Hi

Im doing some baseline lift/drag calculations on a wing. The setup is simple- extruded 2D profile. To speed things up, the wing's width was reduced by 50%, and a symmetry plane was used.

I have read varying opinions on 'if' the function calculator in the post processor yields forces for only that part of the wing in the model (50%), or if it doubles (in this case) lift/drag (to produce results valid for a full model). Most people seem to think the former is correct, and that a doubling of lift/drag results from the function calculator is required. However, I rebuilt my model as a full sized wing, doubled the test area width (no symmetry plane used), and received results nearly identical to that with the symmetry plane.

Does anyone have a definitive answer on this issue? It is not mentioned in any tutorial I can find. I would assume the point of a symmetry plane is precisely to allow for valid results from a sectioned model, but with less processing required.

Best, Roland

The force on the wing is exactly the force calculated on the surface area of the mesh you have given the flow solver. There is no doubling as you describe.

If you double the wing surface, by extruding the mesh twice the distance then the force on the wing surface area should change by a factor of 2 I would think....

Dan

Dan

Thanks for the response. It is consistent with most of the info I can find in the forums, but not with the base experiment I tried. That is, If I generate a wing from an extruded 2d profile at 40" in span, and solve for it at 100mph 25c air, I get effectively the same values in the function calculator for force as if I solve for the same profile extruded 20" span using a symmetry plane. The delta in values was within 10%, which may be attributed to mesh, or to a slight variation in the tunnel I threw together.

I understand that a symmetry plane has great use in reducing calculation time for symmetric models, but as a programmer, I would expect that the function calculator would account for the symmetry plane in its results (i.e. forces in my scenario on lift and drag would double). Here is a thought- if the function calculator treats a sym plane as just a non-interacting wall, would I not see a force perpendicular (that is, some force on the Z axis where X is drag and Y is lift)? Depending on the air speed, this value would transition from negative to positive, but if the sym plane and function calculator interact, these forces should always offset perfectly.

Best, Roland

Hi Roland,

It would be very risky to assume that a symmetry plane means that anything connected to it should be doubled. What if you have symmetry planes on both ends (i.e. periodic symmetry)? What would you do then? Post gives you the value it calculates, nothing more, nothing less. The reason your results did not double must be attributable to the other things you changed.

As a better experiment, go back to your symmetric mesh and mirror it in Pre to create the full geometry. Run this again (to the same convergence level) and calculate the force on the full wing. You will get 2x.

Regards, Robin

Robin

That is a good point. I had assumed that adjustments were made based on the number/geometry of symmetry planes, but your thought on mirroring the existing mesh should help me to validate the function calculator's approach. Im still amazed that my test condition had such a small variance between full and half span on a simple airfoil. Maybe it is just coincidence that some mesh or geometry error yielded a 50% force reduction on the full-scale foil.

Thanks so much for the insight-

Best, Roland

Hi Roland,

The wind tunnel geometry can have a significant effect on the lift. Since you wind tunnel is not exactly the same, I suspect you are seeing this effect. Alternatively, it could be differences in your grid, indicating that you do not have a grid independant solution.

This is actually a common problem, usually resulting in a lack of confidence in a CFD code. People will often go to great lengths to remove boundary effects by moving their boundary conditions very far away from the airfoil, but compare the results to wind tunnel data.

It would be interesting for you to compare the results with slightly different wind tunnel geometries. If you do so, let us know your results.

Regards, Robin

Robin

Well, you were correct about the function calculator. It does compute only for the meshed object, and does not double forces for a sym plane (i.e. wing lift/drag). I suspect that my initial tests were flawed largely by a grid error on a leading edge. I did not have a NACA profile to validate against, so the model with the performance problem went unnoticed. I also suspect your thoughts on boundary layer is a minor contributor, though the wing was operating with about 1x span clearance (I would prefer 2x, but this is a large model with 2M+ cells already).

I re-meshed much more carefully, and found that the results were consistent (that is, the sym plane model generated 1/2 lift, 1/2 drag). Also, as one would expect, in the Sym plane model, there was a large Z axis force, where in the full model the Z value was near zero- easily attributed to assymetrical vortex or spanwise flow.

If you have experience with ICEM CFD, would you think that a more precise grid generator for aerodynamic applications? I have no experience with it.

Thanks again for your help, Best, Roland

Hi Roland,

Good to hear it worked out.

ICEM CFD Hexa is an excellent tool for generating hexehedral meshes for just about any application. The interface takes some getting used to, but I have not seen another tool with more flexibility.

For unstructured meshes, Tetra/Prism is good but I recommend getting CAD2Mesh, which gives you ICEM CFD Tetra/Prism and CFX-Mesh. You will find that CFX Mesh generates a higher quality unstructured mesh and better inflation, whereas Tetra/Prism is more robust on bad CAD.

Regards, Robin