nanoparticles with the solid phases of the hardened cement matrix as well as a clear decrease of the capillary concrete pores after the colloidal nanosilica migration treatment were confirmed. The interaction between the nanosilica and the solid cement matrix evolved with the treated sample ageing, as deduced from MIP results at different times after the treatment finished.


Materials

The study was carried out on mortar samples manufactured with Ordinary Portland Cement (OPC), deionized water and normalized siliceous sand. For the BSE/DSE analysis, mortar samples with calcareous 0/3 sand (97.5 – 99.5 % natural CaCO3) were fabricated to diminish the influence of silicon coming from the sand on the EDX analysis. In both cases, the same experimental procedure was followed: manufacturing, curing and submitted to the same treatment conditions. Cylindrical sections of 7.5 cm in diameter and 1 cm in thickness were fabricated with 0.5 w/c ratio and 1:3 cement sand ratio. The curing process was 7 days in chamber under controlled environmental conditions (98 ± 2 % HR, 21 ± 2 ºC). A colloidal nanosilica suspension with a slightly negative charge was used (30 % in cement weight, 7 nm mean diameter, density 1.2 g/cm3 and pH 8). As reference case, a non‐treated sample was used.

2.2

Treatment conditions

Before treatment, mortar samples were submerged in deionized water and saturated under vacuum conditions for 24 hours. The experimental arrangement of the electrochemical treatments is schematized in Figure 1.

644


Figure 1 Scheme of migration cell set-up.

The saturated mortar samples were located in the centre of the migration cell. Two steel rebars of 6 mm in diameter were located at both sides of the mortar sample and connected as electrodes to the power source for 8 days. Two different electric fields, 6 and 12 V, were tested. The side connected to the negative pole of the power source (catholyte) was filled with the colloidal nanosilica suspension and the other side, connected to the anode (anolyte), was filled with distilled water. During the electric field connection, the effective potential at both sides of the mortar sample and the current passed through the system were monitored to estimate electric resistance by Ohm’s Law. After treatment, mortar samples were wet cured again until testing (28 and 90 days).

2.3

Characterization of treated mortar samples

Treated siliceous mortar samples were characterized at the age of 28 days after finishing the treatment. Different techniques were used to assess the sealing ability of the colloidal nanosilica after migration under the electric field. Non‐destructive measures of electrical resistivity were carried out as indirect parameter of the mortar consolidation by the nanosilica action. Mercury Intrusion Porosimetry (MIP) was used to evaluate the influence of the migration treatment on the porosity and on the pore size distribution of treated mortar. A small portion of mortar from the surface treated with the nanosilica (around 4 mm depth) was used to carry out these measurements. Two ages after treatment, 28 and 90 days, were considered for MIP characterization. Differential Thermal Analysis (DTA) and Thermogravimetric Analysis (TGA) were used on powered samples to assess the nanosilica reactivity once penetrated into mortar. To confirm the nanosilica penetration, mortar samples fabricated with limestone sand were characterized by Back‐Scattering Microscopy with X‐Ray Microanalysis (BSE/EDS) 28 days after migration treatment at 12 V finished. This technique was also used to assess the interaction of migrated nanosilica with the substrate. The C/S ratio of gel C‐S‐H phases was estimated as an indicator of the reactivity between the silica nanoparticle and the cement solid matrix. In this case, a calcareous mortar was used to avoid the influence of siliceous sand in the silicon chemical analysis.

645



Sealing ability of colloidal nanosilica

The electrical resistivity of siliceous treated samples was measured after 28 days migration treatment stopped as a qualitative indicator of the mortar consolidation. The relative increase of resistivity (r) was estimated as

, where 28d is the resistivity at the age of 28

days after finishing the treatment, and 0 is the initial resistivity, before treatment. A clear gain in resistivity was measured after treatment, more significant for higher values of the electric field: an increase of 93% was measured after the 12 V migration test, while an increase or 44%the increase after the treatment at 6 V was estimated. The results obtained with this non‐ destructive indicator were in agreement with the porosity measures after 28 days. The decrease of the total porosity was registered after the migration treatment both with 6 V and 12 V, being higher the diminution when 12 V applied (14% smaller when compared with a reference non‐ treated mortar) than after the 6 V connection (8% smaller than reference case). At the age of 28 days after finishing the treatment, the effect of colloidal nanosilica was highly evident on the pores of smallest size, as shown in Figure 2 (left). High sealing efficiency can be deduced from the significant decrease in 0.05 – 1 m pores. The increase observed on the smallest pores (< 0.05 m) of treated samples could be explained by the decrease of the diameter of sealed capillary pores due to the nanosilica action. No clear effect is distinguished from the different electric power applied during the treatment. In Figure 2 (right) the ageing of the treated sample (12 V) is evaluated, taking a non‐treated mortar as reference case. A decreasing trend of capillary pores (< 1 m) is observed with the ageing of the treated mortar (90 days vs 28 days after treatment finished). Nanosilica continued reacting with time, providing a healing character to the migration treatment by forming new compounds that fill the capillary pores of concrete and increase the sealing efficiency of the treatment with ageing. 100

75