FEA - failure location in tensile test

[puli888]

Hello FE Gurus,Please help me to understand the failure mechanism in FE. Lets consider the simple 1-D tensile test. During actual test,the sample starts necking arround the location where the material is weak within the gauge length( may be porosity or some other material functions). How this can be compared in FE? In FE we use same element property for the entire gauge length.Theoriticaly all element must get same stess and strain wihtin gauge length until the weakest element is identified, and then the necking should occur. My question is how this weakest element is identified in FE?
thanks,
puli

 

[corus]

With large displacement, poisson's effect, and yielding, you should see the same effect in FE.

corus

 

[prost]

Right corus and puli888, you must have the proper mathematical formulation to even get close to modeling reality. In the case puli888 cited, the FE analysis must use a large displacement, large strain formulation (normally requiring a nonlinear iteration to solve the equations) with a material constitutive relationship that allows large disp./large strain. To expect Hooke's law to adequately model the necking process is out of the question!

 

[puli888]

Hello coros/prost,
Thanks,but my question is not about simulating the necking behaviour,rather about deciding the location of necking/failure. Since the cross section of gauge length is uniform,all element with in the gauge must get same stress & strain.Then how come suddenly few elements in one location started necking in FE?
Thanks again,
puli888

 

[sendy]

Hallo puli888,
I think your question is related to bifurcation problem. I agree with you for single homogeneous element test there should not be a bifurcation, unless there's a numerical rounding error, which (i think) may lead to 'somehow' wrong result. However, when you consider higher more elements for the test (i.e. plane strain compression, tensile test) and with proper constitutive equation, I think there might be a bifurcation (or shear band). As far as I know this is not a must, and this shear band depends on the constitutive model that is used, i.e. the elastic and or perfectly plastic (i think) won't give this kind of result. Model that may give this result is the ones consider softening behaviour (what i know is hypoplastic model since i'm studying about it).

I am also interested in when does bifurcation start to occur.

Note: Please don't believe my opinion 100%. Nice posting.

 

[prost]

Sorry for my misunderstanding of the question, which I interpret to be "how can I predict where the coupon will fail prior to the test?" If you have a nice long constant section for the gage section, and the stress load is uniform, than why does the section fail at any particular location? If you tested 100 such coupons, you would observe many different failure locations. What could explain that? Heterogeneity in the material, asymmetry in the load, and imperfections in the test coupon surfaces all could explain these differences. If you tested a bunch of these coupons, you could get an idea of the statistical distributions of failure locations, because these material behavior is stochastic; the material behaviors vary from coupon to coupon. Therefore any model you use to predict where the coupon will fail would have to consider the variation in the material behavior by using statistical distributions of that behavior.

 

[puli888]

Hello Prost,
I understand the physical test very well and agree with you 100% on your explanation. But my question is about the FE simulation. How the failure/necking location is predicted in tensile test(FEA)?...
Thx,
Puli888

 

[prost]

If the behavior is stochastic, as material behavior clearly is, how can it be predicted with other than stochastic models? As far as I have experienced, you can't predict the failure location in the actual test, much less in a FE model, which is an imperfect model of reality.

 

[rb1957]

puli888,

as you noted the real failure location is due to a randomly distributed imperfection within the material. Therefore i don't think you can predict the precise failure location prior to the test. The shank of the test piece is considered to be all the same material, so i'd expect that FEA would predict failure close to the ends, where the test section transitions into the loading fixture.

i think the most you can hope to predict is that the failure load is higher than the standard properties, since these are "knocked down" to ensure that most of the material is stronger than the handbook values.

if you wanted to, i guess you could include a spherical inclusion (0.001" - 0.002" diameter) in your model, which would initiate failure.