When I look at the world of simulation over the last 35 years, many things repeat themselves, albeit in a slightly different form, but the problems and the approaches to solving them are similar.
Something that has never really risen from the mud and made it to the top of numerical creation is the algorithm-based optimization of components. What could be possible here is rarely used. In most cases, the optimization of components is limited to manual variant analyses, 1-3 at most. But only if the pain is high enough. At the end of the day, it's always about the money, rather than what would be technically feasible.
Let's take a look at the optimization of a component in a way that everyone can understand: without technical jargon!
A sculptor working on a block of marble has an ideal model in his head and only has to cut away what is outside the optimum shape. If he accidentally knocks off the nose of the Sphinx, as in Asterix, he's out of luck. He must therefore move closer and closer from the rough shape to the detailed shape, as he can only take away material. If he takes away too much, the figure is no longer worth anything.
An artist who works with clay has an easier time of it. He can take something away, but he can also stick a lump of clay where something is missing.
I can speak to this, as I have taken both a sculpting course and a modeling course. I have attached the proofs. If you have understood the above correctly, you have also understood the principles of optimization using the finite element method:
Topology optimization removes unnecessary material and improves the structure. Shape optimization specifically changes the surface geometry in order to minimize stresses. Size or parameter optimization adjusts material thicknesses, wall thicknesses or cross-sections. Let's leave out size optimization and material optimization, as they don't fit into my nice analogies.
Machining the block with a rough chisel, removing material that is not needed, but the main work now lies in the detailing work until a polished surface is available afterwards. Incidentally, this process would also be suitable for the administration of our country, but that is another topic
The beauty of the computer is that you can now change the method, in effect transforming the block of marble into clay. The rough shape should already be right. Now you trim the whole thing by adding or removing clay. Major changes are now more difficult. Instead, you can easily smooth the surface and make it beautiful.
If you have so-called shell structures (e.g. sheet metal parts in the body), you can again use the example with the clay, where you simply press in dents with your finger so that the whole thing becomes stiffer and doesn't start to rumble like a cheap floor panel on an even cheaper vehicle.
It is therefore quite clever to combine both methods, topology and shape optimization. Is it worth it? It all depends. In the end, it's always about money. If I put in more effort, it has to pay off in cash in the end. Components that end up being painted over with fat safety factors because the standard demands it, or the fear of the most unlikely load case combination dominates, are excluded here.
However, we can already think of customer projects where this effort would be worthwhile. Namely, when it comes to borderline areas where every gram of weight or the highest possible rigidity in the smallest possible space with tolerable stresses is at stake.
I have now been contacted via LinkedIn by a company that has realized the above-mentioned solution (the combination of both approaches): The company FEMopt Studios GmbH with their software called XCARAT.
How well does it work? I don't know (yet). Hence my suggestion:
Do you have a use case where it would be worth the effort to optimize as accurately as possible because other approaches fail here? Let's test the software together on a joint project. In my opinion, far too few new things are tried anyway. Perhaps there are benefits for all sides.
So that you at least get a small technical snack: The attached video shows the topology optimization of a plastic component, including the final implementation. It might look nicer, but we couldn't find anything better in a hurry (which we are allowed to show).
I look forward to your comments and questions.
Yours, Stefan Merkle
PS: The XCARAT software can be embedded in the familiar ANSYS environment.