Different results in composite size optimization and its results file

prcm

Hello,

I'm doing a composite optimization on a model. Did the free-size optimization, edited the ply bundles, the ply manufacturing thicknesses, and the design variables thresholds according to the manufacturing thickness. Included other manufacturing constraints.

The objective of the composite size optimization was to minimize mass with a maximum displacement of 6.38 mm constraint.

The size optimization results converged with no constraint violations, and a maximum displacement of 6.25 mm.

 

With the 7th iteration resulting file did an analysis to check the results and the supposed displacement of 6.25 mm changed to 10.64 mm.

 

I've compared the .out file from the size optimization with the resulting plies created in the .fem file and they are equivalent.

The laminate stacking in the arms and center is symmetric smear in free-size and size optimization and symmetric in the shuffle/analysis. The wheel barrel laminate is not optimized.

What am I doing wrong?

.fem files: https://we.tl/t-lsCU42z3e0

Adriano A. Koga

one question:

 

in your Optimization convergence curve, it is shown ID 917134 for the control node. But when you mention analysis, you're looking at the maximum value at the model, at node 862645.

Are these close to each other?

Looks like in your optimization you're measuring the displacement somewhere else, and not at the maximum.

prcm

They are close (around 30mm apart), both on the end of the loading arm

PaulAltair

Did you share the correct files?, if I rerun the size opt file you sent, it is not able to get anywhere near convergence: (I think due to manufacturing constraints)

prcm

Yes, same files

I redid the size opt since you got those results, but got exactly the same answer. Results converged after 7 iterations, with all constraints satisfied... 

Can it be some problem with my optistruct?

PaulAltair

prcm said:

Yes, same files

I redid the size opt since you got those results, but got exactly the same answer. Results converged after 7 iterations, with all constraints satisfied... 

Can it be some problem with my optistruct?

Possibly, I am running it in 2022.2 (latest release), I narrowed it down to the PLYDRP constraint, it has a TOTDRP limit of 0.5 and this is not achievable when combined with the requested symmetry and balance constraints, if I run without PLYDRP or with TOTDRP set at 5.0 instead of 0.5, then the sizing phase can converge. If I get chance later today, I will run the shuffling that comes from that Size run and see if it is any better.

 

prcm

Paul Sharp_21301 said:

Possibly, I am running it in 2022.2 (latest release), I narrowed it down to the PLYDRP constraint, it has a TOTDRP limit of 0.5 and this is not achievable when combined with the requested symmetry and balance constraints, if I run without PLYDRP or with TOTDRP set at 5.0 instead of 0.5, then the sizing phase can converge. If I get chance later today, I will run the shuffling that comes from that Size run and see if it is any better.

 

I am running in 2021, I'll try and reinstall it, to see if that changes anything.

Thanks, let's see if it is

prcm

Paul Sharp_21301 said:

Possibly, I am running it in 2022.2 (latest release), I narrowed it down to the PLYDRP constraint, it has a TOTDRP limit of 0.5 and this is not achievable when combined with the requested symmetry and balance constraints, if I run without PLYDRP or with TOTDRP set at 5.0 instead of 0.5, then the sizing phase can converge. If I get chance later today, I will run the shuffling that comes from that Size run and see if it is any better.

 

I reinstalled my Hypermesh and Optistruct solver, redid the optimization and got similar results to yours. Some error in the optimization solver last install made it give these wrongful results...

Thanks for your help, nonetheless

PaulAltair

prcm said:

I reinstalled my Hypermesh and Optistruct solver, redid the optimization and got similar results to yours. Some error in the optimization solver last install made it give these wrongful results...

Thanks for your help, nonetheless

I changed TOTDRP to 5.0, and changed the 2 designable stacks to SYM (not SYSMEAR), this gives a converged shuffling deck from sizing in 3 iterations, then that shuffling model is also able to converge, (input SHUF fem attached as 7z)

 

prcm

Paul Sharp_21301 said:

I changed TOTDRP to 5.0, and changed the 2 designable stacks to SYM (not SYSMEAR), this gives a converged shuffling deck from sizing in 3 iterations, then that shuffling model is also able to converge, (input SHUF fem attached as 7z)

 

Nice that it works.

I'll update my model with the stacking sequence, should have only used the SYSMEAR in the free-size, thought I had read that the SMEAR was to be removed only for the 3rd optimization step, but from rereading the documentation I noticed it's on the 2nd step.

And I assumed the TOTDRP was the tapper ratio but it is actually the TOTSLP or PLYSLP...

Thanks, Paul

PaulAltair

prcm said:

Nice that it works.

I'll update my model with the stacking sequence, should have only used the SYSMEAR in the free-size, thought I had read that the SMEAR was to be removed only for the 3rd optimization step, but from rereading the documentation I noticed it's on the 2nd step.

And I assumed the TOTDRP was the tapper ratio but it is actually the TOTSLP or PLYSLP...

Thanks, Paul

Another thing I would do is set your wcomp response as objective to minimise in shuffle phase, mass does not change in shuffle (as plys have mass and just move around in stack), so provides no sensitivity for optimisation