Application of Variational Mode Decomposition and Multiscale Permutation Entropy in Rolling Bearing Failure Analysis

Abstract

The rolling bearing fault test signal has nonstationary and nonlinear characteristics. The feature extraction method based on variational mode decomposition (VMD) and permutation entropy can effectively measure the regularity of the signal and detect weak changes. Since the center frequency of the intrinsic mode function (IMF) of each fault test signal contains more details, this paper further extracts the multiscale permutation entropy feature for each IMF. The training samples and test samples of each IMF are constructed, and then the support vector machine (SVM) and the K-nearest neighbor algorithm (KNN) are used to identify the faults. The test results of the IMF components are used to determine the classification results combined with the maximum attribution index. Compared with the relevant feature extraction, the experimental results show that the method achieves a certain improvement in the accuracy of fault identification. The research results of rolling bearing fault data show that the multiscale permutation entropy and SVM/KNN can more accurately diagnose different fault modes, different fault sizes, and different operating states of rolling bearings.