[Leifheit]

Hello everyone,

I recently set up an Ansys CFX simulation to do a mesh study for a 2D airfoil using the CFX Gamma Theta Model.
The results were quite good for an airfoil without any camber.
Since I switched to an airfoil that does have camber (NACA1408) none of my simulations converge any more.
Since I want to perform a mesh study I have been varying all relevant mesh parameters (Yplus, Cell sizes, Cell expansion rates, ...). I am trying different timesteps at the moment.
Also, since the simulation already produced good results for a non-cambered airfoil there shouldnt be any fundamental mistakes.

Do you guys have any ideas what I could look into to solve the problem ?
Also, if you need any more information please be sure to tell me !

This is only the second time I am using cfd-online so I might not be really good at presenting my problem. Sorry about that !

Best Regards,

Leif

[Leifheit]

Here is some more information ... made a new post since I couldnt find an option for adding attachments after submitting a post.

Attachments are a picture of what my residuals look like and the .out file of the simulation (i cut out most of the timesteps in order to make the file size suitable)

[ghorrocks]

FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

[Leifheit]

Hi ghorrocks,

thanks for your reply. I already read through

http://www.cfd-online.com/Wiki/Ansys...gence_criteria

and also the best practice instructions from Ansys.

Pretty much tried anything except running the simulation in transient mode.
- mesh quality is fine by ICEM check and the same mesh (only a small difference between cambered - non cambered) already produced good results for a non-cambered airfoil
- simulation physics should be fine for the same reason
- experimenting with different time steps at the moment but I do not see a reason why the timestep should have to be so different from non-cambered airfoil simulations to obtain convergence

I am reluctant to do the simulation in transient mode though since I figured there should be a way to obtain usable results via steady type simulations.

[ghorrocks]

Definitely try larger time steps. Also try local time stepping for a bit too. If that does not work you may be forced to run transient.

Transition models can resolve a laminar separation bubble at the transition point. This bubble often jiggles around - and this is why steady state simulations of this type of flow are problematic. So I would not be surprised at all if you end up needing a transient simulation.

[Leifheit]

Thanks a lot for quick and helpful replies !
Will try both and post again when I manage to obtain convergence.