Abstract
This paper summarises the results from a blind-prediction benchmark study for models used for estimating the consequences of vented hydrogen deflagrations, as well as for users of such models. The work was part of the HySEA project that received funding from the Fuel Cells and Hydrogen Joint Undertaking (FCH JU) under grant agreement no. 671461. The first blind-prediction benchmark exercise in the HySEA project focused on vented explosions with homogeneous hydrogen-air mixtures in 20-foot ISO containers. The scenarios selected for the second blind-prediction study focused on vented deflagrations in inhomogeneous hydrogen-air mixtures resulting from continuous stratification of hydrogen during vertical jet releases inside 20-foot ISO containers. The deflagrations were vented through commercial vent panels located on the roof of the containers.
The test program included two configurations and four experiments, i.e. two repeated tests for each scenario. The paper compares experimental results and model predictions and discusses the implications of the findings for safety related to hydrogen applications. Several modellers predicted the stratification of hydrogen inside the container during the release phase with reasonable accuracy. However, there is significant spread in the model predictions, especially for the maximum reduced explosion pressure, and including predictions from different modellers using the same model system. The results from the blind-prediction benchmark studies performed as part of the HySEA project constitute a strong incentive for developers of consequence models to improve their models, implement automated procedures for scenario definition and grid generation, and update training and guidelines for users of the models.
[Display omitted]
•Blind-prediction benchmark study for consequence models.•Vented hydrogen deflagration experiments in 20-foor ISO containers.•Model predictions by computational fluid dynamics (CFD) tools.•Model predictions by empirical or semi-empirical engineering models.