Research on Multivariate Variational Mode Decomposition Method and Its Application to Bearing Fault Diagnosis

The bearing failure diagnosis methods based upon variational mode decomposition (VMD) have been researched extensively in recent years. However, these methods are only capable of dealing with single channel data, in which the amount of information is limited and the anti-interference ability needs to be further enhanced. Recently, as one of extensions of VMD, the multivariate variational mode decomposition (MVMD) was put forward to deal with the multivariate signal. For the sake of monitoring the operation state of equipment more comprehensively and diagnose mechanical failure more accurately, we propose a new bearing failure diagnosis method based upon MVMD. In particular, to select a suitable value of decomposed modes K, a scheme which combines correlation coefficient and kurtosis criterion is innovatively proposed to enhance the performance of the MVMD. The result of experimental verification has demonstrated the enhanced performance of this new method in transient extraction and faults detection of bearings over the original VMD method.