A two-stage Fault Diagnosis Scheme of Partial Demagnetization Faults in PMSMs with Low Labelled Rates

This study proposes a two-stage diagnosis scheme for partial demagnetization faults in permanent magnet synchronous motors (PMSMs) based on a small portion of faulty historical data and one-phase stator current analysis. The proposed method is based on applying anomaly detection in the first stage, and semi-supervised learning (SSML) combining self-training and a support vector machine (SVM) in the second stage to use the healthy, faulty and unlabelled data for improving the overall diagnostic accuracy. The effectiveness of the suggested scheme is evaluated through multiple datasets having different labels ratio, and compared with other supervised machine learning (SML) and SSML algorithms.