Bearing performance degradation assessment based on optimized EWT and CNN

In the process of bearing degradation assessment, problems such as modal aliasing and early failure samples being submerged by normal samples are the main factors that limit the performance of the assessment method. A method is proposed for bearing performance degradation assessment. In this method, optimized empirical wavelet transform (EWT) is used to decompose bearing vibration signal, and the sub-components containing fault information are extracted using frequency-sliced wavelet transform and improved particle swarm algorithm. The STET envelope spectrum of the bearing fault component is input to the convolutional neural network to extract sensitive features. The fuzzy C-means model is used to degradation assessment, which is constructed by the sensitive features of bearing fault-free stage. The result show that the optimal EWT effectively solves the problem of early failure samples being overwhelmed by normal samples. The proposed method have greater sensitivity and stability than original EWT in extracting fault information.

Automated summary

A method for bearing performance degradation assessment is proposed, using optimized empirical wavelet transform and fuzzy C-means model to improve the sensitivity and stability of the assessment method in extracting fault information.