Fault Diagnosis of Rolling Bearing Based on WP Reconstructed Energy Entropy and PSO-LSSVM

Abstract

A fault diagnosis method based on wavelet packet (WP) reconstruction of energy entropy, particle swarm optimization (PSO) and least squares support vector machine (LSSVM) is proposed for non-stationary vibration signals of rolling bearings. Firstly, the vibration signal is preprocessed, followed by 3-layer wavelet packet decomposition, and the energy entropy percentage of the reconstruction coefficient is extracted as the feature vector. Then, the 8-dimensional fault feature vector is reduced to a 2-dimensional feature vector by principal component analysis (PCA). Finally, the 2-dimensional feature vector is taken as the input sample of PSO-LSSVM. In order to diagnose the three fault states of the inner ring, the ball and the outer ring of the rolling bearing, four LSSVM classifiers are established. After the simulation analysis of the bearing vibration data, the diagnostic accuracy rate of the LSSVM multi-classifier group was 100%, which proves the feasibility and effectivity of the method.