[cfxbeginner]

Hi,

I'm trying to model a Darrieus Tidal turbine and I need to define a rotating domain. I have a problem because when I define a rotating domain in a stationary one. The flow is modified by this rotating domain even if the rotating domain is empty...
I use frozen rotor in a steady simulation just to solve this problem. The interface is a GGI interface (Bitmap) but I think that the error is not the GGI.
I join the streamlines and a document with the information of my run and hope someone could help me.

Regards, CFX Command Language for Run.docx

streamlines.PNG

[ghorrocks]

FAQ: https://www.cfd-online.com/Wiki/Ansy...f_reference.3F

[cfxbeginner]

Thank you very much ghorrocks for your answer
Yet I have already read this post and even if I use 'velocity in Stn Frame' the result is not what I expect. The flow should be a uniform flow but it's not the case. Here is the result for the velocity in Stn Frame.
velocity in Stn Frame.jpg

[ghorrocks]

Try activating the alternate rotation model and re-running the simulation. You also might need to run to a tighter convergence tolerance.

[cfxbeginner]

Thank you for your help.

I activate the option 'alternate rotation model' but there is no change. I also focused on the convergence but it seems to converge really well (RMS evolution attached).
Maybe the boundary conditions for the rotating domain are not good for a rotating domain. I impose symmetry for the back and the front ?
The result seems to be that the flow is affected by the rotation of the domain...

Regards ,
convergence.PNG

velocity in Stn Frame_2.PNG

[ghorrocks]

Please show an image of your mesh and geometry. Make sure you show what is happening in the 3rd dimension. Also post your output file.

Try converging to a tighter tolerance and see if that makes a difference.

[cfxbeginner]

I tried to tighten the convergence (RMS=10e-5) and I increased the timescale factor to reach this goal as you advice in an other post. It seems not to improve the resuls.
Sorry for the lack of accuracy of my last reply, I didn't no what could help you understand the problem.
Here are the files expected.
I also checked the 3rd direction, but it is uniform as expected.

rotating_domain.PNG

stationnary_domain.PNG

CFX Command Language for Run.docx

[ghorrocks]

Try tightening it again to 1e-6.

Please show how you have meshed your third dimension.

[cfxbeginner]

I increased the Timescale Factor from 100 to 1000 and obtained a convergence of 1e-6. Yet, the result is the same.
Here are my mesh in the 3rd direction. The third dimension is really thin as I don't need this dimension (0.001 m).

rotating_domain_3rddim.PNG

stationnary_domain_3rddim.PNG

[ghorrocks]

Can you put a mesh in the 3rd dimension - maybe 10 elements rather than the 1 you current have? Give that a try.

(I don't know what is happening but I have a hunch it is to do with the 1 element thick mesh. If you do this test you can see if my hunch is right or not)

[urosgrivc]

Would anithing change if you would switch from simetry to -interface periodicity-?
I think this is similar to changing the sweep number of elements.
And I think it doesent impose a freeslip wall like BC that symetry does

[cfxbeginner]

Is there any boundary condition 'periodicity' or is it an 'Interface model'?
When I add this model, I have an error :
+--------------------------------------------------------------------+
| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| ****** FATAL ERROR ****** The orthographic view transformation fa- |
| iled on domain interface "Domain Interface 1". Failure may be be- |
| cause the rotational periodicity axis was incorrectly specified o- |
| r because side 1 and side 2 of the interface are not exactly peri- |
| odic. Please carefully check if either of these situations occur- |
| ed. |
+--------------------------------------------------------------------+
I think it is because my outer domain (the stationary one) is a parallelepipoid. Do you have any suggestion to solve this problem?

I'm going to properly remesh with several elements in the 3rd direction to test.

[cfxbeginner]

I succeed to properly mesh with 10 elements in the 3rd direction. I increase also the thickness to 0.01m to see well. Nothing happen, I still have the same result...
Hope that the periodicity could improve it if I succeed to impose this one.

A rotating domain doesn't have an impact on the flow, it's just the mesh that turn ? I begin to have doubt about my model. Maybe is it due to the frozen rotor option ? Or the stady state ?

Thank you very much for your help.

[urosgrivc]

O my, what is going on :O

I am geting the same results myself :/ https://drive.google.com/open?id=0Bw...C1yYUh5OFJ2SW8
now this is wierd

before, I ment translational periodicity made vith interface model vhich i have used. But I am confused now I vould expect linear flow even with normal simetry.
I have tightly converged it and i get a similar result for a frozen rotor (vithout enithing in it) simulation as mentioned in previous posts.

Experts please what is going on?

[cfxbeginner]

Thank you urosgrivc to try it.

I think I understand one of my problems but don't know how to solve it.

The mesh for the cylinder (rotating domain) is not at the good place. Indeed, my mesh for the back and the front of this domain are perfect but the cylinder (which is a part of the interface) seems to be too approximate. There is the same problem for the stationnary domain too.

I don't know if it is due to ICEM approximations (between curves and surfs) or my mesh method ....

Urosgrivc : Can you check if you have the same problem please?

Here is a picture of the cylinder for the rotating domain. (In green it's the curve : What I want, in brown-red : it's the cylinder surf, in orange (lines) : the mesh)

Regards,

interface_problem.PNG

interface_problem_2.PNG

[ghorrocks]

I know CFX has a test case they run for all builds which includes flow through a rotating domain, to check the domain rotation does not deflect the flow. So the fact that this case is deflecting the flow suggests something is being missed in this case. I can't think of what right now, all my other ideas did not explain it.

[urosgrivc]

No, my mesh matches node to node exactly, i dint use icem.

Could it be the graphics problem in your case? becouse circle is not round it is a poligon maybe.

[cfxbeginner]

I don't think that it's my problem as we exactly have the same result, the GGI seems to solve this problem. We probably forgot something or don't have the good boundary conditions. Hope someone knows here.

[ghorrocks]

Where is the plan you are using to plot the vectors?

Also please show the results of the run with 10 elements thick.

[urosgrivc]

My result plane is in the midle of the (curently one element thick) domain.
I will try a bunch of things and report more.