Abstract: Rolling bearing fault location and especially the diagnosis of its performance degradation degree can be used for more effective equipment maintenance to reduce downtime rate.For nonstationary vibration signal,a new fault diagnosis method was presented to achieve feature extraction and intelligent classification of different fault positions and performance degradation degree of rolling bearing signal.By using empirical mode decomposition（EMD） method,the vibration signal of each status could be decomposed into a set of intrinsic mode function（IMF） components and a residual component.Empirical mode decomposition method has well adaptability and the result using empirical mode decomposition method is the unique one.Then autoregressive（AR） model could be established for each intrinsic mode function component,and model parameters and variances of remnant which were regarded as feature vectors could be obtained by using Yule-Walker method and Ulrych-Clayton method respectively,then feature vectors were regarded as the input of improved hypersphere multiclass support vector machine classifier for judging rolling bearing fault location and its performance degradation degree.Experiments results show that,by using the presented method,the intelligent diagnosis of rolling bearing fault position and performance degradation degree can be achieved together,and the recognition rate using empirical mode decomposition combined with Ulrych-Clayton parameter estimate method of autoregressive model for feature extraction is more higher.