A fuzzy logic approach to the interpretation of higher order spectra applied to fault diagnosis in electrical machines

Abstract

This paper describes the application of fuzzy logic based artificial intelligence procedures to the development of a novel method for the condition monitoring and fault diagnosis of induction motors. In the proposed scheme, higher order statistical (HOS) analyses are used as a pre-processing procedure applied to a machine vibration signal. Such analyses yield power spectral density, bispectrum, and bicoherence signatures for the vibration characteristics. A combination of data reduction, parameterisation and fuzzy logic procedures is then applied to the HOS signatures to enable diagnosis of the machine fault. Results are presented which demonstrate the effectiveness of the proposed procedure and resulting system for diagnosing a number of induction motor faults. For comparison purposes, the performance of diagnostic procedures developed using artificial neural network (ANN) based and conventional classification approaches are also briefly discussed.