High-Fidelity 3D-CFD Simulation of Hydrogen Injection: Different Modelling Approaches Balancing Accuracy and Computational Efficiency

The computational fluid dynamics (CFD) modeling of the direct injection of hydrogen (H2) through outward-opening injectors currently represents the bottleneck for accelerating the numerically supported H2 internal combustion engines (H2-ICEs) development. Indeed, the high injection velocities, the relatively small flow cross-sectional area and shock structures formation close to the injector (which are typical of under-expanded jets) require very fine spatial and temporal grid resolutions, when compared with those required in the downstream mixing region. Therefore, to improve the trade-off between accuracy and computational cost, the exploration of alternative approaches to model the injector region, capable of reducing runtime while maintaining comparable accuracy, is of paramount importance. Within this framework, the present work proposes a benchmark of different hydrogen injection modeling approaches characterized by decreasing levels of complexity and computational requirement. Large eddy simulations (LES) are first performed and used as reference solutions. Subsequently, detailed Eulerian Reynolds-Averaged Navier–Stokes (RANS) simulations are considered, followed by the application of a novel Lagrangian particle tracking-based methodology for hydrogen injection modeling. The results show that, although both the Eulerian RANS and the Lagrangian-based simulations underestimate the mixing process, the general jet evolution, especially under pseudo-stationary conditions, is reasonably captured. In particular, the Lagrangian approach provides a reduction in computational cost by maintaining reasonable accuracy in downstream mixing prediction, through appropriate calibration of specific model parameters, beyond those of the turbulence model. Its main limitation lies in the inherent inability to accurately resolve near-nozzle flow phenomena, which, however, may be of secondary importance for engine applications compared to the downstream mixing. Overall, this study establishes a consistent reference framework for the comparison of hydrogen injection modeling strategies across different fidelity levels, enabling their application to complex engine-relevant simulations with reduced computational efforts.