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

Gambit/Fluent: limit of nodes, edges or volumes?

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

Reply
 
LinkBack Thread Tools Display Modes
Old   September 29, 2006, 10:48
Default Gambit/Fluent: limit of nodes, edges or volumes?
  #1
Ralf Schmidt
Guest
 
Posts: n/a
Hi!

I want to create a geometry with a very large number of volumes (about 1.000.000). Is there a limit of the number of volumes that can be defined in gambit or that can be read in Fluent??

Ralf
  Reply With Quote

Old   September 29, 2006, 12:00
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #2
Jason
Guest
 
Posts: n/a
If there is, it's way over 1,000,000... I've personally meshed 12million in Gambit (with an amazing amount of patience) and run the same thing in Fluent (parallel!!!). I've heard of people running 30million in Fluent, but I don't know what they used for meshing software.

You may run into a memory limit for 1million cells if you're running a single Fluent node. For a rule of thumb, I typically use 800,000cells/1Gb of Ram using the segregated solver, and half of that for the coupled solver. Also, windows systems (and I heard it's the same for 32bit Linux/Unix, but I'm not sure about that) limits a process to 2Gb of ram, which means for a coupled solver you couldn't run more than 800,000 on a single machine. Besides, if you try running 800,000 on a single machine, you're in for a long wait.

So, yes you can mesh 1,000,000 in Gambit (depending on your system of course... you won't get that done with 128Mb of RAM, but I'm assuming you're using a reasonably high end system with 2Gb+ of RAM). You can run it in Fluent, but if you're running it on a single machine you may run into a memory limit, but even if you don't, it may take a while. So I'd recommend running in parallel.

Hope this helps, and good luck, Jason
  Reply With Quote

Old   September 29, 2006, 12:13
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #3
Ralf Schmidt
Guest
 
Posts: n/a
Hi Jason,

thx for your answer. BUT I had a different question: I don't want to mesh a volume with 1.000.000 cells, I whant to create 1.000.000 different volumes with each of it having one cell.

Ok, thats a little bit confusing, so I will explain:

I have a complex geometry, with many possible variations. Now, I don't want to model the geometry for every variation, that would be lot of work. I want seperate the volume in to a high number of smaller volumes.

Then, I will write a journal file, that defines each of the volumes as solit or fluid (depending of the geometry I want to modell) and defines the corresponding faces as wall (for solit) or internal (for fluid).

I know, that will not represent the flow correctly - the walls will not have the exact shape of the real walls. But the simulation doesn't has to be so accurate, because it is a rough estimate, what happens in the volume.

Thanks for your help

Ralf
  Reply With Quote

Old   September 29, 2006, 15:23
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #4
Jason
Guest
 
Posts: n/a
Wow... never tried anything like that, and really couldn't tell you.

Sorry Ralf, didn't notice it was you (should've recognized the name and realized you've been around here for a while).

If I had to guess, I doubt there's a limit in Gambit, but it could there could be a memory limit due to so many vertices and edges... You could write a gambit journal and try it just to see (brute force trial-and-error method... not very scientific, but it works ). Just create a million cubes (I wouldn't even bother with sizes and locations... just do the same create volume command a million times) and see what happens in Gambit.

I'm not very good when Gambit outputs a mesh though. No matter how I try to automate it, the volume labels just don't seem to match what I was looking for. I could see that as a possible problem when writing your Fluent Journal.

I've never hit a hard limit in Fluent or Gambit (other than 2 force monitors and 1 moment monitor... I still don't get why they did that instead of letting you do 3 and 3...), so I'd guess that you're not limited by any hard limit in Gambit or Fluent, but again I'm not quite sure... and I'd see a memory limit as being the concern because I'm not sure how Gambit stores mesh information vs. geometry information, but I'd think geometry information takes more memory (because the mesh information assumes straight edges and flat faces, but the geometry information can't assume that so there has to be additional information stored along with the geometry to define that).

Best of luck, Jason
  Reply With Quote

Old   September 29, 2006, 16:39
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #5
zxaar
Guest
 
Posts: n/a
I think what you need is immersed boundary method. Try searching for codes those could do it. Currently I have written a code that reads the mesh in fluent format and does the calculation (only LES,DES models). I have plans of extending it into immersed boundary method, hopefully by december I would get it to work. So if you can not solve the prob by that, check with me, you might be lucky.

  Reply With Quote

Old   September 30, 2006, 15:46
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #6
Ralf Schmidt
Guest
 
Posts: n/a
Hi!

Thanks for that idea! Do you think, that the immersed boundary method can be implemented into Fluent? That would be great!

Basically, what I need, is a code, that tells Fluent for every cell (cartesian or not) if it is solit or fluid. So, if the IIM is the right way, I'll keep on trying...

If there are other ways (like a simple udf) please tell me!

Ralf
  Reply With Quote

Old   October 2, 2006, 04:11
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #7
zxaar
Guest
 
Posts: n/a
I do not think it can be implemented in Fluent, one reason is how one would switch of the equations in the solid regions, as fluent wouldn't allow it to happen. Further, even if this is done, immersed boundary method require the interpolation of values to 'interface' cells (between fluid and solid), how it would be done. Its complecated.

By this month, I shall start to do some experimentations with my own code (related to immersed boundary method). Hopefully things would be fine, because I do not have much idea of implementing immersed boundary method. I trying to learn it.

  Reply With Quote

Old   October 2, 2006, 10:53
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #8
Abhijeet karnik
Guest
 
Posts: n/a
Hello Ralf, Jason, zxaar,

Please let me know if you can shed light on a similar issue I have. I Approximately have 10,000 volumes. Its a tunnel, with its wall made up of brick, some of which have been removed to create holes in the wall, so that the gases can pass through it and into the tunnel. So I have to split the wall with the open holes (which are approx. 540 per wall with about 14 walls), and I have more volumes to make up the geometry. Its a Furnace which i'm modeling.

Now the issue, when i split those volumes and later try to save the file Gambit gives me this error - "unable to allocate 158934873 bytes dynamic memory". I have a Dell Xeon with dual CPU's @ 3.8 Ghz and 12 GB Ram. I have increased the page file size to 10 GB.

I havent even tried to mesh the file yet. Does any one of you know what the issue could be? Or does this reflect a limit on the nodes/edges/volumes.

Thanks, Abhijeet
  Reply With Quote

Old   October 2, 2006, 10:56
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #9
Abhijeet Karnik
Guest
 
Posts: n/a
To add to my previous note, I'm using Windows XP, 64-bit.
  Reply With Quote

Old   October 3, 2006, 00:46
Default Re: Gambit/Fluent: limit of nodes, edges or volume
  #10
Sunil
Guest
 
Posts: n/a
Hi:

I am also experiencing a simillar problem. I am trying to mesh a flow volume with perforated plate which has 0.5 inch holes. The cross section at this location is around 140 X 300 inch. I am able to generate the geometry with a lot of effort, but I am having trouble meshing it. (Gambit terminated couple of times saying that free memory is less than 10%). I am using a Linux machine with 2GB memory. I wanted to ask you If increasing the memory will be of any help.

Thanks, Sunil
  Reply With Quote

Old   October 5, 2006, 01:47
Default a little update
  #11
Ralf Schmidt
Guest
 
Posts: n/a
Hi!

I started to do the simulation for a very small case (about 500 cells) and it works fine! The geometry creation journal in gambit works very fine (including setting the names), and changing a volume in fluent to solid (including the walls) works fine as well. Fluent doesn't returnes any error when simulating. It only takes a little bit longer loading the files.

BUT now, I turned to a larger domain (about 100.000 cells) and started the gambit journal file (creating 1. vertices, 2. edges, 3. faces and 4. volumes). I started it the day before yesterday and gambit is still creating the edges at the moment.

I hope, the memory will be enough. Once it happened, that the journal has been stopped, because I didn't reserved enough hard-disk space.

I will keep you updated on that.

Ralf

P.S.: I have a 2 x AMD DualCore Opteron 270 (2Ghz) Sockel 940 FSB1000, 2MB Cache with 8 x 1GB DDR RAM /PC400 ECC reg.Single Rank.

  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
snappyHexMesh won't work - zeros everywhere! sc298 OpenFOAM Native Meshers: snappyHexMesh and Others 2 March 27, 2011 21:11
snappyHexMesh aborting Tobi OpenFOAM Native Meshers: snappyHexMesh and Others 0 November 10, 2010 03:23
SnappyHexMesh for internal Flow vishwa OpenFOAM 21 February 27, 2010 03:47
external flow with snappyHexMesh chelvistero OpenFOAM 11 January 15, 2010 19:43
CFX4.3 -build analysis form Chie Min CFX 5 July 12, 2001 23:19


All times are GMT -4. The time now is 11:21.