Solution convergence in Generative Design

[Anonymous]

Hello, 

I'm having issues understanding how Generative Design works... Based on a model, I ran 2 studies differing only from their objective type :

1. Stiffness increase objective ; Constraints : safety factor (SF)=1 and mass target = M
2. Mass reduction objective ; Constraint : SF=1

The same geometry results from both cases with a mass of ~4*M and a minimal SF of ~10. Firstly, I understand that all constraints cannot always be verified. However, in the case of Study 1., none has even been reached (in other words, the stresses are "low" and the mass in "high").

Anyway, I decided to investigate what happens if I reduce the SF in order to bring the stresses closer to the elastic limit. Since it is not possible to set SF<1, I left it set to 1 and manually increased the elastic limit. It should be equivalent, right ? However it had no effect on the results.

I thought about setting a maximum thickness constraint to force reducing the mass but the solution did not converge at all.

So could anyone please help me understand what's happening here ? How can I manage to reduce the mass closer to my target ? How could I possibly impose a maximum thickness constraint and make it converge ? Why do both objective types provide the same results in my case ? Are some constraint types prioritized with respect to others ?

Thanks a lot and best regards

Raphael

[Anonymous]

Hi again, 

After having kept investigating what could go wrong, I figured out that the problem might be that some preserved geometry bodies don't undergo any load or constraint at some load cases.

But the problem is, the preserved geometry bodies remain the same among the load cases of a certain study. And it is very likely that for some load cases, loads are applied to some bodies but not to others.

I would be very grateful to know your opinion and what you recommand... Try to combine the load cases such that all preserved geometry bodies always undergo a load ? But this cancels the "multi-load case" benefits. Within each load case, apply an arbitrarily small load on all preserved geometry bodies, on which no load is applied yet ?

Thanks a lot in advance and best regards,

Raphael