static stress analysis of topology optimized mesh
The mesh that is kicked out of topology optimization is not useful (within Fusion anyway) for additional FEA. You'll either need to convert the resultant mesh to BRep (mesh to BRep has been covered extensively here on the forum and on YouTube) or you'd need to remodel the shape using the mesh as a guide.
Depending on how organic your form ends up being, you might be better off using TSplines to generate the shape. There is some really neat stuff going on in the industry using freeform type modeling techniques to approximate these optimized forms by using the mesh as an attractor for a TSpline or NURBS surface.
Just as an aside...
An FEA engine breaks your model into little chunks so that it can compute in individual elements instead of the whole shape. These chunks are usually some kind of polyhedral - but the most common type (especially when referring to Autodesk software) is tetrahedral (three equilateral triangles glued together to make a four-sided shape with six edges). Fusion uses tet-mesh exclusively. So if we think about the shape made out of all these little pyramids, it stands to reason that the exterior of the entire model would be made from triangles - just like an STL type mesh. This makes extracting the exterior shape pretty simple from a resultant FEA mesh. The solver looks at each of the little triangles and determines which ones are holding most of the load and which ones aren't contributing much. It then removes the unimportant ones. From that resulting 3D mass, the solver can extract all the exterior surfaces to deliver you a surface mesh (again, like an STL).
When you perform topology optimization in more CAE-centric packages, you have direct access to the result polyhedral mesh. This can be fed back into an FEA solver for additional validation. The downside to this pre-determined mesh is that you are limited to the mesh as-is and can't refine it to remove singularities or concentrations due to large element size near small features.
The automatic BRep conversion of FEA mesh is an arena where a lot of CAD and CAE firms are preparing to square off. This new wave of generative tools has us all realizing just how awful mesh data really is. Lots of companies are trying their hand at this in various ways. The hardest part is reconciling known details with the mesh. For example, if I have an optimized part that had round features for bolt interfaces, the true roundness is gone. The mesh has approximated the curve and left me with a faceted model - one that ends up leaving super nasty surface artifacts. Take a look at this cropped image of a Generatively Designed bow (from an Autodesk Artist in Residence at Pier 9). You can see just how janky that faceted surface is on the machined part. The finishing strategy was an attempt to hide the facets, but they are still very visible.
One strategy that is somewhat of an attempt to keep the wheels greased until good CAD data can be extracted is to leverage boundary approximation tools to run a structural analysis using the STL input. This type of CAE analysis doesn't break the part into small parts but rather looks at the entire design (which also removes the need for defeaturing of complicated parts and assemblies). One external tool that works really well with Fusion in this space is SimSolid. There are other tools on the market that do this as well.
Sorry for the long post!
@I_Forge_KC Awesome Post!
As I understand it the technology that has the potential to be able ot achieve this is called U-Splines.
