Modeling and experimental characterization of large biaxial strains and induced anisotropy in pearlitic rail steel

Doctoral thesis, 2019

Large shear strains accumulate in the near-surface region under the running band of railway rails. In this region, rolling contact fatigue cracks often initiate, causing major problems for the railway industry. However, characterization of the constitutive and fatigue behavior of the material in this region is difficult due to the high strain gradient. The solution proposed in this thesis is to produce highly deformed cylindrical test bars: An axial-torsion test rig is used to predeform the bars in torsion while subjected to axial compressive loading. The obtained material state is found to be similar to that of field samples of rails at a depth between 50 and 100 μm. Using this predeformation method, the evolution of the yielding behavior is evaluated. The predeformed test bars are re-turned and drilled out to form thin-walled test bars, which can be used to measure yield surfaces. It is found that the degree of anisotropy quickly evolves with increasing predeformation and then saturates. Furthermore, the quadratic Hill yield criterion describes the anisotropic yield surfaces well.

To better optimize rail maintenance and material selection, there is an industrial need for a model capable of predicting rail deterioration. An important component of such a model is an accurate material model that captures the relevant physical phenomena. A hyperelasto-plastic framework for finite strain material models is adopted in this thesis. As a first study, the predeformation method was simulated using 2D axisymmetric elements. It is shown that very good results can be achieved by using material models with advanced kinematic hardening laws. Next, an improved simulation methodology for axial, torsional and pressure loading is developed, resulting in an efficient 1D formulation. This methodology includes material removal to simulate the re-machining of the test bars into tubular bars. Using this methodology, 3 different distortional hardening models are evaluated in terms of how well they fit and predict the experimental data. The two phenomenological models perform better than the crystal plasticity model. However, these models should be further developed to improve their predictive abilities.