Extensive experimental investigation and phenomenological modelling of a DI ultra-lean hydrogen light-duty engine: combustion analysis, NOx and unburned fuel emissions

The vehicular sector is focused on reducing engine emissions in response to national and international regulations, targeting CO2 reduction. Carbon-neutral fuels, particularly green hydrogen from renewable sources, offer a promising path forward. In this perspective, this study investigates a 2.5-L 4-cylinder spark-ignition engine directly supplied with hydrogen, designed for the light-duty transport sector. An extensive experimental campaign explores the engine behaviour under various engine speeds and torque levels, working in ultra-lean conditions (relative air/fuel ratio − λ − between 2.15 and 3.30). Experiments evidenced that leaner mixtures and higher speeds lead to longer combustion durations in both early and main stages. They underlined that the combustion centre can assume an optimal timing under medium/low load operations and must be delayed at high loads to avoid knocking cycles. NOx emissions are globally reduced and become relevant only in the conditions where λ is close to 2 (maximum torque curve). The production of unburned hydrogen (uH2) is slightly dependent on the operating conditions and becomes significant only when the combustion is less stable (minimum load). The experimental results are used for an extensive validation of the predictive phenomenology-based combustion model, considering flame propagation enhancements related to turbulence and thermo-diffusive instabilities. Additionally, refined emission models are employed to estimate levels of uH2 and NOx. Simulation results are in good agreement with the experimental global engine parameters and combustion indicators, with average errors below 2 % and 1.6 CADs, respectively. Similarly, the pressure cycles and the related burn rates are correctly reproduced, and the influence on the model predictivity of the thermo-diffusive flame instabilities is evidenced. Finally, the model demonstrates to accurately follow the variations in NOx emission over the engine operating domain, while uH2 levels are adequately captured by the simulations, with higher discrepancies in the operating points with larger combustion cyclic variability.