On the influence of the H2 flame thermo-diffusive instability at engine-like conditions: A CFD-driven study

• G-equation combustion model reliably predicts ultra-lean combustion in H 2 ICEs. • Turbulence flame wrinkling can overcome the thermo-diffusive instability effect. • Wall flame quenching can compensate the instability-induced acceleration. • At low IMEP and Φ, scarce influence of thermo-diffusive instability is observed. The effect of thermo-diffusive (TD) instabilities on combustion of lean and ultra-lean mixtures is one of the most debated aspects in the field of hydrogen internal combustion engines (ICEs). This paper presents a numerical setup for 3D-CFD in-cylinder simulations of H2 ICEs and, by adopting it, the effect of the TD instability is assessed. An initial setup relying on G-equation to represent the flame propagation is presented, with a Damköhler-like correlation for the turbulent flame speed and Verhelst polynomial correlation for the laminar one. Then the simulation setup is improved by three key modifications. In order, a correction factor to account for TD instability, the turbulence-instability interaction and a wall quenching model are progressively added to obtain the final setup. The results at each stage of the framework development are compared with experiments at six operating points, and the discrepancies are analyzed. All the operating points are characterized by lean/ultra-lean mixtures, with equivalence ratios ranging from 0.30 to 0.55. The investigated powertrain is a port-fuel injection (PFI) spark-ignition hydrogen research ICE. Port-fuel injection and resulting homogeneous mixture eliminate the uncertainties related to the stratification and allow the research activity to be completely focused on combustion dynamics. Firstly, the findings highlight that the effect of the TD instability is a slight modification of in-cylinder pressure and combustion indicators at the examined engine conditions. Secondly, the addition of both turbulence-instability interaction and wall quenching of the flame has a similar (and opposite) effect compared to the TD instability one. Therefore, comparing the results provided by the initial and final numerical frameworks, they are similar and both close to the experiments. Despite the fortuitous compensation at the presently investigated conditions, in order to perform accurate simulations, it is necessary to account for both TD instability and wall quenching phenomena. This is even more true considering that, at different conditions, the instability effect may differ and be much more significant.