An optimized multivariate variational mode decomposition for the fault diagnosis of rotating machinery

Various failures are prone to occur in rotating machinery due to the harsh working conditions, thereby making it a vital work to perform accurate fault diagnosis to prevent performance degradation and safety hazards. The presence of multivariate variational mode decomposition (MVMD) provides a good knowledge of how to cope with multichannel data which contains more comprehensive information. In this work, an innovative diagnostic approach based on optimized MVMD is proposed for rotating machinery. Corner-stone of this method is the optimized MVMD, a new approach extracting modes successively with the proper adjustment of initial center frequencies. It achieves the mode decomposition without prior knowledge of the number of modes and initial center frequencies which affect the decomposition results greatly. Moreover, normalized frequency-to-energy ratio is employed as an index for selection of faulty modes. Analysis and comparison results of the experiment data from defective bearing indicates that the new approach shows a prominent superiority in fault identification.