A fault diagnosis method on GWO-SPA and MAE for fault diagnosis of reciprocating compressors

Abstract

Given the non-stationary and nonlinear features of the reciprocating compressor vibration signal, as well as the problems of end-point effect and modal aliasing existing in the current adaptive decomposition method, the fault diagnosis method for reciprocating compressors based on gray wolf optimization-smoothness priors approach (GWO-SPA) and multiscale attention entropy (MAE) is proposed in this paper. The selection of the regularization parameter λ in the SPA algorithm is studied, and the GWO is introduced to improve the decomposition effect of the SPA method. Then, the entropy value of the MAE is calculated for the detrended term components decomposed by GWO-SPA to describe the fault characteristics quantitatively, and perform PCA dimensionality reduction processing. Finally, the optimized feature vector is input into the SVM. Through the experiments show it is proved that the method proposed can effectively extract the fault state features of reciprocating compressors, and realize the accurate distinction of the fault types of reciprocating compressors.