Cerfacs Enter the world of high performance ...

Large Eddy Simulations of hydrogen/air gas explosions in big-scale, complex geometries

ED MEGEP – [Subject to defense authorization]

The prevention of gas explosions is a global industrial safety concern. Every year, gas explosions are the cause of human casualties, material losses and business disruption. Hydrogen is a very energetic fuel with wide flammability limits and a low ignition energy. As such, it presents especially high gas-explosion-related risks. Moreover, its small molecular size, and its high pressure/low temperature storage conditions, increase the likelihood of accidental leaks. Nowadays, in the context of the energy transition, global decarbonisation is at the origin of a rapid growth in hydrogen demand and production. In the next coming years, it is expected that hydrogen will be incorporated, for the first time, in many industrial sectors. This multiplies the risks of hydrogen explosions, with potentially disastrous consequences. Having reliable tools to assess the risk of gas explosions, in particular for hydrogen, in configurations close to industrial safety applications, is of capital importance. Large Eddy Simulation (LES), represents a good compromise between accuracy and cost-effectiveness. It provides a way to understand and reproduce the complex physical processes driving the generation of the overpressure, which is responsible for the destructive effects. Even if LES has already proven its capability to accurately reproduce gas explosions, some challenges remain: i) the multi-scale nature of explosions can make the cost of simulations of larger-scale systems prohibitive; ii) lean hydrogen flames are subject to thermodiffusive (TD) effects, which stem from the peculiar diffusion properties of molecular hydrogen and result in turbulent burning speeds much higher than the ones predicted with classical turbulent combustion modelling approaches for LES; iii) LES and its associated models rely on validation against experimental measurements, which are often scarce or not sufficiently detailed due to the harsh conditions during explosions. This thesis aims at tackling these challenges by developing a LES methodology capable of simulating turbulent gas explosions in complex, larger-scale configurations, especially of hydrogen/air mixtures. First, to reduce the computational cost of the simulations, an Adaptive Mesh Refinement (AMR) method is developed and validated. This AMR method is specifically designed for gas explosions in complex, obstructed geometries. The AMR method uses dedicated sensors and criteria to detect flow features of interest: the turbulent flame brush and the vortical structures in the flow. It ensures that said features are adequately resolved by dynamically adapting the mesh on-the-fly. This allows for a significant cost reduction while preserving the accuracy of the results. Then, the literature is scouted for relevant experimental gas explosion configurations, which are selected according to criteria of scale, available experimental diagnostics, repeatability and control of the boundary and initial conditions. The selected experimental configurations are, then, simulated to showcase the LES methodology, develop and validate new models, and understand the physical mechanisms driving the process of flame acceleration in complex configurations of industrial interest (e.g. obstructed channels and vented chambers). Finally, to be able to perform simulations of very lean hydrogen/air explosions, the turbulent flame speed and fractal properties of flames with TD effects, are studied through 3-D DNS at variable turbulence conditions. A new subgrid-scale turbulent flame speed model for LES is proposed, which accounts for TD effects and their synergistic interaction with turbulence. The model is, finally, validated both in a priori and a posteriori tests of LES of very lean turbulent explosions. Overall, good results are obtained all across the thesis, in agreement with experimental observations. The methodology is proven to be robust and accurate, regardless of the explosion configuration or the operating conditions.

Jury

M. A. Aspden	Newcastle University	Rapporteur
Mme. N. Chaumeix	ICARE, CNRS	Rapporteuse
M. T. Schuller	Université Toulouse 3 – IMFT, CNRS	Examinateur
M. S. Kudriakov	CEA	Examinateur
M. O. Colin	IFPEN	Examinateur
M. T. Poinsot	IMFT, CNRS	Directeur
M. B. Labegorre	Air Liquide	Invité, co-encadrant
M. T. Jaravel	CERFACS	Invité, co-encadrant