The work at hand presents a direct numerical simulation of the flow over periodically ar- ranged hills at Reynolds number Reh = 10, 595. This flow configuration generates a number of complex flow phenomena, including the separation from a curved surface and subsequent reattachment. Since it also comprises well-defined boundary conditions and low computa- tional requirements, the flow case has turned into a benchmark case for the validation and assessment of CFD-codes and turbulence models. In order to be used as a test case, reliable reference data are needed. Direct numerical simulation is one way of providing this reference data which became more popular in recent years due to the increase in computational power. This simulation method resolves all spacial and temporal scales of the flow and therefore does not depend on turbulence models. The computations are performed using the CFD-code MGLET which is based on a finite- volume method. Spacial discretization is done by a second order central scheme. For the simulation, a new Cartesian grid with 331M points and local grid refinement on four levels was created. The assessment of the new grid for the DNS shows a number of shortcomings. Along the walls, the resolution is sufficient to resolve the near wall region and the velocity gradient within. In the bulk region however, the grid spacing is up to 17.4 times the Kol- mogorov length η. Therefore it has to be assumed that not all scales of the flow are resolved within the simulation. The performance of the new grid is validated by running a DNS at Reh = 5, 600. Despite the aforementioned shortcomings of the grid, the simulation results show very good agreement with the available reference data. Additionally, a grid convergence study is included to scrutinize the influence of the grid resolution. Regarding the flow case for Reh = 10, 595, the results show overall the expected flow behaviour. However, there are some deviations from the reference data for the velocity statistics of first and second order. In contrast to other studies no secondary recircula- tion bubbles could be observed. The differences are most likely due to the insufficient grid resolution. Additionally, the data is prone to statistical errors due to the low number of flow-through times that were used for sampling. In addition to the velocity statistics of first and second order, the structure of turbulence was analysed using the anisotropy invariant map.     «

The work at hand presents a direct numerical simulation of the flow over periodically ar- ranged hills at Reynolds number Reh = 10, 595. This flow configuration generates a number of complex flow phenomena, including the separation from a curved surface and subsequent reattachment. Since it also comprises well-defined boundary conditions and low computa- tional requirements, the flow case has turned into a benchmark case for the validation and assessment of CFD-codes and turbulence models. In or...     »