It is demonstrated that in the (near) zero-gravity experiments conducted by Tang et al. (Combust. Flame; 2009, 2011) iron powder aerosols created using the finest powders are optically thick, implying that radiative heat transfer between particles should not be neglected. To test this concept, an iron particle oxidation model has been implemented in OpenFOAM, including a coupling with the P1-model for radiative heat transfer. For flame simulations in which radiation is not included, obtained flame propagation velocities deviate less than 8% with results obtained using Chem1D-Fe and also show a good correspondance with algebraic models for optically thin aerosols. No significant difference in predicted flame propagation velocity is observed between 1D and 3D simulations: contrary to what is seen in gaseous flames, including the curvature of the flame does not increase predicted flame speeds substantially. However, measured flame propagation velocity values exceed numerically obtained predictions excluding thermal radiation by a factor of three to four. To the authors’ knowledge, this discrepancy is exemplary for the difference between experimentally obtained values for flame propagation velocities, and predictions made using numerical simulation tools neglecting radiative heat transfer. Accounting for radiation increases predicted flame propagation velocities, in the absence of confining boundaries, by approximately a factor of 10 which is in line with algebraic models for optically thick aerosols. In 3D simulations for the two finest iron powders in the experiments, including radiation and accounting for the presence of the confining tube wall results in an error of 11% and 35% with respect to measured flame propagation velocities, significantly smaller than predictions obtained excluding thermal radiation. Although these flames are not purely radiation-driven, inclusion of particle-to-particle radiative heat transfer enhances flame propagation velocities in simulations to values that correspond much better with experimental values than if radiation would not be taken into account.

Fuel efficiency improvement and harmful emission reduction are the paramount driving forces for development of gas turbine combustors. Lean-burn combustors can accomplish these goals, but require specific flow topologies to overcome their sensitivity to combustion instabilities. Large Eddy Simulations (LES) can accurately capture these complex and intrinsically unsteady flow fields, but estimating the appropriate numerical resolution and subgrid model(s) still remain challenges. This paper discusses the prediction of non-reacting flow fields in the DLR gas turbine model combustor using LES. Several important features of modern gas turbine combustors are present in this model combustor: multiple air swirlers and recirculation zones for flame stabilisation. Good overall agreement is obtained between LES outcomes and experimental results, both in terms of time-averaged and temporal RMS values. Findings of this study include a strong dependence of the opening angle of the swirling jet inside the combustion chamber on the subgrid viscosity, which acts mainly through the air mass flow split between the two swirlers in the DLR model combustor. This paper illustrates the ability of LES to obtain accurate flow field predictions in complex gas turbine combustors making use of open-source software and computational resources available to industry.