摘要
针对支柱瓷绝缘子振动声学检测信号识别难的问题,提出了一种基于变分模态分解算法的支柱瓷绝缘子损伤BP(反向传播)神经网络识别方法。首先,搭建声振动测试系统并采集完好和带缺陷瓷绝缘子的声振动信号;然后,使用变分模态法对其进行分解及重构,剔除信号的噪声及冗余分量;最后,分别将原始信号及变分模态分解重构后的声信号输入BP神经网络进行损伤识别比较,发现变分模态分解处理后的重构信号验证组识别率得到明显提升。分析表明提出的方法可自动提取并识别声信号,具有正确率高、受干扰噪声影响小等特点。研究工作将为支柱瓷绝缘子的损伤识别提供可行的技术方案。
In view of the difficulty in signal identification in vibration acoustic detection of post porcelain insulators,A BP(backpropagation)neural network identification method based on variational mode decomposition(VMD)is proposed to identify the damage of post porcelain insulators.First,the acoustic vibration test system is built and the acoustic vibration signals of intact and defective porcelain insulators are collected;then,the decomposition and reconstruction are carried out using the variational modal method to eliminate the noise and redundant components of the signals;finally,the original signals and the reconstructed acoustic signals after the variational modal decomposition are reinput into the BP neural network for damage identification,and the recognition rate of the reconstructed signals after the variational modal decomposition is found significantly improved.The proposed method,as shown by the analysis,can automatically extract and identify acoustic signals,is highly accurate and less susceptible to noise interference.The research will provide a feasible technical solution to the damage identification of post porcelain insulators.
作者
罗宏建
张杰
赵洲峰
焦敬品
LUO Hongjian;ZHANG Jie;ZHAO Zhoufeng;JIAO Jingpin(Stae Grid Zhejiang Electric Power Co.,Ltd.Research Institute,Hangzhou 310014,China;Faculty of Materials and Manufactuing,Beijing University of Technology,Beijing 100124,China)
出处
《浙江电力》
2022年第5期69-74,共6页
Zhejiang Electric Power
基金
国网浙江省电力有限公司科技项目(5211DS18003J)。
关键词
瓷绝缘子
损伤识别
变分模态算法
BP神经网络
porcelain insulators
damage identification
variational mode decomposition
BP neural network
