OptiStruct Free Size with strange pattern results

Hello,
I have an optimization model where both mechanical and thermal loads are important. The goal is to optimize the thickness of three SHELL components with a Free Size.
The results show undesired jumps from minimum thickness to maximum thickness in adjacent elements. I wonder if there is anything I can change to avoid this.
Find attached screenshots of the thickness results in the las iteration and the DSIZE card parameters.
Thanks!
Answers
-
Hi Alfonso
Not easy to pass a statement here.
Maybe there is a numerical issue by dealing with very low numbers.
Maybe the model is unstable.
You need to provide more details:
- Does the baseline analysis w/o optimization provide reasonable results?
- What loadcases you are running?
- What is the optimization setup (Responses, Constrains, Objective)?
- Did you look at the optimization's final HyperGraph curves?
How do the Objectives, constraints etc perform over the iterations?
- Are there any Warnings or messages in the .out file which may give a hint?
…
Regards Juergen0 -
Hi Juergen,
Thanks for your time
What do you mean with low numbers?
The geometry of the model consists of a large SHELL mesh where the mechancial loads and SPCs are applied. The previos pictures are just the design space (also SHELL). There are also some SOLID elements near the design space. Contacts between meshes seem right as the analysis w/o optimization is susccessful.
I believe the problematic thickness output is caused by the thermal loads. I have runned analysis separating mechanical and thermal loads and the problem only occurs with thermal loads.
Answering your questions:
- The analysis w/o optimization is fine, the results shows a max deformation of 5.9 mm and max stress of nearly 800 MPa.
- I am running only one loadcase with some mechanical loads and thermal loads (provided with temperatures in each node and a thermal expansion coefficient in the material card).
- These are the parameters of the optimization setup:
RESPONSES
- total Volume of the model
- Von Mises stress design space (PSHELL)
- Von Mises stress nonDesign space (PSOLID)
CONSTRAINTS
- maxstressVMdesign = 650 MPa
- maxstressVMnonDesign = 650 MPa
OBJECTIVE
- minimize Volumefrac
DESIGN VARIABLES
- freeSize: lowerbound = 0.5mm, upperbound = 10mm
- Next images show the curves of the max constraint violation over the iterations and the 3 warnings in the .out. It seems that the strange thickness values that I showed in previous post is causing some stress concentrators.
Thanks for your help
Cheers,
Alfonso
0 -
Hi Alfonso
Watching the curve, it looks to me that the model is physically not able to fulfill the requested stress target.
It is violated from the beginning and gets only worse from there with the change of the design variables.
In such cases Optistruct ignores the specific constraint and tries to find an optimum with the remaining setup. » Infeasible design.
Hint:
In case you have singularities in your model (SPC attachment points. Connection points of RBE2…) you must not select elements attached to such points for stress response in optimization.
Kind regards
Juergen0 -
Hi Juergen
I solved the issue by relaxing the stress constraints so that the zero iteration fulfills it. Now, the constraint violation % graph has some peaks (due to increased penealty factors in the SIMP method) but finally converges to a 0% constraint violation feasible design.
Thanks for your advices.
Kind regards,
Alfonso
0