A novel method for defects marking and classifying in MFL inspection of pipeline. (April 2023)

• Record Type: Journal Article
• Authors: Pan, Jianhua; Gao, Lun
• Is Part Of: International journal of pressure vessels and piping. Volume 202(2023)
• Publication Date: 2023-04
• Subjects: Magnetic flux leakage -- Pipeline inner inspection -- Defect marking and classification -- Improved CLIQUE algorithm -- Finite element simulation; Pressure vessels -- Periodicals; Pipe -- Periodicals; Récipients sous pression -- Périodiques; Tuyaux -- Périodiques; Pipe; Pressure vessels; Periodicals; 681.76041
• DOI: 10.1016/j.ijpvp.2023.104892
• Languages: English
• ISSNs: 0308-0161

Abstract

Abstract: Magnetic Flux Leakage (MFL) testing is the most widely used non-destructive technique for the inner inspection of oil and gas pipelines. Accurate quantification of defects is a long-standing difficulty in the field of pipeline leak detection. Scientific marking and classification of defects is an important prerequisite for accurate quantification. A novel method of marking and classifying defects with MFL signals is proposed, which is aiming at the problem that it is difficult to mark and classify defects accurately due to the tremendous amount of oil and gas pipeline leakage magnetic detection data. An improved CLIQUE algorithm is used to mark the defect areas of segmented pipelines to estimate the number and location of defects. Then the 3D MFL characteristic signals of the marked areas are studied and extracted. The SSA_BP neural network is trained to classify the defects. The effectiveness and accuracy of the proposed method are tested and verified by using finite element simulation defects and actual defects. The results show that the method is more efficient in marking defects and more detailed in marking areas. Highlights: A database was created for defect marking and classification based on finite elements. The article proposes a pipeline defect marking method based on the improved CLIQUE algorithm. The leakage signal signature for defect marking and classification is advanced. The BP neural network was optimized by greedy algorithm to achieve the required training effect.