GMNA reference solution: LPF = 1.2. Regarding the imperfection definition, there are no conclusive definitions for the amplitude, but the GMNIA reference LPF is about 1.0.
Can you share your model, the *.inp would be enough.
Yeah, this one (it was necessary to release rotational DOFs to make contact converge even though it worked in older releases of CalculiX): Contact of two shell pipes - no convergence in newer ccx releases - CalculiX (official versions are on www.calculix.de, the official GitHub repository is at https://github.com/Dhondtguido/CalculiX).
I often have warnings that someone’s file was created in another version (fortunately, it says which one) and either it doesn’t open at all if the versions is very different, opens with some issues due to incompatiboe features or just opens fully. Maybe I had a case that the file was created in the same version I was using to open it (Matej still provides old versions so you can often adapt to that) and there was still an issue, but it’s very rare.
Was it a long time ago ?
GMNA=1.17. I can only estimate it measuring in a pdf. GMNA is one of the values on which there was most agreement among the different analysts.
I can’t find it.
I think there are.
I hope you could still review your submission and help CCX/Prepomax receive a fair overall review in the RR3 Tower Finite Element Benchmark (Part 3 – Software Survey).
The transfer includes the GMNA setup; MNA can be obtained by simply removing the NLGEOM option.
Will be available three days.
EDITED: can–> Could. Sorry about my poor English. I forgot can is rude 
.
3 Likes
crazy, now it also works on my full 3D Solid model, I just removed the rigid body constraint from the bottom, thats it nothing else, I will update my Report on the RR3 and inform Dr. Sadowski that I need a few more days for my Report. Will also upload a video regarding this analysis to my Youtube Channel.
4 Likes
Yeah, it’s just that nasty “rigid body + Nlgeom + shells = non-convergence” limitation of CalculiX. It’s explained by the CalculiX dev here: Rigid-body constraint convergance problems - #25 by dhondt - CalculiX (official versions are on www.calculix.de, the official GitHub repository is at https://github.com/Dhondtguido/CalculiX).
Here you can find a list of various CalculiX limitations with workarounds: Known CalculiX limitations
Ooh, I missed the timing on this one! Congratulations everyone involved, it is always nice to see benchmarks being succesfully tested and even better when CCX gives good results.
When I was using data from old NACA reports, I discovered a good tool to extract more or less precise values from printed plots: PlotDigitizer Online App. Might be useful 
Best regards,
Lucas
1 Like
Personally, I prefer desktop apps, and so I use open-source Plot Digitizer for that: https://plotdigitizer.sourceforge.net/
2 Likes
This discussion is really very interesting.
I have just finished a wonderful course on GMNIA and seeing that PrePoMax can also perform these analyses (with a few minor adjustments and limitations, of course) really surprised me.
Let’s hope that the arch-length method will also be introduced.
I would also be curious to know why Calculix does not support 2D finite elements (I do not consider Calculix’s extrusion to be a solution, as I use 1D and 2D elements to reduce computation and therefore I lose this advantage with extrusion).
I find it really “ peculiar” that I can perform crazy analyses (GMNIA, contact wear, Herz, fatigue, snap-fit, hyper-elastic material, etc.), but I have not implemented 1D and 2D elements, which are the most “dumb” finite elements to use.
Anyway, I am joining the discussion to find out what imperfection value was used (and what deformation was used) and to congratulate you all on the results obtained. It was nice to see how you got around the problems with Calculix to achieve such a good result.
By Enterfea ?
This was planned by CalculiX dev a few years ago (it’s still mentioned here: CalculiX: Preview of scheduled finite element capabilities of CalculiX), but no sign of it yet.
Even though they are internally converted to a single layer of equivalent wedge or hex solid elements, they are still helpful. They provide sufficient accuracy in many cases and are easy to define. Their main disadvantage is that the knot generation is nonlinear and may cause nonconvergence in combination with some other nonlinearities and constraints.
In fact, CalculiX has one true shell user element, but it’s very limited. Hopefully, more such elements will be added in the future.
This whole topic could be a nice tutorial-subject…
I will create a Tutorial regarding this example Tower for my YouTube Channel this month.
3 Likes
Of course, I learned about it from one of your posts here on the forum (I don’t write much, but I visit often).
In my work, I use an approach very similar to that of your tutorials (the 3D model is nice to look at, but I think it’s essential to start with an analytical method or a simplified 2D model), so let’s say I have a very high opinion of your work and your advice. I decided to try the Eurocode course and fell in love with it, so I moved on to the non-linearity course (which was even cooler). He is a really great teacher. You know, courses are often either too theoretical or too practical, but his (and yours too) are perfect.
I am aware of that. One of the “problems” with open-source software is that it is very dependent on the opinions of those who write the code, who, having very limited time, cannot do everything. I imagine that Guido believes that there are other priorities, and that’s fine. The solver is incredible anyway and it’s also mathematically very robust (I recently discovered that in many ways it is better than Code Aster, which surprised me), and with a Pre and Post processor like PrePoMax, I am receiving on my desk more and more jobs done with Calculix.
It’s just strange that it doesn’t have real support for 2D elements.
Can you share your Youtube channel? isn’t that the one with the orangutans? 
If so, the Oceangate video is very, very interesting. I saw it some time ago and, if I’m not mistaken, PrePoMax appears at one point, although I think FEM was created with different software (or am I wrong?).
I am waiting for information on imperfection values (I am recreating the test myself).
Don’t wait, finish the analysis using everything you’ve learned, and then we can see if we’re on the same page. It’ll be more interesting.
my YouTube Channel is here:
1 Like
the problem of rigid body and shell element also been reported in CalculiX forum, out of plane rotational degrees not restrained as Abaqus does. Defining manually in user local coordinate system may solve the problems, an elegant way maybe instead of additional rigid part by material properties.
*edited
probably, i have to expect too much. Trying to keep edge faces to stay plan cannot be succeed, solver divergences early. Latest CalculiX solver v2.23 clearly shown the limitation, it’s aborting this boundary condition setup.
how did you apply the imperfection field to your model? it seems not to work for me, in Buckle Step I export the deformed (File → Export → Deformed Mesh inp) and in FE Model → Model → Tools → Update Nodal Coordinate From File but nothing happens, seems to have something to do with being a hybrid mesh because with only shells or solids it works fine, how do you do it?
I also tried the “Thicken Shell Mesh” Feature now but I can only select the whole shell model and not individual section? how does this work? I just want to apply thicken shell mesh to the upper flange of the tower
*ERROR in gen3dmpc: nodes belonging to
shell elements must not be
subject to a RIGID MPC.
Great, we have a major regression with respect to previous CalculiX versions - now in ccx 2.23 it’s not possible to use rigid body constraint with shells even if analysis doesn’t have Nlgeom.
I guess I’ll have to talk with Guido about it.