Performance Evaluation of Machine Learning Algorithms for Detecting Gas Leakage System

Abstract

Detecting gas leaks in gas plants is a persistent challenge within the Oil and Gas industry, given the prevalence of pipelines for natural gas transportation. Therefore, various traditional techniques are used for gas leakage detection system, however, conventional methods possess various limitations like time consuming, prone to error and can be tedious to work with. Hence, AI based models are preferred for effec        tive gas leakage detection as AI (Artificial Intelligence) based ML (Machine Learning) models generate tremendous advantages in terms of accuracy of gas leakage detection, early detection and being cost effective approaches. Thus present research work focuses on evaluating intelligent models' efficacy in identifying minor leaks in gas pipelines with fundamental operational parameters. The research then proceeds to compare these models using established performance metrics. The ML based models used in the research work are Linear Regression, Logistic Regression, RF (Random Forest) and KNN (K-Nearest Neighbor). The following ML based algorithms are compared and the performance of the model was evaluated using assorted metrics in accordance with different types of damages. Present comparative research work can potentially assist different industrial sectors for identifying gas leaks.