基于相关分析和近似熵的管道泄漏声信号特征提取及辨识方法  

Feature extraction and identification of leak acoustic signal in water distribution pipelines using correlation analysis and approximate entropy

作　　者：杨进[1] 文玉梅[1] 李平[1] 

机构地区：[1]重庆大学光电工程学院重庆大学光电技术及系统教育部重点实验室,重庆400030

出　　处：《仪器仪表学报》2009年第2期272-279,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60804061)资助项目

摘　　要：从泄漏声信号产生机理出发,分析了泄漏声信号具有"不可重复"特征的机理;由于相关函数具有分析时间序列相干结构的能力,且近似熵从统计角度区别时间过程的复杂性,提出将信号相关长度后的自相关函数序列作为特征提取对象,以该序列的近似熵值来量化泄漏信号"不可重复"特征,并将该值作为Elman神经网络输入,辨识泄漏发生。与管外和管内非泄漏固定噪声相辨识结果表明,泄漏发现准确率均高于其他泄漏检测方法,分别达到92.5%和82.5%。According to the generation mechanism of leak acoustic signals, the unpredictability characteristics of leak signal are investigated. The autocorrelation function is used to describe the unpredictability of the leak signal because it has the ability to analyze the coherent structure of time series. The autocorrelation function value for the delay τ larger than the signal correlation length, not the signal itself or its entire autocorrelation function, is used to extract or evaluate the unpredictability degree of the leak signal using the approximate entropy algorithm. The Elman neuralnetwork approach is developed as a classifier, which uses the extracted unpredictability features as the network inputs. The method was employed to identify the leak signals from non-leak noise inside and outside pipelines, and 92.5% and 82.5% correct detection rates are achieved, respectively.

关 键 词：泄漏检测 特征提取 相关分析 近似熵 ELMAN神经网络 

分 类 号：TN911.7[电子电信—通信与信息系统]