Flame Surface Density and Artificially Thickened Flame Combustion Models Applied to a Turbulent Partially-Premixed Flame

Abstract

The Sydney Flame provides an excellent framework for systematic analysis of premixed, partially premixed and non-premixed methane/air combustion. It represents (depending on the axial position from the fuel inlet) non-premixed, partially premixed and fully premixed combustion. We simulate the test case FJ200-5GP-Lr75-57, which represents the partially premixed mode. Large-eddy simulations (LES) are conducted using a flame surface density combustion model (FSD) using one-step chemistry and an artificially thickened flame model (ATF) with a global two-step and an analytically reduced multi-step chemistry reaction mechanism. Unity Lewis numbers are assumed (Le=1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$Le=1$$\\end{document}) at first. The FSD and ATF models are compared using the same computational mesh, numerical scheme and boundary conditions. Both models will be analysed regarding their accuracy and computational efficiency in the regime of partially premixed combustion. A comparison of the results points out the strengths and weaknesses of the FSD models. The FSD model with inclusion of flame stretch effects yields good agreement with mean experimental temperature, CO2 and H2O mass fraction distributions in contrast to the ATF model using the global two-step mechanism, which overestimates downstream temperature, CO2 and H2O mass fractions. The FSD model performs well in all regions, which are dominated first by premixed, then partially premixed and finally non-premixed combustion along the flame. The ATF multi-step chemistry shows good results only in the premixed mode region, while mean temperature, CO2 and mass fractions are overestimated in the non-premixed mode regions at higher mixture fraction values. Including differential diffusion into the transport equations improved the ATF model results in comparison with experiments. A mesh study revealed, that the ATF model performs only after mesh refinement. In particular results are improved in the non-premixed combustion region. In contrast, the FSD model is less resolution sensitive and performs very well for both meshes. An evaluation of the Wasserstein metric provides a quantitative assessment of different ATF model setups on simulation accuracy. Finally, computational times for all simulation setups are compared.