A leak detection method for oil pipeline based on markov feature and two-stage decision scheme. (May 2019)

• Record Type: Journal Article
• Authors: Liu, Jinhai; Zang, Dong; Liu, Chen; Ma, Yanjuan; Fu, Mingrui
• Is Part Of: Measurement. Volume 138(2019)
• Page Start: 433
• Page End: 445
• Publication Date: 2019-05
• Subjects: Oil pipeline -- Leak detection method -- Markov feature -- LS-SVM -- Two-stage decision scheme; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.01.029
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A high-accuracy leak detection method is proposed for oil pipeline. Markov feature can effectively represent pipeline dynamic information. Two-stage decision scheme has a good real-time performance. This method has a good ability on small leak detection. Abstract: In the domain of oil and gas transmission, a high-performance leak detection method is of great significance. In this paper, a novel leak detection method based on Markov feature extraction and two-stage decision scheme is proposed to detect pipeline leak. Different from the traditional feature extraction methods, Markov feature is introduced to extract leak information. By means of a transformation, pressure data can be transformed into a Markov chain. By extracting its dynamic feature, raw pressure data can be effectively represented. Furthermore, a two-stage decision scheme is designed. Utilizing a switching rule, short-term and long-term detection models can be correctly selected to identify pipeline status rapidly and precisely. The proposed method is verified by pipeline pressure data collected from the industrial site and experimental field. Experimental results indicate that the proposed leak detection method has a high accuracy and low false positive rate.