Fault Diagnosis of Spring-Actuated Mechanism Based on Coil Current and Vibration Signal

Due to the fact that the closing or opening action of operating mechanism requires many components to cooperate with each other, it is prone to mechanical failures. At present, most online monitoring systems for operating mechanisms monitor a single state parameter such as coil current, time travel characteristics, vibration signal, etc, few multi state parameter comprehensive monitoring systems have been put into practical operation. This paper proposes a multi state parameter monitoring method based on coil current and vibration signal. The time domain feature of coil current and the sample entropy of intrinsic modal function components of vibration signal are calculated. Principle component analysis is performed for the original feature set which include 11 features of coil current and 5 features of vibration signal. The reduced feature in new space are used to train SVM classifiers and to identify the classification results. The classification accuracy of Fault 1 and Fault 2 is 100%, while classification accuracy of Normal state and Fault $\mathbf{3}$ is $\mathbf{8 0 \%}$, 70%, respectively. It is shown that feature extraction method and SVM can be applied to fault diagnosis of operation mechanism of high voltage circuit breakers.