Generative safety factor vs. Simulation safety factor: an order of magnitude apart
Hi nic,
Looking at the picture you attached, I agree that what you're seeing isn't the stress concentration issue. One quick sanity check -- is the material in the simulation environment the same as the material of the outcome you exported? I'm guessing it is, but it's worth checking as a mismatch in material properties would explain this behavior.
Can you attach the Fusion 360 archive (File -> Export -> Fusion 360 Archive) for your generative setup and the outcome you exported, either to this thread or in a private message, and I'll take a look and see if I can see what's going wrong? Please also tell me the name of the generative study you ran as well as the name of the outcome you exported from Explore.
Thanks,
Ben
Ben Weiss
Senior Research Engineer
Hi nic,
I think we finally know what's going on! It took a lot of digging, but what you're seeing is a bug. Your report and willingness to share your setup with us made it much, much easier to diagnose the problem. Thanks!
It turns out there’s a bug related to the Flip Direction toggle for moments when we transfer loads from generative design to simulation. The issue causes the moment to twist the part in the opposite direction you expect (opposite even than the arrow is pointing). Notice how in the result below our intuition wouldn’t expect the part to twist clockwise when the applied moment appears to be turning it counter-clockwise.
We’ll work on fixing this on our end, but for now you can correct the simulation by editing the Moment load and clicking OK without changing anything. At least on my end, that seems to clean up the bad data and produce the expected simulation result. If you're still seeing issues, try deleting the moment load and re-creating it using the "Vectors" method of input (which bypasses the Flip Direction toggle):
Now I see a more intuitive distortion given your loading conditions:
Let’s compare this new simulation result to the values we expected from generative design.
Displacement is slightly higher in the simulation environment (3.28 mm vs. 3.12 mm, or +5%), but factor of safety is still low (though not nearly as bad as it was before) at 0.6 vs the expected 2.0. However, as noted in the article you linked to in your original post, we may see some stress concentrations near the interface with the preserves. If we look at the regions of the model with safety factor less than 2.0, we see a few red specs near the preserve transitions, and a region of yellow where the T-Spline body got cut by the obstacles. Even these yellow regions are all above a safety factor of 1.5, which is consistent with the level of accuracy I expect given the shape changes in the T-Spline conversion and simulator differences between generative design and the simulation environment.
We've got the bug report filed with the correct team to handle the problem. I can't guarantee it will be fixed in the next release, but at least we have a workaround that gets you back to a sensible solution.
Thanks again, and my apologies that the simulation tool simulated the wrong loads!
Ben
Ben Weiss
Senior Research Engineer
hi to all,
I'm trying the generative design but the result of the static simulation gives a different safety factor, always lower than the limit of safety factor 2 of the generative design, as you can see in the image below many regions, even far from the preserved geometry, show an unacceptable stress level.
Where is the problem?
how can I achieve a good result?
Hi @a.fusi001,
This level of deviation from the target safety factor is likely caused by the slight differences in the shape introduced in the conversion to a T-Spline body, as well as the differences in the simulation mesh used in the generative engine and the simulation environment. Some of that effect is localized at the preserves, but we also see small deviations throughout the body, and your model just happens to look a bit worse than the one in the reference document linked in the original post in this thread.
This is expected behavior; you're not doing anything wrong, and the software is functioning as expected. You could try re-generating the design with a safety factor target higher than what you actually need (perhaps 2.5), or modify the T-Spline of your current design to thicken some of the regions that are showing low safety factor.
As an aside, it's probably best to start a new thread with a question like this so the whole community can see and help.
Thanks,
Ben
Ben Weiss
Senior Research Engineer
