Application of EWT AR model and FCM clustering in rolling bearing fault diagnosis

A fault diagnosis method is proposed, which is based on Empirical Wavelet Transform (EWT), Auto-Regressive (AR) model and Fuzzy C-Mean clustering (FCM) clustering algorithm, in order to solve the problem of fault category is difficult to identify of rolling bearing fault signal. In this method, the original signal of the rolling bearing is decomposed by the EWT, and several AM-FM components are obtained. The AR model is established for each AM-FM component, and the original feature subset is constructed. Then, through the correlation analysis, the four AM-FM components are extremely correlated with the original vibration signal are selected and their AR models are established. Construction of high-dimensional feature subsets based on the auto-regressive parameters of AR model. Finally, using the Locality Preserving Projection (LPP) algorithm to reduce the dimension and enter the low-dimensional feature subset to the FCM clustering, in order to achieve fault diagnosis of bearings. Experiments show that the fault identification method which is proposed in this paper has certain advantages and the fault recognition effect is better.