Aiming at the non- stationary feature of the rolling bearing vibration signal and the fault samples are always in a small number in its fault diagnosis,a rolling bearing fault diagnosis method based on improved Hilbert- Huang transform energy entropy and support vector machine is proposed. Firstly,the vibration signal in different condition is decomposed by improved empirical mode decomposition,and the intrinsic mode functions are obtained and sensitive mode functions are selected by the sensitivity evaluation method. Then,the energy entropy of sensitive mode functions serve as input vectors of support vector machine. Finally,by using support vector machine to identify the rolling bearing fault pattern and condition. The experiment results show that this method can identify rolling bearing fault patterns effectively and offer a practical method for its fault diagnosis.