The advent of Computational Fluid Dynamics (CFD) as an alternative to expensive and dangerous experiments, and the recent emergence of large eddy simulations (LES) in the field of turbulent combustion, have allowed for new perspectives on how to make explosion simulations truly quantitative.

This research, conducted by the European Centre for Research and Advanced Training in Scientific Computation (CERFACS), will study accidental gas explosions in buildings and develop models for turbulent pre-mixed flames. Using previous INCITE allocations, the team developed a unique experimental database created via LES. This database has led to a new experimental campaign which will allow researchers to check the validity of models over a range of scales previously not accessible.

The target configurations for the proposed research on gas explosions are the original small-scale configuration of 0.25 m long, and two larger configurations of 1.5 m and 6 m long. The experimental database includes several diagnostics—pressure signals, flame-front visualizations—for methane, propane and hydrogen. Databases for the medium- and large-scale configurations are more restrictive, considering only propane/air mixtures.

Based on these three experiments of increasing size, a unique set of data is available for validating and up-scaling LES models. In particular, this is a unique opportunity to highlight the limitations of constant-based models for flame wrinkling—the vorticity produced by various flame factors, such as stability and velocity—compared to the benefits that dynamic models may offer.

To perform these computations, CERFACS will rely on the AVBP code, one of the most advanced combustion solvers developed for parallel machines like the ALCF’s Mira. AVBP has proven to be the proper tool to tackle explosion studies to achieve project objectives.