My FEM fails. How to get further?

[Roland]

Dear FEM experts,

I feel a novice in FEM.

This is my model:

Bearing1_Original.png (38.14 KiB) Viewed 2961 times

I want to analyse its strength if an axis (purple) is placed as indicaded and presses downward.
I prepared two approaches, of which none succeeded so far.

• The one based on shells fails with a CalculiX execution error: "ERROR in gen3dnor: size of estimated shell normal in node 7 element 11790 is smaller than 1.e-10." I have no idea what I can do to avoid this error.

Bearing1_Shells.png (8.79 KiB) Viewed 2961 times

• The approach based on a solid yields a few meshing errors. To an extend that CalculiX doesn not like it.

Bearing1_Solid.png (11 KiB) Viewed 2961 times

Could somebody advise a suitable way forward? Improve the shells approach or the solid approach? Or otherwise?

Thanks for your consideration!

Roland

on:
OS: Windows 10 Version 2009
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.24267 +99 (Git)
Build type: Release
Branch: Branch_0.19.3
Hash: 6530e364184ce05ccff39501e175cf2237e6ee4b
Python version: 3.8.6+
Qt version: 5.15.2
Coin version: 4.0.1
OCC version: 7.5.3
Locale: English/United Kingdom (en_GB)

[NewJoker]

Could you move it to FEM subforum ?

1) shell model:

It fails because of inconsistent shell element normals (you can see them when you look at the created mesh). I don’t know how FreeCAD’s FEM module handles those but in CalculiX itself normal direction is controlled by the order in which nodes are specified under *ELEMENT.

2) solid model:

You should simplify it for the purpose of the analysis - remove small details. In this case you get nonpositive jacobian determinant errors. To eliminate them you can try further refining the mesh. If this doesn’t help, here are some other workarounds:
https://wiki.freecadweb.org/FEM_MeshGmshFromShape

[bernd]

moved to FEM

[Roland]

1/ Thanks NewJoker. Can go ahead with your suggestions

2/ Bernd: I posted in the Users forum because I asked for help on usage. I do not see my novice level questions as assisting to FEM Development. But I am glad that my question was noticed. In the near future, all usage questions regarding FEM to Fem Dev?

Greetings from NL

Roland

[NewJoker]

2/ Bernd: I posted in the Users forum because I asked for help on usage. I do not see my novice level questions as assisting to FEM Development. But I am glad that my question was noticed. In the near future, all usage questions regarding FEM to Fem Dev?

This subforum is actually a place for all questions regarding FEM, even though it belongs to the Development group.

To make it clear for future posters (as it can be quite confusing at the moment), could you add a note (to the current About the development of the FEM module/workbench) that this forum is a place for general discussions about FEM module ?

[-alex-]

As @NewJoker said you should simplify your model by removing small details.
Then you can use Draft_Facebinder to get a 2D shell model. Then you can apply 2D contact constraint between the frame and the cylinder.
You have to let a bit of play between frame and cylindre to avoid "fusion" of mesh elements.
Because of contact analysis, do not forget to fully constrain the cylindrical part, otherwise the solver will not converge.
Report here if you face difficulties.

[Roland]

Thanks Alex.

1/ I never used Facebinder. I made those models by Part Fuse (accepting the warning against fusing faces). Upon using Facebinder for the first time I observe that Facebinder does not include some of the shells into the created item. Does this tell me something about the quality of whether or how the faces touch each other?

(Apology if this is too basic for FEM)

2/ I did not use the cylinder to apply a force. Perhaps that is possible.

[Roland]

BINGO!

Problem solved.
.

Bearing1_Shells.png (146.08 KiB) Viewed 2664 times

How? Perhaps:
.

Bearing_FEM.png (92.23 KiB) Viewed 2664 times

.

I shall appreciate your view of cause and solution.

Greetz

Roland

[Roland]

Horrible! I cannot confirm that my diagnosis of the reason for failure is correct.

The solution works anyway!

Roland