CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)

 Anna Tian August 10, 2012 16:43

Question about coarsening the mesh to improve convergence

Hi

I have vortex at some domains of my case. The FAQ suggests me to coarsen the mesh to improve the convergence. But that will decrease the accuracy. Is there any other ways to improve the convergence for this problem?

 sainath.s@tdps.co.in August 10, 2012 23:56

Only way you are going to capture vortices is through a fine mesh. Let me know what you are trying to model. Let me see if i can helpyou out

 Anna Tian August 11, 2012 05:40

Quote:
 Originally Posted by sainath.s@tdps.co.in (Post 376529) Only way you are going to capture vortices is through a fine mesh. Let me know what you are trying to model. Let me see if i can helpyou out
Sorry, I'm not allowed to reveal that geometry. But do you mean that I can solve this problem by refining the mesh and match the mesh with the fluid behavior?

 sainath.s@tdps.co.in August 11, 2012 06:31

If you are planning to capture vortex formation and shedding you need to have a fine mesh such that it is captured. Very small eddies can also be captured using sub grid scale type of modeling.

But there are constraints like you need to have a good computational power. Using a LES or DES model is going to be computational intensive. So plan accordingly. O

If your aim is something else and vortices are formed just as the side effects of the problem then you can refine the mesh only in those are and can go in with a coarse mesh in the other areas.

I really do not know what your problem is, these are just requirements for capturing vortices.

Also go through the FAQ's, you may find some more tips to find effective convergence and accuracy.

 mat_cfd August 12, 2012 01:56

It isnt always required for your residual to converge. If capturing vortex is not important then you can use monitor plots like mass flow, pressure to see if you have achieved fair enough convergence.

 ghorrocks August 12, 2012 08:03

Quote:
 The FAQ suggests me to coarsen the mesh to improve the convergence.
Which FAQ suggests that? This is only one of many suggestions in the FAQ - why do you say a coarser mesh is required in your case?

 Anna Tian August 12, 2012 13:25

Quote:
 Originally Posted by mat_cfd (Post 376633) It isnt always required for your residual to converge. If capturing vortex is not important then you can use monitor plots like mass flow, pressure to see if you have achieved fair enough convergence.
I only care about the pressure drop. I don't care about the vortex at all. But I'm wondering if I don't solve the vortex, could the simulation pressure drop results still be accurate?

 Anna Tian August 12, 2012 13:30

Quote:
 Originally Posted by ghorrocks (Post 376653) Which FAQ suggests that? This is only one of many suggestions in the FAQ - why do you say a coarser mesh is required in your case?
'
A physical instability, such as vortex shedding - the fix here is to use a larger timescale, a coarser mesh in the vortex shedding region, decrease the blend factor (if using hybrid differencing) or use a lower order turbulence model. The first option is preferred as the latter options can have accuracy implications.
'

It's from the FAQ of this forum.

 ghorrocks August 12, 2012 21:57

I wrote the FAQ so if it is unclear I would like to improve it.

My intention was to say that if the vortex is not important in the flow for the results you are interested in then you can dissipate out the vortex by coarsening that region of the mesh. But if this is an important part of the flow - for instance it adds resistance which significantly contributes to the over pressure drop in the simulation - then it is important and you should not get rid of it.

So before you coarsen that region of mesh you have to make sure that it is not significantly contributing to important flow features. Only if it is not then you can coarsen the mesh in that area to get rid of it.

 Felggv August 13, 2012 14:53

What is your problem message? What error does CFX-Solver point?

 All times are GMT -4. The time now is 10:20.