Hydrogen direct-injection engines offer a promising pathway for decarbonizing heavy-duty transportation, but accurate prediction of mixture formation and NO \(_x\) emissions remains challenging due to complex injector dynamics and strong cycle-to-cycle variability. This work presents a comprehensive computational and experimental investigation of supersonic H \(_2\) direct injection, mixing, combustion, and NO formation in a single-cylinder heavy-duty hydrogen engine operated at 1100 rpm and \(\lambda \) = 2.6. A detailed three-dimensional CFD model is developed, coupling a pressure-based injection boundary condition with a realistic Bosch F2 prototype injector needle-lift profile to capture valve-bounce effects. The model is validated against measured in-cylinder pressure, fuel and air mass, and NO emission data. Multi-cycle combustion behavior and NO emission variability are analyzed using the concurrent perturbation method (CPM), with 20 statistically independent realizations at reduced computational cost. Results show that near-spark mixtures with higher fuel concentration accelerate flame propagation and increase peak NO by a factor of two (76 ppm vs. 32 ppm). Simulations reveal that NO forms predominantly in local pockets of high fuel concentration, with turbulent flame speeds of 11–22 m/s during the early combustion phase. Predicted exhaust-port NO levels agree qualitatively with experiments, though unsteady RANS tends to overpredict NO due to limited small-scale mixing. The study demonstrates that resolving the injector flow rather than approximating boundary conditions, combined with CPM can effectively capture hydrogen combustion dynamics and emission variability.
