CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Gambit (http://www.cfd-online.com/Forums/fluent/27379-gambit.html)

Andrew Ooi November 23, 1999 19:26

Gambit
 
I have noticed that I sometimes get messages like

"new face bf???? not used" "node bn??? not meshed"

in the transcript window of GAMBIT 1.1 (running under Win NT with 500 Meg of memory). Are these messages important ? Is it safe to just ignore them ? What do they mean ?

Thank you in advance for you help.


Michael Soso November 23, 1999 23:25

Re: Gambit
 
When exactly do you get these messages? If it only occurs during meshing, it may be that you have two faces defined by the same lines or more than one line between the two points used to define two faces in the same place. Just use the summerize lines and summerize faces button along with the middle mouse button in order to investigate if this is the case.

Andrew Ooi November 24, 1999 06:37

Re: Gambit
 
Thank you Michael for your suggestion.

I guess I should be more specific on what I am trying to do. My domain basically is a cylinder inside a big square box. I've meshed up the cylinder and is happy with the mesh. However, when I try to mesh the region between the cylinder and the square box I get error messages that say

"new face bf???? not used" "node bn??? not meshed"

If I get only a few of these (i.e. one or two), then I seem to be able to export a mesh and read it into FLUENT 5.2.3 without any major problems. If I get many error messages like the above (i.e. hundreds or thousands maybe even millions as I lost count !!) then I have problems reading the exported mesh into FLUENT. Whether I get the error messages or not is dependant on how many mesh nodes I put on the edges or the big square box before I start the meshing routine.

I've checked and the face between the cylinder and the big square box is a common face. i.e. all the faces and edges are connected.

I would really appreciate if anyone can tell me what I am doing wrong.....or rather help me understand the problem. Thank you very much.

Sincerely,

A. Ooi

Michael Soso November 24, 1999 16:13

Re: Gambit
 
I can tell you how to mesh the geometry pretty easily, but first I need to know if the end faces of the cylinder are coincident with the box or if the cylinder is smaller than the box, in which case the end faces would not be coincident.

Mike Henneke November 24, 1999 17:05

Re: Gambit
 
Your description of the problem isn't clear, but I believe these errors occur because the tet mesher in gambit (which is really tgrid) is failing. Many times this is because your edge mesh sizes are very different on adjoining edges. The procedure that tgrid uses is to first try to initialize the tet mesh. When this initialization fails (as evidenced by not meshing all the faces/boundary nodes), tgrid 'perturbs' the surface mesh and tries again (it does this several times, depending on the defaults you have set).

Another thing that will cause tgrid to fail (at least in gambit, I think the old tgrid was more robust) is when you try to create a tet mesh in a volume with a quad mesh on one face. If your cylinder is coopered or mapped, then this may be your problem. If the quad mesh has cells with high skew or large aspect ratios, the pyramiding procedure used by the tet mesher will frequently lead to failure. I don't understand this. Try adjusting the settings in the defaults. There are several having to do with growing pyramids. Reducing the 'pyramid offset' to 0.05 or lower usually leads to a valid mesh, but a poor one. I also like to change the pyramid method to centroid.

I hope some of this is relevant. I would definitely advise you to try to create your mesh entirely in hexes rather than tets. The geometry you describe sounds like it should be decomposable in such a way that mapping or coopering would be successful.

Andrew Ooi November 24, 1999 17:24

Re: Gambit
 
Thank you Michael for your interest in the problem.

The cylinder is smaller than the box and is completely inside the box.

The response posted by Mike Henneke has been very useful in helping me understand what is going on.

A. Ooi

Andrew Ooi November 24, 1999 17:41

Re: Gambit
 
Thank you Mike for your response.

Your description regarding the tet meshing in tgrid is really useful. Just one thing, you mentioned that

"tgrid 'perturbs' the surface mesh and tries again (it does this several times, depending on the defaults you have set)."

I've had a look at the defaults but could not find an option where I can set the number of times tgrid 'perturbs' the surface mesh. Could you please tell me where it is, thanks.

Your guess is correct. My inner cylinder is meshed using hexahedra and the mapping scheme and I am trying to generate a tet mesh in the region between the big box and the inner cylinder. Thank you for the advice regarding the defaults of the pyramid scheme. I'll look into that.

My domain is actually more complicated that a small cylinder inside a big box. Hex meshing will take a very long time.

Sincerely,

A. Ooi

reyman November 30, 1999 11:18

Re: Gambit
 
Andrew, perhaps it is possible to first split a face through the entire domain and remove one half of your geometry if a plane of symmetry is available. There will then be fewer elements and a nice symmetry face(s) to mesh from. A further splitting -i.e. decomposition- of the remaining portion may allow you to hex-cooper your cylinder, and then the pieces of box around it. I agree with Mike that if you can decompose the volumes enough for hexahedrals, you will be better off in the long run. Good luck.

Scott Gilmore December 2, 1999 22:02

Re: Gambit
 
Mike Henneke is correct that these messages appear when the tet mesher fails to initialize. It perturbs the boundary nodes and tries again up to times, and often succeeds on the subsequent attempts. If it successfully meshes the volume, you can ignore these warnings.

Mike is also correct that tgrid can succeed sometimes when gambit fails. This is because tgrid also swaps edges between boundary faces to reduce skewness and Delaunay violations. Gambit does not currently do this swapping, but will in a future release. Relatively few cases need such swapping.

Scott Gilmore December 2, 1999 22:05

Re: Gambit
 
Mike Henneke writes:
> If the quad mesh has cells with high skew or large aspect
> ratios, the pyramiding procedure used by the tet mesher
> will frequently lead to failure. I don't understand this.

This is because the triangles on the tops of the pyramids have very high skewnesses which can prevent tet meshing. Consider a quad face with a 10-to-1 aspect ratio. Connect opposite corners to create diagonals. This is the "top view" of the pyramid that would be created on this face. The triangles are very highly skewed. It is recommended that your quad faces bounding a region to be tet-meshed have aspect ratios of 5-to-1 or lower.

Amal Jugdeo December 7, 1999 04:21

Can Gambit do this ?
 
Is it possible to cancel an undo operation in gambit ( ie, undo an undo operation ) ?

Jonas Larsson December 7, 1999 05:06

Re: Can Gambit do this ?
 
Nope, not in the current 1.1 version. I think that it is coming in a future release though.

Scott Gilmore December 7, 1999 22:24

Re: Can Gambit do this ?
 
Gambit 1.2 has a redo function (undo an undo). It is a right-mouse-button option on the Undo button. Note that Undo has been moved in 1.2 to the lower right corner of the window so that it is now always available. 1.2 is in final preparation for release, and should be available for download later this week or early next. Contact your local Fluent office or distributor for the latest status.

You can accomplish the same thing in prior releases by opening File -> Run Journal..., clicking Current Journal, selecting the second-to-last line in the journal editor, and clicking Step.


All times are GMT -4. The time now is 17:23.