Learning Forum

[Andres Mena]

Hello! 

Previously I asked whether stacking shell elements on top of solid elements (by basically making them share nodes) is okay. As suggested, I used the option NLOC=-1 to ensure that the shell thickness is defined completely in one direction (contact was also modified to account for this). 

I have come across a different issue now in modelling biological tissues because I need to add a small layer on top of these shell elements (already sharing nodes with solid) like the figure shown below:

 

I am not entirely sure if using shell elements is still advisable for this (because the final solid-element layer is supposed to contact first other bodies in certain regions) or if I should just switch everything to solid elements? I am not sure if 3 elements sharing nodes this way is okay from a theoretical perspective(?)

 

[Jim Day]

If the red solid shared nodes with the shell and grey solid, the red solid would not be offset and would therefore be in the incorrect location.

[Jim Day]

If you replaced the shell with a solid, then everything would fit together properly. The concerns would be (a) possible reduction in time step due to the thin solid element(s), and (b) less versatility in controlling number of integration points through the thickess (as is always the case when considering use of solids instead of shells).

[Jim Day]

Instead of thin solid layers, I would recommend thick shell (*ELEMENT_TSHELL) elements. Tshells have the 8-node connectivity of solid elements (but care must be taken so the element normal points in the through-thickness direction), but allow the number of through-thickness integration points to be specified using the variable NIP (like a shell).

[Andres Mena]

Thank you so much Jim! I will use tshell elements to try to solve this issue.

Two quick questions:

You mentioned *ELEMENT_TSHELL rather than using *SECTION_TSHELL. Can I assign my 8-node (solid) elements a tshell section, or do I have to convert the solid elements to tshell elements before assigning them their respective section?

Also, any recommendation on the number of through thickness integration points or should I just leave the default value?

 

[Jim Day]

You have two options for setting up your stackup using tshells. If you intend to use multiple layers of tshells through the thickness of your component, I'd use multiple *PARTs together with *SECTION_TSHELL and *ELEMENT_TSHELL. The other option, easiest if you want to use just a single layer of tshells is *PART_COMPOSITE_TSHELL + *ELEMENT_TSHELL. Either option allows you to vary the material model from layer-to-layer. *ELEMENT_SOLID is incompatible with *SECTION_TSHELL.

[Jim Day]

The syntax of *ELEMENT_SOLID and *ELEMENT_TSHELL are the same, but the connectivity of tshells is more restrictive. You must confirm that the tshell normal direction (normal to the segment defined by the first 4 nodes of the element connectivity) is in the through-thickness direction. You can use LS-PrePost to check those normals by reading the keyword input into LS-PrePost and selecting EleTol > Normals > Tshell.

As for the number of integration points through the thickness, that’s a judgment call that should take the material model and the number of tshell elements through the thickness of the component into account.