Optimization - the wave of the future? 4

[ParabolicTet]

It used to be that computers were too slow to chug throug the dozens of iterations required to run an optimization solve. It was quicker to "manually optimize" the part in a few iterations. Now computers are fast enough to go through an exhaustive study. My observation is that the future direction of meshing tools will focus on those with good paramatric modeling capabilities tied in with a good optimization mesher, like Pro/Mechanica. What do you think? What other tools are good for design and optimization without having to application hop?

 

[Drej]

ANSYS, especially ANSYS Workbench, has some excellent optimisation tools. Workbench has an amazing shape optimisation tool which removes - and animates visually - redundant material within a part. Very impressive.

-- drej --

 

[feajob]

I am agree with ParabolicTet. Structural Design Optimisation through FEA would be a good expertise to get for the real world engineering in future.

I use Optistruct (Altair's product) for topology and shape optimisation of Landing Gear Structures. It works fine for me. But remember that there is no perfect software in this world. You may see some bugs, sometimes!

Airbus used Optistruct to reduce 500kg on Wing Design:

http://www.cimdata.com/newsletter/2003/46/03/46.03.02.htm

Have nice day,
AAY

http://www.geocities.com/fea_tek/asd/

 

[tanfolio]

Drej,

Does ANSYS workbench optimization offer the ability to plot the optimization to review convergence like PRO/Mechanica. For instance can you plot strain energy or von mises stress verses iteration?

Thanks,

 

[mloew]

ParabolicTet,

I agree strongly that optimization, enabled by increasingly faster processor speeds, can now be undertaken in many areas (not only structural analysis) and represents the future of engineering analysis. I must remark, however that the real future belongs to stochastic optimization. Just because a particular problem is optimized, it is not necessarily the best, or most robust, solution.


Best regards,

Matthew Ian Loew
"Luck is the residue of design."
Branch Rickey

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[Drej]

Tanfolio,

Workbench shows the converged solution of the optimised shape, after which you can view all the stress components necessary (normal, principal stress/strains, equivalent stresses etc.). It does not show the stresses at particular 'iterations' - I can't see why you would need that information anyway personally. Are you thinking of mesh convergence rather than shape optimisation? If so then yes, you can plot the convergence vs stress/strain (or whatever you have chosen to converge upon) within Workbench very easily.

-- drej --

 

[tanfolio]

Drej,

Yes, I am thinking of plotting mesh refinement vs stress/strain, etc. I plot p-loop pass vs stress and/or strain energy in Pro/Mechanica to check convergence prior to accepting the results as valid. The company I work for is standardizing its CAD and CAE software to SolidWorks and ANSYS. Since I am not familiar with ANSYS Workbench I was wondering if I could validate convergence properties in a similar manner using h-element mesh refinement.

Thanks,

Steve

 

[Drej]

Steve,

H-element refinement is the only method used in Workbench, so no problems there - in ANSYS, you also have p-method analysis available. In each pass Workbench outputs the convergence variable-vs-iteration pass, which sounds very similar to what you do already in Mechanica.

Cheers,

-- drej --

 

[magicme]

a star to MLoew for his remarks on design robustness versus design "optimization". years ago i witnessed several structures that were very highly optimized to a set of design specs. the designs thus evolved into minimal structures that delighted the design team and project team for there "efficiency". very late in the game the specs were slightly "rethought" based on newly avaliable data.....but the completed designs were not robust enough to withstand any change in specified loads or boundary constraints......an extremely painful and expensive (and project-delaying) re-design (new drawings, new purchase orders, etc etc) was necessary.
if we had been smart enough to understand robust optimization (or stochastic optimization), we wouldn't have looked so dumb in the end.

daveleo

 

[ParabolicTet]

Yes I agree that an optimized design should have a built in 20% factor of safety in case manufacuring can not hold the tolerances or if the design needs to be slightly modified for assembly reasons.

 

[mloew]

ParabolicTet,

Why the arbitrary 20% figure as opposed to a good stochastic optimization study incorporating the noted variability in dimensions, material properties, loads, BC, etc.? This type of analysis is certainly possible today with the appropriate software and data gathering.

Best regards,

Matthew Ian Loew
"Luck is the residue of design."
Branch Rickey

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[chris9]

I think 'intelligent optimisation' is the future. In other words optimising parts to reduce weight/material cost and also taking into account manufacturing processes (welding, forging, casting etc).

 

[mloew]

chris9,

Interesting point. Right now there are products (Altair's OptiStruct, for example) that can introduce manufacturing method specific constraints (draw direction, draft, symmetry, etc.) into the topology, topography, and shape optimization analyses. There is not currently a way to automate topology optimization with semi-material selection (materials and manufacturing process) to the best of my knowledge. This is still left to expert interpretation of the results of the initial topology optimization. The topology optimization step simply identifies the load paths required in a structure. Re-running the analyses with additional manufacturing constraints can assist in determining the "optimal" combination of topology and semi-materials. Again, this process is not yet automated and is still only deterministic optimization. Many possibilities for the future!

Best regards,

Matthew Ian Loew
"I don't grow up. In me is the small child of my early days" -- M.C. Escher

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.