Accidents due to gas explosion in industrial buildings are a major safety issue. Understanding the mechanisms controlling explosions in buildings, typically in offshore platforms, is an economical and physical challenge.

Gas explosions are studied in confined or semi-confined vessels with obstacles where a premixed flame is ignited. The propagation of the flame and the resulting over-pressure are the key parameters that define safety concepts and guidelines for explosion protection. Codes based on the unsteady Reynolds averaged Navier-Stokes (URANS) approach are standard for gas explosion studies at industrial scales but they rely on crude parameterizations.

The recent emergence of large eddy simulation (LES) in the field of turbulent combustion opens new perspectives to make explosion simulations truly quantitative. Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique (CERFACS) recent LES of explosions has revealed the importance of the laminar phase which is often neglected or computed in approximate ways because the codes used in the literature are adapted to turbulent flames and not to laminar ones. A predictive solver must be able, with the same sub-models, to compute both a laminar flame and a turbulent one, something which is almost never done in combustion simulations.

This project aims to use the cutting-edge computational power of the IBM Blue Gene/P and Q systems at the ALCF to achieve this objective. It also suggests that a proper way to investigate the question is to begin by a fully resolved, DNS-type (Direct Numerical Simulation), simulation of the same flow to have a reference solution.