Combustion in Liquid Rocket Engines (LRE) happens in extreme conditions which imply several multi-physics phenomena. For this reason, numerical simulation is used to predict and thus to optimize the engine performances and lifetime. In particular this thesis focuses on two main aspects: turbulent oxy-combustion in diffusion flames of methane at high pressure, and prediction of wall heat transfers. The Large Eddy Simulation (LES) code AVBP of CERFACS is used.
Despite its lower performances, methane is now preferred to hydrogen for future LRE because of its reduced cost and its practicality both for in terms of usage and storage. For numerical simulation, this propellant raises new questions about how to ignite and stabilize the flame. To do so, developing realistic chemistry is a key step. Reduced finite rate chemistry schemes with about 15 species are derived and tested for high pressure and highly strained counterflow diffusion flames. However, even reduced kinetic schemes are still expensive in the context of industrial LES simulations. Therefore a new integration method for the chemical source terms is proposed in order to run reactive simulations closer to the flow time step. It is found that significant computational cost is spared, while keeping the same result accuracy compared to the classical integration. Finally, in order to develop future turbulent diffusion flame modeling, a study on how the mesh resolution impacts diffusion flames is also performed. The development of reduced chemistry allows to study precisely the influence of chemical reactions at the near-wall region in LRE conditions on the wall heat flux. Periodic turbulent channels are computed to compare the resolved and non-resolved turbulent boundary layer, with or without chemical reactions. Results show that the near-wall reactions may have a real impact on wall heat flux, and that wall models should take into account this effect in the context of wall-modeled LES.
Another study is conducted to determine the impact of the coupling between the sub-grid scale model and the wall-law on the wall fluxes prediction. It is shown that the amount of turbulent viscosity at the near-wall region greatly changes the fluxes. A stochastic-based model is proposed in the case of isothermal simulations, in order to improve the results for two common LES sub-grid scale models, WALE and Sigma.
The developed models and analyses of those test cases are then used for the LES simulation of two test rigs: the supercritical 5-injectors GCH4/GOx from ONERA and the subcritical single-injector GCH4/LOx from TUM. Their study particularly focuses on the flame behavior and the wall heat flux comparison with experiment.
|Fabien HALTER||Professor – ICARE, Université d'Orléans||Referee|
|Pierre BOIVIN||Research Director|
M2P2, Université de Marseille
|Michael PFITZNER||Professor – Institut für Thermodynamik, Universität der Bundeswehr München||Member|
|Franck NICOUD||Professor – IMAG, Université de Montpellier||Member|
|Thomas SCHMITT||Research Director|
EM2C, Université de Paris-Saclay
|Philippe GRENARD||Research Director|
|Miguel MARTIN-BENITO||Engineer – CNES||Invited member|
|Didier SAUCEREAU||Engineer – ArianeGroup||Invited member|
|Bénédicte CUENOT||Senior Researcher – CERFACS||Advisor|