Development of a Hybrid Tabulated-Virtual Chemistry Model for Pollutant Formation Prediction
Résumé
Abstract In the context of energy transition, numerical models that can accurately describe pollutant formation are needed for designing novel combustion chambers. However, due to the stiffness of the chemistry describing pollutants, modeling chemical kinetics is a challenging issue. Simplified models such as flamelet-based tabulated chemistry allow a good resolution of the heat release rate and the temperature profile but usually fail in predicting pollutants in practical burners. The virtual chemistry method, recently developed at the EM2C laboratory, is an alternative technique to the tabulated chemistry method. It consists of first building a learning database using detailed chemistry solutions. Then, properties of “virtual” species and reactions are optimized to target key variables of interest. In that way, two virtual schemes are obtained: a main scheme trained to retrieve the main properties of the flame such as temperature and the flame speed, and a satellite scheme trained to retrieve pollutant formation. Although the virtual chemistry method is substantially cheaper than detailed chemistry, further cost reduction can be achieved. To that end, an innovative hybrid method coupling flamelet-based tabulated chemistry for the flame structure and virtual satellite mechanism for pollutant prediction is presented. By tabulating the inputs of the satellite mechanism as a function of the mixture fraction and progress variable, the number of transported variables is reduced by a factor of two. By comparison with detailed chemistry, flamelet-based tabulation, and virtual chemistry, the hybrid model is validated through the simulations of 3 configurations: 1-D freely-propagating premixed flames, 2-D premixed slot burner, and 2-D flame-vortex interactions. Results show the good performances of this model.