Blind deconvolution assisted with periodicity detection techniques and its application to bearing fault feature enhancement

Abstract

Abstract Maximum correlated kurtosis deconvolution (MCKD), multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) and maximum second-order cyclostationarity blind deconvolution (CYCBD) have remarkable performances in extracting periodic impulses. However, these deconvolution methods highly rely on the prior period of measured signal and can only enhance the specific periodic impulses. Aiming at these limitations, six kinds of periodicity detection techniques (PDTs) are introduced to adaptively identify the period of repetitive impulses. Further, PDTs-assisted MCKD, MOMEDA and CYCBD are proposed for bearing fault feature enhancement. The improved deconvolution methods have two characteristics: first, the fault period is automatically identified by PDTs according to the characteristics of the measured signal; second, the impulses of different faults can be enhanced adaptively. The analysis results of simulated and experimental datasets demonstrated the better capability of the proposed methods in enhancing bearing fault features with respect to original deconvolution methods and the fast kurtogram method.