free shape optimization

[api503]

hello,
iam quite new to the optimization field and its a challenge to undertake this certain problem....
i wanted to know the exact procedure to perform a free-shape optimization of a certain component. i am already through with the meshing and static analysis of that component....but i was clueless on how to go forward with the optimization.
i find the help in hypermesh a little complicated to understand.

 

[rickfischer51]

First what software are you using? It sounds like you may have general purpose finite element software, not topopology optimization software like OptiStruct.

There are two approaches. First is a general design optimization where a parameterized model is run through a DOE-type optimization that alters values of the critical parameters. You are starting with a shape and changing proportions. The other approach, which is I think what you want, is to start with a blob of material that represents the part's maximum envelope, then whittling away until you get an optimized shape. This is usually referred to as topology optimization.

Back in the middle '90s I experimented with the element birth and death feature in Ansys. What I did was create a block of elements, apply loads and boundary conditions, run a structural analysis, then go into post and get the strain energy of the elements. I then killed any elements with strain energy below a certain value, and reran the analysis. I iterated this process until a critical level was reached, typically a stress level was exceeded or a certain percentage of the elements were dead. This was simple to do in Ansys with the APDL macro language. I used a mapped mesh of brick or quad plane elements, with no tet or tri elements. This produced a general shape determined by the load path. From there I made a parameterized solid model of this shape, and analyzed it with the Ansys design optimization module. This procedure worked ok, but sometimes produced shapes that were not easy to make with the intended process. Shortly after I left that employer, Ansys added a topology optimization module to Ansys. I've never used it, but I believe it works in a similar manner.

Ansys implements element birth and death by reducing the element's stiffness, I think by 1E6. The element is still there, but just so flexible it contributes virtually nothing to the structural stiffness. At the time, Ansys would not allow you to change material properties after the first solve, but that now has changed. An approach that should work better is to gradually reduce stiffness in increments. Something I worked on was element resurrection, where certain dead elements were reborn. I did this to reduce the effects of checker-boarding. Gradually changing element stiffness would I thing reduce or eliminate checker-boarding.

Alas, things are changing at Ansys. As of R14.5.7, design optimization is complete gone from MAPDL (aka Ansys Multiphysics, Classic, black screen), and topology optimization has been undocumented. Ansys seems intent on removing the Multi from Multiphysics (Monophysics?!!) so they can sell you an add-on that do the same thing. You can always program this with APDL, but there has been talk for years that the marketing buffoons want to take that away too.






Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[GregLocock]

Progressively killing elements should result in a local optimum, it may not result in the global optimum. The OP mentions using hypermesh, so I guess he has access to optistruct. I found the tutorial on the suspension arm was easy enough, where's the issue?



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[api503]

first of, thank you for the replies....
my problem was that i never found a tutorial on "free-shape optimization" that i actually understood, all i found was on topology and free-size....therein lies my problem...and i have gotten it all jumbled.
and yes I have access to optistruct.
what i did was, mesh the component with hypermesh and then check its mesh quality(like skew, aspect and also tet collapse etc) and ran the static analysis in both optistruct and ansys 14.5 which got me quite similar results of stress values. then I was clueless on going forward with it in terms of optimzation and i was told to perform a free-shape optimization, and since i'm new to all this i jumbled stuff up.
I shall follow up on what you have put forward.
also if someone could give me a link to the procedure of it would be much appreciated.
thank you.