Fault diagnosis of rolling bearing based on MPA-VMD and RCMFE

Abstract

Due to the characteristics of non-linear, non-stationary and background noise interference in rolling bearing fault signals, the extraction of fault features becomes very complicated. To solve this problem, this paper proposes a rolling bearing fault diagnosis method based on marine predator algorithm based-variational mode decomposition (MPA-VMD) and refined composite multiscale fuzzy entropy (RCMFE). First, use MPA-VMD to decompose the original fault signal, and select the IMF signal with strong correlation with the original signal for reconstruction; Then use RCMFE to obtain the fault features of rolling bearings, and use the LDA algorithm to reduce the dimensions of the extracted fault features; Finally, beetle antennae search based support vector machine (BAS-SVM) is used to discriminate the fault types of the fault features after dimensionality reduction. Experimental results display that the method proposed in this paper can effectively extract and classify rolling bearing faults, verifying the practicability and validity of the method.