Machine Learning Based Bearing Fault Detection of IM Fed by Variable Frequency Drive Using Motor Current Signature Analysis

Abstract

This work presents machine learning-based bearing defect detection of three-phase induction motor fed by variable frequency drive. Multi band-pass filters, fast Fourier transform (FFT) and machine learning algorithms have been used to detect whether or not the bearing is damaged based on the Motor Current Signature Analysis. The proposed method is developed using acquired stator current data from a simulation model, subjected to healthy and faulty cases under different operating frequencies and motor loadings. The inverter-fed motor monitoring is much noisier than the utility-driven motor, which could hide fault signs and result in an incorrect fault classification. Multi band-pass filters and FFT are applied to extract features from stator current signals and feed them to machine learning classifiers to detect the fault. The results showed that the proposed method could provide an accurate diagnosis of the bearing health of the induction motor.