CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > FLUENT

Gambit - Meshing a parallelogram

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   January 14, 2006, 06:03
Default Gambit - Meshing a parallelogram
  #1
Cyril
Guest
 
Posts: n/a
Hello ! I'm trying to mesh a parallelogram with gambit. There are two horizontal sides, but the others sides aren't vertical (55 for the left one and 40 for the other). See it ? The problem is that the 40 side is considered as horizontal side !! Result : The face mesh (quad) is horible ! (nodes in center, ...)

Is there anybody who knows how to solve this problem ??

(in Simail there is a fonction to select the kind of face... but not in Gambit.)
  Reply With Quote

Old   January 14, 2006, 07:28
Default Re: Gambit - Meshing a parallelogram
  #2
zxaar
Guest
 
Posts: n/a
disect the face into smaller faces and mesh them.
  Reply With Quote

Old   January 14, 2006, 16:06
Default Re: Gambit - Meshing a parallelogram
  #3
Freeman
Guest
 
Posts: n/a
Mmmm, I'm trying to figure it, but I don't know how you have created your geometry. I mean that the problem may be that Gambit has assigned one of the vertex which is made the right side of your parallelogram as Side vertex (you can see this if when you select the edge of the right side, a letter "S" points to the vertex that conforms that edge).

You can try to convert that vertex to "End" type: to do this, go to Mesh->Face and select the icon Set Face Vertex Types. Select your parallelogram in Face, activate End in Type menu, select the edge that represents the right side and try to remesh. If that does not help, try with other types or change the two vertex of the edge...or perhaps it is not the problem xD.

Hope this helps a little. Good luck!
  Reply With Quote

Old   January 16, 2006, 05:06
Default Re: Gambit - Meshing a parallelogram
  #4
MANOJKUMAR
Guest
 
Posts: n/a
Hello Sir, I have a problem regarding mesh. I made a container of 3700mmX2400X1800mm and in that there is a sheet of 0.672mm thick and 770 mm width.First i subtracted the sheet from container and then doing meshing then i said that small area in lagre space. Give solution to solve this problem. Any extra way to solve this problem.

  Reply With Quote

Old   January 16, 2006, 12:16
Default Re: Gambit - Meshing a parallelogram
  #5
Cyril
Guest
 
Posts: n/a
You can see the mesh on this link : http://img66.imageshack.us/img66/118/heberg3ba.jpg

As you can see, I tryed to predict the area of the shock, using the formula angle=arcsinus(1/Mo)=45.

The area around the bullet is good (the mesh indeed). But as soon as I leave the near environment, the mesh is horrible, and I don't understand why.
  Reply With Quote

Old   January 16, 2006, 14:47
Default Re: Gambit - Meshing a parallelogram
  #6
Freeman
Guest
 
Posts: n/a
Well, I don't know if you have realized one thing: if so, sorry and forget it; let's see.

I guess that you've done your "good mesh zone" (I will referer to it as the Bullet Zone BZ) with other more little quadrilateral zones (let's say them LQZ zones), don't you? Then you meshed them one by one and you used Quad & Mapping (perhaps with Thom-Mid smoothing): you could use the Quad and Mapping because every LQZ has 4 vertex and this a "logical parallelogram" (parallelogram with 4 vertex)

The mesh outside the BZ (the bad meshed zone), is also composed of a big parallelogram, but as you can see, it has more than 4 vertex: the side next to the BZ has 14 vertex or so. When you meshed that parallelogram, Gambit didn't find a "logical parallelogram" and set the Quad-Pave mesher instead the Quad-Mapping.

It is possible that if you change the vertex type to Side in these vertexes located in the side of the parallelogram that has the 14 vertex and remesh that zone, you might succeed meshing with the Quad-Mapping. To change these vertexes to Side Type, go to Mesh->Face->Set Face Vertex Types, select the parallelogram in the Face dialog box, and select in the Vertex dialog box the vertexes that are in the edge that has the 14 edges or so (but leave as End type the vertexes in the extreme of this edge!). Then remesh and tell us what happens.

Hope this was clear. Good luck!

I hope you've understand the above;
  Reply With Quote

Old   January 16, 2006, 20:08
Default Re: Gambit - Meshing a parallelogram
  #7
zxaar
Guest
 
Posts: n/a
having seen your mesh there are two three things i wish to say:
[*] Use the structured mesh as much as possible, otherwise evenif you are able to make the mesh you will not be able to converge it with solver due to presence of shock.[*] now if you wish to go with structured mesh, i would further devide the mesh in such a way that i will maximise the meshes with cartesian type elements (90 deg, and no non-orthogonality)[*] This means getting rid of this parallelogram area.[*] you would ask, why??? The reason is that in such calculations gradients play the major role in getting the solution converge. And if your faces are at angles (non-orthogonality, as you will introduce with structured parallelogram), the non-orthogonality terms of diffusion will become important. These terms are decided by gradients and in high gradient areas it will introduce so much of error that you will be asking here after two months what to do to get it converge.
  Reply With Quote

Old   January 17, 2006, 04:39
Default Re: Gambit - Meshing a parallelogram
  #8
Cyril
Guest
 
Posts: n/a
Whow ! I can see that you've understood exactly the problem ! Concerning the Vertex Types... When I select a vertice, it automatically sellect the two edges on each side (I think that's correct). But should I select the 2 vertices at the limits ? (I guess yes, if I wants to select the upper side).

Excuse me if I missunderstood something... I'm a french student and my technical english is not so good sometimes !
  Reply With Quote

Old   January 17, 2006, 05:15
Default Re: Gambit - Meshing a parallelogram
  #9
Cyril
Guest
 
Posts: n/a
Structured mesh... I've never heard of this before. What is it (structured mesh and structured parallelogram) and how to make it ?

Concerning dividing the area with cartesian type elements (and no non-orthogonality), I would be very pleased to do it! The reason why I chose non-orthogonality parallelograms, is that I wanted to improve the calculus time... (I don't need to know what happens behind the bullet, and in front of it). So I tried to mesh just around the shock wave. ()

But I see that's not the best solution... (with the current mesh, 1 iteration=690s! I can't imagine the time with a full area).

Have you got an idea to improve the calculus time ? I've got to study a 30mm shell in a 8-meter radius zone !!
  Reply With Quote

Old   January 17, 2006, 10:39
Default Re: Gambit - Meshing a parallelogram
  #10
Freeman
Guest
 
Posts: n/a
Well, I'm a spanish student and sometimes I understand better the non-technical and plain explanations than the others xD.

When you refer to the 2 vertex at the limits, are you referring to those that are in the lower edge of the parallelogram? If so, you don't; I mean that these 2 vertices must have the End type, as well as those in the upper edge of the parallelogram (this edge only has 2 vertices I think). You must have 4 End Type veertices that will enclose a "logical parallelogram" in order to let Gambit to mesh with Mapping scheme your that parallelogram; the other vertices will be set as Side.

With this configuration, Gambit should let you mesh with the Mapping option: enable the Thom-Mid smoother to have better results if you succeed. If this doesn't solve the problem, I have another idea, but try this first. Good luck!
  Reply With Quote

Old   January 17, 2006, 22:39
Default Re: Gambit - Meshing a parallelogram
  #11
zxaar
Guest
 
Posts: n/a
by structured i mean the mapping scheme.

second, if i were you this would be my approach.
[*] draw the drawing on paper[*] by pen draw lines dividing the whole thing into small portions, keeping as much as possible cells with 90deg angles.[*] then based on this, create the edges in the gambit (of course you would creat vertices first)[*] by these edges create the faces[*] mesh the whole thing with map scheme

this works fine, some time ago i helped a guy(chris) with a very similar problem. I can give you his email if you wish to contact him and learn from his experience.
  Reply With Quote

Old   January 19, 2006, 04:43
Default Re: Gambit - Meshing a parallelogram
  #12
Cyril
Guest
 
Posts: n/a
I'm trying this way. It looks that's better !!

Thanks ! I will tell the results !

P.S : (I would be very pleased to have Chris' mail if you can)
  Reply With Quote

Old   January 19, 2006, 06:38
Default Re: Gambit - Meshing a parallelogram
  #13
Cyril
Guest
 
Posts: n/a
Terrible... I can mesh everything I want now !

Thanks for this nudge in the right direction !

(the only problem now is the solving time !! about month !!! ;-)

By the way, do you know the difference between the different types of quad mesh (map, pave, ...) ?
  Reply With Quote

Old   January 19, 2006, 07:35
Default Mesh Symmetry
  #14
Cyril
Guest
 
Posts: n/a
Another point... The study is about a shell, which has a symmetry axis. Is it possible to mesh the upper side (geometry and mesh)and tell gambit to make a symetry of it ? (meshing is quite long...
  Reply With Quote

Old   January 20, 2006, 10:55
Default Re: Mesh Symmetry
  #15
Freeman
Guest
 
Posts: n/a
You could try to solve your problem telling fluent that your shell's axis is the symmetry axis of the whole problem. To do this, only create a half model and when you set the boundary conditions in gambit, enhable the edge selection (by selecting "edge" in the drop-down menu), select the symmetry axis edge and select as boundary type SYMMETRY.

Now you can go to Fluent and solve this problem as habitual. If you make plot contours, you will only see the half model, but Fluent knows that the problem is symmetric by means of the SYMMETRY edge, so when calculation is finished, go to Display->Views and select as mirror plane the SYMMETRY edge. Now, you will see the whole model ("mirrored")
  Reply With Quote

Old   January 20, 2006, 11:12
Default Re: Gambit - Meshing a parallelogram
  #16
Freeman
Guest
 
Posts: n/a
I'm happy you've solved your problem.

A quad-mapped mesh is a structured mesh that it is suitable when one needs to have the mesh aligned to the flow. The quad-pave is an unstructured meshing scheme that it is managed by gambit as the elements where triangles, but it takes advantage of being a 4 node element and, if the zone you are going to mesh is more or less quadrilateral, the mesh would result much more aligned in comparison to triangles, but allways not as aligned as the mapping sheme, as you saw... but for example, if your domain or a subdomain is triangular and you want to mesh with quad. elements, the onli way is the pave-meshing scheme.

By the way, your calculation is huge! Why do you need domain so big? If I were you, I would try to put a boundary layer with 4 or 5 layers and mesh the rest of the domain with tri elements using a size function in order not to have very small elements far to the shell (it is not necesary).

Then in Fluent, I would start first with 1st order discretization schemes and when the problem seems to be converged, I would switch to 2nd order QUICK to finish the calculation (it is "mandatory" when using tri elements). After that I would Adapt the mesh by pressure gradient option since I get the desired solution.

I hope that was clear. Good luck with your work!
  Reply With Quote

Old   January 23, 2006, 12:04
Default Re: Gambit - Meshing a parallelogram
  #17
Cyril
Guest
 
Posts: n/a
Why I need such a big envronment ??? The project is to study a bullet (or a shell), near a target. Near means between 1 and 8 meters... You understand now... ;-)
  Reply With Quote

Old   January 23, 2006, 13:24
Default Re: Gambit - Meshing a parallelogram
  #18
Freeman
Guest
 
Posts: n/a
Wops, ok!

Just a comment: I think the aerodynamical interaction that the bullet has in the first meter, will be the same than in the 8th one, isn't it? I mean than it is similar to the simulation of a road car or an airplane: you don't have to simulate all the trip, just a domain that ensures the grid would be big enough to capture the flow patterns not having any other influence over it.

I don't have many experience in high Reynolds & compressible flow aerodynamics, so if I'm wrong, please, excuse me and shed some light about it.
  Reply With Quote

Old   January 24, 2006, 04:31
Default Re: Gambit - Meshing a parallelogram
  #19
Cyril
Guest
 
Posts: n/a
I think that the pressure variation won't be the same at 1 and 8m. The shock wave detection is made by microphones, detecting a delta of presure... But I'll check this !

By the way... Thanks for your help !
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Gambit - meshing over airfoil wrapping (?) problem JFDC FLUENT 1 July 11, 2011 05:59
[ANSYS Meshing] Migrating from GAMBIT to ANSYS Meshing David-CFD ANSYS Meshing & Geometry 1 April 1, 2011 05:22
[GAMBIT] Desperate student needs help meshing 3D GAMBIT model - please help! lau06165 ANSYS Meshing & Geometry 1 March 22, 2010 02:09
Meshing problem in GAMBIT Vidya Raja FLUENT 0 May 20, 2006 23:31
Meshing Aircraft with Gambit Ed FLUENT 5 November 9, 2004 06:04


All times are GMT -4. The time now is 05:00.