Performance degradation assessment of train rolling bearings based on SVM and segmented vote method

Abstract

For solving the problem of performance degradation assessment of train rolling bearings, the degradation pattern recognition and assessment method based on segmentation vote and SVM was proposed. Firstly, in order to obtain effective features indexes of bearing performance degradation, with the collected vibration acceleration data, the data was decomposed by LMD and state features were extracted and dimensionality reduced by PCA. And then, on the basis of the feature indexes, the method that combined least squares SVM and segmented vote was developed. Compared with the traditional pattern recognition methods, the new method improves identification accuracy effectively. The experiment results shown that the identification accuracy rates of three stages (normal operation, the primary degradation and the severely degradation) are higher than 95%. Finally, by the precision degradation pattern recognition results, train rolling bearing performance degradation assessment was completed.