CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [GAMBIT] Tunnel bend - neat Hex/Prism volume mesh through Cooper scheme (http://www.cfd-online.com/Forums/ansys-meshing/118712-tunnel-bend-neat-hex-prism-volume-mesh-through-cooper-scheme.html)

A_Prakash June 3, 2013 04:19

Tunnel bend - neat Hex/Prism volume mesh through Cooper scheme
 
2 Attachment(s)
Hello folks,
Attached pics (T1-geo-intro and T2-query) describe my meshing problem.
I have no problem with generating HEX submap mesh for the tunnel. Its at the bends where I face a problem... all thanks to the cable tray!
To keep things simple, I am now meshing the tunnel bend with Tetra TGrid while the straight portions are HEX. My worst element is 0.95 skew - due to tetra meshing. I wish to get better at meshing and would like a 'prettier' hex mesh in the bend portions too.
Can you guys take a look at the attached pics and suggest possible options for hex volume meshing for the bend portion? My attempts with various iterations of Cooper have failed.

Attachment 22368

Attachment 22369

-mAx- June 3, 2013 04:35

1 Attachment(s)
I would try to split differently as what you did.
Thus the triangle based shape, will give you skewness in the sharp angle (I assume)
For preventing that, you could offset your split in your stright portion (check the yellow surface)
Attachment 22370

A_Prakash June 3, 2013 05:41

Thank you MAX.
I agree with you. That would be a better thing to do. It will reduce the skewness of elements in the sharp angle.

Also, do you feel that Cooper would be successful on the volume?? From what I have tried so far, the meshing fails when the algorithm comes to cabletray as it tries to make intervals in Y direction. If there was no cabletray... Cooper scheme works perfect.

The other option is to delete the bend volume. just keep the bottom face mesh and start sweeping in +Y in increments of cable tray with mesh option switched on. But, that's too much work... and calls for too much caffeine :D

-mAx- June 3, 2013 06:16

you may try to split also on cable level.
Each time you have a cable, split along its lower level, and on its higher lever.
Doing this, you will subdivise your domain along Y , and you will have more "layer" and it should fix your issue

A_Prakash June 9, 2013 06:06

Thank you , MAX. What you propose is the final step in solving this issue.
However, I just went back to my geo, and while doing the splits I realised that top quadrant cable trays and bottom corner cable trays are seperated by a very very small distance after taking face projections. But, nonetheless, I will save myself the trouble and go with TET mesh at bends and neat HEX mesh for rest of the tunnel. The results are logical and deadzone is not exaggerated too much using TETs there.
Thanks to you, I learned something new.
However, one quick question: What is the implication in using combination of TET and HEX in terms of solver prediction in say CFX or Fluent. Will there be too much diffusion/inaccuracy when solver suddenly encounters TET elements in a predominantly HEX domain? My peak equation residuals are of the order 10^-2 while the solution itself is converging perfectly. (though I need to double check if peak residuals are exclusively in TET mesh zone... which would mean that there is numerical issue due to TET mesh... nonetheless, for another tunnel I simulated this was not the case..). Thanks again. Cheers!

-mAx- June 10, 2013 07:05

With tetra, you should take care of skewed elements (the max should be around 0.9-0.95)
I am used to work with hex-tetra combination, and have good results
You can also increase hexa ratio, by using tet/hexcore schema (Gambit will insert a core of hexa in your volume)

A_Prakash June 10, 2013 09:03

Yes, skewness is the thing to be aware of. I always preview the worst element and its quality as routine best practice. I read about hexcore in manual but never tried to use it... will try it. Thanks!!

A_Prakash July 9, 2013 04:54

Hello Max,
I have a question.
I am importing several thousand edges from an IGES file. After using the basic import/geometry repair option I am still left with hundreds of short edges which are absolutely tiny.
Is it possible to write a small DO loop or piece of code to search and delete edges that are less than a particular length??
I tried "Clean up Short Edges" from geometry repair menu..yet there are edges on which GAMBIT says it has failed to operate successfully at the end of the clean-up process.

-mAx- July 9, 2013 05:04

Are there not a tolerance option in IGES import?
Else I don't know since I am more experienced with step file (which imports direct a volume).
Dumb question: the iges import gives you surfaces, right?
If you deselect all surfaces in visibility, then you should see all edges and vertices which don't belong to any surface. And then you may select and delete them.

A_Prakash July 9, 2013 05:14

IGES can import vertices, edges, surfaces...all entities basically. I usually use to get my edges from AutoCAD, convert it into IGES and bring it into Gambit. But, this time around I have a too much details in the import and I am not too advanced in AutoCAD.
But your remarks about surface only import has given me an idea... I will get to work right now and see if I can make it work. I will update you. Cheers!

A_Prakash July 9, 2013 08:46

Okay, that 'idea' I just mentioned earlier did not work.
Max, I am using IGES to import edges only. Heal and Make Tolerant options during import manage to filter out edges below 10^-3 unit tolerance (which is default).
My problem has to do with very small edges less than 1 unit, because that is the lowest edge length resolution in geometry that I care about.
I have finally gotten around to doing the process manually with reasonable efficiency by using face create (polygon option)however, I am still curious if I can write a DO loop/code to search and delete edges below a user-specified value. Or maybe there is a way to change the default import tolerance... *there I go! another idea to try* LOL

-mAx- July 9, 2013 08:55

why are you only importing edges?
according to the help, you can prevent importing standalone entities like vertices, edges or surfaces
So if you can import surfaces, I would enable "no standalone vertices" & "no standalone edges"
http://aerojet.engr.ucdavis.edu/gamb...le_import_iges

A_Prakash July 9, 2013 09:35

I have a number of intricate buildings (250 buildings) to simulate for wind analysis in a 1 km by 1km block. This is a learning project.. different from what I usually do. The input data is in AutoCAD (as a top view). The buildings are of variable height. The AutoCAD file has grouped buildings of particular height by a particular color. I have managed to easily remove the hatches and generate building boundaries.
However, I am unable to get AutoCAD to remove insignificant small/sharp corners in the building buondaries (example: a corner with one edge of 0.75 m by 0.01 m meeting at 90 degree angle). If these edges were standalone then I could eliminate them right away as you mentioned (by the way that option has no use in IGES import because IGES imports everything as unconnected you need to do T-connect first thing after import ia completed or opt for virtual clean-up).

A_Prakash July 9, 2013 11:37

Max, my initial query about writing a code for Gambit to filter out edges below certain length has been resolved. Since I work in AutoCAD before taking my geo to Gambit, I figured out that there is a pretty versatile Quick Select option in AutoCAD which can help me accomplish just that and much more... So I think I can call it a day...finally!!


All times are GMT -4. The time now is 16:53.