I simulated a setup of the test section of the Langley 2D low-turbulence wind tunnel per NACA-TN-1283 to the extent possible. Then I simulated an experiment with NACA 2412 airfoil at 0 and 10 degrees AOA and Re=3.1e6.

I got a very close match of the section lift coefficient for 0 AOA, about 0.25. Unfortunately I could not get the section lift coefficient for 10 degrees any closer than 1.000, while per test in the Langley 2D tunnel cl is about 1.25.

I played with top and bottom no slip walls and openings. With walls I got a better result than with openings.

I played with side walls as no slip walls and symmetries. Symmetry was better.

To simulate the AOA I played with velocity components and actual airfoil orientation. Orientation was much better.

I played with coarse mesh and fine mesh. The difference was in the second digit after the comma.

Eventually for the 24" chord I simulated the 0.1" (or 2.5mm) thick boundary layer, split it by 12 nodes per Guidelines for Mesh Generation and applied SST option. Cl has been calculated based on Y-component on the wing. Advection scheme has always been set to High Resolution, the outlet average static pressure to 1 atm. The result was only 0.9997.

So for all the options above the cl varied only between 0.965 and 1.0001, but never got 1.25.

Well, I'm lost. Either I do something very wrong or … the software does?

Hi Andrew, in my opinion the problem is the SST turbulence model. The model usually undervalues lift on airfoils at higher AoA. I have ran into the problem in all CFD codes I have used. Try to switch to another one. Good luck Zbynek

In my experience sometimes the SST model tends to predict too early separation. Andrew, are you seeing large separations in your SST simulation at high AOA? If so, that could explain the poor CL prediction.

I can't see any problems for calculation of Cl. All you have to do is to use proper mesh setting to catch high pressure gradients, because of pressure gradient nature of lift. You don't need to resolve BL in such accurate manner. If you have problems at AOA which is close to stall, it means that you should use transient simulation. Are you going to calculate drag? It's much more difficult

I agree with DAK. The lift prediction should be pretty easy to get right, so long as you have a high quality mesh, a Hex C-grid is best. To get the drag right you'll need to be carefull and make sure the mesh has the correct y+ and enough nodes within the boundary layer. At high angles you should be running a transient simulation as DAK has mentioned, with a small enough time step, so as to pick up the vortex shedding. I've used the SST model recently on a NACA 63-418 and the high and low angle runs converged nicely. James

In fact I see no separation at all. I tried both K-E and SST and for both the airflow looks laminar, i.e. the streamlines are laminar over the entire surface of the wing. Neither switching between K-E and SST does significantly affect the cl value

10 degrees AOA for NACA 2412 is kind of far from stall. It stalls at about 15-16 degrees. And yes, I'm going to calculate drag. In fact I want to get a whole usual set of data for the airfoil for the software validation.

What do you mean with hex C-grid? Do you actually mean hex O-grid?

How fine should it be? I mean with all the variation of the fineness in my case I got variation of the cl only between 0.97 and 1.00. So it's only 3% difference, while the actual value of cl for 10 degrees AOA is 1.25. The difference is more than 20%. I believe there must be another reason, but not fineness of the mesh. Don't you think so?

How do your results for NACA 63-418 match the NACA experimental data?

Here's a sample of my results at AoA = 0.0, i compared my results from Xfoil, which is pretty good at low AoA.

Xfoil Results @ AoA = 0 deg

Cl = 0.35350 Cd = 0.00639 Cdp = 0.00131 Cdv = 0.00508 Cl/Cd = 55.32081 Cm = -0.078

CFX Results @ AoA = 0 deg

Cl = 0.32968 Cd = 0.00923 Cdp = 0.00256 Cdv = 0.00667 Cl/Cd = 35.72826 Cm = -0.07403

Thats about 7% difference in Cl.

James

I absolutely have no problem with AOA=0. In fact the difference between CFX results and NACA experimental data for 0 degrees is close to 0%. That is much better than your 7%. Probably Xfoil is not a very good reference point.

The questions arises for 10 degrees. There the difference is about 0.97...1.00 for CFX against about 1.25 for NACA experimental data.

In my calculations Cl is more different from experiment the more AOA. But all data is within adequate imprecision. But with Cd there are big problems - at AOA = 8 it is 1.5 as large than experiment. The problem maybe in bad model of your airfoil. When increasing AOA transition region appears. And the more AOA the more this region. I've noticed - if I increase resolution in this region, I get diiference about 1.2 in drag.

[Anna Tian]

Originally Posted by James Date
;73862

Here's a sample of my results at AoA = 0.0, i compared my results from Xfoil, which is pretty good at low AoA.

Xfoil Results @ AoA = 0 deg

Cl = 0.35350 Cd = 0.00639 Cdp = 0.00131 Cdv = 0.00508 Cl/Cd = 55.32081 Cm = -0.078

CFX Results @ AoA = 0 deg

Cl = 0.32968 Cd = 0.00923 Cdp = 0.00256 Cdv = 0.00667 Cl/Cd = 35.72826 Cm = -0.07403

Thats about 7% difference in Cl.

James

Your results still have large inaccuracy in Cd. Any experience on correcting that?

[Anna Tian]

Originally Posted by DAK565656
;73887

In my calculations Cl is more different from experiment the more AOA. But all data is within adequate imprecision. But with Cd there are big problems - at AOA = 8 it is 1.5 as large than experiment. The problem maybe in bad model of your airfoil. When increasing AOA transition region appears. And the more AOA the more this region. I've noticed - if I increase resolution in this region, I get diiference about 1.2 in drag.

I tried the mesh refinement. Once Y+<3, the refinement doesn't change the results anymore.