Mechanical Fault Diagnosis of High Voltage Circuit Breaker based on Improved GSO-SVM Algorithm

Abstract

The vibration signal generated by the transmission and impact of mechanical components of circuit breaker has chaotic performances, which is difficult to be analysed by conventional signal processing methods. The phase space reconstruction of vibration signal is worked on, and the signal reconstruction parameters are calculated by mutual information method and Cao algorithm respectively. The vibration signal is reconstructed into a high-dimensional space, and its permutation entropy is calculated as a feature vector. Support vector machine (SVM) is used to identify the failure type of circuit breaker, and PSO improved GSA hybrid algorithm is used to optimize the parameters of SVM so as to obtain high recognition accuracy. The experiment is carried out with the measured vibration signal of the typical operation state of the circuit breaker. The results show that the characteristics of circuit breaker vibration signals can be extracted accurately with the combination of phase space reconstruction and permutation entropy. By using PSO-GSA-SVM, the fault types of circuit breakers can be identified quickly and effectively, and the problems of path distortion, energy leakage and mode overlap of existing diagnosis methods can be solved.