The explicit filtering method for large eddy simulations of a turbulent premixed flame

Abstract

This paper reports large eddy simulations (LES) of a premixed turbulent round methane-air jet flame at an equivalence ratio of 0.8 and unburnt gas temperature of 800 K using the explicit filtering LES (EFLES) method. The nominal 1D premixed flame speed, sL=2.05ms−1 and thermal thickness, δL=300μm. EFLES is formally derived from the approximate deconvolution method for LES computations. The nominal turbulent Reynolds number Ret=34 and the turbulent Karlovitz number Ka=25 places the present flame in the thin reaction zones regime of turbulent premixed combustion. LES and DNS results are obtained by solving species transport equations using a reduced 13-species chemical mechanism. Chemical source terms in the LES are computed using resolved species and temperature fields in all simulations. Conditional means of net heat release rate predicted by the LES and DNS show excellent agreement. The LES results show a marginally thicker flame brush and a shorter mean flame height when compared to the DNS results. Good agreement between LES and DNS for mean quantities is seen upstream of the tip. Larger quantitative differences are seen at the tip due to the difference in flame height between LES and DNS cases, as may be expected. Even at these positions, qualitative features of the DNS results and peak mean values are recovered correctly. Agreement between conditional mean statistics from both LES and DNS is good. RMS fluctuation amplitudes show larger quantitative differences between LES and DNS when compared to mean statistics while still capturing the correct qualitative features of the DNS profiles. The radial profiles of unconditioned RMS fluctuations are broader by O(δL) in the LES results, with increasing downstream distance. The smaller range of resolved scales in the LES causes the resolved flow motions to be more energetic resulting in increased turbulent transport rates for mean and RMS quantities. This results in the observed broadening of flow statistics in the LES when compared to the DNS. With mesh refinement, differences between LES and DNS results for all statistics are found to diminish smoothly.