Rolling bearing fault diagnosis based on improved adaptive parameterless empirical wavelet transform and sparse denoising

Abstract

Empirical wavelet transform (EWT) has been widely used in the fault diagnosis of rolling bearings. However, how to effectively determine the number and boundaries of modes has been a challenging problem. Consequently, a rolling bearing fault diagnosis method incorporating an improved adaptive parameterless EWT (IAPEWT) and an adaptive sparse coding shrinkage denoising (ASCSD) algorithm is proposed to extract periodic impulse features. The IAPEWT method can automatically segment the spectrum and determine the boundaries of the filter bank to achieve the adaptive decomposition of signals. Meanwhile, the ASCSD algorithm can perform sparse denoising on the mode containing the periodic impulses to strengthen the impulse features, thereby improving the fault identification accuracy of rolling bearing. Numerical simulation and real signals acquired from rolling bearings are adopted to verify the effectiveness and superiority of the proposed IAPEWT-ASCSD method by comparison with the classical EWT, EEMD, VMD, and Fast-SC.