Abstract: At present, fault diagnosis method of rotor broken bar of asynchronous motor is based on the method of extracting characteristic frequency from stator current to make diagnosis for rotor state. When asynchronous motor runs with no-load or light load, the characteristic frequency is easy to be influenced by base-frequency leakage so as to be difficult to obtain, meanwhile speed fluctuation has great influence on the characteristic frequency, so it is lack of accuracy according to the characteristic frequency to judge rotor state. To solve above problem, the paper proposed a diagnosis method of using SVM and D-S evidence theory to identify fault of rotor broken bar of asynchronous motor. The method separately extracts characteristic informations of fault of rotor broken bar from current signal and vibration signal of stator on the basis of the extended Park method and the FFT transformation method, uses SVM to make pattern recognition to the status of asynchronous motor and form independent evidences for recognition results and then makes fusion process according to D-S evidence fusion rules, so as to realize accurate recognition of fault of rotor broken bar. The experiment result showed that the method can make accurate judgment for fault of rotor broken bar of asynchronous motor.