exawind / iddes

IDDES turbulence model development & validation for ExaWind suite

ExaWind IDDES turbulence model development

Repositry for storing Exawind simulations with Hybrid RANS/LES models, including:

• Channel flow
• Periodic hill
• Flow around 2D Cylinder

iddes's Issues

Upwinding and cylinder case runs

This issue is devoted to documenting the effect of upwinding parameters on the cylinder runs (also as a test of video uploads in github issues). Feel free to comment below and close the issue when discussion is complete.

Testing alpha_upw

For both the grid0 and grid1 case, the alpha_upw parameter was varied for the velocity component only, and values of 0, 0.25, 0.5, 0.667, and 1.0 were tried:

    - alpha_upw:
        velocity: 0.25
        turbulent_ke: 1.0
        specific_dissipation_rate: 1.0

The turbulent_ke and specific_dissipation_rate were always set to 1.0.

The results show that for alpha_upw>0, the Cd results are relatively consistent, while no upwinding with alpha_upw=0 leads to an outlier result.
For the grid0 case

For the grid1 case:

Flow artifacts near the cylinder were seen in the alpha_upw=0 case

which are not seen for alpha_upw=1

or for alpha_upw=0.25

The Cp distribution is also similar between alpha_upw=0.25 and 1.0:
plotCp

whereas the Cp distribution for alpha_upw=0.0 is obviously different:
plotCp

Testing upw_factor

A similar study was done on the upw_factor. The velocity component of upw_factor was set to 0 for alpha_upw values of 0, 0.25, 0.50, 0.667, and 1.0 for the grid0 case:

    - alpha_upw:
        velocity: 1.0
        turbulent_ke: 1.0
        specific_dissipation_rate: 1.0
    - upw_factor:
        velocity: 0.0
        turbulent_ke: 0.0
        specific_dissipation_rate: 0.0

and for grid1, upw_factor=0 was tested with alpha_upw=1.

The drag comparison with upw_factor=0 also shows a wide variation
plotForcesUPWFactordiff_grid0.ipynb

And is similar to the results for the grid1 case

The wake behind the cylinder is also dramatically different for upw_factor=0
Snapshot_cylinder_run07_iter140_closeup.png

with a corresponding mismatch in the Cp distribution:

The far-field behavior for the grid0 case with alpha_upw=0 and upw_factor=0 is also suspicious:

The unsteady wake behavior for upw_factor=0 is also much different compared to the upw_factor=1 case

cylgrid1new_sst_iddes_Re8p0M_run07_upwfactor0_closeup.mp4

cylgrid1new_sst_iddes_Re8p0M_run01_alphaupw1_closeup.mp4