CAD & composite structures

[johnheritage]

This isn't all CAD related, so if CAD isn't your thing you might still be able to help.

I'm starting to use FEA with my CAD work, but even on the higher end packagaes like Solid Works, it's limited. I can only work with a single material for each component.

I've seen an evolutionary based design optimiser that can run through a database of materials and then, by guided trial and error, determine the best material choices, sizes and shapes - which is a kind of topological part optimisation with material selection as well. In this example, it was designing a flyweel;

http://garage.cse.msu.edu/demos/index.html

(the videos demonstrate the potential of the tool)

I've also seen similar algorithms used to layout PCBs.

I'd be interested to know if anything similar exists for choosing a composite system that is more mainstream. Say for example, I want a panel of X tensile strength with x.y.z possible materials in it and of such and such dimensions.

I'd like to consider composites more for my work, but I'm not happy using them unless I know I'm getting the best I can from them - otherwise the effort begins to make me wonder if it's paying off.

For any given system, there must be ideals for things such as core size and type.

Surely there must be some logic to selecting these parameters beyond hands on trial and error or simply assuming it'll do. Are there any texts I can read online that go into the detail of how to design multiple layer composites and choose each layers thickness? The big aerospace labs must have gotten some kind of art into this by now.

Thanks!
John

 

[MikeHalloran]

Sometimes the problem space gets constrained for you.

Example: I worked in a place that used phenolic for its heat resistance, and glass for its price. They'd been doing hand wet layups for a while, with no complaints.

Then our Marketing Manager bragged to a customer that "Our stuff is so hard that you can't dent it with a hammer."

They tried. They did, rather easily.

Thereupon, we ('we' not including the Marketing Manager, of course) were tasked to:
- Come up with a standard hammer.
- Provide one such standard hammer to the customer.
- Retain one such standard hammer and use it to test production lots of items going to that customer.
- Add plies until that customer's product could reliably withstand application of said standard hammer.
- Trim the substrate extra carefully, and make a trim- checking template and a finished article checking template, because the now thicker laminate was interfering with other features of the customer's product.

Bragging only cost us a hundred grand or so of lost time, extra tooling, development cost, email exchanges, trips to the customer's facility, visits to our facility by the customer, not counting extra product cost and production paperwork.

Sorry; just had to tell the story..







Mike Halloran
Pembroke Pines, FL, USA

 

[kchan711]

Hi John,

If I'm not mistaken, within the new Solidworks Simulation Premium (2009), which replaces COSMOS, it contains an in-depth composite testing simulation. I believe it includes the selection of multiple materials and grain directions as well with common composite examples. It may be something to look into.

-Kyle

Kyle Chandler

http://www.chiefengineering.net

"To the Pessimist, the glass is half-empty. To the Optimist, the glass is half-full. To the Engineer, the glass is twice as large as it needs to be!"

 

[RPstress]

Most laminates are still optimised by hand. Some limited computer optimisation is possible where the ply materials and angles are variables (note that ply thicknesses are rarely varied, as standard materials tend to be procured for a given project), but the problem rapidly gets out of hand for a resonably sized structure. The BigDOT optimiser can go some way to handling the sort of problem size encountered, but you quite often wind up with some sort of two-stage optimisation with smeared (rather than ply-specific) properties in the final runs.

For flywheel energy storage the structure is relatively simple and what needs optimising seems relatively clear. I'd have thought that optimisation could help significantly, even with smeared properties, and those demos seem to show that.

Most optimisers just wrap around a solver and there should be no arbitrary system-imposed limits on what part of the structure can be what material. Once you've got an FE model and have defined what FE parameters are variables, including material properties, the system should do the best it can. (Which isn't really good enough for the very large number of variables you get for varying every bit of every ply, or at least it wasn't the last time I looked into optimisation.)

For your panel strength query, for a simple panel with a single requirement (tensile strength) fairly thorough optimisation on a ply by ply basis should easily be possible. But then again, that should be more or less doable by hand. It's when you get into real structures and your trying to optimise for a lot of loadcases and to minimise a number of properties (not just mass, but a cost-weight compromise with manufacturing restrictions on ply blocking, etc.) that it gets hard for the optimiser to cope.