I've attached a fairly simple model of a glass vessel with a sapphire cap bonded to each end. I have three planes of symmetry so the model represents 1/8 of the full model, with frictionless contact on each plane of symmetry. I have used offset shells throughout to give the required dimensions and thicknesses. Vessel is under vacuum so has atmospheric pressure over the outer surfaces. All seems to work OK with no surprises, but a few questions:
1. What is going on with Open Cracks view? The end cap looks highly fragmented near the join.
2. Are shells appropriate for this analysis?
3. Is the join between the glass tube and the end cap (shell edge to shell edge) fully rigid? I assume it is, as the whole model is effectively edge to edge and is able to transmit bending moments etc., I just wonder how the shells capture the idea that the interface has a finite thickness (they probably don't and it probably doesn't matter).
4. Any other comments on the model?
Thanks,
Dave
Comments
2. Yes except except for stress at the corners.
3. It's only rigid at the connected nodes so it sort of ignores thickness in the same way that the joint between every shell element does. The joint won't be softer than the elements themselves.
4. There's a subtle difference between the internal and CCX solvers with curved offset shells. The internal solver keeps the element size and shape constant regardless of offset (like the Open cracks view) while CCX expands/contracts them to keep them connected to each other which is more correct/agrees with what you'd expect. So on your curved corner, there can be a difference in the amount of material which would affect stiffness.
I would use solid elements because of how thick it is. Then you can forget about all the inaccuracies of shells that become worse with greater thickness, sharper corners, and more offset.