Rolling Bearing Fault Feature Extraction Method Based on MCKD Combined with Sensitive SVD

Abstract

Aiming at the problem that rolling bearing fault features are fairly hard to extract and cannot effectively solve fault, a rolling bearing fault feature extraction method based on MCKD (Maximum Correlated Kurtosis Deconvolution) combined with Sensitive Singular Value Decomposition (SSVD) is proposed. Through the MCKD, the rolling bearing fault signal is denoise processed; then the signal of the denoise is decomposed by the sensitive SVD, and the component signal with rich fault information is filtered by recombination. Finally, the reconstructed signal is analysed by the Hilert Envelope spectrum to obtain the fault feature information. It has been proved by experiments that this method can effectively perform fault diagnosis on the rolling bearing.