Predictive risk-based model for oil and gas pipelines

Among different means of oil and gas transportation, pipelines are considered to be the safest and reasonably efficient. However, reviewing the incidents recorded on oil and gas pipelines in the United States of America (1992-2012) proved that the consequences of failures in these pipelines have been considerable: over 5.5 billion dollars of asset damage, 380 fatalities and 1,500 injuries, in addition to more than 5.5 million barrels of product loss. The domain certainly requires attention. In addition, there is a lack of research available in this crucial field. Therefore, the objective of this research is to develop a risk model for oil and gas pipelines. Data from USA incidents recorded on oil and gas pipelines from 1992 to 2011 are utilized to build the intended model and identify risk factors leading to oil and gas pipeline failure while considering their consequences of failure. An artificial neural network (ANN) with two hidden layers through applying back propagation approach is trained to develop the model which will help decision makers to estimate how significant would be a failure on a pipeline for the identified risk factors.