Intelligent fault diagnosis of induction motors based on multi-objective feature selection using NSGA-II

Abstract

The aim of this paper is to present an intelligent method for fault diagnosis of induction motors which doesn't need any expert to analyze the signals. A new intelligent fault diagnosis scheme based on multi-objective feature selection using non-dominated sorting genetic algorithm II (NSGA-II) is proposed. Firstly, to improve the signal-to-noise ratio, wavelet packet decomposition is performed. Multiple statistical features are then extracted from the decomposed signals. Some of these features contain unhelpful information, so the most superior features are selected using NSGA-II. Finally, the classification of type and severity of faults is performed using a multilayer perceptron (MLP) neural network. The proposed scheme is tested on a bearing fault dataset, and the results show that it, unlike signal processing techniques, is able to detect the faults of induction motor without any expert. It also achieves a better classification rate comparing with the methods based on conventional feature selection algorithms.