[xgarnaud]

Dear all,

I am looking at the implementation of wall models (for now for the SA model) in SU2. The strategy I wanted to use is:

• compute the wall whear stress at the first point off the wall using the wall model
• use it to set nutilda at the first point off the wall, nutilda = kappa*(y+)
• also use it to set nutilda at the wall, such that Density(nu+nut)*du/dy = the wall shear stress given by the wall model

This way the shear stresses at the wall for the NS equations should be correctly computed with minimal modifications to the sources ( in solver_direct_mean.cpp, I only changed the wall shear stress computation to use the value of the laminar+eddy viscosity)


I applied this method to the turbulent flat plate test case, with a fine mesh (y = 1e-6 at the first cell, y+ << 1) and a coarse one (y = 1e-3, y+ \approx 100). Results are attached: it seems that at the first point off the wall the wall shear stress is computed correctly from the velocity and wall distance, and that the value of nutilda is imposed correctly. However, the eddy viscosity in the BL is overestimated and the BL is too thick.

What am I doing wrong?

Best regards

Xavier

u.png

u2.png

nut.png

[fpalacios]

Thanks Xavier,

This is an important implementation in SU2. It would be great if you could work on a GitHub branch (we can give you access). In fact, we started the implementation of the wall functions some time ago in

https://github.com/su2code/SU2/tree/...Wall_Functions

it would be great if we work in a coordinate way. In you are interested, just send me an email fpalacios_at_stanford.edu

Best Regards,
Francisco
SU2 lead developer

[xgarnaud]

Dear Francisco,

The emails I try to send you are not delivered. Could you please email me at xgarnaud_at_gmail_dot_com?
Thank you very much,

Xavier

[OVS]

Hello,

I'm wondering if the wall functions were added to any more recent version of SU2.

Thanks.
Oliver