To make analysis and judgement of main circuit breaker fault with fewer circuit breaker data samples

ap-proximate entropy is proposed as a new circuit breaker vibration signal feature

by performing tests on SF6 circuitbreakers

samples under four states is collected:the normal state

the core jamming

insufficient lubrication

loosebase. Empirical mode decomposition is used on original signal to get a group of intrinsic mode functions as an inputcomponent of approximate entropy algorithm

multiple components of approximate entropy form a feature vector torepresent the original signal. Finally

effectively classify of samples under different states is made by support vectormachine

to verify the validity of using approximate entropy as circuit breaker vibration signal features.