XANSYS Message: 4369 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Rate item:

Subject: Re: Tet meshing in FLOTRAN
Author: Janet Wolf
Date: 1999-09-01 09:48:00

Brad--

5.6 will have tet to hex mesh capabilities, and I've already tested it on
the beta version. The only problem was, I got a warning that pyramid
elements shouldn't be used at a wall; I haven't gone far enough to see if it
REALLY is a problem.

Janet Wolf
ANSYS/Houston

> -----Original Message-----
> From: Brad Lamirand [mailto:blamirand@t...]
> Sent: Wednesday, September 01, 1999 6:35 AM
> To: 'xansys@o...'
> Subject: RE: [xansys] Tet meshing in FLOTRAN

> Stefano,

> I've run several FLUID142 tet meshed problems with varying
> success, although sometimes I was more interested in "trending"
> results to see what happens between differing geometry than
> getting the exact answer. When I NEEDED to get the best answer
> I could, I would take the extra time (sometimes a considerable
> amount of time) to brick mesh. When I needed to do just a
> "quickie", I would try to get away with a tet mesh.

> The real trick to getting a tet mesh to work in Flotran is
> controlling your mesh density and change in aspect ration as the
> mesh transitions away from the walls. This can be almost near
> impossible. The first thing that I always recommended was to mix
> and match your ESIZE,value with your SMRTSIZ,value to get fine
> mesh where it matters and a courser mesh where it don't. Also,
> the layer meshing in 5.5 goes a long way to helping, although it
> makes LOTS of elements.

> The other problem, probably even tougher to overcome, is how many
> more elements a tet meshed problem will take to converge to a
> decent solution compared to a brick mesh. Although Flotran only
> solves for one DOF at a time instead of all seven, it is still
> very easy to get over a million nodes/DOF problem which is of a
> degree of nonlinearity that even the contact analysis pros would
> shudder at. The general rule of thumb that I like to use as a
> first guess is that a tet meshed problem will take about (7)
> times more elements than a brick element meshed problem, although
> this is very dependant on how many boundaries are in your problem.

> Lastly, I have been having very good luck with "sweep" meshing
> volumes; i.e.- quad meshing an area and sweeping that into a
> volume. Areas can almost always be entirely quad meshed (99% of
> the time), and then the real trick is how to project that 2D mesh into 3D.

> ON A SIDE NOTE - HAS ANYONE HEARD IF 5.6 WILL HAVE TET TO HEX
> TRANSITION OF FLUID142 ELEMENTS?

> Hope this helps,

> Brad Lamirand
> Cooper Turbocompressor, Inc.
> Sr. Analytical Engineer
> 3101 Broadway St.
> Buffalo, NY 14225
> Phone: (716) 891-3764
> FAX: (716) 896-1233
> EMAIL: blamirand@t...

> -----Original Message-----
> From: Lucarelli, Stefano [SMTP:stefano.lucarelli@d...]
> Sent: Wednesday, September 01, 1999 5:29 AM
> To: 'xansys@o...'
> Subject: [xansys] Tet meshing in FLOTRAN

> From: "Lucarelli, Stefano"

> Hi group.

> We have to perform an air cooling problem with Flotran but we have little
> experience in CFD. Geometry is quite complex and brick meshing is not
> feasible. Has someone experience with tets in such a case? Results we have
> got seem meaningless.
> We performed some benchmarks on simpler geometries, and we found that
> results coming from tet meshes are very bad; brick meshes, even
> if coarser,
> perform better.
> Any suggestion would be appreciated.

> Thanks in advance,
> Stefano

> Stefano Lucarelli
> Mannesmann VDO AG
> Entwicklung Informationssysteme NFZ
> IC43LE/S
> Tel.: 07721-67-2214
> email: stefano.lucarelli@d...

Posts possibly associated with message #4369AuthorDateScore
4364Tet meshing in FLOTRANStefano Lucarelli1999/09/01 
4365Re: Tet meshing in FLOTRANBrad Lamirand1999/09/01 
4367Re: Tet meshing in FLOTRANRoger Yang1999/09/01 
4369Re: Tet meshing in FLOTRANJanet Wolf1999/09/01 
4371Re: Tet meshing in FLOTRANBrad Lamirand1999/09/01 
4372Re: Tet meshing in FLOTRANDavid Power1999/09/01 
4373Re: Tet meshing in FLOTRANChris Kennedy1999/09/01 
4382Re: Tet meshing in FLOTRANBill Bulat1999/09/01 
4388Re: Tet meshing in FLOTRANRick Morgans1999/09/01