[diwakaranant]

Hi

I am trying to run unsteady flow simulation for NACA0012 airfoil
for Mach = 0.5, Re = 10000 and AoA = 0 degree
The flow is unsteady for this case due to vortex shedding.

But I am not able to get this result in SU2. I am not getting
any periodic oscillations in lift/drag coefficients with time.

I am attaching the .cfg used.

Can anyone tell why is this happening ?

Thanks
Anant

[EMolina]

Hi Anant.

Your configuration file is correct. I think your teste case is incorrect! How can you expect periodic vortex shedding on a symmetric airfoil with AoA=0???? Where did you see this? Can you show to us?

Regards

Eduardo

[diwakaranant]

Hi

Please refer to this paper by M. Braza
http://www.sciencedirect.com/science...45793002001007

Thanks
Anant

[demanosalvas]

Anant,

Your config file looks fine for what you are trying to model. A few suggestions that work very good when trying to model vortex shedding on bluff bodies are:

• Reduce the time steps to make sure that you have around 25 steps per period.
• Sometimes it takes a very long time for the vortex shedding to develop, so its good to insert a disturbance in the flow (run the simulation for a few iterations with a slight angle of attack)
• The size of the elements on the wake could be introducing numerical dissipation (if they are too big) make sure to refine the mesh in the regions where you expect vortex shedding to take place

Hope this helps,


David

[diwakaranant]

Hi

I tried using restart file to induce some disturbance, but still I am not
getting the flow as mentioned in the paper attached in the Post 2 of
this thread.

The config file is attached (without restart option).
I am using the hybrid mesh for NACA0012 given in the testcases
folder ( mesh_NACA0012_lam_hybrid_v3.su2).

Can anyone help ?

Thanks
Anant

[demanosalvas]

Anant,

I noticed that you are using JST, and this scheme can be a bit dissipative. Try lowering the coefficient of the 4th order artificial dissipation term, about one order of magnitude, and if this doesn't work give Roe 2nd order a try.

Have you tried refining your mesh and inserting a disturbance in the flow yet?

Hope this helps,

David

[economon]

Hi guys,

You can achieve the type of results you see in the paper above if you make a few adjustments:

1. I think a c-mesh is better suited for this case, which gives extra refinement in the wake region.

2. Check that you have ~30 mesh points in the boundary layer for this Reynolds number (laminar flow).

3. I would second that the 2nd-order Roe method should be a good choice for this problem.

Please see the attached figure of SU2 results for a shedding NACA 0012 @ M = 0.8 and Re = 10,000. I have posted the files for this case here temporarily if you want to give it a try: http://www.stanford.edu/~economon/dr...m_naca0012.tgz. There are several other small changes that I've made to the config file you'll find in that directory.

All the best,
Tom

[diwakaranant]

Hi Thomas

I tried running the simulation for Mach = 0.8, using the config file and the
mesh file given by you.

The results are not matching to that given in the paper.

In the paper, the time period of the primary oscillation of cl coefficient is around 0.55 seconds, where in SU2 it is much less.

Moreover the secondary oscillation is not observed in SU2.

I am using time step for dual time stepping as 0.0001. So while plotting
cl vs t, I am multiplying no. of external iterations with this time step.

Please look into it. Cl vs time plot is attached.

[economon]

Hi Anant,

I should mention that I did not optimize/investigate any of the parameters in the config file and mesh that I provided above, but rather just wanted to give an example where shedding is observed.

You may want to perform a grid refinement study to make sure that the mesh is suitable for the conditions that you are simulating (again, I would recommend a c-mesh with decent resolution in the wake). You can also check that you are using exactly the same parameters as in the paper above. In particular, you may need to investigate the effect of the limiter coefficient for the 2nd-order Roe method or that the physical time step chosen is adequate to capture the effects that you are expecting.

Cheers,
Tom