[Wowbagger]

Dear Foamers,

I got a question regarding the CD Value of a NACA0012 airfoil calculated with different values of yPlus.

My knowledge about how yPlus affects the simulation is rather limited. I only read about this in before.
Because of this I tried to make it clear for myself and made a testcase.

I made a NACA0012 profile wich is mostly related to only one yPlus value, see picture below as an example with yPlus = 75, expansion Ratio was always near 1,1.
I made this mesh in combination with excel and blockMesh.
I can change the first Layer thickness in an easy manner without a lot of other work around.

Flow_yPlus75.png


I used the kOmegaSST model which should be insensitive with regards to yPlus (http://www.wolfdynamics.com/training...e_2021_OF8.pdf).

I suggest that the outcome of my simulated CD values are fine for yPlus 0,5...3.
I read that the CD result is bigger because of transition effects wich are in the range I have.
(Nasa Exeriment, similar to my Case: Re 5M; Ma 0,15; Cd~0,65)
But I suggested that the value of CD is not drifting that far away in the yPlus range of 5...150.
I know that the the values of yPlus between 5...30 should be avoided. But I have an increasing CD value with increasing yPlus which I think is much too high.

yPlus CD
0,5 - 0,00719
1 - 0,00725
3 - 0,00762
5 - 0,00836
50 - 0,0088
75 - 0,00867
150 - 0,0094


I thought about possible error influences with regards to my boundarys and also the mesh.
See Mesh below, as example for yPlus 75. Other Meshes are very similar but have a bigger cell count in the BL region.
I made a test with a bigger domain at yPlus 1 with the nearly the same result.


Mesh_yPlus75.jpg

k:
    wing
    {
        type            kqRWallFunction; // fixedValue; kLowReWallFunction;
        value           uniform 1e-10;
    }

omega:

    wing
    {
        type            omegaWallFunction;
        value           uniform 1e9;
    }

The convergence is not that good than I would like it, see below.
But I did not manage to get it better.
Anyway, monitoring the inlet and outlet flow and the CD number showing stable values after ~200 Iterations.

Convergence_yPlus75.png
CD-yPlus75.png

For any more related questions I zipped the yPlus75 case into my dropbox: https://www.dropbox.com/scl/fo/2fi79...md1jnf5ae&dl=0
Feel free to ask anything which could help.
I would be very happy for any hint which helps me to understand this topic better.

Best regards
Michael

[Wowbagger]

Why could drag drift that far in relation with yPlus?!

...every idea, every hint, every suggestion is very welcome.


Best regards
Michael

[geth03]

if you want to accurately predict drag you need to accurately predict skin friction/shear stress right at the wall, which you can only do if you resolve the velocity profile accurately -> lower yPlus. the higher the value the higher the deviation.

[Wowbagger]

Hi Geth,

thanks for your answer. Hi read about that in tutorials related to turbulence.
I did not expect a drift so far away. I do not have any experience in that but related to a validation case from Wolf Dynamics (http://www.wolfdynamics.com/tutorials.html?id=149) I thought the results will be better...

[geth03]

Predicting drag accurately is a tough business. 2d airfoil simulation looks like an easy start, but deviations up to 50% is not uncommon. Your mesh needs to be perfect and still you won‘t get perfect results bc of modelling errors (not good enough turbulunce model, assuming turbulence in places where flow is actually laminar and transitioning to turbulent at Re=500k, bad numerics, steady state assumption when flow is unsteady to name a few)

Predicting lift and Cp is easier bc of better predictions of pressure field (most of lift is generated from pressure than from shear stress)

You need to be careful when taking provided examples as tutorials without questioning it bc if you have experimental data you can play around with everything and match quantities of interest somehow for wrong reasons. I read a paper today which showed the simulated skin friction with measurements, they were quite off but in the end the drag matched, the model overpredicted and then underpredicted the skin friction and in the end the integral of that matched the drag coefficient. Perfect result bc of wrong simulation (taking the integral of it smoothed the result) -> can you say that thr simulation is good now or bad? Thats for you to decide …

[Wowbagger]

Thanks a lot Geth,



this helped me a lot in understanding and how to deal with my results.


Best regards
Michael