Direct numerical simulations and data-driven analysis of ignition and combustion in realistic pre-chamber/ engine systems with NH3/H2 blend fuel

To facilitate carbon-free and low emission combustion in practical engine systems, this project investigates the transient ignition and turbulent combustion process in a realistic pre-chamber/engine geometry with NH3/H2 blend fuel. Direct numerical simulations (DNS) are carried out for this system with exascale computation on Supercomputers, generating a vast amount of high-fidelity data. Machine learning techniques are applied to accelerate the chemical kinetic computation in DNS. The realistic geometry is represented by the immersed boundary method. Data-driven analysis is done to investigate in detail the ignition characteristics and the multi-scale features of the turbulent flames. NOx emissions are also investigated. A better understanding of the practical pre-chamber/engine system using NH3/H2 fuel should be finally obtained, which would be useful for both fundamental academic research and practical applications.