Reliability and local sensitivity analysis of rolling bearings based on adaptive Kriging model with dynamic resampling strategy

Rolling bearings, as fundamental mechanical components, may develop surface defects during operation, leading to excessive vibration and compromised equipment performance. To address this issue, this study establishes a reliability and local sensitivity analysis framework for defective bearings. Specifically, an improved spherical defect profile model is developed by incorporating time-varying displacement excitation and contact stiffness excitation. A dynamic model for defective bearings is subsequently validated using the Case Western Reserve University (CWRU) bearing dataset. For enhanced computational efficiency, we propose a reliability and local sensitivity assessment method based on Adaptive Kriging-Monte Carlo Simulation with Dynamic Resampling Strategy (DRS-AK-MCS). Results demonstrate that under varying input dimensions, the DRS-enhanced Kriging model converges within 65, 72, and 78 iterations, achieving errors of 0.0574 %, 0.0471 %, and 0.0018 % respectively. This confirms the method’s effectiveness and robustness. Application of the framework reveals failure probability evolution patterns relative to defect sizes and identifies critical rankings of influencing factors. The proposed approach provides theoretical foundations for bearing optimization design, demonstrating balanced computational accuracy and efficiency with significant engineering applicability.