Shell model design approach

General Questions
1

Hi All,

I am preparing to do some good size deck structure shell models. I would love it if I could use a “tied” contact to automatically connect all the edges without needing to partition everything. It seems to almost work in my test here but there is some non-physical stuff going on at the edge.
Any ideas?

(and yes the model is upside down (negative Z is up)

prepomax_file

new large test2.inp (4.9 MB)

Screenshot 2025-04-29 114656
Screenshot 2025-04-29 1146561465×796 123 KB

stress
stress801×660 105 KB

thanks

Luke

2

I have some questions about your setup:

  1. Did you try merging the shells in CAD software before importing to PrePoMax or creating a compount part in PrePoMax (this may not work though) ? Doing that would eliminate the need for contact/constraints to hold them together. I did it in FreeCAD for this tutorial with a similar model: https://www.youtube.com/watch?v=2tYPiZ5kQHM
  2. Why are you using tied contact instead of tie constraints ? The latter are usually the recommended option in such cases. Then you can also easily check the connections using frequency analysis and that should be done before running the actual load case.
  3. Does the small cube serve only as a mean to apply load to the plate ? If yes, surface partition would be better.
3

Thanks for the reply, I will go check out your demo, hopefully I find out something I am doing wrong.

I found that the tied contact somehow seemed to work better than simply tied connections but that could just be a fluke (neither worked well enough)

As expected I couldn’t convert it into a compound part in Prepomax (which is why I have looked into other options)
I also haven’t found a cad program that would tie all the surfaces. The geometry looks clean but perhaps there is an error somewhere, I will look into that.

In the past I have have manually partitioned every intersection but for larger and more complicated models that is tedious, particularly without partitioning tools like in Abaqus.

Where loads are applied I will naturally use sketched partitions and mesh refinement but that is just one or two locations on a much larger model.

thanks

Luke

4

It might be a matter of proper definition of tie constraints - you may have to define non-default tolerance, maybe enable adjustment and properly select master and slave surfaces. Then tie constraints should be better - more straightforward, versatile and efficient.

If you want to try what I did (and what worked) for that referenced tutorial, use “Fuzzy Union” in the add-on Defeaturing workbench in FreeCAD and follow it with the “Convert to solid” tool in the Part workbench.

1 Like
5

thanks,

The best solution is probably to make a very careful and clean model to start with. At the end of the day it is probably faster and less aggravating.

I will check out Freecad again. I haven’t looked at it lately so hopefully it is more usable than before.

Luke