Evaluation of Presumed Probability-Density-Function Models in Non-Premixed Flames by using Large Eddy Simulation

Four kinds of presumed probability-density-function (PDF) models for non-premixed turbulent combustion are evaluated in flames with various stoichiometric mixture fractions by using large eddy simulation (LES). The LES code is validated by the experimental data of a classical turbulent jet flame (Sandia flame D). The mean and rms temperatures obtained by the presumed PDF models are compared with the LES results. The β−function model achieves a good prediction for different flames. The predicted rms temperature by using the double-δ function model is very small and unphysical in the vicinity of the maximum mean temperature. The clip−Gaussian model and the multi-δ function model make a worse prediction of the extremely fuel-rich or fuel-lean side due to the clip at the boundary of the mixture fraction space. The results also show that the overall prediction performance of presumed PDF models is better at mediate stoichiometric mixture fractions than that at very small or very large ones.