Geometry too complicated?

About the development of the FEM module/workbench.

Moderator: bernd

DE_Michael
Posts: 7
Joined: Mon Apr 06, 2020 4:42 pm
Location: Germany

Geometry too complicated?

Postby DE_Michael » Mon Apr 06, 2020 5:12 pm

Hello, everybody,

in FreeCAD I would like to analyse the strength and deformation of a fish tank and its sub-rack using FEM (see illustrations).

Thereby the problem arises that with increasing complexity of the geometry I cannot create an input-file anymore. For example, I can split the model and successfully calculate the fish tank alone. I can also calculate the sub-rack with a very rough mesh (element size in the range of 100 mm). For all parts in one file or with a slightly finer mesh, the generation of the input file fails.

Can this be due to a limited capacity of my hardware (Windows 10, Intel Core i7-8750H, 32 GB RAM, SSD)?
I have read several posts on the input-file topic in the forum, but didn't found a working suggestion and would therefore be glad about your help.

Kind regards
Michael
Attachments
fishtank.FCStd
(447.77 KiB) Downloaded 8 times
fishtank_FEM.png
fishtank_FEM.png (162.65 KiB) Viewed 465 times
fishtank_3D.png
fishtank_3D.png (864.8 KiB) Viewed 465 times
vocx
Posts: 3962
Joined: Thu Oct 18, 2018 9:18 pm

Re: Geometry too complicated?

Postby vocx » Mon Apr 06, 2020 10:48 pm

DE_Michael wrote:
Mon Apr 06, 2020 5:12 pm
...
I have read several posts on the input-file topic in the forum, but didn't found a working suggestion and would therefore be glad about your help.
If would help to mention which other threads you found, because then we can avoid providing the same suggestion. Alas, you didn't do that.

Be aware that FreeCAD is just a pre-processor for an external finite element solver. CalculiX is the most common solver used. So, where is your problem? In the meshing, input file generation, or solving? It could be the connection between FreeCAD and the Gmsh mesher, or maybe the actual CalculiX solver.

If it's in the meshing stage, you should try to mesh the geometry externally, using Gmsh. If it's in the writing stage of the input file, then it has to be investigated in FreeCAD. I don't think a computer with 32 GB of RAM should have a problem handling big meshes.
Always add the important information to your posts if you need help.
To support the documentation effort, and code development, your donation is appreciated: paypal.
User avatar
Kunda1
Posts: 7319
Joined: Thu Jan 05, 2017 9:03 pm

Re: Geometry too complicated?

Postby Kunda1 » Mon Apr 06, 2020 11:15 pm

Mod edit: moved to FEM subforum
Want to contribute back to FC? Checkout:
#lowhangingfruit | Use the Source, Luke. | How to Help FreeCAD | How to report FC bugs and features
reox
Posts: 845
Joined: Sat Aug 13, 2016 10:06 am
Contact:

Re: Geometry too complicated?

Postby reox » Tue Apr 07, 2020 6:57 am

If you simply want the deformation of the rack, it should be sufficient to model the rack alone and apply a boundary condition for the tank.
This would simplify the model already a lot.
thschrader
Posts: 1618
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Geometry too complicated?

Postby thschrader » Tue Apr 07, 2020 4:05 pm

Hi Michael,
as a starting point for FEM you should begin with a simple model, to
get a good mesh and plausible results. In later steps you can do the refinement.
When starting with the "complete problem" mostly you run into meshing problems
or complicated boundary/loading-conditions. Like your subdivided fishtank for applying the
increasing water pressure. In you FC-model you use rectangular (full) profiles for the frame,
but in your picture there are shown extruded profiles, which have a complex geometry.
Makes meshing harder. But with a i-7 32GB system this should be no problem.
I recommend the newest FC0.19 development snapshot for FEM.
https://github.com/FreeCAD/FreeCAD/rele ... g/0.19_pre
My analysis for playing:
fishtank_TS.FCStd
(415.46 KiB) Downloaded 5 times
aquarium.JPG
aquarium.JPG (85.12 KiB) Viewed 330 times
rexroth_60x60x1000.JPG
rexroth_60x60x1000.JPG (80.73 KiB) Viewed 330 times
DE_Michael
Posts: 7
Joined: Mon Apr 06, 2020 4:42 pm
Location: Germany

Re: Geometry too complicated?

Postby DE_Michael » Tue Apr 07, 2020 4:09 pm

Firstly, thank you for your answers. Here are some more explanations:

vocx wrote:
Mon Apr 06, 2020 10:48 pm
If would help to mention which other threads you found, because then we can avoid providing the same suggestion. Alas, you didn't do that.
So far I have tried to change numerous parameters of the FEM / Solver settings. These include mesh groups, input file splitting, number of CPU's and the matrix solver. Besides Gmsh, I also tried meshing through Netgen and I tried to use the "experimental" CalculiX solver.

vocx wrote:
Mon Apr 06, 2020 10:48 pm
Be aware that FreeCAD is just a pre-processor for an external finite element solver. CalculiX is the most common solver used. So, where is your problem? In the meshing, input file generation, or solving? It could be the connection between FreeCAD and the Gmsh mesher, or maybe the actual CalculiX solver.
As I tried to describe above, the problem is about the creation of the input file. The meshing works and if the input file (with strongly simplified geometry) can be created successfully, the calculation also works.

reox wrote:
Tue Apr 07, 2020 6:57 am
If you simply want the deformation of the rack, it should be sufficient to model the rack alone and apply a boundary condition for the tank.
This would simplify the model already a lot.
As I tried to explain above, the calculation does work if I simulate the fish tank and the sub-rack separately. The question was why it does not work together.
thschrader
Posts: 1618
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Geometry too complicated?

Postby thschrader » Tue Apr 07, 2020 4:26 pm

DE_Michael wrote:
Tue Apr 07, 2020 4:09 pm
As I tried to explain above, the calculation does work if I simulate the fish tank and the sub-rack separately. The question was why it does not work together.
There is a problem with the self_weight. Doubleclicking does not open dialog-window.
No mesh generated. Mmmh..?
self_weight.JPG
self_weight.JPG (22.76 KiB) Viewed 320 times
meshing_gmsh.JPG
meshing_gmsh.JPG (190.65 KiB) Viewed 320 times
DE_Michael
Posts: 7
Joined: Mon Apr 06, 2020 4:42 pm
Location: Germany

Re: Geometry too complicated?

Postby DE_Michael » Tue Apr 07, 2020 5:03 pm

Hello thschrader,

thank you very much for your detailed answers!

thschrader wrote:
Tue Apr 07, 2020 4:05 pm
In you FC-model you use rectangular (full) profiles for the frame,
but in your picture there are shown extruded profiles, which have a complex geometry.
I abstracted the extruded aluminium profiles consciously to rectangular blocks in order to simplify the networking etc. From the extruded aluminium profiles I know the moment of inertia. Therefore, I have calculated the moment of inertia for the rectangular blocks and adapted the Young's modulus of the aluminium profiles accordingly, thus ensuring that the flexural strength (= moment of inertia * Young's modulus) is the same. Hopefully that is just right :D


thschrader wrote:
Tue Apr 07, 2020 4:05 pm
I recommend the newest FC0.19 development snapshot for FEM.
https://github.com/FreeCAD/FreeCAD/rele ... g/0.19_pre
I have tested the developer versoin. In fact it seems to work a little better (the error occurs much later).


thschrader wrote:
Tue Apr 07, 2020 4:26 pm
There is a problem with the self_weight. Doubleclicking does not open dialog-window.
No mesh generated. Mmmh..?
Do you have an idea how I could solve the problem with the self-weight boundary condition?
If I remove this constraint completely, I still can't write the input file.

I get the following error message from the report of the developer version:
ERROR: femelement_table != count_femelements
Error in get_femelement_sets -- > femelements_count_ok() failed!
thschrader
Posts: 1618
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Geometry too complicated?

Postby thschrader » Tue Apr 07, 2020 5:25 pm

DE_Michael wrote:
Tue Apr 07, 2020 5:03 pm
Do you have an idea how I could solve the problem with the self-weight boundary condition?
If I remove this constraint completely, I still can't write the input file.
You use 4 different materials with different densities. There should be 4 entries for
self weight. I will try this tomorrow (I am writing from my private pc without FC0.19...)

And I will try rexroth profiles. Which one?
regards Thomas
DE_Michael
Posts: 7
Joined: Mon Apr 06, 2020 4:42 pm
Location: Germany

Re: Geometry too complicated?

Postby DE_Michael » Tue Apr 07, 2020 8:51 pm

Hello Thomas,

thank you again for you help.

thschrader wrote:
Tue Apr 07, 2020 5:25 pm
You use 4 different materials with different densities. There should be 4 entries for
self weight. I will try this tomorrow (I am writing from my private pc without FC0.19...)
So far I thought that the self-weight boundary condition refers to the whole fusion. I have not yet been able to create them for individual materials. But in principle this approach could lead in the right direction, because this would at least be an explanation why the input file cannot be created for the whole model, but for partial models without problems. I'll look at this intensively again tomorrow.

thschrader wrote:
Tue Apr 07, 2020 5:25 pm
And I will try rexroth profiles. Which one?
Currently I plan to use the following aluminium profiles:
https://easy-systemprofile.de/Profil--N ... rland.html
https://easy-systemprofile.de/Profil--A ... il-96.html
These are certainly not the original Bosch / Rexroth, but they are at least "compatible" and have the same alloy:
Young's modulus for Al Mg Si 0,5: E = 70.000 N/mm^2

45 x 45 mm
moment of inertia Alu-Profile: I_x = I_y = 134.572 mm^4
moment of inertia rectangular block: I_x = I_y = 341.719 mm^4
--> equivalent Young's modulus: (134.572 / 341.719) * 70.000 N/mm^2 = 27.567 N/mm^2

45 x 90 mm
moment of inertia Alu-Profile: I_x = 323.570 mm^4, I_y = 1.250.353 mm^4
moment of inertia rectangular block: I_x = 683.438 mm^4, I_y = 2.733.750 mm^4
--> equivalent Young's modulus: (1.250.353 / 2.733.750) * 70.000 N/mm^2 = 32.016 N/mm^2
[whereby I_y is the the less favourable case]

In the end, the flexural strength (= I * E) of the aluminium profile and the rectangular block should be the same, hopefully.

Best regards
Michael