Hi, I have been carrying out a number of CFX runs on cambered airfoils (ie. lift generating), primarily NACA 23012 and NACA 2412. The runs were carried out at 68m/s, 101325Pa, using the SST turbulence version with transitional turbulence set to fully turbulent. Results have shown that the Drag values are over estimated quite considerably, even at Alpha = 0 degrees by about 50%! See the following image to clarify;

http://img72.imageshack.us/img72/5042/dragplotgy5.jpg

The Standard Roughness and Reynolds 3, 6 and 8.8 million curves have been taken from Abbot and von Doernhoff; Theory of Wing Sections. They have also been backed up using XFOIL and from searching other results on the internet.

I have carried out endless mesh refinement and dependancy studies with no real impact on the values.

I have also tried running the same mesh setup and CFX pre setup on non cambered airfoils, which have resulted in correct values (or compared to Abbott and von Doernhoff at least).

Is there a special method of running CFX on cambered airfoils, or does anyone have any suggestions?

Cheers Mike

Hi,

Do you need the transition turbulence model with Re numbers this high? I suspect just running fully turbulent with no transition model will be OK (unless you are looking at laminar flow airfoils and the NACA 4 and 5 series are not laminar flow airfoils).

To get accurate drag you need to model the boundary layer accurately. What y+ are you using? Can you show a cross section of your mesh?

Glenn Horrocks

I would beg to differ from Glen on the laminar flow. In some ways 4 and 5 digit airfoils are the trickiest to simulate correctly, because they can have quite a lot of laminar flow (for example, on the lower surface of a 5-digit at positive angle of attack), but not the well-defined transition points you would get with an NACA 65 series, for example. So a small difference in transition position would make quite a difference in drag, and you would have a hard time knowing exactly where it was when the experiment was done.

Hi Charles,

If the Re was 1M or less I would agree with you, but with the lowest Re number Mike is considering being 3.3M then the laminar region of the airfoil should be fairly small.

For the purposes of Mike's simulation I would start with a fully turbulent SST simulation and after that is behaving itself I would try a turbulence transition model to see if it showed any significant differences. But I think the fully turbulent simulation is the place to start.

Regards, Glenn Horrocks

I agree with you on the basic methodology, turbulent first, then worry about transition. But I've actually found it quite frustrating to try and match Abbott & Van Doenhoff results in CFD. NACA's "standard roughness" is actually very rough, yet at the other end even at Re=3e6, laminar flow cannot be ignored. For example, XFoil suggests that an NACA 23015 at 3.3 million and 5 degrees incidence could have the lower surface transition as far back as 90% of the chord. Now it's arguable if you will ever see that in practice, but it means that the wind tunnel test results are likely to be quite dependent on just how well polished the model was, and how good the flow quality was. What this means in practice is that the drag target you are shooting at with CFD is not that well defined.

Hi,

If the transition is 90% of chord on the lower surface then yes, transition may well be important and I agree with you in that whether this happens in practise depends on many factors which are difficult to control.

I trust Mike will have fun chasing his drag target!

Glenn Horrocks