Skeletal Reaction Models for Gasoline Surrogate Combustion

Skeletal reaction models are derived for a gasoline surrogate model using a local-sensitivity-analysis based technique. In this technique, the sensitivities of species mass fractions and temperature with respect to reaction rates are estimated using the sparse forced-optimally time dependent (sf-OTD) method. In sf-OTD, the sensitivity matrix is projected onto a low-rank dynamic basis and integrated over time within this low-rank subspace. The sf-OTD is augmented with a sparse-sampling technique termed the discrete empirical interpolation method (DEIM). This method selects dynamically relevant species and reactions at each time step to update their associated sensitivities, thereby enhancing the computational efficiency and the robustness of the sf-OTD. The accuracy and the computational cost of the sf-OTD method are assessed by comparisons against the solution of the exact sensitivity equations. A series of reduced models are developed based on the Lawrence Livermore National Lab's gasoline surrogate model with 1389 species and 9603 reactions. Their performances are compared against the detailed model's predictions of ignition delay and flame speed. The results show that reduced models with 500 and more species are able to reproduce the detailed mechanism's prediction within 15% relative errors for a large range of pressures, temperatures, and equivalence ratios.