Combining engineering and data-driven approaches to model the risk of excavation damage to underground natural gas facilities

In the United States there are thousands of gas pipe miles, long grids, and networks of natural gas lines across the states. Recent pipeline leaks and explosions in various regions have driven the industry to re-evaluate on-going efforts aimed at aggressive pursuit of preventive strategies. Considering that safety and environmental risk is a major issue, particularly in cases where underground gas line damages and other explosions are involved, pipeline accidental risk represents both financial and social interests in the gas pipeline industry. It is possible to, knowingly or unknowingly, damage underground gas services, water services, electrical services, etc. Incidents involving infrastructure damage are far more common than perceived; and these incidents result in hundreds-of-thousands, if not millions, of dollars in repair or replacement. Damages to underground facilities may occur by large construction contractors or by homeowners. The main objective of this research is two-fold: a) to determine the important risk factors contributing to the underground gas pipe damages; b) to identify inputs required for an effective evaluation and assessment of the risk encountered in exchange of information between different parties involved during the repair of underground gas pipelines. Predictive Model will be developed based on machine learning algorithms (Logistic Regression) to be used in predicting the important risk factors affecting the underground Gas Pipe Damages. The research will systematically analyze the risk of underground gas pipeline network damage including; process the data collected from agency, organize/classify the data based on certain parameters, process the data, develop integrated risk model and influence diagram. Next, Bayesian Network will be developed based on the derived important factors, and calculated probabilities for each attribute.