A novel fault diagnosis method for bearing based on maximum average kurtosis morphological deconvolution

Abstract Maximum average kurtosis deconvolution (MAKD) effectively enhances periodic impulses in vibration signals. However, under conditions of random impulse interference, MAKD tends to amplify impulses within a single period. To address this problem, this paper proposes a maximum average kurtosis morphological deconvolution (MAKMD) method. First, on the basis of proposing a time-varying structural element more in line with the characteristics of vibration signals and constructing a new morphological gradient squared operator, an enhanced time-varying morphological filtering (ETVMF) is proposed. Then, ETVMF is introduced into MAKD to eliminate the effect of random impulse. Finally, the diagonal slice spectrum is utilized to detect the coupling frequency of the bearing, which makes the spectrum clearer and more convenient for bearing fault diagnosis. In MAKMD, the effect of random impulse is eliminated and the capability of fault feature extraction is enhanced. To demonstrate the method’s effectiveness and feasibility, experiments are conducted using simulated signals and measured bearing fault data, comparing results with existing deconvolution methods.