A dynamic wear-fatigue coupled iterative model for probabilistic life prediction of rolling bearings

The traditional Lundberg-Palmgren (L-P) theory deviates from actual bearing service life by ignoring wear-fatigue coupling and adopting a constant-geometry assumption. To address this, a dynamic wear-fatigue coupled iterative (DWFCI) model is proposed. It integrates the Archard wear model to quantify raceway material loss and establishes a wear-depth-to-curvature mapping to dynamically update geometric parameters. An hour-level dynamic iterative algorithm enables bidirectional updates between wear evolution and contact stress redistribution. By incorporating S-N curves and Miner’s rule, the model predicts bearing life at the cumulative damage failure threshold. Validation against full-life accelerated degradation tests of LDK-UER204 and MB ER-16K bearings demonstrates the model’s high fidelity: relative errors for the mean life and standard deviation are constrained within 0.0103-0.239 and 0.0392-0.2041, respectively. Anderson-Darling goodness-of-fit tests further confirm statistical significance, with p-values >0.05 indicating no statistically significant difference between predicted and experimental life distributions, whereas the non-wear model yields p-values <0.05. This framework provides a robust basis for bearing prognostics under complex service conditions.