Abstract:In order to solve the problem of rolling bearing fault diagnosis under the complex noise, power frequency and its harmonics interference conditions, to improve the accuracy of diagnosis, the empirical mode decomposition (EMD) and study of the support vector machine (SVM) were carried out, then the corresponding decisionmaking process was given. The feature extraction algorithm was applied based on improved EMD decomposition, and the fault features with obvious intrinsic mode functions (IMF) component were selected for feature extraction. Maximum filter in extend was done on low frequency noise interference, the fault features of the signal were captured, and then the feature sets were input to the SVM classifier to identify. The results show that the method for bearing fault identification has higher accurate rate, which provides a theoretical support for ensure the safe operation of the bearing and fast fault diagnosis.