摘要
针对城市供水管道早期堵塞难以检测的问题,提出了一种基于局部均值分解(LMD)的分量信号特征提取,结合支持向量机(SVM)的堵塞故障识别方法。先对声响应信号进行LMD,得到若干乘积函数(PF)分量,进而采用相关分析法选取有效PF分量,对有效PF分量分别提取能量熵、近似熵和平均声压三个指标的特征,构建分类特征集。最后利用交叉验证(CV)方法优化参数的SVM分类器识别堵塞故障信号。实验结果表明:采取基于LMD特征融合和通过CV优化的SVM相结合的方法可以有效识别供水管道的初期堵塞。与基于LMD特征融合和BP神经网络的方法进行了对比,结果表明:本文方法具有更好的堵塞故障识别效果。
Aiming at the detection problem of blockage within the urban water supply pipeline in the early stage, a method combined component signal feature extraction with support vector machine （SVM）for blockage recognition based on local mean decomposition（LMD） is proposed. The first step of this method is to decompose the acoustic response signal by LMD,so that a number of PF components can be obtained. Then the correlation analysis method is used to select the effective PF components. The characteristics of energy entropy, approximate entropy and average sound pressure of the effective PF components are extracted respectively, so the classification feature sets can be constructed. Finally, the cross validation（CV） is used to optimize the parameters of the SVM classifier to identify the blockage fault signal. The results from the experiments have shown that the method can identify the blockage in the water supply pipeline effectively based on the combination usage of LMD component features and cross validation SVM. In addition, the method is compared with the method based on LMD feature fusion and BP neural network and the results suggest that the proposed method has a better performance on the partial blockage recognition.
出处
《传感器与微系统》
CSCD
2017年第7期57-61,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61563024
51169007)
昆明理工大学引进人才科研启动基金项目(KKZ3201503015)
关键词
供水管道
堵塞物识别
局域均值分解
特征融合
支持向量机
water supply pipeline
blockage identification
local mean decomposition （ LMD ）
feature fusion
support vector machine（SVM）
