Efficient Analysis of Transonic Base Flows Employing Hybrid RANS/LES Methods

Abstract

Scale resolving simulations employing hybrid URANS/LES methods allow for the prediction of transient phenomena of unsteady flows. In order to resolve the underlying phenomena a high spatial resolution especially of free shear layers is required. In addition, small time steps are necessary to obtain the formation of the vortices. Small time steps and high spatial resolution lead to large computing time as well as a great amount of permanent storage for the generated data. Different approaches have been followed to improve efficiency, reduce computational cost and decrease storage requirements. Grid adaptation has shown to be a suitable approach for the refinement of regions where grid resolution is too low. However, proper adaptation criteria are required. It was further found that the high resolution of the shear layer does not allow for a significant speed up of the simulation by employing a time stepping scheme which combines an implicit with an explicit solver. Although grid adaptation successfully reduces the number of grid points leading to a reduction of storage needs, the large amount of data still requires suitable data reduction technologies in order to exploit the results in an industrial context. This can be achieved employing decomposition methods like POD and DMD that describe the flow by a series expansion.