Fault Diagnosis and Separation of PMSM Rotor Faults Using Search Coil Based on MVSA and Random Forests

In this article, a novel rotor fault diagnosis and separation approach based on search coils (SCs) for permanent magnet synchronous motors is proposed. First, the SC setup is introduced along with its working principle and spatial arrangement. Based on this special structure, the local magnetic field distortion caused by the local demagnetization fault (LDF) of permanent magnets or the dynamic eccentricity fault (DEF) can be accurately diagnosed. The fault characteristics of both LDF and DEF are the existence of fractional harmonics in the SC induced voltage signal. Compared to the traditional phase back electromotive force method, these features do not depend on the motor pole-slot combination and are not be hidden due to periodic overlap. Second, to distinguish between two faults with similar characteristics and considering the inherent errors of healthy motors due to processing tolerances, two random forest (RF) models are chosen to gain automated classification. Finally, to simplify the model and improve diagnostic efficiency, the features are screened and the final diagnosis scheme is established. Simulation and experimental results have validated the accuracy and the superiority of the proposed method.