Rolling Bearing Fault Diagnosis Method Based on Improved Variational Mode Decomposition and Information Entropy

Abstract

Due to the complex randomness and nonlinearity of rolling bearing vibration signal, it is challenging to extract fault features effectively. By analyzing the vibration mechanism of rolling bearing, it is found that the vibration signal of local damage defects of rolling bearing has the characteristics of periodic impact and amplitude modulation. The variational mode decomposition (VMD) algorithm has a good advantage in dealing with nonlinear and nonstationary signals and decomposing a signal into different modes. However, VMD has the problem of parameter selection, which directly affects the performance of VMD processing, and causes mode aliasing. Therefore, a rolling bearing fault diagnosis method based on improved VMD is proposed. A new fitness function combining differential evolution (DE) algorithm with gray wolf optimization (GWO) algorithm is proposed to form a new hybrid optimization algorithm, named DEGWO. The simulation results show that the improved VMD method based on DEGWO can adaptively remove the noise according to the characteristics of the signal and restore the original characteristics of the vibration signal. Finally, in order to verify the advantages of the research, the information entropy is extracted from the data of 1000 samples in the bearing database of Case Western Reserve University as the feature set, which is input into support vector machine (SVM) for fault diagnosis test. The results show that the diagnostic accuracy of this method is 96.5%, which effectively improved the accuracy of rolling bearing fault diagnosis.