Virginia Tech Compressor Cascade, comparison of predicted tip gap vortex with experimental data.

All turbulence models are able to reproduce the measured blade loading within the experimental scatter.

The prediction of the flow in the tip gap is improved if Reynolds stress models are used. Especially the mean velocity and the velocity fluctuations show better agreement with experimental data. Nevertheless it is obvious that close to walls, the prediction of secondary velocity as well as Reynolds stresses can still be further improved.

The developing tip gap vortex is well predicted by the Reynolds stress models. This holds true for both mean velocity and Reynolds stresses

The flow through the low-speed subsonic compressor cascade is a demanding test case for turbulence models. The cascade has been extensively investigated by Muthanna and Tang during their respective PhD times at the Virginia Polytechnic Institute and State University in Blacksburg, Virginia. It is built of GE Rotor B section blades with a chord length and blade height of 254mm staggered at an angle of 56.9°. The focus of the investigation is on the flow through the tip gap of 1.65% blade height and the resulting tip gap vortex. The Mach number is 0.073 and the Reynolds number based on the chord length amounts to 400,000.

We simulated the flow using statistical turbulence models of increasing complexity. These are the Menter SST k-ω linear eddy viscosity model, the Hellsten Explicit Algebraic Reynolds Stress Model (Hellsten EARSM) and the SSG/LRR-ω Differential Reynolds Stress Model of Eisfeld. The latter two models account for all the Reynolds stress tensor components by algebraic relations or by solution of partial differential equations. It could be shown that Reynolds stress models lead to improved prediction of the tip gap flow and the passage vortex. More details on the models, the computational method and the results can be found in the proceedings of the THMT 2012 conference.

Morsbach, C.; Franke, M. & di Mare, F.: Towards the application of Reynolds stress transport models to 3D turbomachinery flows, 7th International Symposium on Turbulence, Heat and Mass Transfer, 2012

T106-A Turbine Cascade

Modern low-pressure turbines operate at relatively low Reynolds number values. Hence, their efficiency depends greatly on transition. Unfortunately, the flow inside a real component is very complex due to unsteady and three-dimensional phenomenon. For a better understanding of the different effects, simplified flows must be investigated. In particular, flows through cascades allow to focus on the design’s performance of a blade. For a CFD code, it is then very important to be able to simulate accurately such flows.

The flow through the T106-A turbine cascade is a good example of such a study. The cascade has been designed by MTU-Aero Engines and tested at DLR. Of particular interest is the evolution of the flow when the Reynolds number changes.

(a)

(b)

(c)

Fig.1 : Pressure coefficient distribution over the T106-A cascade.

In order to demonstrate the capability of TRACE to simulate flows over low-pressure turbine profiles, three types of computation have been carried out. The two-equation k-ω turbulence model of Wilcox has been used to simulate fully-turbulent flows. Those results are compared to the DLR’s in-house Multimode transition model with the γ-Reθ Model of Menter and the experiments in Fig 1.

At low Reynolds number (Fig 1.a), both transition models are able to simulate the separation-induced transition on the suction side. On the contrary, the turbulence model alone cannot reproduce the bubble. In Fig. 1(b), the Reynolds number is increased to reach an intermediate value. It can be noticed that a separation bubble is still present on the suction side and its size decreases in comparison with the flow at lower Reynolds number. The transition models produce results in accordance with the measurements, while the turbulence model is unable to do so. The situation at the highest Reynolds number is shown in Fig. 1(c). No separation bubble is reported in the experiment because the flow is fully turbulent. The transition models predict comparable results to the turbulence model, meaning that the transition models can be used with confidence even when the flow is expected to be turbulent.

MTU-B Low-pressure Turbine

Computations are used at the beginning of the design process to assess the quality of the blade’s design. However, in multistage components, the situation is much more complex: unsteadiness and three-dimensional effect must be considered. Consequently, transition models must be able to simulate accurately such flows.

The low-pressure turbine MTU-B illustrates the capability of TRACE to simulate such flows. This turbine has been designed by MTU Aero Engines and tested at the University of Stuttgart. A view of the turbine is presented in Fig.1.

Fig 1: Sideview of the MTU-B Turbine.

The performance of a low-pressure turbine is dramatically impacted by transition. Hence, it is important to be able to simulate it in design and off-design regimes. The Multimode and the γ-Reθ transition models, implemented in TRACE, are used to compute the flows inside MTU-B at different Reynolds numbers. The computed losses are compared to the measurements in Fig 2.

Fig.2: MTU-B: losses evolution with Reynolds number

For both models, the computations show results consistent with the experiments. The efficiency of the turbine decreases when the Reynolds number decreases.

The pressure coefficient distribution along the blades can provide a good indication on how well transition is simulated. Such distributions at different blade heights are shown for vanes 3 and 6. Only the lowest Reynolds number has been considered here.

20 % height

50 % height

80 % height

Fig 3: Pressure coefficient distribution over the MTU-B vane 3

Fig 3 presents the pressure coefficient distributions on vane 3. For all blade heights, the separation bubble observed during the experiments is quite well simulated by both models. Concerning vane 6, the pressure coefficient distributions are shown in Fig 4.

20 % height

50 % height

80 % height

Fig 4: Pressure coefficient distribution over the MTU-B vane 6

On this blade also, the separation-bubble present on the suction side is well reproduced by all models. In particular the occurrence of the separation at the different blade heights is successfully simulated.Hence, it has been shown that the transition models in TRACE can provide satisfactory results when simulating multistage low-pressure turbines

Kontakt

Dr.-Ing. Edmund Kügeler Institute of Propulsion Technology Numerische Methoden Köln Tel.: +49 2203 601-3248