STATISTICAL ANALYSES OF INCIDENTS ON ONSHORE HAZARDOUS LIQUID PIPELINES BASED ON PHMSA DATABASE

摘要

To improve the safety of a pipeline system, engineers use different methods to diagnose the hazardous pipeline accidents. However, most methods ignore the time dependence of pipeline failures. The aim of this paper is to provide a novel approach to analyzing the hazardous liquid pipeline incidents' temporal structure. The database of hazardous liquid spillages from the US between 2002 and 2018 is collected by the Pipeline Hazardous Material Safety Administration of the US Department of Transportation. The result suggests that the whole oil pipeline incident sequence cannot be modeled as a Poisson (random and independent) process, which means that a hazardous liquid pipeline incident is not statistically independent from the time elapsed since the previous event. But the serious pipeline failures are random and unpredictable. The analysis also indicates that the equipment failure, corrosion, material failure and incorrect operation are the four leading failure causes, responsible for most of the total incidents. The study provides insights into the current state of hazard liquid pipelines in the US and baseline failure statistics for the quantitative risk assessments of such pipelines.