VMD Entropy Method and Its Application in Early Fault Diagnosis of Bearing

Abstract

This paper proposes an early faults diagnosis method for bearings based on Variational Mode Decomposition (VMD) and Entropy Theory to monitor the working state of the key components of the high-speed train axle box. Firstly, the box vibration signal is decomposed into detailed signals at different scales by using VMD (Band-Limited Intrinsic Mode Function, BIMF), then the three kinds of entropy are extracted from BIMF and composed into VMD entropy. Finally, the VMD entropy has been input into SVM for training to determine the fault type. This paper is going to take research on the vibration signals of high-speed train axle box under three typical working conditions of normal bearing, cage failure and roller failure. It is concluded that the best VMD parameters of fault identification for high-speed train axle box can effectively improve the recognition rate of entropy in early bearing fault diagnosis by comparing it with EMD entropy. The analysis results show that for a high-speed train running under 200 km/h, the recognition rates under three different working conditions can reach 98.75%, 100%, 98.75% respectively, which proved the validity of VMD entropy for early bearing fault identification of high-speed train.