It really depends on the design itself and what you think is going to be involved in the FEA side of the project. The vast majority of the analysis work I've done for the past 2.5 years has been sheet-metal/shell elements, and over that period of time I've come to several conclusions:
- If you're re-creating geometry (which is not a bad idea for large assemblies or if the model is not native to Creo), it's usually better to create the mid-surfaces with fills/surface extrudes. The typically method for strengthening a weak design is just to up-gauge the material used, which can be a pain if you're shell elements are based on compressing 3D geometry down to a surface. With surface geometry, shell elements are applied manually, so you have direct access to the thickness value tied to the elements (which means you can do sensitivity studies with it). This method also makes it easier when dealing with parts that are connected, because you're thinking in terms of the geometry that'll be leftover after the compression.
- You should also almost always remove the bend radius from the parts and replace it with just an edge. This does not result in singular points and significantly reduces the number of elements created, and has a very minimal impact on the results unless the radius is very large.
- Learn how to use the Pro/Weld tool, as it can really help connecting parts using the "Weld Feature" in Simulate's weld tool. You could probably resolve the issue you're having by using both of these.
- If you ever use the fastener tool for rivits, you should set the fasteners to a "Frictionless Interface". I've done several physical tests to compare measured results against my FEA work, and in all cases I've had excellent agreement with frictionless fasteners for rivits (and poor agreement without the frictionless interface). It should make since too, as the faying surfaces will carry very little shear load due to the low level of friction.