Failure Criteria 1

[anupabq]

Hello Friends,
I am struggling to correlate FEA results with experimental results. I work on polymer parts reinforced with glass fibers. I have stress strain curves for the material till failure. My FEA results are 50% off from the experimental results. The model is setup correctly.
I think my biggest problem is deciding a failure criteria within the software. I don’t know how to come up with a criteria that will decide the load at failure. Do I look at the load when the 1st element reaches the Ultimate Tensile Strength of the material? Or do I look at the load at which a certain percentage (say 15%) of elements cross the Ultimate Tensile Strength of the material. I want to know
1.How you decide based on the FEA results whether the part will fail or not?
2.How much correlation have you got when dealing with polymer parts reinforced with glass fibers?
3.What is the industry practice that works best?
Thanks
Anup

 

[rb1957]

If you're matching to test results ... apply the test load and see what happens in the model. the test result should record what failed in teh real world, which should give you something to focus on in the FEA ... eg, did the adhesive fail? did the glass fiber? did the polymer matrix ?

if there's no failure in your FEA at a real world failure load, then maybe your model doesn't include some design feature (that's inducing the failure).

Quando Omni Flunkus Moritati

 

[blakmax]

Prediction of failure in composites is a bucket of worms. There has been a significant effort by Mike Hinton from Qinetiq to rationally assess a wide range of composite failure criteria and last I heard (several years ago I retired so I haven't kept up with developments) was that most of the usual criteria (Tsai-Hill, Tsai-Wu etc.) all fail to predict failure adequately. The problem is that not only can the fibres fail, under some loading cases so can the resin.

Some models produce unbelievable outcomes. One example is a stress based analysis of single plies where the properties are derived from strength tests and an elliptical failure envelope is generated by plotting longitudinal against transverse stress, with the intercepts with the axes indicating longitudinal strength in tension and compression and transverse strength in tension and compression. An ellipse is drawn through these points. The lack of credibility in htis approach can be easily demonstrated thus: Suppose I wish to avoid transverse cracking in my composite. I simply reduce the allowable transverse tensile strength of my lamina by say 50% and redraw the ellipse. The result is that the biaxial compression strength is predicted to INCREASE! How can a reduction in transverse strength make the material stronger under biaxial compression loading?

One fellow from Boeing (John Gosse) had developed a Strain Invariant Failure Theory (SIFT) and that appears to provide a reasonable agreement with experimental results. I suggest you google Hinton and Gosse and see how you go.

Regards

Blakmax

 

[realtime2]

Partly depends on the type of composite. If its a chopped type
fiber there is a lot of scatter in the material modulus
(e.g.: fig 4 in

http://fde.uwaterloo.ca/Fde/Materials/ShortFiberComposites/Srim/fitting_SRIM_fatigueCurve.html

)

As the other posts mentioned there are a lot of failure modes too:
fiber break, fiber delam (in matrix) due to shear loading, creep,
cyclic creep (even at room temp.), bearing crush, local buckling, tensile,
compression (often barreling), thus check the component failure mode.
In a way it is a problem with similarities to concrete beam design.

Being off by a factor of 2 from strain gage component test results is
probably not that big a deal given the modulus scatter.

Also if it is a repeated loading problem the UTS is not worth much
for design. Fatigue strength is considerably lower.

Hope this helps.