Construction and Verification of a Bayesian Network for Third-Party Excavation Risk Assessment (BaNTERA)

Abstract

According to the Pipeline and Hazardous Material Safety Administration (PHMSA), thirdparty damage is a leading cause of natural gas pipeline accidents. Although the risk of third-party damage has been widely studied in the literature, current models do not capture a sufficiently comprehensive set of up-to-date root cause factors and their dependencies. This limits their ability to achieve an accurate risk assessment that can be traced to meaningful elements of an excavation. This paper presents the construction, verification, and validation of a probabilistic Bayesian network model for third-party excavation risk assessment, BaNTERA. The model was constructed and its performance verified using the best available industry data and previous models from multiple sources. Historical industry data and nationwide statistics were compared with BaNTERA’s damage rate predictions to validate the model. The result of this work is a comprehensive risk model for the third-party damage problem in natural gas pipelines.