Numerical Modeling of Aeroengine Hydrogen Combustion Systems

Kurzfassung

Hydrogen is a promising carbon-free energy carrier for aviation. Its high reactivity, wide flammability limits, and pronounced differential diffusion, however, challenge the chemistry, turbulence-chemistry interaction (TCI), and molecular-transport closures developed for hydrocarbon flames. Their relative importance in a realistic Rich-Quench-Lean (RQL) combustor, and the errors from common industrial simplifications, remain insufficiently quantified. This thesis assesses how the chemistry representation, the TCI closure, the flame-thickening strategy, and the molecular-transport description affect the predicted flame structure of a hydrogen-fueled swirl-stabilized RQL combustor. Large-eddy simulations are carried out with the DLR in-house solver TRACE and validated against the OH* chemiluminescence fields and the injector pressure drop. Five model combinations are constructed so that consecutive cases differ in a single aspect: tabulated Flamelet-Generated Manifold (FGM) versus Finite-Rate Chemistry (FRC); a presumed beta-PDF closure, a dynamic thickened flame, and a hybrid regime-selective thickened flame; and unity-Lewis-number versus mixture-averaged diffusion. The aerodynamic field, the pressure drop, and the lifted flame are reproduced almost identically across all cases, so the differences in flame structure can be attributed to the modelling choices alone. The presumed beta-PDF closure averages very tabulated quantity over the subgrid variance, producing a lower OH* signal. The tabulated manifold agrees with finite-rate chemistry in the partially premixed primary zone but overpredicts OH* at the diffusion-controlled secondary air interfaces, where the premixed manifold assumption fails, as confirmed by the Takeno flame index. The hybrid thickened flame changes the prediction only marginally. Relaxing the unity-Lewis assumption reveals a differential-diffusion effect confined to the hot downstream region, where it decouples the species, temperature, and enthalpy from the mixture fraction at about three percent additional cost. The simplest closure agrees most closely with the line-of-sight OH* field, but this reflects normalization masking its lower signal rather than greater fidelity. The more reliable planar comparison favours finite-rate chemistry, so the choice rests on physical completeness and cost. On that basis, FRC with mixture-averaged transport is the most faithful combination examined, while the tabulated description remains efficient and adequate when the analysis is restricted to the partially premixed primary zone.