Multi-objective maximum correlated Kurtosis deconvolution and its application in rotating machinery fault diagnosis

Abstract Maximum correlated Kurtosis deconvolution (MCKD) has been proven to be a useful way to enhance periodic pulses in rotating machinery fault diagnosis, such as rolling bearings and gears. However, when facing multiple-fault-coupled signals, MCKD is only capable of intensifying the characteristics of one of the faults, which can easily cause missed diagnoses and misjudgments in fault diagnosis. Meanwhile, the correlated kurtosis (CK) index only considers the time-domain feature and ignores the feature in the envelope spectrum of the signal. In order to deal with these problems, a multi-objective maximum correlated Kurtosis deconvolution (MOMCKD) is introduced in this paper. In this method, the fault period set is first established so that deconvolution is no longer limited to enhancing one fault characteristic. Then the idea of the candidate eigenvector set is proposed, and the eigenvectors are identified and screened according to the fault characteristics in the envelope spectrum. In contrast to original MCKD, this method can simultaneously output various fault signal components in a compound fault signal, reducing the complexity of fault diagnosis operations and improving the accuracy of the results. Finally, the good effect and high superiority are proved by handling simulated signals and experimental data.