FEM Non-Jacobian Nodes

[ishnik]

Hi there,

I'm wondering if there are any common gotchas that generate non-jacobian elements that cause Calculux to fail to run - linked is a model I'm working on where I've successfully run FEM on part one of a two-part assembly (succeeded on the loadring at the top) and failed on the assembly as a whole.
Have tried:
- making a Part Compound of the two Bodies, and meshed that;
- making a Part Fusion of the same, meshed that (unrefined boolean);
- making another Part Fusion of the same, meshed that (refined this time).
Still no joy.

Link to my work (just can't shrink it to 1mb...):

https://1drv.ms/u/s!At8bawLbRP2Mgskq4gT ... Q?e=k7c01J

Freecad details are:

Code: Select all

OS: Windows 10 Version 2009
Word size of FreeCAD: 64-bit
Version: 0.20.29177 (Git)
Build type: Release
Branch: releases/FreeCAD-0-20
Hash: 68e337670e227889217652ddac593c93b5e8dc94
Python 3.8.10, Qt 5.15.2, Coin 4.0.1, Vtk 8.2.0, OCC 7.6.2
Locale: English/Australia (en_AU)
Installed mods: 
  * A2plus
  * Assembly4
  * BIM
  * dodo
  * fasteners
  * parts_library
  * PieMenu

[Vision78]

I have switched to using Gmsh instead of netgen as the dialog for it tells you if there are non jacobians when you generate it so you can try different sizes of meshes without having to fail simulations and save time, you could also run a mesh check on netgen but it still takes longer.

Meshes hate small curves as well, simplifying the model may sometimes be a viable option.

Or you could use gmsh and make put smaller elements where needed without having to make the whole mesh small.

[ishnik]

Thanks Vision78,

If it's just a refinement problem I'm glad to know I haven't done anything silly in the model that would explicitly cause this kind of issue. I thought maybe it would be something to do with the multibody compound, and that's why I'd thought to run a refined Fuse to get around it.

What's the best way to refine the mesh - would I need to learn some gmsh gui basics and import it into FreeCAD or is it simpler to do something in FreeCAD itself?

[Vision78]

If it was a multibody issue I doubt it would cry about non-jacobians, I ran one not so long ago and used fusion without an issue using Contact restraint, I' have to log on to check your model more accurately, also netgen mesh check posts the problem nodes as red dots if I remember correctly so you could use that to pinpoint the issue easier

[Vision78]

Scratch that, the might very well be an issue because there isnt a gap between the parts as non-jacobian elements went wild when I separated them

[Vision78]

Using gmsh and switching to first order elements made it work for me, are you possibly missing contact constraints here?

[ishnik]

Thanks for the lead - I also gave it a shot using 1st order elements and gmsh; this worked and I've got a somewhat useful stress plot (attached).
For that run I had deleted the hole details from the loadring up top (not the area of interest in this case) and made a refined Fusion of the two parts.

Were it a Compound, I'd need to explicitly add a contact constraint between the parts of the Compound?

Along the same lines - if I do have a compound, and it fits together like this, then how do you select the contact faces to get them to contact? I haven't been keeping up to date on the FreeCAD development so much but last I looked there wasn't selection-cycling just yet or a button for hicks like me to just add a 'global contact constraint' for all faces that touch to be glued together in the analysis.

[Vision78]

I used a maximum of 40mm element size, seems like you went a bit coarser, I also fused the parts to be just 1 part for computing time (Just writing the .inp file took 440 seconds when contact was put in there) but to see if there would be any slip in the shell or bending realistically you would have to run them apart.

Maximum stress that I got was also around 134Mpa though but it did show a 0,8mm deformation under the holes where the load was set which is concerning.

I'd probably run the simulation in 2 parts using the contact surfaces as constraint surfaces, If neither of the simulations show anything concerning, I'd be happy enough.

There is a Constraint contact there or should be.

https://wiki.freecad.org/FEM_ConstraintContact

[ishnik]

134MPa matches what Calculix says on my machine so that's good to know - I get 132.8 which is close enough.
A little worried by displacement results though - you sure it comes up with 0.8mm? See below mine tells me it's only 0.14mm.

Will rerun with finer mesh max element size just in case that's it...

[Vision78]

134MPa matches what Calculix says on my machine so that's good to know - I get 132.8 which is close enough.
A little worried by displacement results though - you sure it comes up with 0.8mm? See below mine tells me it's only 0.14mm.

Will rerun with finer mesh max element size just in case that's it...

Your mesh seems to have weird optimizations on the side or the holes in it are just hidden in the picture.

Larger mesh elements combined with using 1st order elements might make the elements a bit more rigid and make deformation less accurate.