Bearing fault feature extraction measure using multi-layer noise reduction technology

Abstract

The fault signals of rolling bearings are nonlinear and non-stationary, then it is difficult to extract fault feature of rolling bearings. In order to improve the accuracy of bearing fault diagnosis, a new feature extraction method based on multi-layer noise reduction is proposed in this paper. The proposed method first uses EVMD method to process the original signal, firstly, adding noise to the original signal. Then VMD algorithm is used to decompose the signal multiple times, and several components with more original information were retained and reconstructed. On the basis of the above reconstructed signals, features are extracted by MEMD method. Firstly, setting the Times of EMD measure; After each EMD decomposition, the kurtosis values of IMF components are calculated and several IMF components with large kurtosis values are retained; Finally, several selected components are weighted and fused to form fault feature vectors of bearings. The feature extraction of the proposed method was completed by using the bearing data set of Xi 'an Jiao tong Lei yaguo team. In order to verify the advantage of the proposed algorithm in this paper, the SVM algorithm is adopted to classify the fault features, and compared with the features extraction results of VMD and EMD methods alone, measure is proposed in this paper has higher classification accuracy.