Decision Tree-Based Approach for Defect Detection and Classification in Oil and Gas Pipelines

Abstract

Metallic pipelines are used to transfer crude oil and natural gas. These pipelines extend for hundreds of kilometers, and as such, they are very vulnerable to physical defects such as dents, cracks, corrosion, etc. These defects may lead to catastrophic consequences if not managed properly. Thus, monitoring these pipelines is an important step in the maintenance process to keep them up and running. During the monitoring stage, two critical tasks are carried out: defect detection and defect classification. The first task concerns with the determination of the occurrence of a defect in the monitored pipeline. The second task concerns with classifying the detected defect as a serious or tolerable defect. In order to accomplish these tasks, maintenance engineers utilize Magnetic Flux Leakage (MFL) data obtained from a large number of magnetic sensors. However, the complexity and amount of MFL data make the detection and classification of pipelines defects a difficult task. In this study, we propose a decision tree–based approach as a viable monitoring tool for the oil and gas pipelines.