Flame accelerationin inhomogeneouscombustible gas mixturehas largely been overlooked despite being relevant to many accidental scenarios. The present study aims to validate our newly developed density-based solver, ExplosionFoam,for flame acceleration and deflagration-to-detonationtransition. The solver is based on the open source computational fluid dynamics (CFD) platform OpenFOAM®.For combustion, it uses thehydrogen-air single-step chemistry and the corresponding transport coefficients developed by the authors.Numerical simulations have been conducted for the experimental setupof Ettneret al.[1], which involves flame acceleration and DDT in both homogeneous hydrogen-airmixtureas well as an inhomogeneous mixture with concentration gradientsinanobstucted channel. The predictionsdemonstrate good quantitative agreementwith the experimental measurementsin flame tip position, speed and pressure profiles. Qualitatively, the numerical simulations reproducewell the flame acceleration and DDT phenomena observed in the experiment. The results have shown that in the computed cases, DDT is induced by the interaction of the precursor inert shock wave with the wallclose to high hydrogen concentrationrather than with the obstacle. Some vortex pairs appear ahead of the flame due to the interaction between the obstacles and the gas flow caused by combustion-induced expansion, but they soon disappear after the flame passesthrough them. Hydrogen cannot be completely consumedespeciallyin the fuel rich region. This is of additional safety concern as the unburned hydrogen can potentially re-ignite once more fresh air is available in an accidental scenario, causing subsequent explosions.

The results demonstrate the potential of the newly developeddensity based solverfor modellingflame acceleration and DDT in both homogeneous/inhomogeneous hydrogen-air mixture. Further validation needs to be carried outfor other mixtures and large-scale cases.

A two-layer, reduced order model of high pressure hydrogen jets was developed which includes partitioning of the flow between the central core jet region leading to the Mach disk and the supersonic slip region around the core. The flow after the Mach disk is subsonic while the flow around the Mach disk is supersonic with a significant amount of entrained air. This flow structure significantly affects the hydrogen concentration profiles downstream. The predictions of this model are compared to previous experimental data for high pressure hydrogen jets up to 20 MPa and to notional nozzle models and CFD models for pressures up to 35 MPa. The results show that this reduced order model gives better predictions of the mole fraction distributions than previous models for highly underexpanded jets. The predicted locations of the 4%2lower flammability limit also show that the two-layer model much more accurately predicts the measured locations than the notional nozzle models. The comparisons also show that the CFD model always underpredicts the measured mole fraction concentrations.