摘要
针对传统的高压断路器储能机构故障诊断方法过于依赖主观经验,准确率不高、泛化能力差,提出一种声音振动信号联合构造卷积神经网络(CNN)特征矩阵的高压断路器储能机构故障诊断方法。首先将采集到的声信号通过形态学去除背景噪声,提出基于峭度和包络相似性的时标对位方法保证声振信号的同步性,然后对数据扩容后的声振信号利用皮尔逊相关系数构造二维图像特征矩阵,最后利用CNN对特征矩阵进行训练。实验证明:文中所提出的故障诊断方法与传统方法对比总体诊断准确率高,泛化性能好。
In view of such problems of the fault diagnosis methods of the traditional charging operating mechanism for high-voltage circuit breaker as too much reliance on subjective experience,low accuracy rate and weak generalization ability,a kind of fault diagnosis method of the charging operating mechanism for high voltage circuit breaker based on acoustic-vibration signal construction convolutional neural network(CNN)feature matrix was proposed.Firstly,the background noise of the collected acoustic signals were morphologically removed,and a time-scale alignment method based on kurtosis and envelope similarity was proposed to assure the synchronization of the acoustic and vibration signals.Then,the Pearson correlation coefficient was used for the acoustic and vibration signals after data expansion to construct a two-dimensional image feature matrix.Finally,the CNN was used to train the feature matrix.Experiments show that,compared with traditional methods,the fault diagnosis method proposed in this paper has high overall diagnosis accuracy and good generalization performance.
作者
尹子会
孟延辉
赵智龙
赵书涛
牛为华
徐晓会
YIN Zihui;MENG Yanhui;ZHAO Zhilong;ZHAO Shutao;NIU Weihua;XU Xiaohui(State Grid Hebei Electric Power Co.,Ltd.Maintenance Branch,Shijiazhuang 050071,China;School of Electrical and Electronic Engineering,North China Electric Power University,Hebei Baoding 071003,China)
出处
《高压电器》
CAS
CSCD
北大核心
2023年第9期242-249,共8页
High Voltage Apparatus
基金
国网河北省电力有限公司科技项目资助(kj2020-024)。
关键词
断路器
声振信号
卷积神经网络
时标对位
特征矩阵
数据扩容
circuit breaker
acoustic-vibration signal
CNN
time scale alignment
characteristic matrix
data expansion
