It is proposed that a fault diagnosis method for rolling bearing based on empirical mode decomposition（ EMD） and multivariate statistical process control（ MSPC）,the Hilbert- Huang transformation and principal component analysis（ PCA） are combined effectively in this method. It makes an effective classification for common bearing failure. The main contents are as follows: Firstly,the bearing signal is decomposed into intrinsic mode function under different frequency scale with EMD. Then,the PCA model of normal bearing with the IMFs is established and the control limits of Hotelling T2 of principal component model is calculated.Finally,take IMFs of signal to be detected into principal component model and the Hotelling T,2, value is obtained. The fault detection with T2 value is realized and principal component model. On the base of correlation between IMF functions and various faults,the classification is realized. This study gives an effective method and theoretical support for fault diagnosis and classification.