For running an analysis on FDM printed parts Solidworks Sim would be extremely difficult to use. It's not really designed for running this type of analysis and is at the fairly low end of FEA analysis packages. An FDM analysis should be treated almost like a composite structure and you'd have better luck running something like ANSYS, or NASTRAN but those are by no means the only two. To get good (and accurate) results from FEA you'll need to model each layer, the bonding between each and the internal structure.
I've never done it myself but i've looked into doing it. The mesh required for even a fairly simple part looks pretty massive. Certainly possible and certainly doable, but you'd need to weigh your expectations against how much time it would take to model everything, set up the mesh and then run the analysis. In my experience, given that most FDM prints are pretty quick to build, it always seemed to be more efficient to build a few parts load them up and watch how they fail in a structural test. 20 hours of printing and then an additional 10-15 of setting up and running the tests were faster than running a proper FEA.
For a quick and dirty FEA, i'd use the solid model of my part and only consider loads that acted normal to the Z-Axis of the print. Any loads that act in shear apply a huge FOS. The closer to pure shear the larger the FOS you'd want to have. And any thin, tall parts i'd assume can take no loading outside of pure compression.
Just noticed that this question had a tag on Heat-management
As far as a Heat loads go, it'll depend heavily on the infill type and density you have on your model. Everything from above will still apply, but the loads would then be thermal loads instead of structural. Particularly with FEM at best your first analysis will only be within about 50% of true values and you will then need to update your thermal model with the results from your first verification test.
Depending on the accuracy required from the analysis you could get away with modelling a hollow part with the walls of the part being the thickness of the printed shells. Higher accuracy analysis you'll have to model all layers as they are printed. If you're just looking for a 'will it melt' analysis, thin walls will work fairly accurately assuming you're confident in your material properties.