A roller bearing fault diagnosis method was proposed in which Support Vector Machine(SVM) and Auto-Regressive(AR) model based on Empirical Mode Decomposition(EMD) were combined.EMD method was used to decompose the roller bearing vibration signal into a finite number of Intrinsic Mode Functions(IMFs)

then the AR model of each IMF component was established

finally

the auto-regressive parameters and the variance of remnant were regarded as the fault characteristic vectors and served as input parameters of SVM classifier to classify working condition of the roller bearing.The experimental results show that the proposed approach can classify working condition of roller bearings accurately and effectively even in the case of small number of samples and the atomization of the roller bearing fault diagnosis can be implemented.