Fault detection techniques on active magnetic bearing systems and electrical machines: a review

Abstract

This paper provides a review on fault detection techniques on active magnetic bearing systems and electrical machines. The following non-linear processing fault detection techniques are discussed in this paper: 1) Time domain analysis, 2) Frequency domain analysis, 3) Time-frequency analysis, and 4) Feature analysis. Time domain analysis is discussed and broken up into data collection, time domain features and Weibull distribution. Frequency domain analysis is discussed and broken up into Cepstrum analysis, enveloped spectrum analysis, equi-sampled discrete Fourier transform, high frequency resonance technique, shock pulse analysis and spike energy analysis. Time-frequency analysis is discussed and broken up into short-time Fourier and bilinear transform, which includes Wigner-Ville distribution. Feature analysis is discussed and broken up into artificial neural networks, feature selection and extraction, feature set reduction, fuzzy logic and pattern recognition. This paper focuses only on nonlinear fault detection techniques and ends with a conclusion on the discussed fault detection techniques for active magnetic bearing systems and electrical machines