what structure are you analyzing?
what materials?
are you sizing to unnotched, notched, impact design values?
where will you get your strength properties?
whether you analyze with FE or hand analysis:
- see the papers by Hinton, et al on composite failure criteria
- lamina property based criteria, including the various interactive criteria, do not work well
- max strain criteria using Laminate (not Lamina) derived strength properties is generally used in the aircraft industry
Though unnotched strength analysis is not often common in design, it is a logical starting point when trying to become familiar with strength of composites.
If you are using a lamina based criteria, your FEM may need to consider damage progression (i.e. matrix mode failures may not indicate laminate failure). Since that may not be practical, you could use the max fiber strain (MFS) or truncated max strain (TMS) criteria. They are lamina based criteria, but used at a laminate level. These are popular in the US aerospace industry.
If you are lucky enough to have laminate based allowables, then max strain criterion can be used as SW mentioned. The problem is that this data is not always readily available.
SW and I may differ "slightly" in perspective about lamina based criteria. Use of the TMS criterion seems to do a "fair" job for preliminary anlayis, provided the failure is fiber dominated. Where the lamina based criterion did very poorly, as demonstrated in the papers by Hinton, where layups that may not have been fiber dominated. Of course, there may be no practical reason to use those sorts of layups, but it does point out the fact that the lamina based failure criterion are probably fundamentally incapable of generalized strength prediction.
Failure criteria for composites is a notorious minefield. John Hart-Smith published extensively on this and convinced me that Tsai-Hill, Tsai-Wu and the many variants of those failure criteria are nothing more than mathematical gymnastics. Papers presented at ICCM12 and ICCM13 by Mike Hinton assessed the veracity of failure criteria on real materials with real load cases and assessed the predicted values against known experimental results. The correlation for many theories was at best deficient, and in some cases alarmingly unconservative. The major deficiency is that many failure theories ignore that fact that for some conditions failure is by fibre fracture (some suggest that this is shear dominated) but in other conditions failure is by resin fracture. These are two entirely seperate failure modes and trying to fit them into one failure curve really is stretching things.
An example of where the Tsai-Hill or Tsai-Wu theories fall down is this. I undertake tests to determine the lamina (individual ply) strength in tension and compression for loads aligned with and perpendicular to the fibre direction. I then set up a failure envelope for that lamina on the basis of the mathematical expressions provided by these theories. That should look like an elipse which passes through my experimental points.
Now let me enforce a condition that I never want to see transverse tensile cracking (matrix fracture) so I reduce the allowable transverse tensile strength by say 50%. I redraw the ellipse and find that the biaxial compression stength has increased because the smaller axis to fit the reduced transverse tensile strength causes a corresponding increase in the larger axis to make the curve fit the new data points. I actually asked Stephen Tsai how this could possibly be, and his response was that "the mathematics says so". The ICCM papers showed that for compression compression loads, these failure theories were unconservative by a significant error.
In more recent times I have heard that Jon Gosse from Boeing has developed a Strain Invariant Failure Theory (SIFT) which actually addresses both resin and fibre failure modes. I retired some years ago and have not kept up with recent developments, so I suggest you google FIST or Gosse and see how you go.
Good comments. I am a bit surprised by Tsai's response since it is well known that all of the lamina based criteria have significant shortcomings, not just his. That is probably a testament to complexity of the problem and why we might bypass lamina based criteria in favor of laminate based.
As far as SIFT goes, it is interesting. I think what makes it attractive is that it attempts to use the fundamental physics rather than simply curve fitting data. I have witnessed a few heated discussions about it's capability and generalized use. It would appear the jury is still out. However, it seemingly has the most long term potential for use in generalized solutions, which is what everyone ultimately wants. It is even now published in some of the later composite books, indicating that it is gaining ground. That being said, its use does not seem common in current applications (at least from what I have observed).
Perhaps the only real way to feel confident is to use the laminate based allowables and max strain (as SW mentioned). Anything short of this can work to some degree, but is probably only good for well designed layups, preliminary analysis, and/or with knockdowns to address shortcomings.
All that being said, this discussion is somewhat academic. If you are designing for holes, fasteners, future bolted repairs, or damage tolerance, none of the lamina failure criteria have much value.
Well said ESPC. I have always leant towards laminate strain based failure criteria, mainly because the stresses in each layer vary significantly while strains by definitions of continuity must be consistent between layers. I also tend to rely on the max strain method the same as you and SW, even though for some load cases it is over conservative. We have always adopted a policyt of keeping max load at limit load below 40% of the laminate strength (based on max strain) because there is some data that shows such a policy avoids fatigue at holes, slots etc.
It's a bit long, but the relevant conclusions are on p.58/59.
"For fiber dominated laminates, max stress, max strain, and Hashin-Rotem failure criteria outperform other criteria (including Tsai-Hill/Tsai-Wu). These criteria are insensitive to variations in matrix strengths (Y and S) which are very difficult to obtain in situ."
"For fiber-dominated lamiantes, maximum stress, max strain, and Hashin-Rotem failure criteria should be used to obtain most reliable laminate strength predictions".
What is meant with fiber dominated laminates? Is a volume fiber fraction of 50% and matris fraction of 50% either fiber or matrix dominated? I have a laminate with 50/50 ...?
