Life risk prediction and evaluation of electric drive input shaft bearings under naturalistic driving conditions

Electric drive input shaft bearings operate under highly variable speed and torque conditions, and their fatigue risk is governed by both load magnitude and the temporal organization of driving and regenerative braking states. This study proposes a hybrid physics-based and data-driven life risk assessment framework using preprocessed 1 Hz motor state, speed, and torque data from 87 battery electric vehicles under naturalistic operating conditions. A standardized shaft bearing model was established to convert motor-side operating data into bearing loads through helical gear force decomposition and support equilibrium analysis. Corrected equivalent dynamic loads and instantaneous damage intensities were calculated to construct an eight-state representation combining damage severity and operating direction. Vehicle-level Markov features were extracted for operating pattern clustering and life risk index construction. The results show that the vehicles can be divided into three interpretable operating clusters. Cluster 1 represents a low damage stable pattern with a higher proportion of mild damage states and relatively low bearing loads. Cluster 2 represents an intensified medium to high damage pattern, characterized by more frequent high damage residence, longer persistence, and stronger drive regeneration switching, together with higher loads and shorter corrected life indicators. Cluster 3 shows a drive dominated moderate damage pattern, with risk characteristics between Clusters 1 and 2. The proposed life risk index provides stable relative risk ranking and distinguishes vehicles with different bearing life risk levels. The sensitive operating region forms a broad speed-dependent band, indicating that different combinations of speed and torque can produce disproportionate fatigue damage. The framework provides an interpretable tool for fleet-level bearing risk screening and duty cycle simplification.