A diagnosis method of compound faults of rolling bearings based on VMD optimized from multi-dimensions

In view of the difficulty in accurately extracting composite fault features of rolling bearings, this paper started with three aspects of impact, complexity and cyclostationarity to set up a new fault feature index of signal ( Kurtosis-SES E -complexity ) based on kurtosis, square envelope spectrum negentropy ( SES E ) and complexity parameter to embody comprehensive description of fault information contained in the signals. Meanwhile, the decomposition level of variational mode decomposition (VMD) was determined with the maximum of Kurtosis-SES E -complexity as criterion. The two modal components, IMF max1 and IMF max2 , corresponding to maximum and sub-maximum of Kurtosis-SES E -complexity index were chosen to determine the sensitive component signals which can better reflect the fault information, and prevent the loss of fault information. Finally, the chosen component signals were reconstructed in the form of equal weight, and the fault characteristics of rolling bearings were extracted by using the frequency spectrum of autocorrelation function of reconstructed signals for fault identification. For verify the validity of the presented method, the bearing fault data collected under multiple states was verified and analyzed and the presented method was contrasted with the widespread method of adaptively determining the decomposition level of VMD based on kurtosis. The result of the comparison has proved that the Kurtosis-SES E -complexity index set up in the paper can represent bearing fault information more comprehensively, adaptively determine the decomposition level of VMD, and extract compound fault information of bearing effectively.