I'm using CFD 2013 to try to do an analysis on switchgear and its enclosure, however the solve quits unexpectedly after "Radiation Problem Set Up Complete." The progress bar here get to 99% and then quits. If i watch the task manager, the resources of the computer are used 100% (CPU and PF memory) until the progress bar gets to 99%, then it dips off and eventually goes down to 0% CPU and low memory usage. Not sure if that is helpful.
I have simplified the model several times to get it to work, but nothing seems to do the trick. Also, when the mesh is autosized, it uses about 6 million mesh, but our computers here supposedly can only run 2 million elements. (8 gigs of ram, running windows Xp x64), so i edit the mesh to get it close to 2 million, and then apply and spread the changes. I have no clue whats wrong or how to go about correcting it.
Thanks in advance, Austin
I tried to attach the .cfz file, but the website would not allow me.
Hi Austin,
Have you tried running without radiation?
What is your actual mesh count once meshed?
You may well be running out of memory.
Regards,
Jon
I realized today that it is failing in the "Computing View Factors" section of the solve. I have simplified the model as much as i can and am running as much mesh as my machine should be able to handle. This model is pretty complex compared to the usual CFDs that we do. Running without radiation isnt really an option because the results needed have to do with the radiation.
What could be causing it to fail at 99 percent every time when computing view factors?
Memory. How many elements do you have?
The last attempt this was the readout:
2123452 Total Elements, 1765809 Fluid Elements , 357643 Solid Elements
However, I have done it with less elements and it still failed. Somewhere aroudn the 1.4 million mark
Would you suggest trying to run it with less elements? I know the fewer i use the less acurate the analysis will become.
Assuming the model is corectly set up. It does sound like memory, which likely means a machine upgrade.
You could try an older radiation model, which should use less memory, by using this flag set to 1:
rad_model_1
Feel free to share the file too, could test it on a beefier machine here.
Kind regards,
Jon
Okay. Ill try to run with that flag set.
I'd love to see if it works on a beefier machine. How do i share the file with you? When i attach the file it says "
The contents of the attachment doesn't match its file type."
Hi Austin,
Test that flag first.
Is it fails, can you share the CFZ file? It may be too large for te forum, which I think has a 2MB limit - if so use 360 or Dropbox?
Thanks,
Jon
To create this file we need go to the File menu_Save As_Save Share file_Support type. It will have an extension of .cfz
The file is already compressed, so you do not need to zip it up.
For future reference to everyone, the .cfdst file has very little use on its own. About the only time I will open it up is if I need to determine if someone is running off a network drive which isn't a good idea.
If the file is too large to attach you could always upload it to your 360.autodesk.com account (similar to dropbox) and provide a public lnk.
The model did still fail even with the flag set. I'm not even sure if it was a memory issue this time. The PF memory was at 7.92 gigs until it dropped off again at 99%. Usually the PF memory is at like 10 or 11 gigs page filed. Have you had a chance to try it on a better machine?
Lets take this step by step.
Issue one that we need to fix first is to add inlet/outlet extensions to all the opening in the model. Also, since we don't know which is going to be inlets and outlets we need to use a film coefficient instead of temperature for all the openings. Make the lengh of all these about 2-3 x Hydraulic Diameter.
We might even want to consider putting this in a large box and doing a bucket approach analysis on this unit instead since we have a significant number of opening. Eitherway, we should never have flush inlets/outlets like you have now especially when running with heat transfer turned on. I vote for doing a bucket analysis instead. What do you think?
I just got the model to run by using around 600,000 elements. I am getting out IT guy to upgrade my computer to 24GB of RAM when he get back from corporate.
Could you explain a few of the things you've suggested to me? I just started using CFD two weeks ago when i started my co-op here. The project engineer ive been working with was puzzled by this as well.
Anyway, Why is it a bad idea to have flush openings? And usually we would put temperature and pressure on the bottom vents and then use only pressure at the top vents. Why do you suggest using a film coefficient instead of a temperature?
Also, I'm not sure that bucket approach would work (at least with my current set-up) because it would raise the mesh, which would probably cause the solve to fail in the same place, if not earlier in the meshing phase of the solve.
Just looking to get some clarification on your suggestions. I appreciate all the help thus far
Austin
Royce,
Not sure if it just slipped by since these are forums, but could i get you to clarify about the need for extensions on the openings as well as using film co-efficient instead of temperature for the openings?