Assessment of combustion models of a self-regenerative flameless oxidation burner

Diluted combustion, also called Flameless Oxidation, has been a major interest in the advanced combustion technologies since it satisfies apparently opposed objectives. Indeed, high dilution reduces temperature gradients and allows to preheat combustion air without rising the NOx emission. This technique also modifies flame characteristics as well as heat transfer inside the furnace.
The present study is oriented on the modelling of this particular combustion mode and is applied on a pilot-scale furnace, provided with a 200kW natural gas FLOX REGEMAT® 350 burner. The objective is to reproduce the time average behaviour of the system thanks to the numerical models available with the commercial CFD code FLUENT 6. The validation is performed by comparison with experimental data resulting from test carried out previously, for furnace temperature varying from 1100 to 1300°C.
The burner and the pilot-scale furnace have been modelled with a fine hexahedral cells grid. The k-? model is chosen for the turbulence and the discrete ordinates model is applied for the radiation. The thermal and prompt NOx are predicted in a post-processing computation.
Combustion models are assessed to compare their ability to predict the heat transfer and the location of the reaction zone inside the furnace.
The 'PDF-Equilibrium' model resolves a transport equation for the mixture fraction with a beta-shaped probability density function to integrate the turbulence/chemistry interaction.
The 'Eddy-Dissipation/Finite Rate' and 'Eddy-Dissipation' models solve transport equations for each species involved in a 1 or 2-step reaction mechanism, and differ by the way they compute average reaction rates (by taking into account chemical kinetics or not).
All combustion models give good predictions outside the reaction zone, but there is an overestimation of temperature in the near burner zone, especially with standard 'diffusion controlled' combustion models (PDF-equilibrium and Eddy-Dissipation). The Eddy-Dissipation/Finite Rate model improves prediction because it takes chemical kinetics into account. However, the smooth increase of temperature and the location of the reaction zone do not agree with experimental observations.
Computed NOx values well reproduce measurements although corresponding temperature fields do not seem totally correct.
Further work will focus on the implementation of detailed reaction mechanisms with the 'PDF-Flamelet' and the 'Eddy-Dissipation Concept' models.