Sanity check of NACA 0012 gives wrong coefficient of lift

A subforum specific to the development of the OpenFoam-based workbenches ( Cfd https://github.com/qingfengxia/Cfd and CfdOF https://github.com/jaheyns/CfdOF )

Moderator: oliveroxtoby

thschrader
Veteran
Posts: 3129
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by thschrader »

BobHope wrote: Fri Jan 13, 2023 9:30 pm ...
Running the 3D model now with a fairly efficient mesh and an improved volume mesh in the wake.
...
With the 3D-case and using SimpleFoam I still get residual convergence. Seems there is a stable
vortex at the wingtip. The CoL is 0,73, compared to 1,01 in the 2D-case.
The result seems correct, because with the 2D-case you simulate a wing with infinite length
(the air can only pass over/below the wing), but the 3D-case simulates a wing with finite aspect ratio
(the air can pass over/below and left/right of the wing), so the 3D-CoL is lower.

Runtime with 1,97 mio cells and simplefoam approx 30 min. Meshing goes fast.
3D_case_simplefoam.JPG
3D_case_simplefoam.JPG (94.99 KiB) Viewed 1589 times
BobHope
Posts: 22
Joined: Fri Dec 16, 2022 9:10 am

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by BobHope »

I found this full size test with a complete wing with an aspect ratio of 6

https://ntrs.nasa.gov/api/citations/199 ... 091723.pdf

The NACA at 0012 look like this :
Full Size Windtunnel tests.png
Full Size Windtunnel tests.png (651.54 KiB) Viewed 1489 times

Converting and solving variables from the reference :

NACA 0012

87 feet ps = 26.5 M/S

Re 3.3x10^6


q = (1.2 kg/m^3) / 2 * (26.5 m/s)^2
= (1.2 / 2) * (26.5^2)
= 0.6 * 702.25
= 421.35 N/m^2

Reference area

6 foot x 36 foot = 182.88 cm x 1097.28cm = 200619.36 or just over 20m square.
Results are good, the Cl of 0.76 at 10 degrees angle ( green line above ) agrees well with the real windtunnel results, and so does the Cd

13:07:33 forces ReportingFunction write:
Sum of forces
Total : (552.2152 0.88156894 6165.8845)
Pressure : (485.04306 0.87345659 6172.4242)
Viscous : (67.172144 0.0081123454 -6.5396431)
Sum of moments
Total : (33820.04 -3027.8889 -3027.621)
Pressure : (33855.953 -3027.2752 -2658.9246)
Viscous : (-35.912867 -0.61368162 -368.69647)
writing force and moment files.
13:08:08
13:08:08 forceCoeffs ReportingFunction001 write:
Coefficient Total Pressure Viscous Internal
Cd: 0.068073866 0.0082805897 0.059793276 0
Cd(f): 2.3502242 2.3227871 0.027437122 0
Cd(r): -2.2821503 -2.3145065 0.032356154 0
Cl: 0.76009424 -0.00080616902 0.76090041 0
Cl(f): 0.1726802 -0.20772798 0.38040818 0
Cl(r): 0.58741404 0.20692181 0.38049224 0
CmPitch: -0.20736692 -0.20732489 -4.2028382e-05 0
CmRoll: 2.3161873 2.3186468 -0.0024595158 0
CmYaw: -0.20734858 -0.18209816 -0.025250416 0
Cs: 0.00010867467 1.0000426e-06 0.00010767463 0
Cs(f): -0.20729424 -0.18209766 -0.025196578 0
Cs(r): 0.20740292 0.18209866 0.025304253 0
BobHope
Posts: 22
Joined: Fri Dec 16, 2022 9:10 am

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by BobHope »

Now I have something that looks like it approaches reality thanks to Thomas, I am testing the limits and common aerodynamic effects to see how they are modeled.

The first test is to take the working wing above and stall it, and see what the results are. Stalls and turbulent conditions are notoriously hard to model in CFD, so lets see how we do.

The first model I built with the same mesh as the model above, but with the wing now at 20 degrees. NACA 0012 stalls around 18 degrees.
The simulation became unstable and did not converge at all.

I then tried reducing the mesh complexity. To my surprise, I was able to obtain reasonably accurate data ( Cl: 0.7486035 ) at 200mm mesh with only a 0.3 surface refinement with no boundaries. Looking at Paraview, I could see a turbulent, detached vortex had formed behind top wing skin. I suspect it is this vortex that breaks the simulation, therefore it may be worth removing volume refinement from areas of models that may stall.

09:50:48 forceCoeffs ReportingFunction001 write:
Coefficient Total Pressure Viscous Internal
Cd: 0.31431869 0.0032964317 0.31102226 0
Cd(f): 2.4420516 2.2879999 0.1540517 0
Cd(r): -2.1277329 -2.2847035 0.15697055 0
Cl: 0.7486035 -0.00046503492 0.74906853 0
Cl(f): 0.069031774 -0.30519454 0.37422631 0
Cl(r): 0.67957172 0.30472951 0.37484222 0
CmPitch: -0.30526997 -0.30496202 -0.00030795032 0
CmRoll: 2.2848923 2.2863517 -0.0014594244 0
CmYaw: -0.95868109 -0.94862498 -0.010056109 0
Cs: 0.00070168837 -9.3709998e-06 0.00071105937 0
Cs(f): -0.95833024 -0.94862967 -0.0097005795 0
Cs(r): 0.95903193 0.94862029 0.010411639 0
NACA 20 degree stall.png
NACA 20 degree stall.png (688.27 KiB) Viewed 1398 times
BobHope
Posts: 22
Joined: Fri Dec 16, 2022 9:10 am

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by BobHope »

I cut a chunk out of the Cylinder mesh volume refinement where the stall vortex/separation occurs.

This worked well, and the simulation converged and gave good Cl and Cd measurements at higher refinement rates.
User avatar
oliveroxtoby
Posts: 810
Joined: Fri Dec 23, 2016 9:43 am
Location: South Africa

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by oliveroxtoby »

BobHope wrote: Tue Jan 17, 2023 10:21 am NACA 0012 10m Full Size 25 degree should stall.FCStd

The simulation became unstable and did not converge at all.

I then tried reducing the mesh complexity. To my surprise, I was able to obtain reasonably accurate data ( Cl: 0.7486035 ) at 200mm mesh with only a 0.3 surface refinement with no boundaries. Looking at Paraview, I could see a turbulent, detached vortex had formed behind top wing skin. I suspect it is this vortex that breaks the simulation, therefore it may be worth removing volume refinement from areas of models that may stall.
I suspect the divergence is due to the fact that cfMesh produces an invalid mesh in this case (use the Check Mesh function). It can struggle when there are a lot of boundary layers. A vortex shouldn't cause divergence, although it may well prevent the residuals falling below a certain value.
BobHope
Posts: 22
Joined: Fri Dec 16, 2022 9:10 am

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by BobHope »

Thanks Oliver, I think you are right. The stall scenario works with different mesh parameters. Thanks for the hint about boundary layers.

No CFD today sun is shining, and it's time to make a real planes fly :-)
IMG-20230118-WA0000.jpg
IMG-20230118-WA0000.jpg (330.87 KiB) Viewed 1204 times
BobHope
Posts: 22
Joined: Fri Dec 16, 2022 9:10 am

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by BobHope »

Reference density.png
Reference density.png (26.34 KiB) Viewed 1133 times
Hi Guys, I noticed something in the Reporting Coefficient Function,

There is a value "Reference Density" in the Data of the Reporting function that is not exposed via the GUI.

Should this be changed to 1.2KG to match the density of Air ?

Or is this value not important ?
User avatar
oliveroxtoby
Posts: 810
Joined: Fri Dec 23, 2016 9:43 am
Location: South Africa

Re: Sanity check of NACA 0012 gives wrong coefficient of lift

Post by oliveroxtoby »

BobHope wrote: Thu Jan 19, 2023 2:35 pm Reference density.png

Hi Guys, I noticed something in the Reporting Coefficient Function,

There is a value "Reference Density" in the Data of the Reporting function that is not exposed via the GUI.

Should this be changed to 1.2KG to match the density of Air ?

Or is this value not important ?
This is only exposed in a compressible analysis. In the incompressible case, OpenFOAM already divides the pressure by density, so it isn't needed or used.
Post Reply