Hello, this seems similar to the post a couple below this one except that I'm not getting a nice visible little error code. I started off having issues with the auto surface refinement taking forever to run. So I decided to make a region around the area that I know will cause issues. This caused an "out of memory" error upon surface smoothing. Now I've added a second region located elsewhere on the model in an attempt to decrease the total number of elements. Unfortunately, now it won't even get to surface smoothing before the solver crashes.
I've attempted to attach the cfz file, but I doubt it'll go through as it's 240MB.
Computer has 24 Gb of RAM and over 300Gb of free hard-drive space.
Solved! Go to Solution.
Solved by Jon.Wilde. Go to Solution.
Hi,
Can I ask why you think a particular region will cause meshing issues? Have you considered fixing this in CAD rather than increasing the mesh count?
the total volume is 3m by 3m by 2m and I have a section that's 1.4m x 1.4m that has 6mm holes in it. When I analysed this section on its own (previously) the gap refinement kept bugging me until I set the smallest possible element to 7e-8. Even then there were still a few corners that didn't get meshed correctly.
This section can't be simplified because one of the 2 purposes of this analysis is to determine whether they do a good enough job at straightening the flow or not.
Hi,
This sort of model should be fine, so long as you are working within the memory confines of the machine. What is your estimated element count? You should be OK to mesh up to about 12-15m elements.
Are you using automatic meshing, with surface and gap refinement on?
We could take a look at the model if you can share a cfz.
According to the automatic tools dialog box it's over 600 million elements. I can't get the 2nd region to accept a value above 5 or so.
I began with automatic meshing. Then I turned on surface refinement and waited 2 weeks. It hadn't finished by the end of that so I canceled it and decided to set a region where I knew the issues would occur. Then I added another region in an attempt to force some of the elements to maintain a larger size (and thus reduce total element count).
I'd like to share it but the cfz file is 240MB and this forum system doesn't seem to like files that large.
Wow, 2 weeks. You must be very patient!
Typically this should not take more than a few minutes, if it does, something is wrong.
How about I set up a Dropbox folder, happy to pm you a link, so that you can share the cfz.
Cheers,
Jon
Hi,
I sent one over this morning 🙂
hm, then how do I get to the PMs? (it hasn't sent stuff to my e-mail address in a long time)
Edit:
oops, typo. Could you please re-send the link? Sorry for the inconvenience.
Ah, I sent you an invite directly from Dropbox to your email, maybe check your junk email from this morning.
Please could you also share the native CAD geometry?
Thanks!
Jon
it turns out having a student account and a regular account does funky stuff with email addresses. It has now been updated to one that will actually get through to me. Could you please re-send it? (it should be showing a @gatech.edu address) Apologies again.
Sure, i'll add that to the upload.
Hi Colin,
I have taken a look at this. I can see the issue.
I doubt very much that we could even run this model as it is, there is far too much detail in there with the numerous holes in the filter. It would require such an inordinate amount of PC power.
The best suggestion I can make is that we replace that part with a solid in CAD and then make it a resistance in CFD.
These are customisable parts that act like filters with none of the detail, so we would still have the same pressure drop over it, but we just would not see the localised velocity effects.
How does that sound?
Kind regards,
Jon
Hi Colin,
This is exactly what the resistance materials are for 🙂
1) If you set this up using a Free Area Ratio in the main flow direction 9based on the ratio of holes/solid) and a high (1e8) constant resistance in the other 2 directions then this would be fairly representative. The flow would be straightened as you would expect - it would be uniform over the area though.
2) Yes, this would be very representative.
Bear in mind that you must apply a good uniform mesh to the part - at least 3-4 elements from inlet to outlet.
I would have thought this is the best/only approach to take here.
Kind regards,
Jon
Can't find what you're looking for? Ask the community or share your knowledge.