[Nick R]

Hi guys,


I'm modelling flow around NACA OO21 using cfx. The Reynolds number is 120,000. My main problem is that SST gives me a much lower lift coefficient than expected (about 0.3 less) compared to experimental data. The K-epsilon model is a little better but not that accurate either.
My mesh which is structured is created in ICEM and the yplus is lower than one in cfx contours on the foil's surface.
The angle of attack is 10 but for lower angles there's still that discrepency.
I wonder if you can suggest ways to improve the results. What turbulence model and settings or solver control parameters would you recommend I try. Please let me know if there's any important parameter that would impact lift. Many thanks.

[alastormoody11]

hi,

you might want to check whether your wall y plus values are within the recommended ranges, if not you need to modify your mesh and then rerun your simulation.

Also since you desire to extract the lift coefficient from your simulation you need to plot against the iterations and ensure the variation in the lift coefficient is within reasonable limits when your convergence criterion are met which I am guessing, you are using the default values.

Also you might want to check the refernce pressure from which Cl is calculated.

[Nick R]

Thanks for the response. I calculate lift coefficient once the solution has converged using the force option in post. But if you're suggesting I monitor my forces then I'll do that.
I was also thinking perhaps I should use a "transitional model" but I'm not familiar with the option. The default setting for onset Reynold is 260. Should this be changed? Is this sufficient?
Also what do you think about switching to a transient model instead of a steady one. I'm quite inexperienced so it would be good to know what's worked for others before. Thanks again.

[alastormoody11]

Hi,

I was suggesting that you use the variation of Cl as it would be a better indicator of convergence, since it might be a good idea to iterate the solution till the residuals quiet low compared to the set defaults.

You can also the check the same using the vertical force, but I think you can also monitor Cl if you create a CEL expression in CFX pre, check that when the solution converges(acc. to default values) whether the Cl or forces are still oscillating over a large range, if so iterate it further till the variations drop to within an acceptable range. Also you should refer the convergence criterion section on CFD -wiki.

Also monitor the force variation wrt time if the variation is periodic, then your case is transient, but I think your case is steady. I had simulated a NACA0018 airfoil, and had got reasonably accurate values for Cl using a steady state approach.

Most importantly check whether your wall y plus values are in accordance with the recommended values stated for your chosen turbulence models.

I would not recommend using the transition model until you have run out of options, and changing the threshold value for Re would be a bad idea.

[ghorrocks]

Most importantly check whether your wall y plus values are in accordance with the recommended values stated for your chosen turbulence models.

A better and more general approach is to do a sensitivity study of mesh y+ versus lift, drag or whatever is important to you. Simply obtaining some arbitrary y+ target is not alway sufficient.

[Nick R]

Thanks. I'll keep trying!

[selbee]

This obviously has not been updated in a while but wondering if you ever achieved a more accurate presentation of the CL values?

[ghorrocks]

There is nothing on this thread which is not covered in the general FAQ on accuracy: https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

The FAQ which follows on publishable results is also useful.