关键词: 变分模态分解, 核模糊C均值聚类, 样本熵, 故障识别

Abstract: A novel approach for fault identification of rolling bearings was proposed. The method integrated VMD denoising and KFCM clustering. Firstly, the measured vibration signals were processed to obtain VMD parameters. Secondly, the vibration signals were decomposed by VMD to obtain a series of band-limited intrinsic mode function(BIMF) components. And the effective BIMF components were screened out and superimposed into the reconstructed signals. Thirdly, the sample entropy and the root mean square value were calculated and coalesced as a feature vector, and the feature vector sets of test samples and that of training samples were obtained. Fourthly, the feature vector sets of training samples were analyzed by KFCM clustering method to obtain the clustering centers of the four types of signals. Lastly, depending on the principles that the Euclidean distance among feature vectors of test samples and cluster center of training samples was minimum, the fault types were recognized. In addition, the vibration signals were decomposed with the empirical mode decomposition(EMD) method and recognized by KFCM clustering method. Compared with the method based on the EMD denoising and KFCM clustering, the proposed approach may obtain better fault identification results.

Key words: variational mode decomposition（VMD）, kernel fuzzy C-means（KFCM） clustering, sample entropy, fault identification