The only time shells are not appropriate are when you are trying to simulate a part which has a very thick cross section.  In those instances you have to use solids.

Shells will give you better results, especially at the connection points between the different pipes.  Mostly because you can actually model the joint, instead of just connecting beam end points.

Lastly, if you have a very large structure that either takes more memory to run than your current computer has or it takes so long that you could build/test the structure before the simulation is done then using beam elements would be a good way to idealize part of the structure while simultaneously lowering the computational overhead.

Hope this helps,

Steve

 

Stephen Seymour, PE
Seymour Engineering & Consulting Group
www.seymourecg.com  

Thanks Steve,
I know I can mix shell with beam elements, but are there any requirements for making connections between portion of beam elements and portion of shell elements? Do you have to manually connect the two or it's automatically done?
For example, long pipe can be represented by beams and plates that are welded on the sidewalls of the pipe can be shells, but how is that connections between the two handled in Mechanica. Thanks.

Typically the connections will be automatic.  In WF2 I had some issues getting shells and beams to join correctly.

The biggest issue I have seen is the connection between shells/beams and solid elements.  Solid elements only have 3 degrees of freedom (DOF), while shells/beams have 6.  Many times people (even myself) fail to account for this and have solution failurs due to rigid body motion.

From what I understand the connection between beam and shells occurs at the element nodes or any explict datum points that are shared between the beam and shell.  If the stress or deflection at the connection locations isn't realistic, then you might have to see some additional datum points or use the AutoGEM mesh constrols so seed some additional nodes at the shell/beam interface (assuming the shell/beam interface is not a simple endpoint of beam to surface of shell).


Keep in mind though that when you create a beam as point to point you do not have much control on the number of beam segements or the number of nodes between your point to point locations.  Because of this I typically create all my beams initially as datum curves in Pro/E and then establish beam elements on the datum curves.

One of the best options you have at your disposal for determining if the connections are working is to perform a modal analysis.  Use the modal analsyis and solve for high frequencies.  Then observe the deflection animations in the results viewer.  Connections that did not take will show up pretty readily.  Plus, a modal analysis set to a convergence of quick check can most of the time be much faster than a static solution set to be SPA or MPA.

Steve

Steve
 

Stephen Seymour, PE
Seymour Engineering & Consulting Group
www.seymourecg.com