HyperMorph optimization runs only for one iteration
Hello,
I’ve recently been trying to learn topology optimization using HyperMorph and stumbled across the following static analysis problem. I have a simple, 2D fitting with applied loads and constrains. I would like to minimize mass by manipulating the inner and outer radii, as well as four linear dimensions while still not exceeding 600 Von Mises MPa. The solver runs for only one iteration, then stops with a comment that it has reached feasible design. The design however is nowhere near perfect – the maximum Von Mises stresses come to ~200 MPa. I’ve tried manipulating the bounds, checking for collapsed elements but thus far everything was in vain.
Below I’ve attached my model. I’d be really grateful for some hints and suggestions. Thank you.
Answers
-
If you look in the output it states that the run has stopped due to excessive element distortion, you have 2 issues, 1st the bounds of your first desvar (inner radius) allow the hole to go to a negative radius, and 2nd, the design is able to meet the requirement very easily, so very large shape changes are needed to get up to the 600MPa, I reduced all your bounds to +/- 0.3 (and disabled element quality checking), that converges quickly, so I applied those changes and ran it a second time
After that, it again hits the bounds of the desvars (so a soft convergence), the max stress is up to 544MPa then, but you can see the elements are quite distorted.
You should review your initial design vs the constraints and start with a smaller starting point
I attached my mods (QuestionPS.hm) and also a version with a refined mesh that starts smaller and ends up closer to your bounds (QuestionModified.hm)
My final point, is, at the outset you said you are looking at topology optimisation, but this isn't topology optimisation, it is shape optimisation. Typically you would do topology optimisation first to get a basic loadpath, then interpret that and do shape optimisation to fine tune it.
1 -
Paul Sharp_21301 said:
If you look in the output it states that the run has stopped due to excessive element distortion, you have 2 issues, 1st the bounds of your first desvar (inner radius) allow the hole to go to a negative radius, and 2nd, the design is able to meet the requirement very easily, so very large shape changes are needed to get up to the 600MPa, I reduced all your bounds to +/- 0.3 (and disabled element quality checking), that converges quickly, so I applied those changes and ran it a second time
After that, it again hits the bounds of the desvars (so a soft convergence), the max stress is up to 544MPa then, but you can see the elements are quite distorted.
You should review your initial design vs the constraints and start with a smaller starting point
I attached my mods (QuestionPS.hm) and also a version with a refined mesh that starts smaller and ends up closer to your bounds (QuestionModified.hm)
My final point, is, at the outset you said you are looking at topology optimisation, but this isn't topology optimisation, it is shape optimisation. Typically you would do topology optimisation first to get a basic loadpath, then interpret that and do shape optimisation to fine tune it.
Fully understood. Thank you so much!
0
Altair Employee