INNOTRACK Deliverable D4.3.5, Simulation of material deformation and RCF

Rapport, 2009

Testing of rail steels has a long tradition in the railway industry both in the form of full-scale and scaled tests. However test results from different test stands have largely been self-standing, which have made quantitative comparison between different test rigs and towards operational scenarios difficult. The main purpose of this deliverable is to investigate the possibility of bridging this gap by numerical simulation aimed at predicting rolling contact fatigue (RCF) initiation. Two modelling approaches are employed: Methods that aim at relating the evaluated contact stress distribution directly to the RCF initiation life are denoted “engineering approaches”. In contrast “finite element (FE) based approaches” evaluate the (elasto-plastic) material response owing to the acting contact stresses and relates the fatigue life to the resulting stresses/strains in the material. To obtain reliable RCF life predictions is found to currently be on the brink of what is possible. The report takes the approaches outlined above as far as possible and reports intermediate results, complications etc. In the report prediction based on engineering approaches sets out from previously derived contact stress distributions [5]. These are complemented by simulations of dynamic wheel–rail interaction in order to evaluate possible additional longitudinal tractive stresses due to the rolling on the wheel on the rail. For the case studied, these were found to be comparable to the lateral tractive stresses. It should here be noted that an evaluation of wheel–rail dynamics based on Hertzian contact theory is not sufficient for high-resolution evaluation of the contact stress distribution for the case studied. Further such an analysis is extremely sensitive to the geometry of the contacting surfaces. Supported by these simulations it was presumed that the contact conditions corresponded to full slip for all the three test rigs simulated. Based on this a shakedown map based fatigue index was evaluated and a plot of the fatigue index versus the fatigue life in a log-log diagram was shown to give a reasonable match to a straight line, which would indicate a Wöhler-like fatigue life relationship. If this relationship holds for further scrutinization it would be a major step forward for fatigue life prediction and comparisons between test rigs and towards operational scenarios. However the current analysis contains major uncertainties most notably in estimated wheel–rail friction. The FE based approaches sets out by calibrating an elasto-plastic material model of the rail steel towards experimental stress–strain data. FE-models of the two full-scale test rigs are then developed and load cycles featuring the loaded wheel rolling over an representative rail section are carried out. The stress–strain response in highly loaded material points on the rail surface are then evaluated and quantified. Methods to predict resulting fatigue life are then described both based on the presumption that the dominating fatigue mechanism is low-cycle-fatigue (in the sense that fatigue life is governed by the strain range) or ratcheting (in the sense that fatigue life is governed by accumulated plastic strain). The simulations highlighted the fact that the studied combination of very local high deformations, large displacements, high contact pressures and interfacial shear stresses, conformal contact and need for sophisticated constitutive models is currently extremely cumbersome to simulate. Reliable results for one test rig for two load conditions have been obtained after tedious work of tuning analysis parameters. These results show a high sensitivity also for rather low variations in applied lateral load magnitudes. Further the results show that plastic deformations in the FE-simulations introduce a level of “smoothing”, which levels out very high stress concentrations. On the other hand this also makes a priori identifications of critical material points difficult. The results finally show that the use of non-linear hardening seems crucial and that the proposed methods of fatigue assessment seem credible. In addition simulations of the second rig stumbled upon a bug in the commercial code, which resulted in an erroneous evaluation of the contact stresses. Although this rendered these results useless it highlighted the sensitivity and complexity of the problem at hand. Measures are currently taken to find a work-around for the identified bug. In summary, the current report is in our opinion a major step forward in comparing different bench tests to each other and to operational conditions by numerical simulation. Based on the results this should indeed be feasible although very complicated. However the main benefit of the report lies perhaps in the identification of the limitations of today’s top-notch approaches and simulation toolboxes. Finally a word of caution is needed: As the simulations here and in [5] clearly show, the problem at hand is inherently sensitive. Consequently the accuracy of predictions (be it numerical, empirical, experimentally-based etc) will always be limited.