Development of Probabilistic Reliability Models for Pure and Blended Hydrogen Pipelines

Abstract Growing demand for hydrogen and blended hydrogen / natural gas pipelines has introduced challenges that require engineering assessments and controls to ensure safe pipeline design, conversion, and operation. Quantitative risk assessment using probabilistic reliability models is a leading engineering assessment technique that is uniquely suited to addressing these challenges, even where limited historical experience and regulatory guidance exist. This paper presents developments that have been made to adapt the probabilistic reliability models commonly used to estimate the failure rates of conventional natural gas pipelines for application to pure and blended hydrogen pipelines. These developments are based on the latest research on the adverse effects of hydrogen on pipeline material performance (i.e. hydrogen embrittlement). Hydrogen embrittlement has two primary effects on pipeline steel properties: a reduction in fracture toughness, and an increase in fatigue crack growth rate. Both effects are expected to increase the predicted failure rate of pipelines containing crack-like flaws. The extent to which these effects may significantly compromise the integrity of a pipeline asset is dependent on the extent of the embrittlement (as determined by the susceptibility of the pipeline material and the degree of hydrogen exposure) and the pipeline operating conditions (which influence both nominal and cyclic stress states). In any assessment, significant uncertainty may exist regarding the precise nature of hydrogen embrittlement mechanisms, and the material properties and existing damage conditions of the asset. However, probabilistic reliability models provide a means of systematically accounting for these uncertainties and can provide a quantitative basis to reduce those uncertainties as research advances. This paper will discuss the use of empirical models to predict the extent of hydrogen-induced fracture toughness reduction and fatigue crack growth rate increase, the results of which have been characterized probabilistically herein. These models have been incorporated in a simulation of bi-directional crack growth (based on a variable amplitude, cyclic loading histogram input) to predict the time to reach a critical crack size. The probabilistic reliability model repeats these crack growth simulations in a Monte Carlo simulation to predict future failure rates for a prescribed evaluation period. A demonstration of the model predictions is also presented in this paper through a comparison of the predicted failure rates for a representative set of pipeline scenarios with varied hydrogen blend ratios and baseline predictions with a pure natural gas product. Finally, a description is given of the remaining gaps to be addressed in future research.