Research of circuit breaker intelligent fault diagnosis method based on double clustering

Abstract

According to the energy variation of the mechanical transmission in the process of circuit breaker operation which is characterized by acoustic and vibration signals, a new method of high Voltage circuit breaker mechanical fault diagnosis was proposed in this paper. This method combined Density Peaks Clustering Algorithm (DPCA) fused Kernel Fuzzy C Means (KFCM) and support vector machine (SVM). It is an intelligent method of double clustering. Vibration and acoustic signals are decomposed by Local Mean Decomposition. Three product function components with the largest correlation of the original signal are filtered. And the characteristic entropy can be extracted by approximate entropy. DPCA is utilized to get the best peak density clustering decision and optimize the initial clustering center of KFCM. The fault training samples is pre-classified and input SVM. And the fault classification result of the circuit breaker can be received by mesh optimization algorithm. Finally, the DPCA-KFCM and SVM method in the fault diagnosis of the circuit breaker is verified by the typical failure test of the circuit breaker, the loosening of the pedestal and the refusal of the circuit breaker, which improve the accuracy of the fault diagnosis greatly.