Unsteady Simulation (DNS & LES)
By Dr Jean-Philippe LAVAL
The main course objective is to provide students with an overview of the methods used for direct numerical simulation of a flow. The requirement in terms of grid refinement as well as in accuracy of the numerical methods will be discussed. The aim is to show the possibility as well as the limitation of DNS in several flow configurations. As DNS are usually associated to very large simulations, parallelisation will be discussed. An introduction to parallelisation languages such as MPI and Open MP will also be offered.
In a second part the aim is to provide a thorough knowledge in the theory of large eddy simulation and to show how different it is from a Reynolds Averaged Navier Stokes (RANS) modelling approach. In order to understand the concept of Large Eddy Simulation, the energy transfer between scales will be investigated. The different concepts of subgrid scale models (turbulent viscosity, scale similarity, de convolution based models) will be also presented and several sub grid scales models will be derived and investigated in detail. Different composite methods such as RANS-LES modelling or Detached Eddy Simulation (DES) will be introduced with a focus on the precautions for use of such methods.
Vortices in a DNS of converging-diverging channel flow
(click to see the movie)