Institute of Aerodynamics and Gas Dynamics
- DFG SPP 1276: Discrete-continuous hybrid models based on integral conservation laws
- Compressible Turbulence simulation and modelling
- Multiscale numerics
- Large Eddy Simulation (LES) with Discontinuous Galerkin Methods (DG)
- High performance computing (HPC)
Turbulent flow characteristics manifest themselves in a wide spectrum of non-linearly interacting temporal and spatial scales. These complex physics pose serious demands on high-fidelity simulations in terms of numerical accuracy as well as high-performance computing aspects. Discontinuous Galerkin (DG) methods score favorably in bothfields and are thus considered viable candidates for the efficient computation of relevant turbulent flow situations. In particular, their inherent separation of flow scales renders them attractive for scale-dependent modelling approaches, aimed at reducing the number of degrees of freedom of the multiscale flow field in a physically sensible way, thereby allowing the computation of otherwise not treatable massive-scale flow simulations.
During the last year, our research group has focused on developing and implementing a DG framework for the highly efficient parallel computation of Direct Numerical Simulations (DNS) of turbulent flows, complemented by the associated postprocessing tools. As validation and benchmark test cases, this framework has been applied to the simulation of homogeneous isotropic decaying turbulence of high Reynolds number, a standard canonical flow situation for the evaluation of code performance and subgrid scale models of Large Eddy Simulation-type (LES) schemes, a turbulent round jet flow and super- and subsonic channel flows.
In the further course of our research, we will combine the high-order DG schemes with suitable modelling approaches for the unresolved scales. A special focus will be set on understanding and controlling the interactions of numerical properties of the DG scheme with the small-scale models and on finding numerically consistent subgrid scale formulations. The investigation of these aspects will draw from and rely heavily on the newly-created highly efficient DNS framework for basic research and validation. Since turbulent flow fields are inherently unsteady and three-dimensional and thus implicate very memory- and CPU-demanding simulations, the efficient implementation of high performance computing (HPC) strategies and massively parallel computations on large grids will be crucial. The work will by complemented by the application of the framework to aeronautical and meteorological flows of current research interest, including the scientific visualization and interpretation of complex, multiscale data.