Quick start

Case setup

Assume that you’ve already set up a case for the classical wall-resolved LES. To convert it to WMLES you need to do the following:

  • Add libWallModelledLES.so to the loaded libraries in the controlDict.

  • Go into the nut file, and set up wall models as the boundary conditions at the walls. A minimalistic setup for an explicit Spalding law-based algebraic wall model is given below.

1type           ExplicitWallModel;
2Law
3{
4  type         Spalding;
5}
  • In your 0 directory, you should add a new volScalarField, hSampler, see the Sampling section for details. For a quick start, set the value of hSampler to uniform 0 at the wall, and use zeroGradient at all non-wall patch boundaries. This will lead to sampling from the wall adjacent-cell, which is very robust, but inaccurate.

The settings above are not optimal, but should get your case running. Of course, you should never run your WMLES on a wall-resolving mesh. Instead, we recommend using a meshing strategy presented in Grid construction.

Miscellaneous tips

  • In regions where the TBL is attached, set hSampler to be the distance to the second consecutive off-the-wall cell centre. In other regions, set it to 0, i.e. sample from the wall-adjacent cell.

  • Use a mildly diffusive numerical scheme. One can consider LUST as a conservative option, but it is still very diffusive. On a good mesh, running with linear is a good option, as long as the simulation is stable.

  • The WALE model is a good first choice for SGS modelling. Don’t use implicit LES on a WMLES mesh.

  • If you use \(h = 0\), use an algebraic wall model in integral formulation, i.e. the LOTWWallModel with e.g. the IntegratedReichardt law.

  • Prefer explicit variations of wall models when they exist, e.g. ExplicitWallModel subclasses. They are more robust and cannot fail inside a root-finder iteration.

Publicly available cases

There is a number of cases that use the library available on the web. These can serve as good examples on how to set up your simulation!