Development of a LES method for characterising the impact of mixture preparation on combustion stability in H2 SI engines

Europe has outlined its plan to pursue its ecological transition, known as the ``Green Deal'', through the implementation of a series of political initiatives. Different road transport technologies support this transition, such as electric propulsion powered by batteries and fuel cells, or internal combustion engines running on hydrogen. The DELHYCE industrial and academic partnership (``DEsign of Low emission and efficient HYdrogen internal Combustion Engines''), led by the PRISME laboratory, Stellantis and Renault Trucks, aims to develop a scientific methodology for the design of internal combustion engines that use hydrogen. This methodology involves smalldisplacement Diesel engines for commercial vehicles and larger-displacement engines for trucks. The goal is to optimise the efficiency of internal combustion engines while minimising NOx emissions and controlling the abnormal combustion characteristics that are often found in these engines.

Current hydrogen (H2) internal combustion engines (ICEs) are commonly designed based on existing Diesel models, due to their robustness and their ability to cope with the resulting  high in-cylinder peak pressures. An onboard storage of gaseous H2 at 700 bar is typical for such applications. Still, high injection pressures (greater than 100 bar) would compromise the use of the maximum H2 mass available in the tank, or require an H2 compression stage prior to injection. Therefore, low- or medium-injection pressures up to around 40 bar are mostly preferred, either with a port fuel injection (PFI) or a direct injection (DI) approach. Combustion is initiated by a spark. The PFI strategy usually limits engine power output, since the presence of low density H2 during compression translates into volumetric efficiency penalties, and can also cause backfire events. DI strategies allow optimising the volumetric efficiency and thus power output, by injecting H2 into the combustion chamber at later timings. They also allow avoiding backfire events. The limited momentum of the gaseous jet (as compared to the liquid sprays found in gasoline engines), combined with comparatively short mixing durations can lead to unwanted inhomogeneities and thus locally rich and lean zones, while ideally a homogeneous mixture is expected. Rich zones may favour the appearance of abnormal combustion phenomena, and lead to NOx production. Extremely locally lean zones can result in reduced combustion speeds, or even in misfires, especially if they appear around the spark plug at spark timing. Counterbalancing such negative effects can require measures such as increased intake aerodynamics, exhaust gas recirculation (EGR) and air dilution strategies, or port water injection. However, this can result in additional complexity and lead to a reduced combustion efficiency. In particular, running ultra-lean mixtures to avoid generating locally rich zones can lead to increased levels of cyclic variability that can reduce the range of operation of the engine.

From a computational perspective, DI of gaseous H2 and its combustion cause a number of specific challenges. Injection pressures around 40 bar in general lead to under-expanded jets. The latter are characterised by a complex system of shocks appearing near the injector nozzle, which are  related to very small space and time scales. This makes their numerical resolution complex or even impossible, due to the excessive computing times it would require to resolve all scales.  Reynolds-Averaged Navier-Stokes (RANS) simulations of such jets reveal that their predictions depend on the employed numerical schemes, the grid resolution, the turbulence models, the equation of state and the modelling of turbulent mixing. Similarly, RANS simulations of the internal flow within hydrogen ICEs often overestimate the mixture stratification, depending on the turbulence model used. An alternative to RANS is Large-Eddy Simulation (LES), where part of the instantaneous turbulent spectrum is resolved. LES of gasoline ICEs has shown to yield accurate predictions of mixture preparation, cycle-to-cycle variability (CCV) and abnormal combustion. However, few published studies have applied LES to DI H2 ICEs, and for those, no experimental validation has been found in the literature so far.

In this context, the main objective of the present PhD thesis is to develop a LES methodology, based on the CONVERGE CFD CODE, that is able to predict the mixture formation inside DI-SI ICE running on gaseous H2, to validate its predictions against experimental findings in High-Pressure High-Temperature (HPHT) cells and single-cylinder engines, and to apply it to the study the impact of mixture inhomogeneities on the combustion variability under overall lean conditions. In a final step of the PhD, the developed LES method could also be used to quantify the importance of mixture inhomogeneities in the appearance of abnormal combustion, using state-of-the-art LES combustion model for H2 developed at IFPEN in other PhDs.

Mots-clés : Moteurs thermiques, hydrogène, jets sous-détendus, LES, combustion