Pipeline Failure and Deterioration Models

The utility managers and other authorities rely on location-specific pipeline failure information to develop and implement long- and short-term management plans. The pipeline failure or deterioration depends on multiple pipe dependent, time-dependent, and pipe and time-dependent factors. For this, substantial efforts have been made to develop pipeline failure or deterioration prediction models using different statistical models like regression-based, survival analysis-based, and Bayesian inference-based models. The performance of these models depends significantly on the quantity and quality of available data. Most prediction models focus on larger utilities where sufficient amount of pipeline failure and other related information is available. Small and medium-size utilities often suffer due to technical and financial resource shortages and data scarcity. It is essential to consider model uncertainties for pipeline failure or deterioration prediction model development. In this study, a brief review of existing regression-based, survival analysis-based, and Bayesian inference-based pipeline failure prediction models are presented. To demonstrate the applicability of different pipeline failure and deterioration prediction models, the pipeline failure data of the City of Calgary is considered.