In the modern process of the aerodynamic design of multistage compressors and turbines for jet engines as well as for stationary gas turbines, 3D-CFD plays a key role. Before building the first test rig several designs have been investigated using numerical simulations. To understand the characteristics of the individual components it is necessary to simulate their behavior in a multistage simulation and investigate for example, the single stage maps of the compressor in order to understand how the load is divided between the different parts of the compressor during throttling. Increasing computing resources allow ever more details to be incorporated in a 3D simulation. In former times only single blade rows were investigated with a high resolution of the boundary layers, whereas in multistage configurations wall functions were state of the art. Today we are able to apply Low Reynolds resolution even for multistage configurations, so the designer is required to include more and more geometrical details into the simulation. One important such feature is the fillets of rotor and stator blades. Fillets reduce the flow deflection at the endwalls and therefore the loading of the downstream blade rows. This effect is accumulated in a multistage simulation. In this paper a 15-stage compressor with additional inlet and outlet guide vane designed for a stationary gas turbine was investigated with a modern CFD tool by using a real gas approach for two speedlines. Two simulations were done: first a clean configuration with tip and hub clearances but without blade fillets; in the second simulation all rotor blades and the cantilevered stator blades were additionally modeled with fillets. The comparison of the overall global values with measurement data shows a better performance of the simulation with fillets, especially by throttling the compressor. A deeper look into the compressor shows different loads for a considerable number of single stages. The analysis of the steady multistage simulations shows that the numerical stability is reached in different regions of the machine.

Fundamental concepts for roughness modeling have been further explored and advanced. A basic understanding of the effect of distributed roughness on fully developed turbulent boundary layer, its possible influence on transition, and the mechanism of local spanwise roughness on transition has been achieved. Predictions with a refinement around a spanwise roughness element have been conducted in comparison to TATMo’s turbine cascade investigated at VKI. 3d-computations document the status in comparison to the T106C measurements with spanwise roughness for all Reynolds-numbers with two different transition models. Additional validation work shows the reproduction of accurate behavior of influence of height, location, and shape of the roughness element on pitchwise averaged loss and exit angle at midspan. Beside the correct reproduction of flow quantities for the spanwise roughness element, the right assessment of distributed roughness on surfaces of an industrial configuration is important. Because a high grid resolution very near the wall on all surfaces is not always possible, the problem can be solved with the help of wall-functions. The results of the application document the significance of rough wall-function modeling for tubomachinery.

This article describes how to extend the dsicrete adjoint method to functionals that are evaluated on arbitrary rotational control surfaces that intersect the flow domain at a position specified by the user, e.g. the pressure loss coefficient of a single blade in a multi-stage configuration. The definition and implementation of the mixed-out states on such surfaces is revisited. The calculation of the corresponding right-hand sides in the adjoint system is explained. These techniques can be used to specify functionals that quantify the deviation of the radial distribution of the flow angles, relative mass flow, etc. from a given target distribution. Sensitivity studies using the conventional approach, i.e. by means of finite differences of many steady solutions, are compared to results based on the adjoint method. The applications demonstrate that the agreement between adjoint and conventional sensitivity predictions is excellent, if the exact definition of the surface functionals is taken into account.