Crashing problem for challenging meshing problem

Hello all,

I am looking for a workflow for a set of parts that are layered such that adjacent layers touch. I am looking to simplify the geometry for meshing by compositing the small surfaces to ensure a high quality mesh. I’m looking to make this model a non-manifold so I need to imprint and merge the model. Should I do the compositing of the surfaces first or should I imprint and merge the model before I compost the small surfaces? My current workflow is:

import step “file.stp” heal #for each of 7 volumes
merge tolerance 0.008
imprint tolerant volume all merge
#composite each surface by selecting the entire surface (top surface than bottom surface, one at a time for each volume)
vol all scheme tetmesh
vol all size auto factor 5 propagate
mesh vol all

I then get the following warning and the parts never mesh. I have let the program sit an entire day and it doesn’t respond…

If I try to mesh the model one piece at a time, I can get the first two parts meshed, but the 3rd causes a crash. Not sure what is going on.

This is the model and an exploded view for reference.

Screen Shot 2022-01-07 at 10.02.09 AM

Kind regards,

Ron

If you have overlapping surfaces, it could be that your CAD geometry does overlap. Check the thin volumes to make sure a surface isn’t protruding through a surface on the other side. If the CAD geometry is fine, it may be that the surface mesh on those thin volumes is overlapping at the given size. Try building just the surface mesh and look for triangles sticking through another surface. They would most likely be on the composited surfaces of those thin volumes.
“mesh surf all” instead of “mesh vol all”

thanks for the quick response Mark, the CAD geometry should have no overlaps. Each adjacent surface is the result of a boolean subtraction from the adjacent surface, so the parts should meet perfectly. I was reading in the manual that to make a high quality mesh, small surfaces should be dealt with in a number of ways, for which compositing appeared to be a reasonable option for my case. However, this now seems to create problems when meshing the volumes, like you said, producing overlaps. Is there a different strategy for trying to mesh such a complicated set of parts that you would recommend? When I mesh them one volume at a time I still get overlaps and am unable to refine the mesh, with an ERROR: Volume 1 is not meshed… although it has been meshed (green layer). Even with what I see with the current mesh, I now Abaqus is will have issues with the mesh regarding mesh quality, creating distorted elements and negative Eigen values.

Any thoughts?

RP

Hi Ron,

If you do a draw tet in volume 1 you will see that volume 1 has a surface mesh but no volume mesh.

The composite step creates a new surface id. I would do a merge all after compositing and drop the merge keyword from the imprint tolerant step.

It shouldn’t just hang. I have seen cases where the machine starts to swap because you run out of memory and that can cause a hang. Are you hitting a low memory situation?

Thanks,
Karl

Hi Karl,

I gave your workflow adjustments a try and unfortunately not much I have tried seems to be able to mesh this geometry. Is it possible to mix hex and tet meshes or must you use one or the other…
Happy to upload the models and see if you are able to get them to mesh. I have also been working with the hex mesh and web cutting, which works for building a nice mesh for smooth surfaces, but still have trouble even with that in Abaqus (see hex mesh).

Im at about a loss now. I have 4 models to try to mesh and have only been able to mesh the smooth surfaced model, using tets, which worked fine. But meshing the models with the surface features has proven to be quite challenging. I am wide open to any advice, suggestions, help etc…
Here are what the 4 models look like…

RP

Hi Ron,

As to mixing hexes and tets, yes, as long as you are comfortable with having pyramid elements (PENTA) in your analysis.

Go ahead and upload one of the models and I will take a look.

Karl