Rolling Bearing Diagnosis Based on Adaptive Probabilistic PCA and the Enhanced Morphological Filter

Early fault diagnosis of rolling element bearing is still a difficult problem. Firstly, in order to effectively extract the fault impulse signal of the bearing, a new enhanced morphological difference operator (EMDO) is constructed by combining two optimal feature extraction-type operators. Next, in the process of processing the test signal, in order to reduce the interference problem caused by strong background noise, the probabilistic principal component analysis (PPCA) method is introduced. In the traditional PPCA method, two important system parameters (decomposition principal component k and original variable n) are usually set artificially; this will greatly reduce the noise reduction performance of PPCA. To solve this problem, a parameter adaptive PPCA method based on grasshopper optimization algorithm (GOA) is proposed. Finally, combining the advantages of the above algorithms, a comprehensive rolling bearing fault diagnosis method named APPCA-EMDF is proposed in this paper. Experimental comparison results show that the proposed method can effectively diagnose the vibration signals of rolling element bearing.