[Sony107]

Hi All,

I am currently running a simulation with Ansys CFX for the NACA 0012 airfoil. I’ve observed some discrepancies in the results:

When the trailing edge is closed and an unstructured mesh is used, the lift and drag coefficients do not yield accurate values.

However, when the trailing edge remains open and edge sizing is applied to the attached edge, the results are much closer to expected values.
MeshFile.jpg
Do you have any insights into why this might be happening? Additionally, is it acceptable to proceed with the simulation in this way?

[ghorrocks]

Proving that your results are independent of the mesh is one of the fundamental aspects of CFD. Your comments suggest you are not mesh independent, so you need to do more work here. Some comments:

* I am assuming that this model is running in the attached flow regime, with a fully turbulent boundary layer. If this is not correct please let us know as that will affect the meshing strategy.
* For simple airfoil models like this the C grid meshing strategy is usually optimum. This allows you to refine the mesh in the airfoil wake region. Your current mesh quickly expands the mesh in the wake region, so you will not resolve this nicely.
* For a simple airfoil mesh I would also do a pure hex mesh. That way you can easily control the mesh spacing in both directions.
* If you refine the meshes you already have you should, theoretically, achieve mesh independence. But a bad mesh will require a crazy fine mesh to achieve it, and you will probably be stopped by round-off errors before you achieve it.
* I would suggest you do a mesh independence study on each of your meshing strategies - that will show you which one is converging to mesh independence quickest.

Finally: What do you mean by and open and closed trailing edge? Can you show an image of what you mean?

[Sony107]

Thank you for your response.

1.I am using the SST turbulence model and ran the simulation for a Reynolds number of 2x10^5. I did not use any transition model.
2.I have already implemented the C-type grid meshing strategy and conducted a mesh independence study, where I obtained good results.
3.I try to create a structured hex mesh. However, in my simulation, it is needed a circular domain inside another outer domain while allowing the airfoil position to vary. When I change the airfoil's position, the mesh quality deteriorates, resulting in poor simulation outcomes (see Attachment 1).
4.For the unstructured mesh, I used two variations of the airfoil geometry: one with the trailing edge closed (from the airfoil plotter, see Attachment 2 and 3), which produced poor results, and another with the trailing edge left open. In the latter case, I connected the trailing edge using a line and refined the line, which yielded better results (as shown previously).

Could you provide suggestions on creating a high-quality structured mesh for my case, for the case shown in Attachment 1?

Thank you for your guidance!

1.jpg

2. Unstructured_joint mesh.jpg

3. zoom of 2.jpg

[ghorrocks]

Some points about your latest post:

1) Most airfoils at Re=2e5 have a significant length of laminar boundary layer transitioning to a turbulent boundary layer somewhere mid-chord. You might need a turbulence transition model to capture this as you will be influenced by both laminar and turbulent sections. You have modelled the flow as fully turbulent, that is likely to over estimate the drag and stall at higher AOA.

2) Your post #3 meshes do not appear to have any inflation layers. You are going to have problems getting accurate results without inflation layers in this Re regime.

3) Your inflation layers in post #1 appear to go from very fat to very thin at the trailing edge. You probably want this to be a more even resolution along the chord length.

4) Do the tutorials for your meshing package for advice on how to generate hex meshes.

5) I would recommend you read the CFX documentation, CFX Solver Modelling Guide, section 16.3 on mesh resolution near the wall and mesh quality.

6) If you are using wall functions you will probably want a y+ value of 10-20 on the wall, or if you are using a turbulence transition model you will need a y+ approximately 1 or less. This can guide you on how fine to make your airfoil mesh.