Multi-source information fusion to identify water supply pipe leakage based on SVM and VMD. Issue 2 (March 2022)

• Record Type: Journal Article
• Authors: Wang, Zhoufeng; He, Xiangqi; Shen, Huiling; Fan, Shangjie; Zeng, Yilong
• Is Part Of: Information processing & management. Volume 59:Issue 2(2022)
• Publication Date: 2022-03
• Subjects: Leakage recognition -- Feature extraction -- Information fusion -- SVM; Information storage and retrieval systems -- Periodicals; Information science -- Periodicals; Systèmes d'information -- Périodiques; Sciences de l'information -- Périodiques; Information science; Information storage and retrieval systems; Periodicals; 658.4038
• DOI: 10.1016/j.ipm.2021.102819
• Languages: English
• ISSNs: 0306-4573

Abstract

Highlights: The paper proposes a multi-source information fusion identification method based on VMD and SVM. The paper uses VMD to decompose the acoustic vibration signal of water supply pipes and proposes a criterion for the selection of IMF components. Three feature vectors for multi-source information fusion into a new feature vector, which is fed into the SVM classifier for leakage identification. Abstract: In order to solve the problem of the low leakage recognition rate of water pipes due to operating conditions influence in practice, a multi-source information fusion recognition method based on VMD and SVM is proposed. In this method, it firstly uses VMD to decompose the acoustic vibration signal of water pipes, and then a principle of IMF component selection is proposed. The IMF component selection is selected to extract the kurtosis vector of VMD, the sample entropy vector of VMD, the center frequency vector of VMD. Because the different eigenvectors to the sensitivity of different operating conditions have a great gap, the three eigenvectors become a new eigenvector by multi-source information fusion, which is finally input into SVM classifier for leak recognition. The comparison of experimental results show that this method can effectively recognize the signals of water pipes leak and other operating conditions. The recognition accuracy rate reach 98.75%, which is 1.04 times higher than SVM sorting technique, 1.18 times higher than that SVM classification recognition accuracy based on the sample entropy vector of VMD, 1.14 times higher than that SVM classification recognition accuracy based on the kurtosis vector of VMD, and 1.11 times higher than SVM classification recognition accuracy based on the center frequency vector of VMD.