A probabilistic assessment method for fatigue reliability of bonded insulated rail joint under random rolling contact

Abstract

Bonded Insulated Rail Joints (BIRJs) face significant challenges due to uncertainties in load conditions and dynamic wheel-rail interactions, which affect their fatigue reliability. This study proposes a fatigue reliability assessment method for transient rolling contact in BIRJs. To reduce computational cost, a surrogate model based on the Whale Optimization Algorithm (WOA) and Artificial Neural Networks (ANN) is used, incorporating a multi-error control (MEC) strategy to enhance prediction accuracy of fatigue damage. The WOA-MEC-ANN approach is essential as it efficiently maps multiple fatigue model responses and better captures the uncertainties in wheel-rail contact, while maintaining accuracy. Shear cracks, primarily located near the endpost, are identified as the main failure mode based on the J-S fatigue model. Sensitivity analysis reveals that lateral contact position and speed are the most influential factors on fatigue life, which also significantly impact wheel-rail contact behavior. These findings provide important insights for predicting fatigue life and crack initiation in BIRJs, offering valuable references for real-world engineering practices and operational maintenance strategies.