基于融合特征和残差神经网络的10 kV高压断路器机械故障声纹识别方法  

Voiceprint Recognition Method for Mechanical Faults of 10 kV Circuit Breaker Based on Fusion Feature Residual Neural Network

在线阅读下载全文

作  者:段梵 李先允[1] 单光瑞 陈兰杭 杨凯 DUAN Fan;LI Xianyun;SHAN Guangrui;CHEN Lanhang;YANG Kai(School of Electrical Engineering,Nanjing University of Engineering,Nanjing 211167,China;State Grid Zhenjiang Power Supply Company,Jiangsu Zhenjiang 212003,China)

机构地区:[1]南京工程学院电气工程学院,南京211167 [2]国网镇江供电公司,江苏镇江212003

出  处:《高压电器》2025年第3期205-213,共9页High Voltage Apparatus

基  金:国网江苏省电力公司科技项目(J2023085)。

摘  要:针对传统10 kV高压断路器故障诊断方法过于依赖主观经验、准确率不高、泛化能力差的问题。提出了一种基于声学特征的10 kV高压断路器常见机械故障识别方法。首先,以(ZN63)12/630A型高压户内真空断路器为研究对象,设置常见的8种机械故障,采集其分合闸时的声音作为检测信号;其次,将采集的故障声纹信号进行预处理,提取故障声纹信号的梅尔倒频谱系数(MFCC)特征、色度特征(chroma features)以及一维平均能量和频谱质心,并利用Fisher比舍弃贡献率低的分量,构成融合特征;最后以提取的融合特征作为诊断依据,构建基于残差神经网络的10 kV断路器机械故障诊断模型。结果表明文中方法对10 kV高压断路器常见的8种机械故障诊断识别准确率为99.99%。可作为当前检测手段的有效补充,提高高压断路器综合检测和潜伏性缺陷识别能力。In view of such issues as relying too much on subjective experience,low accuracy and weak generalization ability of fault diagnosis method of traditional 10 kV high-voltage circuit breaker,a kind of common mechanical fault identification method for 10 kV high-voltage circuit breaker based on acoustic features is proposed.Firstly,type(ZN63)12/630A high-voltage indoor vacuum circuit breaker is taken as the research object,8 kinds of common mechanical faults are set and the sound during opening and closing operation is collected as the detection signal.Then,the collected fault voiceprint signals is preprocessed,the Mel cepstrum coefficient(MFCC)features,chromat features,one-dimensional average energy,and spectral centroid of the fault voiceprint signals are extracted,and Fisher ratio to discard components with lower contribution rates is used to form fusion features.Finally,the extracted fusion features is taken as the diagnostic basis,a residual neural network based mechanical fault diagnosis model for 10 kV circuit breaker is constructed.The results show that the accuracy for diagnosing and identifying 8 common mechanical faults of 10 kV high-voltage circuit breaker by the method proposed in this paper is 99.99%.It can serve as an effective supplement to the current detection methods,improving the comprehensive detection and latent defect identification capabilities of high-voltage circuit breaker.

关 键 词:10 kV高压断路器 声纹识别 融合特征 残差神经网络 故障诊断 

分 类 号:TM561[电气工程—电器] TP183[自动化与计算机技术—控制理论与控制工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象