Abstract: Support sector machine (SVM) has stronger generalization ability than artificial neural networks and can guarantee that the local optimal solution is exactly the global optimal one. Meanwhile, it can solve the learning problem of smaller number of samples. According to the situation that it is hard to obtain enough fault samples and the non-stationary characteristics of gears fault vibration signals, a gears fault diagnosis method based on empirical mode decomposition (EMD) and SVM is proposed. Firstly, vibration signals are decomposed into a finite number of stationary intrinsic mode functions (IMFs), then the AR model of each IMF component is established, finally, the auto-regressive parameters and the variance of remnant are regarded as the fault characteristic vectors and served as input parameter of SVM classifier to classify working condition of gears. The experimental results show that the proposed approach can classify working condition of gears accurately and effectively even in the case of smaller number of samples.

Key words: AR model, EMD, Fault diagnosis, Gears, SVM