Fault Feature Extraction of Rolling Bearing Based on Improved VMD-SVD Joint Denoising and Synchrosqueezing Wavelet Transform

Bearing faults are the main contributors to the failure of motors. To solve the problem of extracting fault features from bearing signals with low signal-to-noise ratio, a method combining improved Variational Modal Decomposition (VMD) -Singular Value Decomposition (SVD) joint denoising with synchrosqueezing wavelet transform is proposed. The proposed method is based on the analysis of VMD-SVD theory. In view of the existing problems of VMD-SVD, the reconstruction part of VMD is cancelled, so that the noise is eliminated to the maximum extent before reconstruction. Using RSVD instead of SVD, and PMI index instead of energy index during reconstruction, make the method more suitable for fault diagnosis of rolling bearings. Finally, the effectiveness and superiority of the proposed method are verified by simulation of motor bearing and measured data.