Cerfacs Enter the world of high performance ...

Numerical study of hydrogen/air deflagrations instratified mixtures

MEGEP (Mécanique, Energétique, Génie civil & Procédés) – [Subject to defense authorization]

https://youtube.com/live/EoeZJHtwBfw?feature=share

Hydrogen is a promising energy carrier in the transition toward carbon-free energy systems. However, its physicochemical properties, wide flammability limits, very low ignition energy, and high diffusivity, pose major safety challenges. In confined environments, accidental leaks can lead to the formation of flammable H2air mixtures, whose combustion may evolve from a slow deflagration to a detonation. A central phenomenon governing this transition is flame acceleration (FA), whose understanding and prediction are essential to avoid destructive overpressures. An often overlooked factor that further complicates explosion behavior is mixture inhomogeneity. Following an H2 leak and delayed ignition, the low gas density promotes the accumulation of rich pockets near the ceiling, creating locally highly reactive zones. In this context, computational fluid dynamics (CFD) emerges as a powerful tool for risk assessment, offering a safer alternative to experimental approaches while providing detailed spatial and temporal information on all relevant flow and combustion quantities. This thesis aims to develop and validate a large-eddy simulation (LES) strategy for stratified H2/air deflagrations in confined environments, with the dual objective of improving the predictive capabilities of numerical tools and gaining deeper insight into the mechanisms driving flame acceleration. A first modeling strategy, based on the state of the art, relies on detailed chemical kinetics capable of capturing the local variations of flame properties induced by mixture stratification. Applied to the GraVent explosion channel, this approach shows good agreement with experimental data but reveals a systematic overprediction of flame acceleration in lean mixtures. Lean H2/air flames are characterized by subunity Lewis numbers, which induce a strong sensitivity to flame stretch. This work demonstrates that the use of the Thickened Flame (TF) model to simulate non-unity Lewis number flames introduces an artificial amplification of stretch effects, leading to inaccurate predictions of lean H2–air deflagration propagation. To overcome this limitation, the Stretched–Thickened Flame (S–TF) model is developed. Calibrated a priori from detailed chemistry and validated on canonical laminar flames, this formulation accurately and efficiently reproduces the flame's response to stretch over a wide range of equivalence ratios. Applied to the GraVent configuration, the S–TF strategy achieves excellent agreement with experiments while drastically reducing computational cost. Furthermore, the physical analysis of the simulations highlights the influence of stratification on the mechanisms of flame acceleration. Counterintuitively, stratification does not always enhance FA; its effect strongly depends on geometry. Reactivity gradients along the flame front induce significant flame-surface elongation, promoting acceleration and giving rise to two distinct propagation modes: (i) a small-surface mode, dominant in rich regions and strongly driving FA, and (ii) a large-surface mode, slower but increasingly prevalent as the flame expands into leaner zones.

Jury

Pr. Agnes JÖCHER	Technical University of Munich	Reviewer
Dr. Pierre BOIVIN	CNRS/M2P2	Reviewer
Pr. Thierry SCHULLER	Université Toulouse 3	Examiner
Dr. Nabiha CHAUMEIX	CNRS ICARE	Examiner
Pr. Ashwin CHINNAYYA	CNRS-Université de Poitiers-ISAE ENSMA	Examiner
Mr. Etienne STUDER	CEA	Invited member
Dr. Lucien GALLEN	AIRBUS	Invited member
Dr. Omar DOUNIA	CERFACS	Thesis supervisor
Dr. Quentin DOUASBIN	CERFACS	Thesis co-supervisor