A modeling framework for rolling contact fatigue analysis in rough rolling elements
Mostafa Gargourimotlagh is a PhD student in the department Surface Technology and Tribology. (Co)Promotors are prof.dr.ir. M.D. de Rooij and dr. D.T.A. Mathews from the Faculty Engineering Technology, University of Twente.
Rolling Contact Fatigue (RCF) is a primary failure mechanism in rolling elements across critical engineering applications such as rail-wheel systems, bearings, steel rolls, and gears. With increasing demands for higher load-bearing capacities, accurate failure prediction and optimized maintenance strategies are essential. The complex nature of RCF stems from the nonlinear interaction of multiscale parameters, material inclusions, surface roughness, lubrication regimes with transient, multiaxial loading conditions. Given the impracticality of experimentally evaluating all parameter combinations, numerical modeling is indispensable for both prediction and the development of design and material innovations.
This study emphasizes surface roughness as a key factor influencing RCF behavior. Originating from machining processes, surface roughness impacts traction, adhesion, and localized stress fields near the contact zone. Its inherently multiscale and anisotropic nature, particularly pronounced in directionally machined surfaces such as those produced by hard turning or grinding, necessitates modeling approaches that surpass traditional 2D analyses. To this end, a 3D boundary element method (BEM) is employed, using surface topography as direct input from real measurements. This enables realistic simulation of asperity interactions based on datasets exceeding one million points, with sub-micron resolution. Since surface topography directly governs the contact pressure distribution at the interface of a rolling pair, incorporating it accurately is vital for reliable predictions. The proposed contact model introduces three key features. First, it achieves high mesh resolution and a fully 3D simulation framework that allows efficient and detailed calculation of contact pressure fields. Second, the proposed methodology decouples the contact model, considering surface topography at the micro-scale and the stress model that accounts for material and geometrical nonlinearities in the bulk of rolling elements. Third, the model incorporates scale-dependent hardness by distinguishing between asperity-level and bulk hardness, recognizing that local deformation behavior cannot be captured accurately using bulk material properties alone. Surface plastic deformation is handled efficiently by allowing plastically yielded nodes to settle into the contact plane after convergence, maintaining computational efficiency despite the fine resolution.
The transient running-in phase, though often overlooked, has a significant impact on the evolution of RCF. During this early period, surface roughness, residual stress, and hardness undergo notable changes, which in turn alter the stress–strain response and influence crack initiation and propagation behavior. To capture this effect, the contact model was utilized to analyze over 30 experimental RCF tests totaling more than 50 million cycles. This effort led to the development of a trend line characterizing the evolution of contact characteristics during running-in, linking initial surface plasticity to systematic changes in roughness statistics (e.g., RMS, skewness, kurtosis) required to reach steady-state behavior. The model’s predictive accuracy was validated in two key ways. First, surfaces with no predicted plastic deformation showed negligible topographical changes through subsequent rolling passes, consistent with purely elastic contact. Second, asperity-level features initially predicted to deform plastically were progressively worn away during testing, demonstrating a transition toward elastic shakedown, where the contact response stabilizes. These findings confirm that accurate modeling of the running-in phase is essential for realistic RCF life predictions, and highlight the importance of accounting for transient plasticity in surface evolution models.
The experimental RCF results obtained in this study demonstrate that surface topography and material hardness have a critical influence on plastic deformation, crack initiation, and propagation in steel rolling elements. Under identical loading and testing conditions, fatigue wear emerged as the dominant degradation mechanism, primarily governed by asperity-level features. Surface roughness affects the orientation and location of crack initiation, while hardness and evolving surface features control crack growth rates. The results also show that RCF and wear in the tested materials are closely linked, where crack initiation and propagation significantly influence both wear rate and the formation of wear particles. Although the tests were conducted under pure rolling or low traction conditions (traction coefficient as low as 0.07), measurable wear was observed. This indicates that the wear mechanism cannot be fully described using traditional Archard’s wear law and instead requires a more nuanced understanding of RCF behavior. Furthermore, these insights were incorporated into a combined BEM/FEM modeling framework, which used a hardness-dependent stress–strain approach to simulate materials with a similar chemical composition but varying hardness. The simulations revealed that the interaction between roughness and hardness determines whether failure occurs in the low- or high-cycle fatigue regime. The results demonstrate that higher hardness is beneficial in the high-cycle fatigue regime (lower surface roughness), where the material fatigue limit dominates. In contrast, at higher roughness levels, the system shifts to the low-cycle fatigue regime, where higher hardness may accelerate crack initiation due to the asperity-driven contact stresses. A critical surface roughness threshold was identified for crack initiation suppression. As hardness increases from 526 HV to 762 HV, this threshold shifts from 0.1 𝜇𝑚 to 0.3 𝜇𝑚. Beyond 700HV, the threshold saturates at 0.3 𝜇𝑚, indicating minimal additional benefit from further hardness increase. These findings underscore the necessity of jointly optimizing surface finish and hardness to improve RCF performance.
Finally, to accurately predict crack propagation, the framework integrates both manufacturing-induced and plasticity-induced residual stresses. These are updated dynamically over rolling passes, reflecting the evolving material state. Incorporating the running-in phase further enhances predictive accuracy by accounting for early-stage transformations in surface and subsurface conditions. This comprehensive approach enables realistic modeling of crack growth, providing insights into how different surface textures, such as hard-turned versus sandblasted, affect crack initiation sites and propagation paths.
