Aiming at the problem of the type and the location of mechanical failure of high-voltage disconnectors aredifficult to be effectively identified

a fault diagnosis method based on Relief F algorithm is proposed to optimize themulti-channel vibration characteristics and utilize BP neural network fusion decision. First

simulating the failurethrough the tests and the install multi-point vibration sensors in the disconnectors body and the operating mechanismto collect the vibration signals of different locations. Then the vibration signals collected by a plurality of sensors areconducted with EMD to obtain the intrinsic modal function. And calculating the energy moment and integrating theenergy-moment of the multi-channel signals. Finally

the Relief F algorithm is used for extracting the principal com-ponents to construct the input feature vector of the BP neural network

so as to realize the fault type and the positiondiagnosis. The experimental results show that the multi-channel sensor feature fusion has a better adaptability andclassification ability for the different faults of disconnectors than the single-channel signal feature extraction has

which can diagnose the type and the location of the fault and improve the diagnostic accuracy.