Comprehensive multiphysics model of an induction motor for generating synthetic data under diverse bearing faulty conditions

Abstract

This study proposes a comprehensive multiphysics model for an induction motor (IM) that can generate high-fidelity synthetic fault data to mitigate data scarcity in data-driven fault-detection methods. The proposed model integrates three physical characteristics to replicate bearing faults. First, a two-dimensional electromagnetic model calculates the electromagnetic force based on the magnetic flux density. Second, a three-dimensional structural dynamic model generates dynamic responses of the IM under various health conditions. Finally, a bearing model calculates bearing forces while considering ball deformation to replicate outer-race, inner-race, ball, and mixed faults. Subsequently, the bearing forces are fed back into the structural dynamic model, thus ensuring accurate dynamic response replication. Extensive validation through experiments under different bearing failure modes demonstrates the accuracy of the model in replicating IM dynamics with varying fault severities. This capability highlights the potential of the model in generating high-fidelity synthetic fault data for fault detection methods.