Improved Methods of Laminar Flamelet Model for Compressible Flow

Abstract

To improve the computational performance of the laminar flamelet model for compressible flow (the compressible flamelet model), two formulations of the heat flux term are proposed: form1 achieves efficient calculation of the spatial gradient of the mass fraction of each chemical species, and form2 eliminates the dependence of the calculation process on the number of chemical species. Based on these formulations, three methods are proposed that use linear interpolation (lerp) or an artificial neural network (ANN) for their flamelet tables: form1-ann, form2-lerp, and form2-ann. First, it will be shown that the accuracy of the ANN in the proposed form1-ann and form2-ann methods is sufficient for numerical simulations. Then, to evaluate the form2-lerp and form2-ann methods and show that they can greatly improve the computational performance of the conventional method, numerical simulations will be conducted for the scramjet test-engine combustor of the German Aerospace Center, DLR. The calculation time of the form2-lerp method is reduced by about 0.878 times, and the memory usage is increased by about 2.97 times compared with the values for the conventional method. The calculation time of the form2-ann method is reduced by about 0.946 times, and the memory usage is reduced by about 0.508 times compared with the conventional method.