Identifying location of turbulence model shift in DES or SST models
 

Hello,

Is there a standard or convenient way of identifying where in my domain the solution shifts from one turbulence model to another when using for example a DES model or the k-w SST Model?

For the k-w SST Model the shift should be close to the walls, but how close?
For DES models the shift from LES to RANS should be controlled by the mesh element size, but if we have a gradual increase in element size, can we identify exactly where the shift is?

I believe this may be important when doing a Mesh Sensitivity Analysis. If a mesh is so coarse that it does not utilize the LES part of the DES turbulence model, can it really be considered the same model and representative for the MSA?

Thank you for any input!

Depends on which particular implementation/software you are using. There is usually a blending function that keeps track of how much LES-like and how much RANS-like the solution is in each region. You have to figure out how to retrieve the value of the blending function for whatever you are using. SST model also uses its own blending functions to switch from epsilon to omega.

DES: grid induced separation around foil
 

Hi Jonas and all foamers,



I'm not sure if you need to visualize DES Region, but we also investigate which model controls which region of the mesh to make our simulation converge correctly.

We are simulating wind tunnel experiment of a flow around an airfoil at 85000 Reynolds number and try to recover the experimental results measured with pressure sensors on the center line of our experimental airfoil.

We have tested komega-SST DDES model with 2 types of meshes varying the axial length of 1st layer cells from 3.1 mm to 8.2 mm. Unfortunately, we encounter a first layer treated by the LES model instead of komegaSST model in the foil boundary layer for both cases... This phenomenon leads likely to GIS (Grid Induced Separation) because we observe early flow separation on the suction side of the airfoil. It is likely the reason of our bad results you can see in the chart attached :( (DES_results_10deg.png).
Just as well for our faith in CFD (;-)), DDES-SA model with the same mesh (1st layer axial length = 3.1 mm) gives good results in agreement with the experiment :).

Does somebody has any hint to prevent the DDES model to switch in LES for the 1st layer? Do we have to modify any shield function or do we still have to enlarge our mesh cells in axial direction?

Thank you for your advice.

Clément