[mecarlg]

Hi All,

I'm having major problems with Ansys Mesh and some highly skewed elements (0.9999). I really need some help!

The geometry is a jet-powered vehicle which is fairly complicated but it is extremely mesh friendly i.e. no slither faces or tiny gaps. I have tried both a tetrahedral mesh and the cutcell method with appropriate face sizing throughout.

Both meshes look fine, however there are some highly skewed elements which appear to be near faces but not on them. The cutcell mesh has 4 million elements and the separate tetrahedral mesh has 10 million. There are only about 25 elements with skew in the range 0.9-1.0 however one element is at 0.9999. There seems to be no way of controlling this...

I have tried changing the grid density, relevance centre but having no luck. I am using Ansys Mesh version 14.0.0. Simulations run in Fluent 14.0.0 just diverge.

I have tried pressure and density based solvers, lowering under-relaxation factors, changing discretisation schemes and even used very small air velocities through the domain - all diverge.

Any suggestions or advice? I never ever had these problem with Gambit but this new generation of software is a constant source of frustration! I am really tearing my hair out so would appreciate any advice.

Many thanks,

Carlos.

[taxalian]

Hi Carlos,
I don't know about Ansys Mesher (did you mean ICEM CFD?).
Anyway you can may be try to perfom grid smooth/swap operation in Ansys Fluent. If you go to Grid Menu --> first do grid check, if the mesh check succeeds. Then Proceed to Grid Menu --> smooth/swap operations.
This might help not sure!
Can you still do not use Gambit to do your meshing or it is no more available in Ansy v14.0.
Or you could try to find some advanced settings to tweak your mesh in the new Ansys Mesher.

Good luck and regards.

[Far]

I would suggest you to make surface mesh in Gambit and volume mesh (tetra prism) in ICEM.

Gambit will give you very good surface meshing and manual control over every thing.

[mecarlg]

Originally Posted by taxalian

Hi Carlos,
I don't know about Ansys Mesher (did you mean ICEM CFD?).
Anyway you can may be try to perfom grid smooth/swap operation in Ansys Fluent. If you go to Grid Menu --> first do grid check, if the mesh check succeeds. Then Proceed to Grid Menu --> smooth/swap operations.
This might help not sure!
Can you still do not use Gambit to do your meshing or it is no more available in Ansy v14.0.
Or you could try to find some advanced settings to tweak your mesh in the new Ansys Mesher.

Good luck and regards.

Hi Taxalian,

Thanks for the reply. I don't think smooth/swap does much with elements which are so highly skewed but I will try when in the office tomorrow. I will try the polyhedral "convert skewed elements" function which may aglomerate the nasty cells. Unfortunately Gambit was withdrawn by ANSYS for academic customers so we HAVE to use their software. I don't know how to use ICEM and not enough time to learn now so need to get ANSYS Mesh working.

Thanks for tips though, will try them tomorrow.

Cheers,

Carlos.

[mecarlg]

Originally Posted by Far

I would suggest you to make surface mesh in Gambit and volume mesh (tetra prism) in ICEM.

Gambit will give you very good surface meshing and manual control over every thing.

Hi Far,

Thanks for the advice. Unfortunately we don't have licenses for Gambit so I can't do the surface mesh there. Also not sure how to use ICEM and I know it is difficult. There isn't enough time for this so I will try looking at advanced options in Ansys Mesh as Taxalian suggests.

If anybody know a magic "recipe" of settings in Ansys Mesh please let me know!

Cheers,

Carlos.

[foamcfd]

Hi Carlos,
Few elements with poor quality shouldn't be a problem. I ran simulation succcessfully with following approach
(1) Try to convert skewed cells to poly in fluent
(2)Switch off secondary gradients
(3)Use lower value of temperature explicit URF (not implicit URF from GUI). It can be done using TUI. Let me know if you don't know how to do that.
(4) If it is natural convection then you can try reversing the URF for pressure and mom (i.e use 0.7 0.3 instead of default 0.7 0.3)
(5) Use node based gradient method

This should work.