Leak detection method of liquid-filled pipeline based on VMD and SVM

被引：8

作者：

Zhao, Si-Liang ^{[1
]}

Liu, Shao-Gang ^{[1
]}

Qiu, Bo ^{[2
]}

Hong, Zhou ^{[1
]}

Zhao, Dan ^{[1
]}

Dong, Li-Qiang ^{[1
]}

机构：

[1] Harbin Engn Univ, Coll Mech & Elect Engn, Harbin, Peoples R China

[2] CSSC Jiu Jiang Fire Equipment Co Ltd, Dept technol, Jiujiang, Peoples R China

来源：

URBAN WATER JOURNAL | 2023年 / 20卷 / 09期

基金：

中国国家自然科学基金;

关键词：

Leak detection; variational modal decomposition; Spearman correlation coefficients; support vector machine; FAULT-DIAGNOSIS; SUPPORT; ENTROPY; EMD;

D O I：

10.1080/1573062X.2023.2251952

中图分类号：

TV21 [水资源调查与水利规划];

学科分类号：

081501 ;

摘要：

In order to solve the problem of inconspicuous leakage signal characteristics under external noise interference, a leakage detection method based on the combination of variational modal decomposition (VMD) and support vector machine (SVM) is proposed. The method first calculates the spearman correlation coefficients (SCC) of multiple intrinsic modal components (IMFs) obtained by VMD with the source signal, then extracts the energy and central frequency features of IMFs with larger SCC, and finally performs leak detection using the SVM classifier. The experimental results show that the VMD-SVM method can effectively perform leak detection with an accuracy of 98.27%. The accuracy of the VMD-SVM method proposed in this paper is improved by 6.5%, 5.63% and 10.39% compared to the time-frequency (TF) feature SVM, empirical modal decomposition (EMD) feature SVM and wavelet (DWT) feature SVM, methods, respectively. In addition, feature sensitivities are analyzed to reduce model complexity while ensuring accuracy.

引用

页码：1169 / 1182

页数：14