| description abstract | Pipelines are integral to global energy infrastructure, facilitating the secure transport of large volumes of oil and hazardous materials across continents. Despite their critical role, these systems are vulnerable to incidents frequently resulting in significant disruptions and economic losses worldwide. Previous studies have primarily used bivariate statistical analyses, which overlook complex variable interactions, while few have employed Bayesian networks, often depending on subjective expert input. This study employed association rule mining on the Pipeline Hazardous Material Safety Administration incident database to overcome these gaps. The proposed methodology aims to uncover intricate associations between incident causes, pipeline characteristics, failure modes, and shutdown durations, which previous studies have overlooked. This study’s key findings demonstrate that the interplay between multiple background factors exacerbates the severity and likelihood of incidents, offering new insights into failure dynamics. For example, the analysis revealed that older pipelines with a large diameter installed in underground areas significantly increase the likelihood of mechanical punctures and medium-term shutdowns due to excavation damage. Understanding these factors is expected to substantially benefit pipeline planning, design, and operations. By identifying and addressing the causes of shutdowns, stakeholders can implement preventative measures, thereby promoting safer and more reliable energy transport systems. | |
