Uncertainty variations and quantification of motor bearing time-varying resistance–capacitance characteristics

• A large-scale dataset of over ten million bearing current breakdown conduction events is analyzed. • The influence of motor operating parameters on bearing time-varying resistance–capacitance characteristics is revealed. • The inherent uncertainty in bearing time-varying resistance–capacitance characteristics is identified. • The uncertainty in resistance–capacitance characteristic parameters variations is quantitatively characterized. • An interpretable framework for constructing bearing constitutive models is proposed. During the operation of generators and drive motors, current breakdown conduction through bearings can cause electrical erosion, thereby reducing motor reliability and performance. This phenomenon results from mode-switching between the resistance–capacitance (RC) characteristics of the bearing. However, the transient variations of these time-varying RC characteristics remain insufficiently understood. To elucidate the uncertainty variation laws of bearing time-varying RC characteristics under realistic motor operating conditions, this study analyzes more than ten million bearing current breakdown conduction events across over five thousand transient and steady-state motor operating conditions. The influences and mechanisms of four key motor operating parameters—rotation speed, torque, rotation speed change rate, and torque change rate—on the time-varying RC characteristics (including equivalent resistance, equivalent capacitance, and RC mode-switching frequency) are systematically investigated. Based on these analyses, the boundary conditions of RC mode-switching are redefined, and a mixture density network is employed to model the mixed probabilistic distributions of RC characteristic parameters. Accordingly, a constitutive modeling framework based on interpretable descriptor surrogate models is developed to quantitatively characterize and probabilistically predict the uncertainty variations of bearing time-varying RC characteristics under both transient and steady-state operating conditions. This framework achieves high predictive accuracy while maintaining interpretability, simplicity, reliability, and generalizability. The proposed approach provides a theoretical foundation and modeling framework for the design and optimization of motor and bearing systems, offering potential strategies to suppress bearing currents and mitigate electrical erosion in generators and drive motors.