Buckling

About the development of the FEM module/workbench.

Moderator: bernd

fandaL
Posts: 440
Joined: Thu Jul 24, 2014 8:29 am

Re: Buckling

Post by fandaL »

thschrader wrote: Wed Feb 08, 2023 4:39 pm File for testing:
DNV_B2_bucklingFrame.FCStd
I just downloaded the file, recreated the mesh (no modifications) and run it. I got buckling factors which seems ok:
6.210960
7.131222
8.551764
mode1.png
mode1.png (49.63 KiB) Viewed 1480 times
mode2.png
mode2.png (45.41 KiB) Viewed 1480 times
mode3.png
mode3.png (60.45 KiB) Viewed 1480 times

Code: Select all

OS: Windows 10 Version 2009
Word size of FreeCAD: 64-bit
Version: 0.21.30398 (Git)
Build type: Release
Branch: master
Hash: b3dfdc568a92810449bf47bc73b6784d65dfa206
Python 3.10.6, Qt 5.15.4, Coin 4.0.0, Vtk 9.1.0, OCC 7.6.3
Locale: Czech/Czech Republic (cs_CZ)
Installed mods: 
  * Defeaturing
  * nurbs
Using ccx 2.20 using 6 threads
fandaL
Posts: 440
Joined: Thu Jul 24, 2014 8:29 am

Re: Buckling

Post by fandaL »

and almost the same results with

Code: Select all

OS: Windows 10 Version 2009
Word size of FreeCAD: 64-bit
Version: 0.20.1.29410 (Git)
Build type: Release
Branch: releases/FreeCAD-0-20
Hash: f5d13554ecc7a456fb6e970568ae5c74ba727563
Python 3.8.10, Qt 5.15.2, Coin 4.0.1, Vtk 8.2.0, OCC 7.6.3
Locale: Czech/Czech Republic (cs_CZ)
Installed mods: 
  * Defeaturing
  * nurbs
which uses ccx 2.17
thschrader
Veteran
Posts: 3129
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Buckling

Post by thschrader »

fandaL wrote: Thu Feb 09, 2023 10:25 am ...
I just downloaded the file, recreated the mesh (no modifications) and run it. I got buckling factors which seems ok:
6.210960
7.131222
8.551764
Hi fandaL,
thanks for having a look at this.

I just downloaded my own file, re-run with no changes, and get the correct results (using 12 threads on intel i7).
I am feeling a little bit lost... :?

However, the stuff works.
In my posted file the 2D-meshing was set to DelQuad elements, when using Automatic, I get this:
2D_Automatic.JPG
2D_Automatic.JPG (14.96 KiB) Viewed 1444 times


@ Laurie:
I will extend the frame to the full geometry and see what happens.
Good proposal, thanks.

my_own_file.JPG
my_own_file.JPG (85.85 KiB) Viewed 1444 times
Laurie Hartley
Posts: 522
Joined: Mon Feb 27, 2017 5:33 am
Location: Australia

Re: Buckling

Post by Laurie Hartley »

thschrader wrote: Thu Feb 09, 2023 12:37 pm
I made a small video of what is going on in your model Thomas.
it is interesting to see how the top horizontal flanges and the outer vertical flanges behave as the buckling sets in in the "arch" supports.

Another thing that I don't understand is why the outer end of the frame to which the force of 76420 N (7.8 tonnes :shock:) is applied is not displaced in the vertical direction as there is no restraint to the Z axis.

https://www.dropbox.com/s/8oqvbtkb5extj ... 0.mp4?dl=0
thschrader
Veteran
Posts: 3129
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Buckling

Post by thschrader »

Laurie Hartley wrote: Sat Feb 11, 2023 4:40 am ...
Another thing that I don't understand is why the outer end of the frame to which the force of 76420 N (7.8 tonnes :shock:) is applied is not displaced in the vertical direction as there is no restraint to the Z axis.
Thanks for the video!

Running a buckling or frequency analysis is based on a different mathematical approach than
running a static analysis. The resulting buckling or eigenfrequency shapes of the structure must not necessarily correspond
with the static deformation. When ramping up the max slider to 1000000 you can see that there is a vertical deformation,
but compared to the buckling shape of the flanges it is very small.
You can account for preloading by adding a "static-step" and the parameter "perturbation" to the input-deck,
see txt files in zip. The preloading results in a drop of the buckling factors.

I ran a stability analysis with a UPE-100 profile and compared the results to Euler-theory and the DIN-manual.
There are huge differences in the system-utilization. Euler gives 25%, DIN-manual 40%!
Dont trust FEM alone... See also files in zip.
FreeCAD_Smath_ccxInput_files.zip
(209.44 KiB) Downloaded 53 times

What totally confuses me is the fact, that the results depending on the number of cpu I activate.
I tested the UPE-100 calculation on two different machines (intel i5 + i7 with 2 and 6 cores).
When using 6 cores on i7, I get these odd buckling shapes below the first correct eigenvalue.
multi_cpu.JPG
multi_cpu.JPG (116.35 KiB) Viewed 1321 times

EDIT:
UPE-100 calculation was done with FC0.20.2 stable release portable

Code: Select all

OS: Windows 10 Version 2009
Word size of FreeCAD: 64-bit
Version: 0.20.2.29177 +426 (Git)
Build type: Release
Branch: (HEAD detached from 0.20.2)
Hash: 930dd9a76203a3260b1e6256c70c1c3cad8c5cb8
Python 3.8.10, Qt 5.15.2, Coin 4.0.1, Vtk 8.2.0, OCC 7.6.3
Locale: German/Germany (de_DE)
Installed mods: 
  * BIM 2021.12.0
  * CfdOF 1.21.3
  * Curves 0.5.8
  * Defeaturing 1.2.0
  * dodo
  * fasteners 0.4.4
  * Manipulator 1.4.9
  * Plot 2022.4.17
  * Reinforcement
  * WorkFeature
Laurie Hartley
Posts: 522
Joined: Mon Feb 27, 2017 5:33 am
Location: Australia

Re: Buckling

Post by Laurie Hartley »

thschrader wrote: Sat Feb 11, 2023 9:58 am
Thanks for the answer to my query about displacement Thomas.
I am away from my computer till Monday but look forward to opening your files then.
cad1234
Posts: 331
Joined: Fri Dec 28, 2018 11:11 am

Re: Buckling

Post by cad1234 »

Another great Freecad FEM Thread!!
Thanks Laurie, thschrader and fandal.
Great to have you on board and so much opportunities for me to learn..
Many thanks!!
thschrader
Veteran
Posts: 3129
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Buckling

Post by thschrader »

Laurie Hartley wrote: Thu Feb 09, 2023 5:33 am ...
Perhaps it is a question of understanding the model. If it was a "goal posts" frame the buckling would make sense.
Here is the complete frame.
DNV_B2_bucklingFrame_extended.FCStd
(23.06 KiB) Downloaded 49 times
You must apply a load of 494 kN to get a z-deflection of 10 mm in the middle of the frame.
Next step: how to apply imperfections to the frame?

Example: Mastan2 can calculate the buckling eigenform and after that you can use the
scaled eigenform as new geometry for the static calculation.
https://www.mastan2.com/

That means we must use the deformed (buckled) mesh as input geometry for the statics.
When I remember right there was a discussion about this in the forum, but I cant find the post
Laurie Hartley
Posts: 522
Joined: Mon Feb 27, 2017 5:33 am
Location: Australia

Re: Buckling

Post by Laurie Hartley »

thschrader wrote: Sat Feb 11, 2023 9:58 am
I ran a stability analysis with a UPE-100 profile and compared the results to Euler-theory and the DIN-manual.
There are huge differences in the system-utilization. Euler gives 25%, DIN-manual 40%!
Dont trust FEM alone... See also files in zip.
_________________________________________________________________________________________________

What totally confuses me is the fact, that the results depending on the number of cpu I activate.
I tested the UPE-100 calculation on two different machines (intel i5 + i7 with 2 and 6 cores).
When using 6 cores on i7, I get these odd buckling shapes below the first correct eigenvalue.
I played around with my spreadsheet and discovered that if I applied the recommended design factor of 2.1 the Johnson formula produced a critical load of 127.36% or 40% utilisation - Is this mere coincidence once more???

The spreadsheet is attached with a .txt extension added. Change cell B4 from the drop down box - Theoretical to Recommended and this will change the results.

If I change cell B4 back to Theoretical the Euler Load is 198.45kN or 25% utilisation

We haven't talked about the "slenderness ratio" but in this model it is greater than the "critical slenderness" ratio (I hope I have this the right way round!!) so that is why the Euler result is called b24. (EDIT - in cell B24)

I also ran your U100 file on my machine and came up with a slightly higher buckling factor 4.17 - see screenshot. 5% higher than your machine.
EDIT; I also notice it twists.

Code: Select all

CPU:       Info: Quad Core model: AMD Ryzen 7 3700U with Radeon Vega Mobile Gfx bits: 64 
           type: MT MCP arch: Zen family: 17 (23) model-id: 18 (24) stepping: 1 
           microcode: 8108102 cache: L2: 2 MiB 
           flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm bogomips: 36730 
           Speed: 1152 MHz min/max: 1400/2300 MHz boost: enabled Core speeds (MHz): 1: 1152 
           2: 2351 3: 1286 4: 1130 5: 1256 6: 1131 7: 1691 8: 1696 
Attachments
UP_100_LH MOD.xls.txt
(66 KiB) Downloaded 58 times
Laurie Hartley
Posts: 522
Joined: Mon Feb 27, 2017 5:33 am
Location: Australia

Re: Buckling

Post by Laurie Hartley »

thschrader wrote: Mon Feb 13, 2023 4:29 pm
Laurie Hartley wrote: Thu Feb 09, 2023 5:33 am ...
Perhaps it is a question of understanding the model. If it was a "goal posts" frame the buckling would make sense.
Here is the complete frame.
I have downloaded and inspected your full frame model - it seems my suggestion doesn't throw any light on the subject.
I did find the image of the displacements in the third set of buckling results quite intriguing - see attachment

So your original question regarding the first model remains unanswered. "why are the top flanges buckling?"

OK - I will lead with my chin again. In the "olden days" when I was a young construction worker :D preparing to lift or place a heavy load with a crane I always looked for the "weak spot" in the set up as a sort of risk assessment.

When I look at this set up and ask myself where is the weak spot my answer is - in the webs at the points where the force first comes into contact with the flange "arches". (like the outermost point of support in a cantilever).

Could it be that the buckling first sets in in the webs subsequently causing buckling in the top flanges?
Attachments
Displacement_Full Frame.png
Displacement_Full Frame.png (106.53 KiB) Viewed 1016 times
Post Reply