A modification of hybrid RANS-LES methods of DES-type for flows around airplane wings

Kurzfassung

Turbulent flows around airplane wings cannot be simulated using wall-resolved LES methods for the next thirty years due to the very high Reynolds numbers. On the other hand, state of the art statistical turbulence models (RANS models) often cannot give quantitatively satisfying predictions at the boundary of the flight envelope, where flow separation occurs. The most popular hybrid RANS-LES approach considered in the aerodynamics community is the so-called detached-eddy simulation method by Spalart (1997) DES aims to model regions of attached boundary layer flow in RANS modus and regions of separated (or detached) flow in LES modus. Despite a later improvement of the method referred to as delayed DES (DDES), see Spalart (2006), the method needs careful investigation before it can be applied to full aircraft configurations. In this talk we present a modified DES method which addresses the following issues. The DDES method tries to detect the boundary layer and to enforce the RANS modus there. However, as a first shortcoming, the identified boundary layer can be much too thin in the region of a strong adverse pressure gradient. This was shown for the flow around an airfoil at high angle of attack in Probst (2010), causing a much too early flow separation. The most challenging flows around airplane wings are flows with incipient separation, where the separation region is thin and lies inside the boundary layer. In DDES, however, the region of separated flow then lies inside the RANS region, and DDES behaves like a RANS model, which is seen as a second drawback. A third shortcoming of DDES is that, after the flow entered from the RANS region into the LES region, the generation of turbulent content takes a travelling distance in streamwise direction of several boundary layer thicknesses. This can cause poor prediction of flow reattachment. The aim is therefore to trigger the development of turbulent content more rapidly by using some forcing. Finally, a proper spatial resolution in the LES region needs to be ensured. We use a so-called single-grid estimator based on the resolved turbulent kinetic energy, see Knopp & Zhang (2008). Note that the third and fourth goal go beyond the aims in Spalart (1997) and Spalart (2006).