Structural Integrity and Reliability Assessment of Railway Axles and Wheelsets - Researches at PoliMi Stefano Beretta Politecnico di Milano, Dipartimento di Meccanica
[email protected] March 2013
Abstract Railway axles are a longstanding problem for structural integrity, since they are designed against fatigue limit but their very long service life (30 year or 3·106 [km]) exposes them to agents (impacts, corrosion) that can trigger the growth of cracks that severely reduce the service life. In the following a statement of the problem and a summary of the researches at the Department of Mechanical Engineering at Politecnico di Milano are presented. Similar exploitations for other wheelset components are also mentioned.
1
Statement of the problem
Railway axles are commonly operated over a service life of 30 years or more which refers to a very high number of loading cycles in the order of some 109 . They are designed for an infinite life with admissible stress levels which correspond to generous safety factors applied to full-scale fatigue properties of materials (see EN13103, EN13104). Nevertheless, some failures occur due to corrosion, ballast impacts [1, 2]. It should be emphasised that railway axles are usually safe. In the foreword to a special issue of the Journal of Rail and Rapid Transit from 2004 the editor noted: "their safety record is generally second to none." [3]. Nevertheless, railway axles are safety-critical components. When an axle breaks there is a high risk of derailment with potentially disastrous consequences. According to the Railway Safety Performance Report of the European Railway Agency (ERA) of 2011 [4] the number of broken axles in the European Union between 2006 and 2009 was in total 329. Related to the number of about 1.66 × 1010 train kilometres over these four years this refers to one fracture event per 50.45 Millions train kilometres: for an average train size of 40 axles (which is an average value over the last years) the number of axle failures per axle kilometres
1
would reduce to approximately one fracture event per two Billions of axle kilometres. This means that improving the reliability of axles needs the careful implementation of new design concepts(i.e. the adoption of damage tolerance) or new monitoring strategies (see Bruni [5]).
2
Damage tolerant design of railway axles
In order to avoid the rare failures, which are related to a number of factors not easily predictable, the solution is to define a suitable inspection strategy supported by the damage tolerance analyses [6]. The application of damage tolerance concepts to railway axles has been the subject of an intense European cooperation which begun in 2004 with a Special Issue of the journal J. Rail Rapid Transit and the constitution of TC24 (Structural Integrity of Railway Structures) of ESIS - European Structural Integrity Society, where the technical issues about the development of the damage tolerance of railway axles have been discussed in these years. According to Zerbst [6] the key factors for a damage tolerance analysis are: i) component geometry and dimensions, ii) location and orientation of potential cracks or crack-like defects, iii) loading including secondary loads such as thermal or residual stresses, iv) deformation characteristics of the material (its stress-strain curve) and crack resistance of the material, v) fatigue crack extension characteristics (da/dN − ∆K curve) of the material, vi) initial crack size due to the non-destructive inspection (NDI) detection limit or service experience, vii) initial crack shape, and viii) probability of crack detection (POD) versus crack size characteristics of the NDI technique applied. The research activity of the research group led by Prof. S. Beretta has addressed some of the above points and the results are summarized in the following sections. Some other points are currently being analyzed.
Component geometry and secondary loads A number of SIF solutions for cracks on smooth axles were initially determined within the EU project WIDEM [8]. The analysis of SIF at the typical axle geometrical transitions was then developed in cooperation with Prof. U. Zerbst (BAM - Berlin, Germany). The main results of the activity are: i) the press-fits of wheels and gears induce, in the groove between them (see Fig.1), positive residual stresses which are very detrimental for propagation lifetime [7]; ii) a compendium of SIF solutions for common axle transitions [9].
Crack growth properties and algorithm Characterization of the axle steels EA1N and EA4T begun under the EU project WIDEM. The first results onto EA1N showed that the growth rate and thresholds obtained with small SE(B) specimens had a
2
F
F
Figure 1: A railway axle: FE model for a crack at groove between the wheel and the gear press-fittings [7]. wide scatter, especially in the region of near-threshold propagation [10]. Then the data on small scale specimens were significantly different to crack growth data obtained onto full-scale axles: in order to overcome this problem new SE(T) specimens with the same constraint of the full-scale axle were designed [11]. The latest findings (within the EU project WOLAXIM) show that this scale effect is less pronounced when compression-precracking is adopted for tests on small scale specimens. The new SE(T) specimens were then adopted for other research projects for investigating the crack growth retardation with different axle steels: the analysis of the tests carried out within WIDEM onto EA1N steel show that full-scale axles (tests at low ∆K levels) show a significant retardation, while tests onto SE(T) specimens at higher ∆K show almost no retardation [12].
Design optimization: ’One million miles axle’ One of the expected results of WIDEM was to obtain a new axle design on the basis of the ’damage tolerance’ tools developed within the project. A simple concept derived from ’damage tolerance’ for a reliable axle design is to assure that during service lifetime a prospective crack could only grow of a few millimeters, so that failure probability and to the number of NDT inspections are drastically reduced (this design concept had been named one-million mile axle). In order to see the effect of changing axle design onto prospective propagation lifetime, the stress spectra that had been measured for a freight train have been multiplied by a K-factor (K > 1 implies
3
increase in the stress levels and therefore corresponds to decrease axle flexural modulus) in order to determine the stress levels at which there is a limited propagation, comparing this result with traditional fatigue damage calculations. The results are summarized in Fig. 6 (initial crack size a = 2 [mm]): it can be seen that K factors corresponding to traditional fatigue optimization imply a propagation lifetime much than 100, 000 [km]: this clearly show that ’damage tolerance’ is a much more stringent and rigorous criterion than traditional fatigue [13]. New design criteria which improve and formalize this concepts are currently being developed within the EU project EURAXLES.
Initial defect size and damage One the main factors in the initiation and incubation of a crack in railway axles is corrosion. A resarch project begun in 2005 has shown that the presence of rainwater is able to induce, onto A1N steel, the onset of corrosion-fatigue [14], the propagation of multiple small fatigue cracks under the combined action of cyclic loads and the mild corrosive medium, even at stress levels much lower than current design limits [15] (see Fig. 4). Similar results were also obtained for A1T steel within the RSSBT728 project (a reserach contract by RSSB to a consortium formed by: Deltarail-TWI-PoliMi) onto the A1T steel. In this project, the S-N curve obtained with small scale specimens was validated with fullscale tests. These tests also showed that corrosion-fatigue damage can be detected by the presence of small cracks and corrosion-craters: this idea is being exploited within the EU project WOLAXIM. The RSSB-T728 has also dealt with the statistical description of other kind of surface damages (impacts, surface deterioration), even if a fully
(a)
(b)
Figure 2: Crack growth rate of EA1N steel [11, 12]: a) small scale specimens vs. full-scale; b) data obtained with the new SE(T) specimens.
4
Figure 3: Prospective lifetime for a heavy-haul freight axle (initial crack size a = 2 [mm]) increasing the stress level by a factor K: comparison between propagation lifetime and traditional fatigue damage [13]. developed probabilistic model for does not exist yet.
Figure 4: S-N diagram for A1N under rainwater [15].
3 Current researches on axles structural integrity The research topics that are currently under development are: • Improvement of the POD curve A research project has been launched with LucchiniRS and Gilardoni: the scope of the project is to improve the performances of bore-probes currently adopted
5
(a)
(b)
Figure 5: Full-scale fatigue tests for A1T axles exposed to rainwater: a) detail of test set-up; b) the axle surface is covered with corrosion craters which have small cracks at the bottom [16].
for the NDT inspection of the railway axles. The reasearch has first dealt with preparing a set of cracked axles with natural cracks (nucleated under fatigue) in the size 1 − 10 [mm]. We are currently analyzing the response of the bore-probes in order to determine the POD as a function of defect area, instead of the crack depth, and the improvement of the current probes with computer-simulated response of the cracks in the critical regions of the axles. The final step of the research will be the definition of a new calibration sample. • New methods for inspection and reliability simulation of axles Within the EU project WOLAXIM [17], the research team of PoliMi is working on: i) cooperation with TWI for the development of a new optical device detecting the early stages corrosionfatigue damage and the validation of the device with full-scale experiments; ii) development of a new probabilistic model for the STRUREL software (in cooperation with Prof Straub TU Munich) able to incorporate the latest findings in crack growth modelling and to take into account both the occurrence of impacts and the onset of corrosion-fatigue damage as the events triggering crack propagation. • Improving axle residual lifetime A possibility to improve axle reliability is the adoption of a surface rolling able to induce compressive residual stresses which can retard (or inhibit) crack propagation. This is the subject of the project MARAXIL - Manufacturing Railway Axles with Improved Lifetime in coop-
6
Figure 6: Simulation of UT response of crack in critical axle locations. eration with LucchiniRS and IWM (Freiburg - Germany) where we are: i) simulating and measuring the residual stresses induced by rolling; ii) simulating crack growth in the rolled axles; iii) comparing the results with full-scale experiments; iv) exploiting the benefits of the process on life-extension of high-speed and freight axles [18]. • Improving axle design The know-how in modelling axle life and reliability is currently being exploited in the EU project EURAXLES [19] in terms of revising the design rules in the standards (EN13103, EN13104) in order to define new design rules, compliant with Eurocode, that can help designers in incorporating reliabiliity calculations at a design stage. The EU project SUSTRAIL [20] will be devoted to develop new freight axles with an improved lifetime able to increase the current axle capacity and to improve life cycle costs also with adoption of CBM concepts [5]. • Improving crack growth algorithm The latest experiments show that there is a significant retardation of crack growth applying typical load spectra of real axles in service both on SET specimens and on full-scale axles. Two different research projects with LucchiniRS and Deutche Bahn on two different axle steels are devoted to characterize this retardation: crack growth calculations considering this retardation would surely extend the
7
inspection intervals (and reduce the life cycle cost) for different applications.
4
Applications to other wheelset components
In terms of evaluation of the structural integrity and reliability we have been working with the group of Prof. Bruni for the assessment of different wheelset components following the scheme: a) simulation of the dynamic behaviour in different service conditions; b) combinations of simulations for obtaining a service spectrum; c) statistical extrapolation of the load spectrum using the ’statistics of extremes’; d) evaluation of prospective service life in terms of fatigue damage or crack propagation. In Fig. 7, a bogie frame that was analyzed following the above described scheme: the evaluation of prospective lifetime suggested the adoption of reinforcements for some critical areas.
Figure 7: A railway bogie: elements 5 − 8 were reinforcement plates in critical areas. The dynamics → assessment scheme was also adopted for evaluating the damage [21] (see Fig. 8) of a wheel due to RCF at the wheel-rail interface. In this case the service dynamic loads and contact positions along the wheel were analyzed to calculate the distribution of contact stresses along the wheel surface and then the consequent fatigue damage. A similar approach was also adopted for the ratchetting cumulative damage over the surface of a rail [22].
References [1] R.A. Smith and S. Hillmansen. Monitoring fatigue in railway axles. In Proc. 13th Wheelset Congress, 2001.
8
Figure 8: Assessment of the fatigue damage of a railway wheel. [2] Transportation Safety Board of Canada. Main track derailment: Canadian national train No.G-894-31-14. Railway Investigation Report R01Q0010, 2001. [3] S. Hillmansen. Editorial to the Special Issue. IMechE, Part F: J. Rail Rapid Transit, 218(4), 2004. [4] ERA. Railway Safety Performance in the European Union. Technical report, http://www.era.europa.eu/DocumentRegister/Pages/Railway-Safety-Performance-in-the-EuropeanUnion-2011.aspx, 2011. [5] S. Bruni. New Condition-Based Monitoring Techniques for High Speed Railway Vehicles. Technical Note, Jan. 2012. [6] U. Zerbst, M. Vormwald, C. Andersch, K. Madler, and C. Pfuff. The development of a damage tolerance concept for railway components and its demonstration for a railway axle. Engng. Fract. Mechanics, 72:209–239, 2005. [7] M. Madia, S. Beretta, and U. Zerbst. An investigation on the influence of rotary bending and press fitting on stress intensity factors and fatigue crack growth in railway axles. Engng. Fract. Mechanics, 75:1906–1920, 2008. [8] http://www.widem.eu. [9] M. Madia, S. Beretta, M. Schodel, U. Zerbst, M. Luke, and I. Varfolomeev. Stress intensity factor solutions for cracks in railway axles. Engng. Fract. Mechanics, 78:764–792, 2011. [10] S. Beretta and M. Carboni. Experiments and stochastic model for propagation lifetime of railway axles. Engng. Fract. Mechanics, 73(2627-2641), 2006. [11] M. Carboni, S. Beretta, and M. Madia. Analysis of crack growth at R = -1 under variable amplitude loading on a steel for railway axles. J. ASTM Int., 5, 2008.
9
[12] S. Beretta and M. Carboni. Variable amplitude fatigue crack growth in a mild steel for railway axles: Experiments and predictive models. Engng. Fract. Mechanics, 78:848–862, 2011. [13] S. Beretta, M. Carboni, and S.Cervello. Design review of a freight railway axle: fatigue damage versus damage tolerance. Materialwissenschaft und Werkstofftechnik, 42:1099–1104, 2011. [14] S. Beretta, M. Carboni, A. Lo Conte, and E. Palermo. An investigation of the effects of corrosion on the fatigue strength of AlN axle steel. IMechE, Part F: J. Rail Rapid Transit, 222:129–143, 2008. [15] S. Beretta, M. Carboni, G. Fiore, and A. Lo Conte. Corrosion– fatigue of A1N railway axle steel exposed to rainwater. Int. J. Fatigue, 32:952–961, 2010. [16] M. Carboni, S. Beretta, and A. Loconte. Research on corrosion fatigue of railway axles. Insight, 53(361-367), 2011. [17] http://www.wolaxim.eu/project/overview.jsp. [18] http://maraxil.mecc.polimi.it. [19] www.euraxles.eu. [20] www.sustrail.eu. [21] A. Bernasconi, P. Davoli, M. Filippini, and S. Foletti. An integrated approach to rolling contact sub-surface fatigue assessment of railway wheels. WEAR, 258:973–980, 2005. [22] S. Foletti, S. Beretta, and G. Bucca. A numerical 3D model to study ratcheting damage of a tramcar line. WEAR, 268:737–746, 2010.
10
