I promised you that I would make a summary document of the different episodes.
Some episodes are already on the forum: but the synthesis, completes the episodes and corrects typos and spelling. If you find any typos or errors, don't hesitate to point it out so that I can modify the PDF.
Happy reading
PS: I wrote this document to make up my own mind on the subject, so it may include prisms due to my perception of things. Be Indulgent
Hello Zozo_mp, Thank you again for sharing, I just finished reading what you just wrote, it sums up well what AI will or will not be in the near future. It's a very nice approach, which scared me, but thanks to you less ... For the moment, to date... To be continued. @+. AR.
As for 3Dxperience, I personally think that all this has been pushed by marketers whose one and only objective is to make users totally captive (once all your data is in THEIR cloud, try to change software...). And once well captive, long live the waltz of increases, limitations of services... For the sharing of comments between departments, if you are not in the big brother company with a design office in France, marketing in the USA, the subcontractor in India and the assembly plant in China, I don't really see the point: a good physical meeting in front of a video projector showing the state of the study advantageously replaces this tool (knowing that you can also do this kind of meeting remotely with common tools such as Teams or any other online meeting software).
If anyone has an example of a real 3Dxperience payout, please don't hesitate to share it.
I feel like we're going to have a quick laugh here: Auto-Generate Drawings (BETA) is probably the most exciting new AI functionality. It generates a drawing from your selected CAD and automatically adds standard views, section views, dimensions and annotations, BOMs, and revision tables. While it is not intended to complete your entire drawing for you, it takes care of most of the setup of your drawing, allowing you to add and customize for a polished final drawing Go Engineer's test (short video) almost looks like the pile of odds of the automatic import. The AI just seems to add the 3 or 4 views needed for this fairly simple piece. Managing to put enough distance between the views to avoid overlapping dimensions/annotations does not seem to be very difficult to achieve.
AURA (BETA): this is the SW help where you have to make a sentence instead of looking for 2 words
And although we pay dantesque maintenance fees, we are not entitled to the ' Selection Helper' ' Skecth Helper' or the ' Mate Helper' if we don't use their online tool→ We pay more and more for less and less new features (except for bugs of course)
On the other hand, if they manage to do something that holds up on the ' Mesh-to-parametric CAD ' I would then lower my hat (this would mean doing a finite element topological study of a part → volume capable (this is the current stage) → real parametric part (so a priori modifiable) under SW). When I see the current level of function recognition on imported basic prismatic or revolving parts, they have their work cut out for them to properly define the biomechanical forms released by the topological study (spoiler alert: 99.999% chance that the model is immodifiable and almost impossible to rate).
My friend @froussel Exceptionally, I find you a little harsh on topology. Indeed, you can set that you want to make the object in CNC for example. Which is quite cotable. On the other hand, it is recommended to simplify the first model to avoid untimely radii and kilometers of machining with a ball mill Personally, what I find interesting are the solutions he finds that he would not have thought of in my vision of the world. And then I reinterpret in classical design.
I deviate from AI because tolopological opimization deserves a subject in itself, knowing that topological optimization is a rich man's sport and that is only useful when weight is the first determinant.
In practice, the SW2020 function recognition works very poorly (on a fairly basic part imported in step or parasolid for example: commercial part not too complicated to be able to easily recreate a family of parts from a size). The few times I've tested it, it's very slow (selection function by function) for a bad result in the best of cases (and no result at all in the worst). So since it already doesn't work well for simple parts, I don't dare to imagine the result on a topological optimization output.
