Number of nodes limitation - Calculix error 201

[Jckcdt]

@Jckcdt: Same problem or do You have a different outcome?

Cheers,

PJ

Hello,

Even if it's not related to the problem presented in the topic, I had to deal with this one too. For my experience, when you see these green dots, its in fact a matter of jacobian and mesh refinement.

Jonathan

[spiderc3]

Hello Bernd, hello Jckcdt,

thank You for Your reply and the kind offer to have a look at the geometry. You are right, the mesh was not good. An extruded mesh worked a lot better and the results are faster! Thank You again!
Cheers,

PJ

[bernd]

how did you create this quad mesh?

[aomanchuria]

I'm going to guess that he took the model into gmesh and then using the profile he used extrude, recombine and then... I forgot the details, but I think I need to do something like this for one of my components.

[NYScott]

Back to the original topic:
I am also having an issue with the CalculiX 3221225477 Error. I am running v0.19 but I see there is now a later Pre-release. I'm running:
FreeCAD_0.19.22960-Win-Conda_vc14.x-x86_64

I have experience with other FEA Analysis tools but have only recently started using the FEM tool in FreeCAD. I'm already using a reduced size part, a "very coarse" Netgen mesh and without second order nodes. In other software I would normally:
- Associate a different size mesh to specific parts of the model
- Split the model on a line of symmetry and use a frictionless /sliding surface contact (allows motion of a surface but only on the plane of the constraint).
I haven't figured out how to do either of those things in FreeCAD yet.

The links to the dll's, which someone mentioned in an earlier post, no longer seem to be active links. My next plan step was to see if the dll's exist in the version I'm running and if not I've already downloaded FreeCAD_0.19.23578-Win-Conda_vc14.x-x86_64 but I haven't tried to switch to it yet.

I decided to post because it seemed someone thought this problem might be unique and second I know of potential set-up solutions, above, if they are also possible within FreeCAD.

Any advice is appreciated,
Thank you,
Scott

[NYScott]

Hello,
I reduced my model size further and I am still getting the 3221225477 error. I thought I would add a capture which shows the mesh settings. If the limitation is 350 000 nodes as I think I recall reading I am getting closer at least...

If I could split the model along the center of the cylinder and protrusions using a frictionless sliding plane on that surface I believe would still yield valid results. I just don't see a sliding surface constraint. There is a displacement constraint so possibly a sliding surface can be created by specifying zero displacement but ONLY in one axis....

Again any suggestions are welcome.
Thank you,
Scott

[NYScott]

I decided to “save as” in order to experiment but while retaining my original file/version. I saved the file as a different name and I reduced the protrusions by 1. When I ran the Netgen mesh routine I received the following warning which prevented going further:
FemMeshShapeNetgenObject::execute(): Access violation

I had just reduced from 14 to 10 protrusions within the same file and without issue so this time, reducing from 10 to 9, I thought perhaps access was locked due to a temporary file saved for auto recovery. I opened another copy of FreeCAD, same version, and I cleared one auto recovered file and I tried again but received the same message. I did search prior to clearing and the only instance listed on my C: drive when I searched for “fc_recovery_file” was a single file and from a prior day.

The model is parametric so I decided to simply update it. I went back to the file I used as the base for the "save as". During this process I've I learned a few things I thought I would share:
1) It meshed perfectly so I suspect the new file was still related to the old file and the older file was blocking access to the newer version hence “Access violation”.
2) The Analysis updates every time a parametric geometry change is made and most times after you make a change, especially changing the protrusions in an array, the constraints shift to nonsense faces because the numbering changes each time you change the geometry. In my case it delays each step by around 10-15minutes. To avoid this delete the analysis before making any changes; bite the bullet and re-create the analysis again. It made me wonder if there was a way to simply force the analysis to deactivate and not auto-recalculate. The combination of recalculate and auto-recovery with my model can take significant time as I mentioned; 10-15minutes after simple adjustments which normally take seconds.
3) The protrusions are the smallest mesh, most node intense, but removing one protrusion reduced the number of nodes by roughly 5000 and I was about 18400 over what I understand to be the limit of 350 000.

My computer is a relatively weak machine for this but I had a friend run the same analysis when I was getting the 3221225477 error and her machine with 32GB RAM i7 @ 2.60 GHz "ZBook" took almost exactly the same time as my machine took with 8GB RAM i5 @1.70 GHz 1.90GHz. Both were around 90 minutes.

I’m using the “Default” solver and I did see that in the preferences you can increase the number of processors but only if you use “spooles”… not sure what “spooles” is.

I also read in a thread from 2015 that auto recovery could be switched off. In my case I think the auto recovery function is roughly doubling the time it takes while the analysis is active. So I wish I could either disable the FEM auto-recalculate or if not possible the auto recovery. This is the thread which lead me to believe auto-recovery can be turned off:
https://forum.freecadweb.org/viewtopic.php?t=12345

Scott

[NYScott]

I found where 350 000 nodes was referenced and they made a point to say approximately. I was successful in running 363 325!

Possibly there are already existing solutions but to summarize things which would have helped me:
- 1) Ability to turn the Analysis only auto-recalculation OFF (Helpful to have the geometry only update so you can see your change)
- 2) Ability to turn off the auto-recovery function
- 3) Adding a Sliding / frictionless surface constraint which would have allowed splitting the model in half on a symmetric plane
- 4) I didn't mention this but I struggled to add an internal surface pressure constraint. If you literally cannot see the surface you cannot associated it as I know. I thought possibly I could do it with the "wireframe" view but I could not because faces cannot be "touched" in the wireframe view. I thought I could make one of the endcaps, temporarily, less than 360 degrees but the surfaces change number once you restore the full 360 degree geometry so that does not work unless you were extremely lucky. I was able to load through an opening in the end of one of the protrusions and I reversed the pressure on the endcaps. Adding a frictionless surface constraint would solve this issue as well (open access on that surface).

Notable learnings:
a) Running a "saved-as" file with a new name seems to maintain some reference to the prior geometry so you may see an "Access Violation" message.
b) Very coarse mesh in Netgen and without second order nodes still results in what seems like a very useable mesh
c) This was my most complicated models for FEM/FEA and I had trouble initially with fillets I added. In some programs not having a radius would have caused a big issue but FreeCAD handled it well in my opinion.
d) The CalculiX 3221225477 error was avoided with 363 325 Nodes.

If anyone has a suggestion regarding the 1-4 items especially I would still like to expand what I am able to do.

Best regards,
Scott

[Gimbal]

I only want to add that I stumbled upon the same CalculiX 3221225477 Error the other day. I upped the number of nodes just to give the computer something to chew on but the only result was this error. The number did not fit inside a signed 4 byte int.
A powerful PC hits this limit within a few minutes so that is a bummer.

[NYScott]

Gimbal,
Did you check to see how many nodes you have? It's listed in the detail of the mesh. I agree it seems like a low limit but if you haven't tried unchecking "secondary" and if you reduce the mesh "Very coarse" seemed to produce a much finer mesh than I expected and the results seem valid.

My biggest problem now is I am seeing the "Access violation" again. I wanted to run several variations of the same basic model(common with FEA to evaluate changes based on the original). Partially due to the time the file takes to make even simple changes I also wanted to keep all of the cases modeled (mainly if I wanted to change the loading or constraints it would be easier to revise without recreating the model).

In a prior post I suggested the issue was related to a saving the model as a different file name. I thought that was the issue so I copied the model within the same file and it worked fine until the second copy (first copy plus original was fine but the error occurred while running the second copy/third case analysis). My plan is to create a smaller model to see if the copies are really the issue because with the auto-save and or auto-recalculate this larger model takes a long time to even make simple geometry changes. I would like to turn-off the auto-save but I don't see it in "preferences" where it apparently existed circa 2015.

Scott