Self-Adaptive Multivariate Variational Mode Decomposition and Its Application for Bearing Fault Diagnosis
Qiuyu Song, Xingxing Jiang, Shuang Wang, Jianfeng Guo, Weiguo Huang, Zhongkui Zhu
Abstract
In actual engineering scenarios, multichannel datasets that contain complete information contribute to better accuracy of bearing fault diagnosis. Multivariate variational mode decomposition (MVMD), as an extension of variational mode decomposition (VMD), can deal with multivariate signals. However, the performance of MVMD is affected by initial parameters, i.e., the number of decomposition modes, the bandwidth balance parameter, and the initial center frequencies (ICFs). To overcome the difficulty of initial parameter selection, a self-adaptive MVMD is proposed, where the number of decomposition modes and the ICFs are determined adaptively on the basis of the convergence tendency in MVMD. The bandwidth balance parameter of each extracted mode is also optimized adaptively in the process. In addition, the normalized frequency-to-energy ratio is used as the evaluation index to identify faulty mode. Final results of experiments pave the way for a new method in bearing fault diagnosis with prominent superiority.