Rolling Bearing Fault Diagnosis Method Based On LMD Entropy Feature Fusion

Since each entropy feature has some defects in feature extraction, it appears that it is impossible to use one entropy feature to completely extract the time-frequency features of rolling bearing failure. Starting from the information entropy fusion theory, using nonlinear dynamic parameter entropy as a feature, a rolling bearing fault diagnosis method based on local mean decomposition (LMD) entropy feature fusion is proposed. First, use LMD to decompose the original fault signal to obtain multiple PF components, calculate the kurtosis value and correlation coefficient of each PF component, and choose the appropriate PF component to reconstruct the signal. Then, the approximate entropy and singular spectrum entropy of the reconstructed signal after LMD decomposition are calculated respectively, and the entropy feature fusion is performed to obtain complementary rolling bearing fault features. Finally, the fused entropy features are used for fault diagnosis through the Random Forest (Random Forest) algorithm. The simulation results show that the accuracy of the method reaches 98.3%. The study of this method can provide an effective theoretical basis for the fault diagnosis of rolling bearings in rotating machinery.