Sparse-Lagrangian MMC modelling of the Sandia ethylene sooting flame

Abstract

Reliable soot predictions in turbulent flames remain challenging due to the need to use relatively large chemical mechanisms and the presence of slow kinetics of the soot species that necessitate the use of an advanced combustion sub-model but prevent tabulation for all species in particular for soot precursor species such as PAHs. The joint probability density function (PDF) approach offers a “model-free” closure for the chemical source term but its computational expense typically hinders the incorporation of detailed soot mechanisms. In this study, a sparse particle method called ‘Multiple Mapping Conditioning’ (MMC) is used. The number of stochastic particles can be reduced by almost two orders of magnitude and large-eddy simulations of a turbulent ethylene flame with a detailed sectional soot model become feasible. Predicted concentrations of gaseous species and temperature agree well with experimental data and indicate an accurate modelling of the turbulent mixing process and gas phase reactions by MMC-LES. MMC-LES with the detailed sectional soot model and a second MMC-LES using a two-equation model provide accurate predictions of the zones where soot is formed and capture the onset of oxidation very well. Simulated peak values of soot volume fraction differ depending on the model, with values being around three times too large for the two-equation model while the detailed sectional model gives very decent agreement everywhere except in the very rich region along the centreline where soot volume fraction are overpredicted by up to 50%. The sectional model yields reasonable results for the aggregate size distributions everywhere in the flame and also the primary particle sizes predicted by the two-equation model agree with expected values, but a quantitative assessment is difficult as corresponding measurements are not available.