In order to improve fault diagnosis accuracy and efficiency

a novel approach based on multiply feature se-lection techniques and K-nearest neighbor algorithm（KNN）is proposed for fault diagnosis of high voltage circuitbreaker operation mechanism. Firstly

16 features

including peak current values

critical time points

currentchange rates and different stages lasting time of coil current

are extracted from opening/closing coil current. Then

3 feature selection methods based on different evaluation criteria are applied to compute comprehensive importance ofthe aforementioned features and determine the optimal feature subsets subsequently. Finally

a fault diagnosis modelbased on the optimal feature subset and KNN is established for fault diagnosis of high voltage circuit breaker opera-tion mechanism

whose diagnosis accuracy is assessed with cross validation technique. Experimental results showthat the proposed approach can eliminate redundant features

simplify structure of fault diagnosis model

enhance di-agnosis efficiency and obtain desirable fault diagnosis accuracy with few critical features.