A virtual chemistry model for soot prediction in flames including radiative heat transfer

Modeling soot formation is a very difficult task which has been assessed by the numerical combustion community. Numerical simulations of turbulent sooting flames remain very challenging due to the com-plexity of soot formation phenomena, which implies development of elaborate soot models often too computationally demanding, or simplified soot models that are limited to a small range of operating con-ditions of interest. Here an innovative optimized global approach called virtual chemistry is proposed. It consists of a mathematical formalism including virtual species and virtual reactions, whose thermochem-ical parameters are optimized in order to describe a combustion system. Thermochemical properties are trained through a genetic algorithm, employing a learning database made of reference flame elements. This makes it possible to reproduce multiple combustion regimes and operating conditions. The virtual chemistry approach reproduces then the structure of hydrocarbon-air flames, as well as predicting spe-cific user-defined pollutant species. The objective of this article is to extend the virtual chemistry method-ology for soot formation prediction. As the radiative heat losses are important in sooting flames, a new radiative virtual model is also developed to account for them. The final reduced chemistry consists of 12 virtual species and 6 virtual reactions considerably decreasing the computational time compared to de-tailed chemistry models. Simulations of 1-D and 2-D laminar ethylene-air sooting flames are performed to evaluate the virtual soot and radiative models. Temperature, soot volume fraction and radiative heat losses are well predicted and are in good agreement with the reference data. (c) 2021 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Automated summary

Modeling soot formation is a challenging task, and the innovative virtual chemistry approach provides a new optimized global method for simulating turbulent sooting flames. This approach optimizes thermochemical parameters to describe combustion systems, trains thermochemical properties using genetic algorithms, reduces computational costs, and effectively predicts combustion system characteristics.