Research and Application of Intelligent Diagnosis Technology in Permanent Magnet Generator for Stress Demagnetization Fault

Abstract

Fault diagnosis before its existence and the complete shutdown is essentially critical for the whole industry. Fault diagnosis based on condition monitoring methods and artificial intelligence techniques are very potent. This paper assesses the machine-learning-based processes using air gap flux and stator current for eccentricity, magnet broken, and stator inter-turn short circuit faults in Permanent Magnet Generator (PMG). To apply machine learning, features are extracted via Discrete Wavelet Transform (DWT) technique for faulty and healthy conditions. Afterward, the classification learner toolbox in MATLAB is used to investigate various machine learning classifiers. The six fundamental classifiers comprising 23 sub-classifier algorithms are trained, whereby 16 out of 23 algorithms have achieved a perfect accuracy of (100 percent) while two have acquired an accuracy of more than 60 percent. The results indicate that air gap flux has performed better than stator current for fault diagnosis.