Development of Liquid Hydrogen Leak Frequencies Using a Bayesian Update Process

To quantify the risk of an accident in a liquid hydrogen system, it is necessary to determine how often a leak may occur. To do this, representative component leakage frequencies specific to liquid hydrogen can be determined as a function of the normalized leak size. Subsequently, the system characteristics (e.g., system pressure) can be used to calculate accident consequences. Operating data (such as leak frequencies) for liquid hydrogen systems are very limited; rather than selecting a single leak frequency value from a literature source, data from different sources can be combined using a Bayesian model. This approach provides leakage rates for different amounts of leakage, distributions for leakage rates to propagate through risk assessment models to establish risk result uncertainty, and a means for incorporating liquid hydrogen-specific leakage data with leakage frequencies from other fuels. Specifically, other cryogenic fluids like liquefied natural gas are used as a baseline for the Bayesian analysis. This Bayesian update process is used to develop leak frequency distributions for different system component types and leak sizes. These leak frequencies can be refined as liquid hydrogen data becomes available and may then inform safety code requirements based on the likelihood of liquid hydrogen release for different systems.