Abstract
Reducing the computational cost of chemical kinetics is essential to implement detailed reaction mechanisms into realistic numerical simulations. The present study introduces an artificial neural network (ANN) that can predict the chemical source terms of each species for the given species mass fractions and temperature, replacing the conventional chemical terms based on Arrhenius rate equations. The ANN was trained using numerical solutions of opposed-flow flames that can cover a wide range of combustion problems. The OPPDIF code and a detailed reaction mechanism for hydrogen and air with 9 species and 19 reactions were used to generate a training dataset comprised of species mass fractions, temperature, chemical source terms. A physics-guided loss function that considers mass conservation of elemental species was employed. Using the trained ANN, a modified OPPDIF, named OPPDIF-ANN, was prepared by replacing the CKWYP with CKWYP-ANN evaluating the chemical sources via the trained ANN. For multiple global strain rate conditions, the solutions using ANN-based source terms were proven to be identical to those using Arrhenius source terms.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
