A Novel Method of Fault Diagnosis for Rolling Bearing Based on Dual Tree Complex Wavelet Packet Transform and Improved Multiscale Permutation Entropy

Abstract

A novel method of fault diagnosis for rolling bearing, which combines the dual tree complex wavelet packet transform (DTCWPT), the improved multiscale permutation entropy (IMPE), and the linear local tangent space alignment (LLTSA) with the extreme learning machine (ELM), is put forward in this paper. In this method, in order to effectively discover the underlying feature information, DTCWPT, which has the attractive properties as nearly shift invariance and reduced aliasing, is firstly utilized to decompose the original signal into a set of subband signals. Then, IMPE, which is designed to reduce the variability of entropy measures, is applied to characterize the properties of each obtained subband signal at different scales. Furthermore, the feature vectors are constructed by combining IMPE of each subband signal. After the feature vectors construction, LLTSA is employed to compress the high dimensional vectors of the training and the testing samples into the low dimensional vectors with better distinguishability. Finally, the ELM classifier is used to automatically accomplish the condition identification with the low dimensional feature vectors. The experimental data analysis results validate the effectiveness of the presented diagnosis method and demonstrate that this method can be applied to distinguish the different fault types and fault degrees of rolling bearings.