Efficient modeling of the filtered density function in turbulent sprays using ensemble learning

Summary: Machine learning techniques have been successfully applied to predict the conditional scalar dissipation rate in turbulent spray flames, showing excellent agreement with actual data. The artificial neural networks not only improve the closure models but also help identify important features affecting local dissipation rates.

Summary: This study proposes a data-assisted approach utilizing random forest classifiers for local and dynamic submodel assignment in turbulent-combustion simulations, demonstrating promising results in predicting different quantities of interest and improving simulation accuracy. The method shows reasonable classification performance in assigning different combustion models even when targeting multiple quantities of interest, and it outperforms traditional calculations in predicting temperature and CO mass fraction in a posteriori assessments.