Machine Learning-Driven Chemical Reactor Network Modeling of the Sandia-D Flame
Researchers have developed a machine learning-assisted framework to create more efficient chemical reactor network (ERN) models for turbulent combustion simulations. This approach uses principal component analysis and k-means clustering on computational fluid dynamics data to identify flame regions, which then initialize a reactor-network graph. This initialization is further refined using gradient descent with Cantera simulations, achieving a significant speedup over traditional solvers while maintaining reasonable accuracy for maximum temperature predictions. AI