An application of support vector machines for induction motor fault diagnosis with using genetic algorithm

Abstract

This paper introduces a technique for diagnosing mechanical faults of induction motors by using support vector machine (SVM) and genetic algorithm (GA). Features are extracted from the vibration time signals and selected by using GA with a distance evaluation fitness function. All SVM parameters are also obtained simultaneously by the same GA. The SVM is studied with two types of kernel functions, the radial basis function and the polynomial function. Four motor conditions are investigated with the chosen SVM classifiers. The classification results have high accuracy for the chosen feature set and SVM parameters. © 2008 Springer-Verlag Berlin Heidelberg.