声信号特征优化断路器机械故障诊断算法

Mechanical Fault Diagnosis of High Voltage Circuit Breaker Optimized by Acoustic Signal Characteristics

为提高断路器机械故障的诊断准确率,并解决传统基于振动信号方法面临的接触安装局限、低频本振和高频干扰等问题,提出基于声特征优化提取的机械故障监测算法。算法首先将声信号划分为信号区间并以K-S显著性检测提取故障信号差异较大的信号区间,然后提取其LDA优化凸显的GFCC特征,以提高特征对故障状态的描述能力,然后采用特征加权优化的SVM算法对高压断路器的机械故障进行诊断监测。以四种模拟机械故障取得的声信号数据进行实验,结果表明,算法对四种故障的诊断准确率均值达到94%以上,验证算法的有效性。

To improve the accuracy of the diagnosis of mechanical faults of circuit breakers,and to solve the problems of contact installation limitations,low-frequency local vibration and high-frequency interference faced by traditional vibration signalbased methods,propose a mechanical fault monitoring algorithm based on the optimal extraction of acoustic features. The acoustic signal is first divided into signal intervals,and the signal intervals with large differences in the fault signal is extracted by KS significance detection. and then,the GFCC features highlighted by its LDA optimization is extracted to improve the ability of the features to describe the fault state. At last,the feature-weighted optimized SVM algorithm is used to diagnose and monitor mechanical faults of high-voltage circuit breakers. The results of the experiments with the acoustic signal data obtained from four kinds of simulated mechanical failures,show that,the average value of the algorithm′s diagnostic accuracy for the four types of faults is over 94%,verifying the effectiveness of the algorithm.