基于振动信号分析的高压断路器机械故障诊断

摘要

In allusion to problems of low accuracy rates of current mechanical fault diagnosis methods for high-voltage circuit breakers,this paper proposes a new diagnosis method combining improved wavelet packet decomposition,Hilbert spectrum analysis,particle swarm optimization (PSO)and optimized support vector machine (SVM)to make simulating experiments and analysis on spring fatigue and high proportion jam fault of the closing pawl of mechanical failure of high-voltage circuit breakers. Research results indicate time frequency characteristics of vibration signals can better reflect mechanical states of high-voltage breakers,especially the feature classification method based on PSO-SVM is more effective which is able to greatly improve accuracy rates of current mechanical diagnosis methods for high-voltage circuit breakers.%针对高压断路器机械故障诊断方法准确率较低的现状,提出将改进的小波包分解(wavelet packet,WP)、Hilbert谱分析、粒子群优化(particle swarm optimization,PSO)算法和优化后的支持向量机(support vector ma-chine,SVM)相结合的诊断方法,对弹簧操动式高压断路器进行弹簧疲劳和合闸挚子卡涩故障(在高压断路器机械故障中占比较高)的模拟实验和分析.研究结果表明,振动信号的时频特性能较好反映高压断路器的机械状态,尤其基于PSO-SVM的特征分类方法分类效果较好,能大大提高现有高压断路器机械故障诊断方法的准确率.