Research on bearing fault diagnosis method based on SCVMD and CGLF under various rotating speeds

Abstract

To solve the problem of bearing fault diagnosis at different speeds, a fault diagnosis method based on single-component variational modal decomposition (SCVMD) and coarse-grained lattice feature (CGLF) is proposed by analyzing the influence mode of speed transformation on frequency spectrum. First, the central frequency of the main resonance band of the signal is extracted based on SCVMD to eliminate the problem of spectrum shift caused by speed change. Then, the signal fragments are intercepted from the original signal spectrum to construct CGLF. Finally, a deep convolutional neural network (DCNN) is established to solve the sideband shrinkage problem caused by speed change and used to construct the mapping relationship between CGLF and category labels. In the experiment, bearing fault experimental platform dataset is used for the algorithm verification, and the final recognition rate is 98.3%. It proves that the method can effectively achieve bearing fault diagnosis at different speeds.