Pipeline Leak Detection System Using Machine Learning

Leak detection is a critical concern across various sectors due to its potential environmental, economic, and safety implications. This paper presents a machine learning-based approach aimed at enhancing leak detection strategies, focusing on predicting the distance and diameter of leaks in pipelines. Leveraging a comprehensive dataset sourced from experimental setups featuring deliberate leaks, our model utilizes pressure and flow measurements to accurately identify and localize leaks. Through rigorous evaluation of multiple machine learning algorithms, including K-Nearest Neighbors, XGBoost, Decision Tree, Random Forest, and Support Vector Machine, we identify XGBoost for predicting leak distances and Random Forest for estimating leak diameters as optimal choices. The proposed approach demonstrates promising results in accurately detecting and quantifying pipeline leaks, offering valuable insights for ensuring pipeline integrity and safety.