The vibration signal of the circuit breaker contains a lot of mechanical information. In order to identify the fault of the circuit breaker more accurately, this paper proposes a method based on variational mode decomposition and support vector machine. Firstly, the local extremum method is used to determine the appropriate VMD decomposition modal number. Secondly, the signal can be decomposed into multiple tight modes by VMD, and calculate the singular values of each mode as feature vector to train fault model by SVM. Finally, the test signal is diagnosed by SVM, and different faults are successfully diagnosed.