To multi-faults of a gearbox, the strong fault feature vibration of gearing always interfered with rolling element bearing weak fault vibration. The traditional methods are difficult to extract the weak bearing fault feature. In order to solve the problem, a vibration analysis scheme for multi-faults vibration of gearbox based on discrete random separation（DRS） is proposed. In the proposed method, the resonance parameters of the gearbox vibration is obtained based on the fast kurtogram algorithm at first, built a bandpass filter that included resonance parameters, and the envelope is extracted by Hilbert transform. Then, converted envelope to the angle domain based on the equi-angle increment resampling to remove speed fluctuation. Next, the SANC is performed to separate the envelope in the angle domain. As a result, the gear fault corresponding envelope component and the bearing fault related envelope component are obtained; Finally, the spectrum analysis is carry out on the two envelope components respectively for the fault diagnosis information. Test results show that this method can be used to extract the gear and bearing fault features effectively.