Cerfacs Enter the world of high performance ...

Adaptive mesh refinement methods for large eddy simulations of gas explosions

The global energy demand continues to rise, and is largely met through combustion, using fossil or renewable fuels. These fuels, often stored in enclosed environments, pose a significant hazard in the event of a leak. The ignition of a premixed gas cloud can lead to an explosion, causing rapid flame front propagation and generating dangerous overpressures that threaten both human life and infrastructure integrity. To understand and prevent these explosions, various experiments are conducted, ranging from laboratory tests to industrial-scale simulations. However, extreme conditions of temperature and pressure make it challenging to obtain accurate diagnostics experimentally.

Numerical simulation, especially Large Eddy Simulation (LES), complements these experiments by providing a better understanding of combustion and turbulence phenomena at stake. LES has already proven effective in replicating the dynamics of deflagrations and the associated overpressures in small domains. It also allows for precise diagnostics at every point within the computational domain. However, the large dimensions of industrial installations raise challenges for a complete numerical resolution of the physical phenomena involved. An homogeneous discretisation of the entire computational domain would be too costly in terms of return time and computational resources. Therefore, mesh adaptation, particularly dynamic adaptation, is used to refine the discretisation in regions of interest that evolve during the calculation. This technique helps reduce mesh size and computational costs by tracking predefined phenomena of interest during their propagation.

This thesis focuses on the development and validation of an adaptive mesh refinement (AMR) method for LES simulations of deflagrations, based on instantaneous physical criteria relevant to explosions. The proposed method, called “Turbulent Flame Propagation-AMR” (TFP-AMR), reproduces the transient dynamics of turbulent flames and vortical structures in the flow, and relies on the unstructured AMR library kalpaTARU. The method relies on criteria derived from the physical characteristics of deflagrations, minimising reliance on user-dependent parameters. In particular, a vortex selection criterion is derived from flame/vortex interaction theory. A specific mesh adaptation triggering criterion is also developed to ensure that regions of interest remain within a refined mesh zone throughout the transient propagation process.

The methodology is validated on fundamental cases representative of the essential physical bricks of the problem, such as flame propagation, vortex propagation, and flame-vortex interaction. Finally, the method is applied to deflagration configurations. A semi-confined obstructed chamber is first considered, with extensive parametric variations regarding the chamber geometry and the initial mixture properties. A fully confined obstructed channel is then considered, where deflagration can reach high-speed regimes with shock waves forming ahead of the flame front. Comparisons between experimental and simulation results demonstrate that the TFP-AMR method achieves accurate results at a lower computational cost compared to static mesh reference simulations, while requiring minimal parameter adjustments.  These application cases validate the method robustness and effectiveness for such applications.

Jury

M. Christian TENAUD	EM2C, CentraleSupélec	Rapporteur
M. Laurent BRICTEUX	Université de Mons	Rapporteur
M. Marc BELLENOUE	ISAE-ENSMA	Examinateur
Mme. Laura GASTALDO	IRSN	Examinatrice
M. Cédric MEHL	IFP Énergies nouvelles	Examinateur
M. Vincent BLANCHETIERE	GRTGaz	Invité
M. Laurent GICQUEL	CERFACS	Directeur de thèse
M. Thomas JARAVEL	CERFACS	Invité – Co-encadrant de thèse