An improved Autogram and MOMEDA method to detect weak compound fault in rolling bearings

Abstract

When weak compound fault occurs in rolling bearing, the faint fault features suffer from serious noise interference, and different type faults are coupled together, making it a great challenge to separate the fault features. To solve the problems, a novel weak compound fault diagnosis method for rolling bearing based on improved Autogram and multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) is proposed. Firstly, the kurtosis index in Autogram is modified with multi-scale permutation entropy, and improved Autogram finds the optimal resonance frequency band to preliminarily denoise the weak compound fault signal. Then, MOMEDA is performed to deconvolute the denoised signal to decouple the features of compound fault. Finally, square envelope analysis is applied on the separated deconvoluted signals to identify different type faults according to the fault characteristic frequencies in the spectrums. The proposed method is performed to analyze the simulated signal and experimental datasets of different types of rolling bearing weak compound faults. The results indicate that the proposed method can accurately diagnose the weak compound faults, and comparison with the analysis results of parameter-adaptive variational mode decomposition algorithm verifies its effectiveness and superiority.

Keywords

• compound fault diagnosis
• rolling bearing
• autogram
• multipoint optimal minimum entropy deconvolution adjusted