Mesh-Independent Large-Eddy Simulation with Anisotropic Adaptive Mesh Refinement for Hydrogen Deflagration Prediction in Closed Vessels

The use of high-fidelity simulation methods based on large-eddy simulation (LES) are proving useful for understanding and mitigating the safety hazards associated with hydrogen releases from nuclear power plants. However, accurate modelling of turbulent premixed hydrogen flames via LES can require very high resolution to capture both the large-scale turbulence and its interaction with the flame fronts. Standard meshing strategies can result in impractically high computational costs, especially for the thin fronts of hydrogen flames. For these reasons, the use of a recently formulated integral length scale approximation (ILSA) subfilter-scale model, in combination with an efficient anisotropic block-based adaptive mesh refinement (AMR) technique is proposed and examined herein for performing LES of turbulent premixed hydrogen flames. The anisotropic AMR method allows dynamic and solution-dependent resolution of flame fronts and the grid-independent properties of the ILSA model ensure that numerical errors associated with implicitly-filtered LES techniques in regions with varying resolution are avoided. The combined approach has the potential to allow formally converged LES solutions (direct numerical simulation results are typically reached in the limit of very fine meshes with standard subgrid models). The proposed LES methodology is applied to combustion simulations of lean premixed hydrogen-air mixtures within closed vessels: a problem relevant to hydrogen safety applications in nuclear facilities. A progress variable-based method with a multi-phenomena burning velocity model is used as the combustion model. The present simulation results are compared to the available experiment data for several previously studied THAI vessel cases and the capabilities of the proposed LES approach are assessed.