Open AccessLeak diagnosis of pipeline based on empirical mode decomposition and support vector machine
Author(s) -
Atik Faysal,
M. S. N. A. Adhreena,
E. Vorathin,
Z. M. Hafizi,
Wai Keng Ngui
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1078/1/012023
Subject(s) - hilbert–huang transform , support vector machine , radial basis function kernel , radial basis function , polynomial kernel , sigmoid function , pipeline (software) , kernel (algebra) , fault (geology) , computer science , pipeline transport , engineering , artificial intelligence , pattern recognition (psychology) , kernel method , mathematics , artificial neural network , environmental engineering , filter (signal processing) , geology , combinatorics , seismology , computer vision , programming language
The pipeline is used as a medium of transportation in global gas and oil industries, providing the most efficient, convenient and transportation method for natural gas and oil from downstream to upstream production of the economical mode of the power station, refineries, and domestic needs. However, the pipeline leakages become a major concern as their failure may contribute to operational and economic loss as well as environmental pollution. This paper proposed a system to detect pipe fault at different locations. Empirical Mode Decomposition (EMD) was applied for feature extraction using energy and kurtosis. The one-against-one (OAO) and one-against-all (OAA) multiclass SVM with radial basis function (RBF), polynomial and sigmoid kernel functions were implemented in order to classify the multiple fault locations from the extracted features. RBF kernel function recorded the highest classification accuracy for both OAO and OAA approaches with 97.77% and 96.29%, respectively, followed by slightly reduced accuracy for sigmoid whereas significantly low accuracy for the polynomial kernel. The outputs were further analysed to justify the performance of the classifiers. From all the cases, it was observed that OAO-SVM with RBF kernel performed the best for pipe fault diagnosis.