Mr.Hyde wrote:
@Epic-o: I used an old version of CosmosWorks the SolidWorks module. This also is able to handle anisotropic material data, but it did not work the way I tried to operate it... But I also had problems to define the properties in the right planes for every tube. A seminar for this software would have been great, I think
.
Finite element analysis of composite parts is actually quite difficult. Most codes (AKA programs) don't handle anisotropic materials well. They may let you define the materials properly, but indicating the reference direction (the 0-degree direction) has always been quite a challenge. Only a few codes are good at this.
Nastran has long been dominant in composites-dominated aerospace, largely because there are a wide variety of tools for indicating material direction. A few other codes are catching up, though. I'm an ANSYS user, and I've been quite pleased with ANSYS' new composites tool, ANSYS Composites Prep/Post. ABAQUS has a composites module that looks reasonably good, though I haven't played with it myself. There are also some code-independent tools (e.g., FiberSIM) for "laying up" your part; you then import your "laid up" mesh into whatever FEA tool you like.
Different engineers like different FE codes, and they can get quite passionate about their preferences. Personally, I'm not like that. I like ANSYS, and I know it well, but you can do a perfectly good analysis in any reasonably sophisticated code. Unfortunately, CosmosWorks is not especially sophisticated. If you can get your hands on a different code--especially one with a composite materials preprocessor--I'd highly recommend it.
Another thing to keep in mind is that your material properties are probably optimistic (unrealistically high). Most composite material properties are determined via tensile tests of unidirectional or woven plies. The test specimens (AKA coupons) for these tests are typically cut from flat plates. The problem is that flat plates are
very easy to manufacture.
A frame like, say, a Cervelo S3 is typically laid up by hand into a female mold and then autoclaved under pressure. This introduces a huge number of variables. There are a vast number of in which the real-world frame can deviate from the FE model. Voids, resin pools, errors in fiber angle and material condition all contribute. The result is that the FE model is an idealized form of the real-world part--and to a much greater extent than would be true for a part made out of, say, 7075-T6 aluminum.
I worked on one project that involved a carbon-fiber riser bar. In the prototyping stage, we were designing the bar to survive a 1000 lbf (4.45 kilonewton) load applied in the grip areas. The parts we got back from the factory failed with a standard deviation of 20%. That is, some of them failed at 1200 lbf and some at 800 lbf, and a few bars failed at higher or lower loads. This is a much bigger variation than you'd get with an aluminum bar.
Applying FEA to composites is still worthwhile, but it's important to understand the quirks. And as with any FE analysis, you'll need to validate your model with physical testing. An unvalidated FE model is a dangerous thing.
If I can answer some questions about FE modeling of composite structures, I'd be glad to do so. This looks like an interesting project; good luck!
Jason
FAQ
Search
Login
Register
. As you might have read, I will post news about that in my 
)
. You really do a great job, beautiful bikes and a lot more output then I have 
