Adaptive sparse denoising and periodicity weighted spectrum separation for compound bearing fault diagnosis

The compound fault diagnosis of rolling bearings has become a hot topic. In this study, a novel method based on adaptive sparse denoising (ASD) combined with periodicity weighted spectrum separation (PWSS) is proposed to diagnose compound faults in rolling bearings. Specifically, ASD reveals fault types and PWSS separates compound faults. First, ASD determines regularization parameters adaptively using the proposed compound frequency multi D-norm, thereby denoising the raw vibration signal and revealing fault types. Then, PWSS constructs the time-frequency spectrum (TFS) and uses the fault periodicity from ASD to determine the time occurrence positions of the repetitive impulses. With this time occurrence position information, a weight matrix is constructed to reweight the TFS. Finally, through the reweighted TFS, PWSS extracts and separates repetitive impulses from compound faults. The performance of the proposed method is validated in both simulation and experimental studies. The results demonstrate that the proposed method can successfully diagnose and separate the compound faults.