Optimization of Kinetic Mechanism for Ethylene Combustion Based on Machine Learning

To address the dual challenges of high computational costs in detailed ethylene kinetic mechanisms and insufficient prediction accuracy in minimized ethylene kinetic mechanism, a two-stage machine learning model integrating genetic algorithm (GA) and radial basis function (RBF) interpolation was developed. The kinetic parameters in the minimized ethylene kinetic mechanism are optimized by the developed model using experimental data of ignition delay times (IDT) and laminar flame speeds (LFS). The two-stage model combines global parameter exploration with local refinement, balancing efficiency and accuracy. Results demonstrate that the optimized mechanism significantly reduces prediction error for both ignition delays and flame speeds, while improving accuracy in key species concentration distribution. This study validates the two-stage optimization method for kinetic mechanism refinement, providing a high-precision ethylene mechanism for scramjet numerical simulation.