Open AccessA new recognition method for oil pipeline leakage using PCA and SOM neural networks
Author(s) -
Beilei Ji,
Hong Zhang,
Shen Liu,
Zhang Ke-zheng,
Wei Zhang,
Dong Zhang,
Xiaoben Liu
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/783/1/012167
Subject(s) - leakage (economics) , principal component analysis , artificial neural network , petroleum engineering , artificial intelligence , pipeline (software) , pipeline transport , computer science , environmental science , pattern recognition (psychology) , acoustics , engineering , environmental engineering , mechanical engineering , physics , economics , macroeconomics
Oils are mainly transported by pipe in long distance for its high efficiency. While oil pipe leakage will cause serious social and environmental consequences, e.g. fire even life lost, water and soil pollution. Thus it is important to recognize pipe leakage at initial stage in engineering practice. In this research, a negative pressure wave based detection method was established for pipeline leakage recognition. Suitable parameters of negative pressure wave signals with significant difference for different working conditions were selected. Principal Component Analysis (PCA) method was conducted to reduce the dimensions of the negative pressure wave vector. Self-organizing map (SOM) Neural network was finally adopted to identify the signals for different working conditions. The proposed method was validated by experimental data, which shows that the methodology gives a high recognition rate, which can be referenced in pipe monitoring in engineering practice.