Probabilistic model development and risk management strategies for pipelines with anomalies

Metal pipelines are exposed to many threats such as corrosion and cracking during their service life, which can weaken pipeline integrity and cause leaks or ruptures, leading to significant safety, economic and environmental consequences. For corrosion, many models have been developed to predict failure pressure of pipeline with an isolated defect, and only a few models consider interacting defects. On the other hand, the existing models for predicting failure pressure of pipeline with crack defects are very limited. Several model performance evaluations have shown that most of the existing models for either corrosion or cracking-like defects are conservative since they are intended for design, but this may lead to unnecessary repairs and maintenance when they are used in pipeline risk management programs. The goal of this research is to develop a quantitative risk assessment for deteriorating pipelines with corrosion and crack-like defects based on reliable probabilistic failure prediction models. To achieve this goal, the research includes four objectives: (1) to develop a probabilistic failure pressure model for pipelines with an isolated corrosion defect, (2) to develop a probabilistic interaction rule and failure pressure model for pipelines with a colony of corrosion defects, (3) to develop a probabilistic failure pressure model for pipelines with a single crack-like defect, and (4) to build a framework of determining expected life-cycle cost of deteriorating pipelines, which can be used for optimal inspection and maintenance planning. For the first three objectives, three comprehensive databases are established, respectively. Each database consists of experimental and numerical data collected from literature and newly generated numerical data obtained from finite element analysis using ABAQUS. The pressure prediction models are developed by either utilizing the existing models as independent variables or adding correction factors to existing models. The framework for determining expected life-cycle cost of deteriorating pipelines is developed based on a decision tree model using analytical methods to evaluate events and with the consideration of the impact of inspection and possible repair on the failure time evaluation. This study results in probabilistic prediction models for failure pressure of pipelines containing interactive anomalies that provide unbiased predictions with reduced variability and an understanding of the propagation of prevailing uncertainties in the prediction models for quantitative risk management of pipelines.