基于EMD和SVM的高压断路器机械故障诊断方法研究

高压断路器操动过程中,声波信号的变化反映了断路器机械状态,声波信号特征提取直接关系到故障诊断的准确性和实用性。提出一种基于经验模态分解(empirical mode decomposition,EMD)的状态特征提取方法,根据高压断路器分闸过程的物理特点,将分闸声波信号分为分闸前期、分闸中期、分闸后期三个阶段,各阶段声波信号等时间分段,形成等时间分段能量熵以反映声波信号的时间、频率和能量变化。将能量熵输入支持向量机(support vector machine,SVM),采用"一对其余"策略进行多级分类,可诊断出断路器的正常运行状态和典型故障。通过对ZN28A型真空断路器正常状态、拐臂润滑不足、缓冲器有多余无效撞击和其它故障的测试实验,基于EMD和SVM相结合的诊断方法,有效提高了小样本下诊断断路器机械故障的准确性。

Research machinery fault diagnosis of high voltage circuit breaker based on EMD and SVM

During the operation process of the high voltage circuit breaker,the changes of acoustic signals reflect the machinery state of the circuit breaker.The extraction of the acoustic signal feature will directly influence the accuracy and practicability of the fault diagnosis.An extraction method based on EMD is proposed.The break-brake acoustic signals are divided into three stages according to the time: early stage,middle stage,later stage.Each phase of the acoustic signal is sected by time,forming longitudinal energy entropy to reflect frequency change of the signal.The normal operation state and typical fault of the circuit breaker can be diagnosed by adopting the "one-to-rest" strategy for multistage classification after the input of the horizontal and longitudinal energy entropy into the support vector machine.Through the testing experiment of ZN28A type vacuum circuit breaker under normal,turn arm insufficient lubrication,extra invalid hit of the buffer and other faults,diagnosis method of EMD with the combination of SVM can effectively improve the accuracy of diagnosing the circuit breaker mechanical problems in the small-sample.