Refer to the attached image extracts from CAD and Femap. My typical approach to simulate such models is to import the solid objects into Femap using *.sat format and then mesh with solid finite elements with midside nodes (priority is to have a hex mesh for all objects; if some objects cannot be hex meshed using simple Femap options then I go for tet mesh).
I have the following three queries-
If I had a structure like this I would do the following:
1. Are you talking about volumetric locking? AFAIK this phenomena occurs in incompressible materials, with Poisson`s ratio 0,5 like rubber. Not steel or other metals/ceramics/concrete.
2. This is basics of structural mechanics, not FEA.
Two dimensions less than third -> beam.
One dimension less than two others -> shell.
All dimension are equal -> solid.
However, all depends from task - what you want to calculate. If you want estimate global stiffness - use plate. If you study some small part of your structure (for example in scientific/research problem), then use submodeling - plates and solids in zone of interest.
So now you have two votes for shell model and zero for solid model.
95% of structures can be properly solved with beams and shells (ships, planes, cars, buildings, bridges etc.).
It is not common practice to mesh every structure with solids. I see people who mesh beam with 0,5M tetra to get beam deflection from point load.