I count " adaptive algorithms " as Artificial Intelligence (but I understand your disagreement).
Example:
Adaptive mesh adaptation (this feature automatically adjusts the density of the mesh in areas where the solution requires more precision) or Time " step " adaptation: This feature allows the solver to automatically manage the size of time " steps " during the transient simulation.
Even if it's not at the level of " machine learning ", the " program " makes a choice (not always the right one, but it does...), and it's this feature that I count as part of AI.
… The linear static study module includes an option that allows the solver to adapt the mesh during resolution in regions of the model with high strain energy errors. This results in a more accurate solution with little user intervention.
There are two types of adaptive mesh refinement methods: adaptive h and adaptive p. This guide explains the differences between these two methods and how to enable them.
Some FE calculation software (including SOLIDWORKS Simulation Professional & SOLIDWORKS Simulation Premium) incorporate a kind of intelligence: modification of the model by successive material removal in order to optimize the volume/weight. It's not necessarily real AI but it should still help converge towards a lighter/less expensive solution (especially in aeronautics / space, but it must also work for moving parts on machines). https://www.solidworks.com/fr/media/topology-optimization
Well, no! Calculation of the strength of materials (FEM/FEA) to date ditto, no AI. The calculation is performed via the finite element method (FEM), which is a deterministic solver.
Ditto for topological optimization No AI, I want to prove that when the mode of transition between Soidworks CAD and the topology requires retelling it a good part of the design. Prohibited area (he must not remove material in these areas) Force to be applied in particular in the axle holes or on the supports, torsion etc (teaching irer of the PEF sim. In addition, you have to tell him what type of machining you are going to use, 2 or 3 axis milling or 3D printing, not being able to specify which one if it is by metal deposition. Solidworks communicates the form and that's it. Like solidworks with simulation. If you have more precise information don't hesitate
The important thing for me before using an AI is to have the knowledge that is required of the AI, in order to know how to interpret the result. For a calculation, know how to do it by hand or via software, for code, know how to debug... It should not be forgotten that some studies have shown that more than 50% of AI answers are partially or totally wrong So saving time yes, but knowing how to verify the veracity of AI is mandatory for me, otherwise you risk a lot of things, especially in design.
It's true that if you only take GPT or Mistral there are errors or hallucinations, but if you have a critical eye on the result you can save time. I'm not talking about coding a macro or a program like a plug-in, an add-on.
A little more before I show you how to save a lot of time in the construction of simple parts or a little more complex and ASM (not in the way of the salespeople who do the big lion trick and their effects Whaouuuu without tomorrows). It's as if you were making this part yourself, but much faster and with position proposals, more advantageous functions in terms of cost, etc.
PS: Note that I only mention CAD, simulations and topological optimization. Someone who is familiar with PLM and ERP would surely see a lot of advantage in seeing what AI can bring because these are the fields.
Hi @froussel Topological optimization has the disadvantage of being a separate trèscherware from Solidworks and Simulation. The future is a transparent transition between these functionalities spread over three software programs that do not communicate with each other and finally the same functionalities in permanent interaction with each other. I will give an example that shows this underlying trend.
