Intelligent identification method for meshing impact faults in electric vehicle reducers based on transmission error modal decomposition and multi-feature fusion

Abstract

The transmission error (TE) signal of electric vehicle reducer gears directly reflects meshing impact fault characteristics. To address the complexity of TE signals, this paper proposes a gear meshing impact fault diagnosis method based on multi-feature fusion. First, variational mode decomposition (VMD) is used for adaptive multi-scale decomposition of the TE signal. A comprehensive evaluation index combining transient energy, sample entropy, and envelope spectrum significance is constructed to select key fault-sensitive modes. Then, wavelet packet decomposition extracts detailed energy features from the selected modes, forming multi-dimensional feature vectors rich in fault information. Finally, principal component analysis (PCA) reduces feature dimensionality, and a support vector machine (SVM) classifier achieves accurate identification of meshing impact faults under different conditions. Experimental results show the method has high accuracy, good noise resistance, and strong condition transferability, providing an effective and reliable solution for electric vehicle reducer fault diagnosis.