基于改进ELMD和多尺度熵的管道泄漏信号识别  

Pipeline leakage signal identification based on improved ELMD and multi-scale entropy

作　　者：郝永梅 杜璋昊 杨文斌 邢志祥 蒋军成 岳云飞[2] HAO Yongmei;DU Zhanghao;YANG Wenbin;XING Zhixiang;JIANG Juncheng;YUE Yunfei(School of Environmental and Safety Engineering,Changzhou University,Changzhou Jiangsu 213164,China;Branch of Changzhou,Jiangsu Special Equipment Safety Supervision and Inspection Institute,Changzhou Jiangsu 213161,China)

机构地区：[1]常州大学环境与安全工程学院,江苏常州213164 [2]江苏省特种设备安全监督检验研究院常州分院,江苏常州213161

出　　处：《中国安全科学学报》2019年第8期105-111,共7页China Safety Science Journal

基　　金：江苏省重点研发计划专项项目(BE2018642);江苏省研究生科研创新项目(KYCX18_2622);常州市科技支撑计划(社会发展)项目(CE20185024)

摘　　要：为预防城市管道泄漏事故,准确提取管道泄漏信号的特征,首先提出一种改进的总体局域均值分解(ELMD)与多尺度熵的管道泄漏信号识别方法,通过峰值波形匹配延拓法处理端点处的信号,减弱端点处信号分量的畸变、失真;然后对管道原始泄漏信号进行ELMD分解,得到一系列乘积函数(PF),计算各PF分量的多尺度熵值,根据熵值的大小筛选出含有主要泄漏信息的PF分量,消除背景噪声的影响;最后构建反向传播(BP)神经网络,并识别泄漏信号。结果表明:该方法减少了分解后的误差,能够实现管道泄漏的检测,与未改进的ELMD方法相比,泄漏信号的识别率更高。This paper is conducted with the aim of preventing urban pipeline leakage accidents and accurately extracting the characteristics of pipeline leakage signals.Firstly,an improved ELMD and multiscale entropy pipeline leakage signal identification method was proposed.The signal at the end point was processed by using the peak waveform matching extension method so as to attenuate the distortion of signal components.Secondly,ELMD decomposition of the original leakage signal was carried out to obtain a series of product functions values,the value of whose components were calculated through multi-scale entropy.The PF component containing the main leakage information was screened according to entropy value to eliminate the impact of background noise.Finally,a BP neural network was constructed to identify leakage signals.The results show that the proposed method,reducing errors after decomposition,is able to detect pipeline leakage,and it works better in recognizing leakage signals compared with unmodified ELMD method.

关 键 词：城市管道 总体局域均值分解(ELMD) 多尺度熵 反向传播(BP)神经网络 信号识别 

分 类 号：X944.4[环境科学与工程—安全科学]