Composite size optimization with discrete design value

Altair Forum User
Altair Forum User
Altair Employee
edited October 2020 in Community Q&A

Hello,

 

I am performing size optimization of a composite structure containing several zones. The objective is to minimize the mass while respecting displacement, composite failure and buckling constraints. I have twelve design variables representing different orientation thickness in different zones. Each design variable is linked to a discrete design value in order to account for a realistic ply thickness (multiple of my unitary ply thickness). My properties (PCOMP) link linearly one design variable to my components.

I am running my job through a nonlinear quasi static optimization (NLSTAT) because of an elastomeric component in my model.

The analysis is going well except that I can’t understand the way optistruct actualize my design variables:

-          -I set initially my design variables to my upper bound

-          At iteration 0 all constraints are satisfied

-          But then the solver only decreases all the design variables through the different iterations. This makes constraints violation more and more important.

-          At the last iteration all my design variables have reached my lower bound and the design is qualified unfeasible while all the constraints were satisfied at the beginning of the analysis.

How does optistruct determine the design variable for the next iteration? I suppose there is an algorithm to converge toward an optimum solution but in my case it looks like my design variables just decrease from upper to lower bound. So I suppose there is a problem in the way I set my optimization problem. Do you have any clue of what that problem could be?

Thank you very much for your help.

Adrien

Tagged:

Answers

Categories