Modeling of Material Anisotropy in Rolling Contact Fatigue
Rolling contact fatigue (RCF) is the primary mode of failure in tribological contacts like rolling-element bearings (REBs), gears, and cam-follower systems. RCF processes have a crack initiation phase followed by a propagation and coalescence phase, resulting in spalls that lead to catastrophic failure. Crack initiation is a highly localized process that is strongly influenced by the inhomogeneity of the material microstructure. Therefore, a microstructure-sensitive model is required to simulate the damage evolution and failure due to RCF loading. This document presents the development of a microstructure-based finite element (FE) framework for RCF, which accounts for the inhomogeneity of bearing steel microstructure by using an explicit definition of polycrystal topology and material anisotropy. The granular topology of the bearing steel microstructure is described using randomly generated Voronoi tessellations. A cubic elastic material definition with a random spatial orientation is specified for each Voronoi grain to simulate the material anisotropy. The Voronoi grains generated using this approach were used to model the critically stressed microstructural volume in RCF loading. A domain size study was conducted to estimate the minimum number of grains that need to be contained by the critically stressed volume such that the macroscopic material response of the polycrystalline aggregate matches the linear elastic material properties of bearing steel. The estimated critically stressed volume was then embedded into a semi-infinite domain for the FE simulation of RCF line contact loading. The RCF domains developed were then subjected to a moving Hertzian pressure over the surface to simulate a bearing load cycle. A boundary averaging scheme was used to estimate the effective stresses along the grain boundaries of the Voronoi cells. Due to the anisotropy of the polycrystalline material, local stress concentrations occur at the grain boundaries as compared to isotropic models. The resolved grain boundary stresses were used to predict critical locations for RCF crack initiation, which closely match observations from RCF bench test data. Since RCF failures typically exhibit subsurface locations for the first crack initiation, the model uses the critical resolved shear stress (RSS) reversal along the grain boundaries and the corresponding subsurface location of the maxima as the driving parameters for RCF fatigue failures. The parameters from the model were fit into a Weibull distribution to estimate the stochasticity in initiation life. The Weibull predictions corroborate well with experimentally measured RCF life scatter. The framework was then extended using a coupled damage mechanics - cohesive element method (DM-CEM) to individually model the crack initiation and propagation phases in RCF. An explicit definition of the grain boundaries was incorporated using cohesive elements. Damage is initiated at the grain boundaries by degradation of the cohesive elements and the rate of damage/degradation is used to characterize the evolution of fatigue life. The rate of damage was calculated at each grain boundary using a fatigue damage law based on the RSS reversal parameter. The model is able to simulate the crack initiation and the propagation/ coalescence phases in RCF, with distinct life estimates for each phase. This model framework is further extended to investigate the effects of lubrication conditions in RCF by integrating an elastohydrodynamic lubrication (EHL) model to simulate the pressure load with the DM-CEM model. Further improvements to the fatigue life predictions using the DM-CEM model are made by coupling it with a crystal plasticity (CP) based submodel approach to predict the crack initiation life in RCF. CP-based metrics are used to correlate the microplasticity developed under RCF loading with the formation of fatigue micro-cracks and the corresponding initiation life estimations. The resulting final spall patterns and RCF life estimates were found to match well with experimental data available in the open literature.