Large eddy simulation/probability density function modeling of a turbulent jet flame

In this work, we develop the large-eddy simulation (LES)/probability density function (PDF) simulation capability for turbulent combustion and apply it to a turbulent CH4=H2=N2 jet flame (DLR Flame A). The PDF code is verified to be second-order accurate with respect to the time-step size and the grid size in a manufactured one-dimensional test case. Three grids ð64 64 16; 192 192 48; 320 320 80Þ are used in the simulations of DLR Flame A to examine the effect of the grid resolution. The numerical solutions of the resolved mixture fraction, the mixture fraction squared, and the density are duplicated in the LES code and the PDF code to explore the numerical consistency between them. A single laminar flamelet profile is used to reduce the computational cost of treating the chemical reactions of the particles. The sensitivity of the LES results to the time-step size is explored. Both first and second-order time splitting schemes are used for integrating the stochastic differential equations for the particles, and these are compared in the jet flame simulations. The numerical results are found to be sensitive to the grid resolution, and the 192 192 48 grid is adequate to capture the main flow fields of interest for this study. The numerical consistency between LES and PDF is confirmed by the small difference between their numerical predictions. Overall good agreement between the LES/PDF predictions and the experimental data is observed for the resolved flow fields and the composition fields, including for the mass fractions of the minor species and NO. The LES results are found to be insensitive to the time-step size for this particular flame. The first-order splitting scheme performs as well as the second-order splitting scheme in predicting the resolved mean and rms mixture fraction and the density for this flame. 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.