I was analysing stress in some lifting clamps when I discovered that Inventor returns very different results for a simpel geometry depending on it being a part or an assembly. How come?
Snapshot #1 shows the geometry analysed as a single part.
Snapshot #2 shows the geometry analysed as an assembly of two parts. The weld is a combined with the arm.
Snapshot #3 shows The geometry analysed as an assembly of three parts - plate, arm and weld.
2 and 3 are quite similar but the max stress in 3 is nearly twice the max in 2.
1 is completely different with a lot more spread out stress and a huge max stress
Why this difference?
If you could attach the model files (assembly + components) with the simulations saved, it would help us to investigate further.
I believe that this is the 'as expected' behavior of any FEM package, since they are different geometries, number of bodies, etc. and they will yield different simulation results (at least for stresses, displacments may remain similar).
For example, if everything is modeled as one part, there are no 'assembly' contacts. If you are closely inspecting the stresses at the interface of two mated faces (a bonded contact in an assembly simulation) the stress distribution and values could be very different than if there was no interface since the body is one 'lump'.
Sharp inside corners are what we term a stress singularity, meaning the stress goes to infinity the closer you get to the root. This is also the case in real-life as well. The stress at a singularity should not be taken as the true stress, since as the mesh is refined to have smaller and smaller elements, the stress will increase dramatically until it reaches infinity. In this case a fillet should be placed there to eliminate the singularity.
Hope this helped clear things up a bit.
Have you tried changing the 'Separate Stresses across Contact Surfaces' solver option? Or, perhaps the assembly option is not using curved mesh, where the part file is, so you would get different results. i would say curved meshing will give better results for this geometry but I can't see your refinement - the mesh is probably way too coarse based on some of the gradients. Make sure you are comparing apples to apples.