Abstract
Risk is defined as a product of the probability of a hazard causing an adverse event combined with the severity (or consequences) of that adverse event. Accordingly, pipeline risk management should integrate both the concepts of failure frequency and the potential consequences for each hazard scenario. Pipelines’ risk assessment is particularly challenging because pipelines cover extended geographic regions and there are numerous threats to pipeline integrity. Consequences are not always easy to evaluate depending on many parameters such as type of product being transported and terrain. Furthermore, physical models and statistical data cannot completely capture the complex time-based phenomena leading to failures. On the other hand, risk ranking or scoring methods often appear to be arbitrary in terms of how the final risk score is computed. To solve this problem, a pipeline risk assessment model has been created using a probabilistic graphical model called a Bayesian Network. The model calculates separately internal and external corrosion risks, manufacturing and construction risks, natural hazard risks, third party damage risks and maintenance/operation error risks. Moreover, these risks might lead to a “loss of containment” which has four different consequences: financial impact, environmental impact, effects on health and safety, and public outrage. The Bayesian network model is illustrated through a specific example involving internal corrosion. The geographically-based risk assessment allows the user to prioritize effectively pipeline inspections and repairs.
