Dynamic predictive analysis of the consequences of gas pipeline failures using a Bayesian network

Abstract

Modern natural gas pipeline failures constitute devastating disasters, as they can result in cascading secondary crises. Therefore, reduction of buried gas pipeline's reliability, has become a major concern among stakeholders and researchers in recent years. This study employs a dynamic Bayesian network to investigate the consequences of natural gas pipeline failures. We consider seven parent nodes—age, diameter, length, depth, population, time of occurrence, and land use—and twelve consequence factors to analyze the overall losses stemming from pipeline failure. The proposed model can handle both static and dynamic systems using quantitative and/or qualitative data. To demonstrate the applicability and effectiveness of our developed model, we analyze the gas pipeline network of Regina in Saskatchewan, Canada. The results show that age and diameter are the two most important and sensitive parameters. The developed Bayesian network model will aid decision-makers in effectively managing and improving the reliability of their assets.