Fault diagnosis of bearings based on improved sparse decomposition via DTCWT and GA

In order to solve the problem that it is difficult to extract the fault signal characteristics of roller bearing with low signal-to-noise ratio, an improved sparse decomposition method based on dual-tree complex wavelet transform (DTCWT) and genetic algorithm (GA) was proposed to achieve deep noise reduction and reconstruct fault features. Firstly, the dual-tree complex wavelet was used to decompose the bearing vibration signal. The optimal component containing the impact feature was selected via the maximum kurtosis criterion. Then, in view of the low computational efficiency of sparse decomposition when processing high-dimensional complex signals, genetic algorithms were used to optimize the optimization process based on the matching pursuit (MP) algorithm to improve the efficiency of signal reconstruction. In addition, the termination criterion based on spectral entropy of the residual signal was proposed to select a reasonable number of iterations. Simulation and experimental results show that compared with the traditional sparse decomposition, the improved method is more suitable for strong noise signals and can adaptively extract fault features. © 2021 Editorial Board of Journal of Huazhong University of Science and Technology. All right reserved.