Robust multivariate polynomial modeling of Laminar Flame Speed of Hydrogen–Oxygen–Steam mixtures supporting simulation applications in sustainable combustion

This study presents the development and validation of mathematical models for reproducing the Laminar Flame Speed (LFS) of Hydrogen–Oxygen–Steam (i.e. ) mixtures over a wide range of operating conditions. A data-drive black-box approach, employing multivariate polynomials and model selection criteria, was used to identify two models of comparable complexity and accuracy. Specifically, the adjusted coefficient of determination ( R adj 2 ), Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were employed, although R adj 2 and AIC both selected the same model. The necessary LFS data for model identification and validation activities were generated through extensive simulations conducted using the commercial Advanced Combustion Toolset (ACT) in GT-Suite (Gamma Technologies). The Lutz kinetic mechanism was chosen over the GRI-Mech mechanism due to its superior agreement with experimental data, providing a robust foundation for simulating mixtures combustion. The resulting dataset encompassed a comprehensive three-dimensional parameter space, covering initial temperatures from 373 to 973 K, pressures ranging from 1 to 90 bar and steam molar fractions from 0 to 0.6, yielding 9009 data points. The developed models exhibited exceptional accuracy in predicting LFS of mixtures across the entire spectrum of operating conditions. Their simplicity and predictive capabilities make them particularly valuable for accelerating numerical simulations in energy applications involving oxy-combustion of hydrogen. The models developed in this work serve as robust and efficient tools for analyzing combustion processes , thereby promoting the study and development of sustainable, zero-carbon energy solutions. • Accurate and efficient LFS models for mixtures combustion. • LFS models reproduce detailed kinetic mechanism of mixtures combustion. • Comprehensive LFS dataset for robust model training. • LFS models have high accuracy and low computational cost. • LFS models are valuable tools for hydrogen oxy-combustion simulations.