The purpose of this project is to show and highlight the effects and challenges of scale-resolving turbulence models on the simulation of the 1.5-stage LSTR. The Reynolds number in the LSTR is equal to a real high-pressure turbine so that nearly all flow phenomena as well as computational efforts are identical. The main purpose of this project is the derivation of a highly accurate numerical setup and its validation against experimental data. However, due to problems regaarding the test rig and the strong dependency on its data, which is usually used as a boundary condition for the CFD, not all simulations could be conducted. Therefore, only a first comparison could be drawn between the standard RANS simulation, the more sophisticated URANS simulation and the most high-fidelity SBES simulation. The LSTR setup as well as the first numerical results can be seen in Fig. 1 as a 0D comparison of selected performance parameters as well as radial profiles of the flow quantities in the rotor inlet (ME02) and outlet (ME03) plane. As the flow enters the turbine, it interacts with the hub (RIDN) and casing (RODN) side platform film cooling ejections of the first Nozzle Guide Vane (NGV1). Within the NGV1 passage the flow is accelerated and turned in circumferential direction to provide angular momentum to the flow that can be extracted by the rotor. In ME02 (Fig. 1, top right) the first deviations can be seen between the RANS and the URANS simulations, that both use the k-ω-SST turbulence model. The only difference between these approaches is the treatment of time. While the time is neglected in the RANS (steady-state), it is fully resolved in the URANS, and therefore effects of stator-rotor-interaction. These seem to affect the flow mostly in the lower 20 % of the channel and seem to cause a higher pressure loss as well as higher turning and therefore a reduced mass flow. Higher deviations occur between the URANS and the SBES simulation. Compared to the RANS approaches, the SBES and its LES formulation in the free stream appear to predict a different mixing of the main with the cooling flow of the trailing edge (TE) slot ejection, which can be seen in the more wavy appearance of the flow quantities in the region from 20 % to 85 % channel height.

In ME03 (Fig. 1, bottom right) the deviations between the different approaches become far more pronounced. This can mainly be drawn back to the fact that the simulation of a rotor in turbomachinery is subject to high numerical uncertainties due to the assumptions made in steady-state RANS simulations. Therefore, high deviations can be seen between RANS and URANS, mainly in the mass flow redistribution μ₂₁ in the lower 40 % of the channel. This shows that in order to be accurate in the simulation of a high-pressure turbine, the higher computation costs of a URANS seem to be justified. The deviations between URANS and SBES are highest in the rotor tip region above 70 % channel height. Due to the radial gap between rotor tip and the casing, a vortex rolls up, that causes high losses and blockage. The region of this Tip Leakage Vortex usually shows the highest amounts of turbulent viscosity μ_t and therefore a portion of modelling in the flow. Due to the LES formulation in the SBES this modelling can be reduced by a significant amount.