Our project aims at improving the understanding of deflagrations and detonations in hydrogen-air mixtures. A special focus lies on the flame acceleration phase and the sudden transition to detonation which is a critical issue for nuclear reactor safety considerations. Whereas in the past homogeneous mixtures have intensely been investigated, concentration gradients perpendicular to the main direction of flame propagation are taken into account here. In case of a core meltdown, this represents a more realistic scenario of hydrogen distribution within a reactor containment dome [1]. Once a reliable prediction in terms of flame speeds and pressure loads is possible, the structural response of the containment can be evaluated.

Besides experimental research, numerical simulations are a promising way to study the combustion process. The latter approach is discussed in this paper to make a contribution in the analysis of an explosion channel experiment on laboratory scale. Responding to an idea raised in [2], two different obstacle configurations are examined. Thus, the orientation of obstacles (which are necessary to promote flame acceleration) relative to the concentration gradient is varied. Three-dimensional computations were performed to capture the full geometrical complexity of the problem. Moreover, two-dimensional and threedimensional simulations are checked against each other. We want to demonstrate the potential of flame surface analysis to identify the relevant differences. To study the DDT (deflagration-to-detonation transition) process in detail is not the goal using the simplified but computationally very efficient approach described. Unlike simulation techniques requiring high spatial resolution, the analysis can therefore be applied to real-world scenarios.