Pipeline leak detection based on empirical mode decomposition and deep belief network. (January 2023)

• Record Type: Journal Article
• Authors: Yan, Yulin; Hu, Zhiyong; Yuan, Wenqiang; Wang, Jinyan
• Is Part Of: Measurement and control. Volume 56:Number 1/2(2023)
• Page Start: 396
• Page End: 402
• Publication Date: 2023-01
• Subjects: Cyber-physical system -- deep belief network -- empirical mode decomposition -- leak detection -- support vector machine; Automatic control -- Periodicals; Engineering instruments -- Periodicals; Production engineering -- Periodicals; 629.8
• DOI: 10.1177/00202940221088713
• Languages: English
• ISSNs: 0020-2940

Abstract

Leak detection of an oil pipeline can prevent environmental and financial losses. A method for the cyber-physical system of pipeline leak detection is proposed based on the empirical mode decomposition (EMD) and deep belief network (DBN). Experiment data are acquired from an oil pipeline company. The EMD is suitable for noise removal and signal reconstruction from raw pressure signals, and the reconstructed signals are used to establish a DBN model of pipeline leakage. Our proposed method obtains higher-recognition-accuracy results (98% accuracy) and can more effectively identify leak detection than the twin support vector machine (TWSVM), support vector machine (SVM), and back-propagation neural network (BPNN).