Julie Armengaud LMDC

Influence of Aggregate Size Distribution on Silica Fume Efficiency in Dry-Mix Shotcrete Julie ARMENGAUD*, Marc JOLIN**, Antoine GAGNON**, Géraldine CASAUXGINESTET*, Martin CYR*, Bernard HUSSON* *LMDC, Université de Toulouse, INSAT, UPS, France ** Centre de recherche sur les infrastructures en béton (CRIB), Université Laval, 1065 Avenue de la Médecine, Québec, QC G1V 0A6, Canada Abstract The dry-mix shotcrete technique is a process in which dry or slightly damp constituents of concrete are introduced into a machine and conveyed pneumatically through a hose to the nozzle where water is added. This technique is used in various civil engineering projects but generates high material losses due to rebound. These losses induce overconsumption of material, which is damaging for the cost of the work and for the environment. Thus, reducing rebound is an important objective in drymix shotcrete research. However, shotcrete placement depends on the operator and on many parameters that interact with each other to a greater or lesser extent, which makes the study of shotcrete quite complex. This paper focuses on interactions occurring at the mixture design level. The goal of the study was to analyse the role of the aggregate size distribution on the efficiency of supplementary cementitious materials (SCM), and more precisely silica fume (SF), from a rebound, strength and porosity point of view. It has already been shown that the granular phase plays a significant role in rebound, as does the presence of SCM, but the study of the interaction of these two parameters could bring new insights in terms of rebound and durability. For this purpose, six mixture designs were prepared. Three were formulated with the same proportion of cement (OPC) by mass but with different proportions of sand and gravel. The second set of three mixtures was identical to the first, except that the binder included 8% of silica fume. This allowed three objectives to be attained: 1- validate results from the literature regarding the effect of the aggregate size distribution on rebound; 2- validate and measure the rebound reduction brought about by the addition of silica fume; 3- study the combined efficiency of optimized aggregate size distribution and silica fume. All mixes were shot at different consistencies in order to have an overview of the W/B ratios achievable with a given mix. For a normal consistency, results on rebound showed that there were almost no differences between the three OPC based mixtures (maximum 3.7% difference of rebound). On the other hand, when silica fume was added, it became clear that, when more sand was used, the addition was more efficient in reducing rebound (maximum 15.6% difference of rebound with SCM). In terms of strength, the finest granular distribution with the addition of silica fume gave the highest strength at 28 days (56±2MPa). However, porosity was lower for a coarser mix with silica fume. These tendencies varied with the spraying consistency. For instance, a reduction of water content (leading to a dry consistency) induced an increase in rebound for a fine granular distribution with silica fume, while the strength was higher for a coarse aggregate distribution with SF. This means that SF efficiency is dependent on the aggregate size distribution as well as on consistency. In order to optimize the use of SF in dry-mix shotcrete, and depending on the properties expected, aggregate size distribution and spraying consistency must be considered. 1

Introduction

In this paper, rebound of dry-mix shotcrete is studied. Shotcrete is a “mortar or concrete, pneumatically projected onto a surface at high velocity” and rebound is defined as the “proportion of shotcrete material that ricochets off the receiving surface” (ACI Committee 506 2008). As shotcrete can be used in a wet or a dry process, a clear distinction must be made between the two methods. The wetProc. of the 11th fib International PhD Symposium in Civil Engineering Aug 29 to 31, 2016, The University of Tokyo, Tokyo, Japan

1

11th fib International PhD Symposium in Civil Engineering

mix shotcrete process consists of the pneumatic placement of concrete that is batched with water before being introduced into the delivery hose whereas dry-mix shotcrete is a process in which dry or slightly humidified constituents of concrete are introduced into the machine and conveyed pneumatically through a hose to the nozzle, where the water is added. The main difference lies in the location where water is introduced and thus the time of contact with water through the water ring before the material reaches the surface. In both cases, it is the high velocity imparted to the flow that induces consolidation of the material on the surface. Dry-mix shotcrete is used in various civil engineering or construction projects, from tunnelling or anchored retaining walls to the repair or construction of buildings. Unfortunately, the dry-mix process can lead to high losses of concrete due to rebound. Such losses induce overconsumption of material, which is damaging for the cost of the work and for the environment. Thus, reducing rebound is an important objective in dry-mix shotcrete. During the dry-mix shooting process, the nozzleman is in charge of the water adjustment, and thus of the substrate texture (Jolin, Beaupré, and Mindess 1999). This texture has an influence on the impacting conditions of the material flow and, by extension, on rebound and in-place composition (as rebound tends to create a richer in-place mixture). But rebound is also dependent on many technical parameters such as air flow, orientation of the hose, and temperature (Puri and Uomoto 1999; Parker, Fernandez-Delgado, and Lorig 1976; Gérômey 2003), and mixture properties such as binder content, amount and size of coarse aggregates, and type of admixture (Pfeuffer and Kusterle 2001; Parker, Fernandez-Delgado, and Lorig 1976; Morgan and Wolsiefer 1991; Jolin and Beaupré 2004). Shotcrete properties depend on a chain of phenomena governed by many input parameters and resulting in a complex settling of the concrete. This article will focus on interactions at the mixture design level. It has already been shown that the aggregate size distribution plays a significant role in rebound (Jolin and Beaupré 2004), as does the presence of SF (Bindiganavile and Banthia 2000)(Jolin and Gagnon 2005), but the study of the interaction of these two parameters could bring new insights in terms of rebound and durability. 2

Test program

In order to analyse the influence of aggregate size distribution on SCM efficiency, six mixture designs were prepared and shot with a dry-mix shooting machine at the Université Laval Shotcrete Laboratory. All technical parameters where kept constant; only water and mixture designs were modified. 2.1

Experimental setup

2.1.1 Shotcrete equipment The shooting took place at the Shotcrete Laboratory of Université Laval, at a facility designed for shotcrete studies. The machine used was an Aliva-246 rotating barrel with a 38 mm inner diameter hose. Input air pressure was 700 kPa. The nozzle was a double-bubble hard rubber type with a 32 mm diameter exit. Oven dried, pre-bagged mixtures were introduced dry into the machine. 2.1.2 Rebound measurement Rebound was measured by shooting onto a vertical rebound panel fixed to a load cell. The shooting machine and the dry material were placed on a scale. A water flow meter was used to indicate the amount of water added to the dry material (the full description of the setup can be found in Jolin (1999)). The data were recorded as a function of time by an acquisition system. The rebound was then calculated as: Rebound (%)=

–

( )

( )

x 100

2.1.3 Consistency measurement In addition to the rebound value, the consistency was evaluated with a penetrometer, also called a proctor needle, that measured a static contact stress (P). This measurement has been used in many studies with different shapes of indenters (Figueiredo 1999; Jolin 1999; Armelin and Banthia 1998; Prudêncio, Armelin, and Helene 1996). It is a very convenient tool for comparison as the W/B ratio cannot be evaluated before the shooting because it is the nozzleman who adjusts the water content during the operation. In our case, the needle was a 6 mm diameter flat indenter. It was pushed into the 2

Advanced materials

Influence of Aggregate Size Distribution on Silica Fume Efficiency in Dry-Mix Shotcrete

substrate and the stress was indicated by a spring system. The value of P was the maximum stress and has been defined as proportional to a yield stress by Johnson (1985). 2.1.4 Evaluation of Water Binder ratio The total water content of fresh shotcrete was evaluated after each shooting. A sample was taken from the rebound panel, weighed, and dried in an industrial microwave oven until the water had totally evaporated. The water content w was calculated as: #=

$$

$

$$

( )–

$$

( )

(%)

The binder content (b) was evaluated by a decantation test (Bolduc 2009). A sample was taken from the in-place shotcrete, weighed, rinsed on an 80 µm sieve and kept in a stove until its mass stabilized. After drying, the sample was weighed again and, using the water content value, it was possible to calculate the mass variation before and after sieving at 80µm, which gave the mass of binder b. This implies the hypothesis that all particles under 80µm are part of the binder. The water/binder ratio was then determined as w/b. 2.1.5 Evaluation of strength and porosity Strength and porosity were evaluated on cylindrical core after 28 days of curing at a relative humidity >95%. The procedure used was ASTM 1604 for the compressive strength of cylindrical concrete specimens 75 mm in diameter and ASTM C642 for boiled water absorption and volume of permeable void. 2.2

Materials and mixtures

100 90 80 70 60 50 40 30 20 10 0

100 90 80 70 60 50 40 30 20 10 0 0,08 Aggregate Refusal

Fig. 1

0,8 Sieve size (mm) Sand Refusal

% Refusal

% cumulative passing

Six mixtures were tested in this study. Three were formulated with the same mass proportion of cement (OPC) but with different proportions of sand and gravel. The second set of three mixtures was identical to the first one except that the binder included 8% of silica fume. Angular coarse aggregates with a dmax of 10 mm and sand with a dmax of 5 mm were used. The granular distribution is presented in Figure 1. The silica fume used met the ASTM C1240-10a and CSA A3001-08 specifications and was provided by Silicium Québec. All dry mixtures presented in Table 1 were batched and packaged in a specialized plant. All the aggregates were oven dried to prevent any hydration in the 30 kg bags prepared.

8 Aggregate

Sand

Particle size distribution for aggregates and sand. % cumulative passing and % refusal.

As the nozzleman is responsible for adjusting the amount of water when shotcrete is implemented, each mixture was shot three times with a different amount of water each time. This gave an overview of the range of consistencies.

Julie ARMENGAUD*, Marc JOLIN**, Antoine GAGNON**, Géraldine CASAUX-GINESTET*, Martin CYR*, Bernard HUSSON*

3


Sand (% total mass)

OPC (% total mass)

Name ↓

Coarse aggregate (% total mass)

Silica Fume total mass)

Fine

15

63

22

-

Medium

24

54

22

-

Coarse

38

40

22

-

Fine + SF

15

63

19.2

1.8

Medium + SF

24

54

19.2

1.8

40

19.2

1.8

Constituent →

38 Coarse + SF Table 1: Mixtures used in the study 3

Results

3.1

Consistency

(%

The consistency (P) and water/binder (W/B) ratio of each mixture were determined after each shooting. Figure 2 shows the relation between W/B and P for mixtures with and without silica fume. A linear relationship can be observed to exist between the two values and mixtures with silica fume can be seen to allow a wider variation of water content. The driest consistencies correspond to a low W/B and a high penetration stress and the wettest to high W/B and low penetration stress. In order to make the results easier to exploit, each mixture was classified in a consistency category: dry, intermediate or wet. During the shooting, dry consistency was obtained by lowering the water content until the mixture reached an unreasonable amount of rebound. The wettest consistency was obtained by adding water just before the point when the mixture started to slough off the wall. The third, intermediate, consistency was between these two. 0,50

Without SF With SF

Water / Binder

0,45 0,40 0,35 0,30 0,25 0,20 0

Fig. 2 3.2

2 4 P Static contact stress [MPa] ←Wet / Dry→

6

W/B as a function of static penetration stress (P) for mixtures with and without silica fume Rebound

Figure 3(a) presents the rebound versus consistencies for mixtures without silica fume. It can be seen that for Fine and Medium mixtures, rebound decreased slightly with the wettest consistency. However, for the Coarse mixture, there was an inverse tendency. Losses increased when more water was added. This observation was, in fact, due to the expulsion of in-place mortar when coarse aggregates hit the surface. These losses were not attributable strictly to the rebound phenomenon, but were also due to expulsion of material that did not rebound in when it was first applied. For the driest consistency, it seems that the finer the mixture was, the higher were the losses. This might have been due to the smallest particles not having enough energy to penetrate the stiff substrate.

4

Advanced materials


For the wettest consistency, if the Coarse mixture was not taken into account (where the expulsion phenomenon was predominant), the Medium mixture seemed more efficient to reduce rebound than the Fine mixture. For the intermediate consistency, there were few differences between mixtures. Figure 3(b) shows the rebound versus consistency for mixtures with silica fume. The global tendency is a decrease of rebound when consistencies are wetter. No expulsion was observed with mixtures containing silica fume. However, for the wet consistency with Medium+SF, the air flow was slightly higher, and the comparison was thus difficult. For mixtures Fine and Medium, there seemed to be a threshold effect in rebound diminution as there was an abrupt drop in rebound (around 15%) between dry and normal consistencies while, with Coarse mixture, there was a progressive decrease in rebound as the consistency became wetter. This raises the question of stability of a mixture under water variation, as the adjustment of water was more difficult for Fine and Medium mixtures than for Coarse mixture. For a normal consistency, Fine and Medium mixtures with silica fume appear to lower the rebound more than Coarse mixture, but the range of Intermediate consistency was small for Fine and Medium mixtures.

Dry Fine (a) Fig. 3

Medium

Wet Coarse

26%** 19%

14%

42% 41% 34%

50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0%

24% 28% 25%

Rebound

30%

40%* Intermediate Consistency

22%

32% 28% 31%

50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0%

36% 30% 25%

Rebound

As for mixtures with no SCM, rebound was higher for Fine mixture with SCM in the case of a dry consistency. For the wettest consistency, there was less difference among mixtures, but the Medium mixture had higher rebound. During the shooting of Medium mixture, the air flow was slightly higher than for the other shooting and, as the air flow has an influence on rebound (Armelin 1997), this might explain the rise. A comparison between Figures 3(a) and 3(b) reveals that, at dry consistency, mixtures with silica fume show increased rebound. On Figure 2, mixtures with silica fume appear to be drier than mixtures without silica fume, which could explain the rise in rebound with silica fume. Usually, silica fume is known to reduce rebound but previous results were based on mixtures having intermediate to wet consistencies (W/B estimated at 0.39 for Morgan and Wolsiefer (1991), and W/B ratio of 0.5 to 0.55 for Pfeuffer and Kusterle (2001)). In that range of consistency, the efficiency of silica fume in diminishing losses was also verified in our study.

Dry Fine + SF

Intermediate Wet Consistency Medium + SF Coarse + SF

(b) (a) Rebound versus consistency for mixtures with no silica fume *the value includes loss of material due to debonding of material that did not rebound when first applied (b) Rebound versus consistency for mixtures with silica fume ** this value cannot be compared with others because the air flow was slightly higher.

The results at dry consistency might be explained by the fact that silica fume is known to consume more water than cement (Cyr, 1999; Pfeuffer and Kusterle, 2001). This might lead to a stiffer in-place mixture and thus higher rebound. This property also means that more water can be added before reaching wettest stable consistency compared to a mixture without silica fume. As Pfeuffer and Kusterle (2001) explained, this will form a plastic suspension with the binder, favouring adhesion of the impacting flow of material, and thus lowering rebound at the wettest consistencies. 3.3

Porosity and compressive strength Julie ARMENGAUD*, Marc JOLIN**, Antoine GAGNON**, Géraldine CASAUX-GINESTET*, Martin CYR*, Bernard HUSSON*

5


3.3.1 Porosity On Figure 4, porosity is presented according to mixture and consistency. It appears that the lowest porosity occurred for a Coarse mixture with silica fume at dry consistency. For a given consistency, Coarse mix with silica fume was always better. This might be due to compaction induced by coarse aggregates combined with a filler effect of silica fume. Nevertheless, when mixtures with 100% OPC were compared, it was the Medium mixture that showed the lowest porosity. This means that the Medium aggregate grade ensured the best compaction with the 100% OPC binder. 16%

Consistency Dry Intermediate Wet

14% Porosity

12% 10% 8% 6% Fine

Fine + SF

Medium

Medium + SF

Coarse

Coarse + SF

Mixture Porosity versus mixtures for different consistencies

Fig. 4

When silica fume was used, it can be noted that the variation in porosity was higher than without SCM. This may be related to the high variation of W/B that can be achieved with silica fume, and thus the high variation in porosity. However, mixtures with no silica fume did not show a clear variation of porosity with consistency; this might be explained by the smaller variation in W/B. 3.3.2 Compressive strength

Compressive strength [MPa]

Figure 5 shows compressive strength after 28 days of curing at >95% humidity versus consistency for the 6 mixtures. It can be noted that strength was always higher for dry consistencies whether silica fume was used or not. The highest value was obtained at dry consistency for the Coarse mixture with silica fume (when porosity was lowest) and the lowest values were obtained for wet consistencies, also with silica fume. As proposed for the results on porosity, this might be due to the large variation of W/B achievable for mixtures with silica fume. The value obtained for wet coarse mixture was relatively high and did not follow the global trend of decreasing strength when the consistency became wetter. This might be due to the particular placement that occurred for this mixture, as explained in section 3.2, which might have led to greater compaction. For dry and normal consistencies, silica fume was beneficial to the strength but, at wet consistency, the large amount of water added impaired strength. 65 60 55 50 45 40 35 Fine

Fine + SF

Medium Medium + SF Mixture Dry

Fig. 5 6

Intermediate

Coarse

Coarse + SF

Wet

Variation of 28-day compressive strength with mixture and consistency. Advanced materials


4

Discussion

Shotcrete is a composite material, the compaction is ensured by an impacting flow of material that also results in rebound and loss of material. This particular material can be studied by separating the aggregate phase and the paste phase. The interaction between these phases has been highlighted in this study by changing the properties of the paste with SCM and modifying the aggregate size distribution. Results on rebound show that there is a combined influence of aggregate size distribution and silica fume. It has been shown in section 3.2 that the efficiency of silica fume is dependent on the aggregate size distribution and on consistency. The rebound phenomenon has been described by Armelin (1997) for a single particle impacting a substrate, and depends on both substrate properties and the energy of the impactor. When considering the spraying process as a whole, where a multitude of impactors are involved, it is the energy of the spray that must be taken into account (Ginouse and Jolin, 2014). The framework that can be formulated to explain the results obtained in this study is that the type of binder and the amount of water act on the properties of the substrate while the aggregate size distribution leads to variation in the energy of the spray impacting the substrate. Reducing rebound is then a question of balancing these parameters. If the substrate is stiff, a large amount of energy will be necessary to penetrate it and reduce rebound, as shown on Figure 3(a): for dry consistency the lowest rebound occurs for the Coarse mixture. This aggregate size distribution may be the one that develops the most energy with the air flow used. Following this reasoning, the Medium mixture seems to give better results with intermediate to wet consistencies when no silica fume is used, while Fine mixture is more suitable with normal consistency when silica fume is used. This raises the question of the properties that can be expected of the substrate according to the energy developed in the spray. When silica fume is used, it seems that the optimal aggregate grade differs from that found with 100% OPC binder at a given consistency (the lowest rebound occurs for Fine mixture instead of Medium mixture at normal consistency). The most explicit example is provided by the Coarse mixture behaviour, which changes completely with silica fume. The mixture without SCM leads to high losses due to the expulsion phenomenon, while the same mixture with silica fume does not induce decohesion at any consistency. Silica fume is known to increase the water consumption at constant workability for cast concrete (Cyr, 1999) or at constant consistency in the case of shotcrete (Pfeuffer and Kusterle, 2001). This indicates that, at the same consistency as the reference mixture, a mixture with SCM will induce different interaction with water and give the substrate the capacity to absorb the energy of the impacting flow differently. This “energy consumption capacity” is indirectly considered in the theory of Armelin (Armelin and Banthia 1998) through a stress parameter measured with the dynamic penetration of a single impactor. From a rebound point of view, this means that an initially mediocre mixture (with an aggregate grade leading to high rebound) can be corrected by choosing an appropriate binder. At the same time, an initially good mixture, Fine or Medium for instance, might see its range of rebound increase because of a modification of its rheological properties. The optimization of the mixture by adapting the binder and the aggregate grade should also take the compaction phenomenon into account. Setting a goal of simply reducing rebound can interfere with porosity and compressive strength. This has been shown in the case of wet consistency, where the best porosity and compressive strength results were obtained using a Coarse mixture with silica fume for dry consistency, which was not the case for rebound. Some authors (Gérômey, 2003) have stated that higher rebound is a condition for low porosity but, as shown on Figure 4, for dry consistency, Coarse mixture with silica fume has lower rebound than the Fine and Medium mixes, and also has the lowest porosity. Thus compaction is not related only to rebound. For further research, the question of the energy of the flow has to be considered in greater depth. A few studies have already started (Ginouse and Jolin, 2015), and a link should be made with the fresh properties of the mixture under impact. 5

Conclusion

Shotcrete is a composite material, the compaction of which is ensured by an impacting flow of material that also produces rebound and loss of material. It can be studied by considering the aggregate Julie ARMENGAUD*, Marc JOLIN**, Antoine GAGNON**, Géraldine CASAUX-GINESTET*, Martin CYR*, Bernard HUSSON*

7


phase and the paste phase as separate parts that interact with each other during the placement and have a combined influence on rebound. It has been shown that the influence of the aggregate size distribution is dependent on the paste properties. These properties can be modified by adding silica fume and changing the water/binder ratio. Rebound can be reduced by adjusting the aggregate grade to the paste properties and vice versa. However, it has also been shown that seeking to reduce rebound might result in impaired porosity and strength. A possible way forward would be to determine the properties expected of the paste as a function of the energy of the impacting flow. Acknowledgements The authors are grateful to King Shotcrete Solutions. References Armelin, Hugo Sogayar. 1997. “Rebound and Toughening Mechanisms in Steel Fiber Reinforced Dry-Mix Shotcrete.” PhD diss., University of British Columbia, Canada. Armelin, Hugo Sogayar, and Banthia, Nemkumar. 1998. “Mechanics of Aggregate Rebound in Shotcrete (Part I).” Materials and Structures 31: 91–98. Bindiganavile, Vivek, and Banthia, Nemkumar. 2000. “Rebound in Dry-Mix Shotcrete : Influence of Type of Mineral Admixture.” ACI Materials Journal 97 (2): 1–6. Bolduc, Louis-Samuel. 2009. “Etude des Propriétés de Transport du Béton Projeté.” M.Sc., Université Laval, Canada. Cyr, Martin. 1999. “Contribution à la Caractérisation des Fines Minérales et à la Compréhension de leur Rôle joué dans le Comportement Rhéologique des Matrices Cimentaires.” PhD diss., INSA Toulouse, France, Université de Sherbrooke, Canada. Figueiredo, A. D. 1999. “Rheological Behavior of Dry-Mix Shotcrete.” Paper presented at the Second CANMET/ACI International Conference, High-Performance Concrete - Performance and Quality of Concrete Structures SP-186. Michigan : American Concrete Institute: 113–28. Gérômey, Sylvie. 2003. “Evaluation des Paramètres d’Obtention de la Qualité des Bétons Projetés utilisés dans des Soutènements Provisoires, des Revêtements Définitifs et des Renforcements d’Ouvrages.” PhD diss., INSA Lyon, France. Ginouse, Nicolas, and Jolin, Marc. 2014. “Effect of Equipment on Spray Velocity Distribution in Shotcrete Applications.” Construction and Building Materials 70. Elsevier Ltd: 362–69. doi:http://dx.doi.org/10.1016/j.conbuildmat.2014.07.116. Ginouse, Nicolas, and Jolin, Marc. 2015. “Investigation of Spray Pattern in Shotcrete Applications.” Construction and Building Materials 93. Elsevier Ltd: 966–72. doi:http://dx.doi.org/10.1016/j.conbuildmat.2015.05.061. Johnson, K. L. 1985. Contact Mechanics. Cambridge University Press. Jolin, Marc. 1999. “Mechanisms of Placement and Stability of Dry Process Shotcrete.” PhD diss., University of British Columbia, Canada. Jolin, Marc, and Beaupré, Denis. 2004. “Effects of Particle-Size Distribution in Dry Process Shotcrete.” ACI Materials Journal 101-M15 (March-April): 131–35. Morgan, D R, and Wolsiefer, J T. 1991. “Silica Fume in Shotcrete.” Paper in CANMET/ACI International Workshop on the Use of Silica in Concrete. Washington, DC. Parker, Harvey W., Fernandez-Delgado, Gabriel, and Lorig, Loren J. 1976. “A Practical New Approach to Shotcrete Rebound Losses.” ACI-SP54- Shotcrete for Ground Support. Pfeuffer, Markus, and Kusterle, Wolfgang. 2001. “Rheology and Rebound Behaviour of Dry-Mix Shotcrete.” Cement and Concrete Research 31 (July): 1619–25. Prudêncio, Luiz R, Armelin, Hugo Sogayar, and Paulo Helene. 1996. “Interaction between Accelerating Admixtures and Portland Cement for Shotcrete : The Influence of the Admixture’s Chemical Base and the Correlation between Paste Tests and Shotcrete Performance.” ACI Materials Journal 6 (93): 619–28. Puri, U. C., and Uomoto, T. 1999. “Numerical Modeling- A New Tool for Understanding Shotcrete.” Materials and Structures 32 (May): 266–72.

8

Advanced materials