A Fault Diagnosis Method for Rotating Bearings Based on EWT Multi-Scale Entropy and PSO Algorithm to Optimize SVM

The mechanical fault diagnosis results of the rotary bearings are mainly determined by the feature vector and classifier used. In order to obtain more significant signal characteristics, a new mechanical fault diagnosis method is proposed. First, the empirical wavelet transform (EWT) method is used to decompose the rotary bearing vibration signals into several intrinsic modal functions (IMFs). Secondly, by applying a Hilbert transform to each IMF component, the frequency characteristics of signals are illustrated in a time-frequency representation. Thirdly, the multi-scale entropies (MSEs) of components being highly correlated with the original signals are calculated to construct the eigenvectors of rotary bearing signals. Finally, the optimal support vector machine (SVM) classifier is established by particle swarm optimization (PSO). The experimental results show that the diagnosis rate of unknown samples is as high as 95% by using the method proposed in this paper.