Investigation of the Very Large Eddy Simulation Model in the Context of Fluid-Structure Interaction

Turbulent flows play a dominant role in most technical applications, for instance in the automobile industry or in aviation. Therefore, the correct prediction of these flows is a very important task. The numerical simulation of three-dimensional time-depended and chaotic turbulent flows requires a very high computational effort, what remains an essential problem for the simulation of such flows despite the increasing computational power. On this account the hybrid models for the numerical simulation of turbulent flows were developed. These techniques combine advantages of established turbulent basic models, such as Large Eddy Simulation (LES), Reynolds averaged Navier-Stokes equation (RANS) models and direct numerical simulation (DNS), with the goal to produce the correct results with reduced computational effort. The focus of the present study lies in the investigation of a relatively new hybrid modeling technique, the so-called very large eddy simulation (VLES) strategy, firstly on stationary grids and afterwards on moving grids. This investigation requires an extension of the turbulence modeling part in the in-house code FASTEST, from the Institute of Numerical Methods in Mechanical Engineering at the TU Darmstadt, with the k-ε, k-ω and ζ-f VLES models. To ensure the correct performance of the VLES model, the verification of these RANS methods with the method of manufactured solution was realized. After the verification procedure, the newly implemented VLES models were systematically validated on stationary grids. Three different flow configurations were selected to demonstrate the ability of the VLES approach to predict different types of turbulent flows, which occur most frequently in technical applications: a channel flow, a flow with massive separations and a flow over a bluff body. The k −ε and ζ −f VLES methods have shown reasonably good agreement with the reference data for all three test cases. The k − ω VLES model yielded very good results for the cylinder flow and for the two-dimensional periodic hills flow, while in the prediction of the fully developed turbulent channel flow this method demonstrated weaknesses. To improve the results of the k − ω VLES model, this method was modified by means of the introduction of a new filter width, the so-called IDDES, in the formulation of the VLES model. This modification leads to a significant improvement of the results already on a quite coarse grid. After the validation on stationary grids the VLES method was investigated on moving grids. Therefore two test cases were calculated: a flow over a forced oscillating circular cylinder and a flow over a tandem of an oscillating and a static asymmetric airfoil. The VLES method demonstrated the ability to simulate turbulent flows with strongly moving structures. The character of investigated flows was captured very well for both test cases. The last step was an investigation of the VLES approach in the context of a fluid-structure coupling, where the high computational costs play an essential role. To this end, two FSI test cases were investigated. In both cases, the VLES method demonstrated the capability to capture a variety of turbulent structures in these configurations. Other parameters available from the reference experimental data are captured very well by the VLES model. In summary, the ability of the VLES model to predict different kinds of turbulent flows correctly was demonstrated by means of different test cases on stationary as well on moving grids. For all simulations, a relatively coarse grid, in comparison to this required for the correct LES calculation, was applied. The application of this hybrid turbulence model is very promising because of its capability to predict flows with mild and massive separations more accurately than RANS methods and due to the reduced computational effort in comparison to LES approaches