Physics - Informed Machine Learning for Reduced Space Chemical Kinetics

Abstract

Modeling detailed chemical kinetics is a primary challenge in combustion simulations. Recent machine learning (ML) approaches aim to accelerate chemical kinetics integration, though often limited to simpler reaction mechanisms. This study presents a framework to enforce physical constraints, such as total mass and elemental conservation, into ML model training for chemical kinetics of complex mechanisms in reduced space. Our method leverages a strong correlation between full and reduced solution vectors, using a small neural network for accurate and physically consistent mapping. This enforced constraint in ML training for reduced-space kinetics is demonstrated with CH4 oxidation, where our Deep Operator Networks-based solutions are accurate and physically compliant