FEA Material Prop Requirments?

[oharag]

It's been a while since I've performed an FEA analysis. Usually I'll perform an analysis on material info provided by the FEA software.

I'm trying to perform an FEA analysis on a plastic part (in the linear range), and the Material Prop Sheet does not have a poisson's ratio. Is this a problem? Can you get away from not having a p.r. to perform an FEA analysis? I always thought you needed this propoerty.

I'm going to perform an analysis using Ansys 10 inside SW 2007. What are the min req to perform an analysis?

The material is supplied by UMG ABS Ltd (I believe out of China). It is a plating grade ABS that is licensed from GE Plastics. I do not know's it's equivalent on the GE side.

http://www.umgabs.co.jp/en/data/index.htm

http://www.umgabs.co.jp/en/data/ed/3001m.pdf

I checked some GE Cycolac ABS data on Matweb. The poisson's ratio is also not supplied there.

 

[GregLocock]

1) Why not run your analysis with different levels of PR. Then you'll know how sensitive your answers are, and whether you should be bothered. Real values of PR lie in the range -1 to +.5

2)First two google hits say 0.39 to 0.42

Cheers

Greg Locock

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

 

[Jordonlaw]

I agree with Greg both on the sensitivity runs and typical values of plastic PR. I run a fare bit of plastic FEA and have always used values in that range.

 

[oharag]

I received another data sheet from UMG showing Poisson's ratio. It is 0.392.

Thanks for the info.

So it's absolutely necessary to have Poisson's ratio inorder to perform an FEA analysis? I wonder how Matweb can export out data to solvers without having the PR listed?

 

[prost]

If you don't set the Poisson ratio, it will assume zero (0.0) if you are lucky, and some random number if you are not lucky, won't it?

If the Poisson ratio is not set, and it assumes zero, then it probably won't act much like a part made of plastic. Since plastic generally has a Poisson ratio close to 0.5, I'd guess you'd get some pretty odd behavior if you don't set the Poisson ratio (PR) and it assumes PR=0.0. What will happen? Say you pulled (with tension, that is) on a flat piece of this plastic. If the PR was set to 0.0, the bar would not contract at all; very odd behavior, don't you think? Aphysical even.

 

[prost]

It would be interesting to me to see what happens if you set the Poisson ratio to 0.5. I assume you have the equations correct; if you set the PR to 0.5, and it doesn't bomb (due to divide by zero), then I'd say you have the equations wrong.

 

[GregLocock]

rubber has a PR close to 0.5, demosntrably structures made from rubber don't bomb in the real world

Cheers

Greg Locock

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

 

[prost]

The question isn't whether the reality is stable, the question is whether the numerical method, in this case, the finite element method, is stable. A finite element calculation in which you blindly use the equations of linear elasticity WILL bomb if you set PR=0.5 (because there is a (1-2*PR) term in the denominator of the elasticity matrix used in the calculation of elements in the stiffness matrix); as you increase PR from 0.45 to 0.499 then to 0.49999 etc. you get the familiar "element locking" effect; purely numerical, nothing to do with reality. If PR is close to or equal to 0.5, as it is with rubber, you must resort to all sorts of 'tricks' such as penalty methods, mixed methods, etc. to mitigate the numerical problems.

 

[cameprak]

If youngs modulus and shear modulus values r available with material property database, poisson ratio must(may) be calculated automatically. isn't it. Am i wrong?!

cameprak

 

[rb1957]

cameprak ... that's just what i was thinking as i read thru the thread. i'd've thought codes would've been smart enough to fill in the blanks (given E and G, calc v). I have done some analysis where i tripped the element stiffness to be shear effective only, in which case i put in all three vlaues and NASTRAN nicely told me what the values were inconsistent (E=1E4, G=4E6, v = 0.3) but it'd carried on regardless.

 

[cbrn]

Hi,
yes, as E, G and nu are correlated (in other terms, only two are independent), then a FE engine won't use the three of them.
What the FE needs to know is the "rigidity to shrink" and the "rigidity to elongate" (OK, I express it very roughly), and this info can be provided by either of the two moduli together with "something" which undirectly expresses their ratio (i.e. Poisson ratio), OR by the two moduli altogether (without nu). If you provide the three altogether, then either you have them consistent or the FE program will have a built-in "preference" about how to use them. YOU MUST know how your FE works IF you want to specify the three values altogether, or you will most likely get unexpected / unwanted results !!!

Regards

 

[prost]

Given the calculation of G, shear modulus, from E, Young's modulus, and PR, Poisson ratio, is a straightforward calculation (G=0.5*(E/(1+PR))), anybody know why MilHDBK5 values of G, E, and PR don't seem to be similiarly connected? For instance, take Table 3.2.3.0(b1), for 2024-T3 sheet. E=10,500 ksi, PR=0.33, G=4000.
Calculate 0.5*(E/(1+PR))=3947 ksi. Or you can calculate PR from E and G: PR=0.5*E/G-1=0.3125

These E and PR results are probably the means of the E and PR from many many tests, right? (2024 is a oft used material, I ASSUME that many many tension tests for E and PR have been run and used in MIL5). Is G measured or calculated? I can think of a few reasons why the G calc. from MIL5 is not the same as the G quoted, but no one seems to really know:
1) 'G' can be measured directly with the right kind of strain gage arrangement--if this is so, G is measured, not computed from E and PR.
2) Roundoff error
3) Calculating G from E and PR for each individual test, then taking average of those calculated Gs. Obviously
G=1/N *sum(0.5*E/(1+PR), i=1 to N) is not equal to
G=(1/N)*sum(E)/(1/M *sum(PR)), where there are N tests to measure E and M tests to measure PR. There's an additional complication--there might be different number of tests to measure E than those to measure PR.
4) typo--not likely, every G I have tested by calculating from E and PR fails this test--my immediate conclusion is that they are all this way.

 

[gfbotha]

Hmmm, interesting thing this about the compatibility of the values listed in say, MIL5.

I tend to think that different experimental methods & error also might play a role (even more so when measuring PR? - smaller lateral strain range, etc.).

I would say the accuracy of PR is the less important parameter for normal structural materials & case studies but very important when analysing rubber applications.

Regards