A weighted nonconvex sparse representation with high-pass filter function for fault diagnosis of rolling bearing

Abstract

Vibration signal analysis is one of the most effective and convenient method for fault diagnosis in rolling bearing. A challenging problem is how to extract the fault features from the noisy signal accurately. In this paper, a novel sparse representation algorithm, a weighted nonconvex sparse representation with high-pass filter function (WNCSR-HPF) is proposed for bearing fault feature extraction. WNCSR-HPF is developed based on a weighted nonconvex sparse regularization term, which can remove the noise interference and promote sparsity. Moreover, an adaptive setup method of regularization parameter is proposed for improving the applicability of WNCSR-HPF. The majorization-minimization (MM)-based algorithm is developed for solving the objective optimization problem in this paper. A simulation signal and a bearing vibration signal are used to illustrate the effectiveness of WNCSR-HPF for fault feature extraction. The experimental results show that WNCSR-HPF has the good performance on the fault feature extraction.