Novel variational mode decomposition method for rotating machinery fault diagnosis based on weighted correlated kurtosis and salp swarm algorithm.

The mechanical vibration response in engineering is the superimposition of multi-frequency characteristic information. Therefore, it is of great necessity to utilize signal decomposition methods to extract fault characteristics for ultimate diagnosis. In the traditional variational mode decomposition (VMD) methods, the decomposition parameters (i.e. the mode number and quadratic penalty factor) are determined according to the principle of convenience and experience. This behavior reduces the performance of VMD methods to a great extent, and limits their decomposition accuracy and feature extraction capability. To resolve this problem, a novel VMD method for rotating machinery fault diagnosis is developed in this article. Firstly, a measurement index called weighted correlated kurtosis (WCK) is constructed by combining correlated kurtosis and Pearson correlation coefficient. Secondly, taking the maximum WCK as the goal function, salp swarm algorithm is utilized to find the optimum parameters. Lastly, the feature extraction is performed according to the selected sensitive mode possessing the maximum WCK. Two experimental examples demonstrate the effectiveness of the developed VMD method on mechanical vibration signal processing and fault diagnosis. Furthermore, by comparing with other two typical VMD methods, the superiority of the developed method is verified. [ABSTRACT FROM AUTHOR]