Fault Diagnosis of Bearing Based on Wavelet Packet Transform-Phase Space Reconstruction-Singular Value Decomposition and SVM Classifier

Abstract

The excellent features of bearing vibration signal are helpful to obtain accurate diagnosis results for the failure of bearing. In this study, the feature extraction method of bearing vibration signal based on wavelet packet transform-phase space reconstruction-singular value decomposition (WPS) is presented to improve the traditional feature extraction method of bearing vibration signal based on wavelet packet transform-singular value decomposition (WS). In the proposed feature extraction method, singular value decomposition is performed for phase space reconstruction signal of each wavelet packet coefficient’s reconstructed signal of bearing vibration signal. The dynamic characteristics of a certain frequency range can be reflected by phase space reconstruction for wavelet packet coefficients’ reconstructed signals of bearing vibration signal. Support vector machine (SVM) is a machine learning method based on structural risk minimization principle, and SVM classifier can solve the classification problems with small training samples, high dimensions, and nonlinearity. Thus, the SVM model of bearing is established by the features of bearing vibration signal based on wavelet packet transform-phase space reconstruction-singular value decomposition in this study. The experimental results show that the feature extraction method of bearing vibration signal based on wavelet packet transform-phase space reconstruction-singular value decomposition is better than the feature extraction method of bearing vibration signal based on wavelet packet transform-singular value decomposition, and SVM established by the features of bearing vibration signal based on wavelet packet transform-phase space reconstruction-singular value decomposition (WPS-SVM) has a stronger fault diagnosis ability of bearing than SVM established by the features of bearing vibration signal based on wavelet packet transform-singular value decomposition (WS-SVM).