This research concerns the Large Eddy Simulation (LES) of turbulent combustion in
both the premixed and the non-premixed regime. Non-premixed hybrid bluff-body/swirl
flames are simulated by means of a steady flamelet model (Flamelet-LES) based on
detailed chemical kinetics. LES of lean premixed twin flames stabilised on a turbulent
opposed jet (TOJ) burner are carried out using an algebraic Flame Surface Density model
(FSD-LES) and a newly developed model based on Linear Eddy Mixing (LEM-LES).
Isothermal swirling flow at a medium Reynolds and swirl number is simulated first and
the LES model is shown to accurately predict the velocity statistics and the complex
flow field governed by vortex breakdown and two recirculation zones. The Flamelet-
LES model is subsequently used to simulate a low speed swirling methane flame and
the capability of the model to predict downstream recirculation, vortex breakdown and
central jet precession in the presence of heat release is demonstrated. The simulation of
two high speed hydrogen/methane swirl flames with the Flamelet-LES model shows that
some quantitative predictions of this challenging test case for combustion simulation can
be achieved, while the overall predictions are not satisfactory. The flamelet approach is
found sensitive to minor errors in the mixing field which strongly affect the simulation
results due to the highly non-linear mixture fraction/density relationship.
Non-reacting simulations of turbulent opposed jet flows at moderate Reynolds number
are performed and compared to experimental reference data. The inclusion of the flow
field inside the nozzles into the computational domain is shown to yield accurate predictions
of the velocity statistics between the nozzles. For these predictions the detailed
knowledge of the initial jet development region near the turbulence generating plates is
vital and provided by PIV measurements inside a glass nozzle. FSD-LES of the twin premixed
TOJ flames show that the velocity statistics, both along the burner axis and the
stagnation plane, can be predicted to high accuracy. However, the algebraic flame surface
density model employed in the present study requires the adjustment of a model parameter
and as a result, predictions of the turbulent burning velocity cannot be achieved. The
comparison of two different interpretations of the FSD model show that a formulation
using an additional diffusion term allows for a better resolution of the premixed flames
in LES than the original formulation without diffusion.
A complex LEM combustion model is first developed as a Stand-Alone approach to
simulate premixed combustion and subsequently coupled to LES. The LEM-LES model
requires a number of sub-models to represent the effects of sub-grid stirring, finite-rate
chemistry, sub-grid expansion, 3D convection (splicing) and flame propagation which
are described in detail. The LEM-LES model is – to the author’s knowledge – the
first attempt to simulate premixed flames with finite-rate chemistry in incompressible
turbulent flow. Preliminary results from the application of LEM-LES to the premixed
twin TOJ flames are reported and show a high sensitivity to the 3D convection model
and the requirement to improve the splicing procedure for premixed flames in anisotropic
turbulent flow. The difficulty to accurately resolve the turbulent flow field by LES while
simultaneously accommodating a premixed flame of finite thickness on the LEM sub-grid
is found to be a challenge for the LEM-LES of premixed TOJ flames.