Machine learning model to predict the laminar burning velocities of H2/CO/CH4/CO2/N2/air mixtures at high pressure and temperature conditions

Abstract An empirical model based on machine learning is developed for predicting the variation of the laminar burning velocities of H2/CO/CH4/CO2/N2/air mixtures with volumetric fractions as the independent variables at different elevated mixture temperatures and pressures. The proposed model is derived partly based on the measured burning velocities of syngas-air mixtures at elevated temperatures and pressures using diverging channel method, and partly established from the predictions using the FFCM-1 detailed kinetic model. The experiments at elevated pressures and temperature strongly agree with the predictions of the FFCM-1 kinetic model for PG1 (H2/CO/CO2/N2 = 15/15/15/55) syngas composition. Based on the detailed analysis of the experimental results, a power-law correlation considering the α, β variations is proposed: Su = Su,o * (Tu/Tu,o)α0+α1 (1−Pu/Pu,o) * (Pu/Pu,o)β0+β1 (1−Tu/Tu,o). Machine learning model (multiple linear-regression) was trained for the variables (Su,o, αo, α1, β0, β1) in the power-law correlation to enable the prediction of laminar burning velocity at various pressure and temperature conditions. The empirical model was developed with mole fractions of various components (H2/CO/CH4/CO2/N2) in the syngas composition and equivalence ratio as independent variables. The developed model was intended for low-calorific value syngas mixtures, and it performs exceedingly well without solving detailed governing equations, detailed chemistry, and transport equations. The proposed model is accurate for a wide range of syngas-air mixtures reported in the literature. A detailed comparison showed that the empirical model accurately predicts the laminar burning velocity with error X H 2 0.70 , 0.25 X C H 4 X C O 2 X N 2