A Check Valve Fault Diagnosis Method Based on Variational Mode Decomposition and Permutation Entropy

Abstract

Aiming at the problem that the vibration signal of the check valve has background noise and low fault recognition rate, a signal characteristics extraction method based on variational mode decomposition and permutation entropy was proposed. The extreme learning machine was used for fault recognition. Firstly, the check valve vibration signal was decomposed by the variational mode decomposition, and the intrinsic mode functions were obtained in different scales. Secondly, the permutation entropy of each intrinsic mode function was calculated and used to compose the multiscale feature vector. Finally, the high-dimensional feature vector was input to the extreme learning machine for check valve fault diagnosis. The comparison is made with EEMD and LCD (local characteristic-scale decomposition). The experimental results show that the method can effectively identify the fault type of the check valve.