To detect circuit breaker’s potential mechanical fault and determine whether a circuit breaker can be normally used

the mechanical vibration signal of a circuit breaker is analyzed

and a vibration signal processing methodisput forward.In this method

wavelet packet analysis is applied to the signal to calculatethe energy spectrum of the waveformin every scale

and the cross-correlation function is employed to calculate the delay time of the signal in each faulty condition relative to the signal in normal condition. Adding the characteristic quantities into the modified support vector machine

named feature weighted support vector machine（FWSVM）

to perform fault diagnosis of circuit breaker. It is shown that some of the mechanical faults of high-voltage circuit breaker can be effectively analyzed by this method.