Composite failure

[sonia518]

I'm using Patran 2012/MSC nastran to calculate total damage for a composite tube but when i got the result i don't found total damage for progressive failure of a composite.
I just found failure indice, critical components reserve factors and margins of safety. Should i using MD nastran instead of MSC nastran 2012 or a subroutine?? Does anyone could help me plz.

 

[RPstress]

I assume you're modeling a polymer matrix with continuous glass or carbon fibers. By progressive failure do you mean matrix cracking (often called 'first ply failure')?

No code that I know of can treat matrix and fiber separately unless they are modeled separately (which needs a lot of elements).

The usual method is to deduce from the stress state whether the matrix or fiber is going to fail. With a simple maximum stress or strain failure criterion is easy to distinguish matrix failure from fiber failure. These simple failure criteria are incorporated in basic Nastran. A more sophisticated failure criterion can used with a postprocessor (such as a modified maximum stress or strain; these typically still assume a somewhat different criterion depending on which quadrant the stress or strain is in). Modern versions of Nastran seem to have a lot of nonlinear failure criteria options. The problem with the usual so-called sophisticated failure criteria (such as Tsai-Wu) is that they don't distinguish between matrix failure and fiber failure.

A nonlinear methodology can be used to remove the matrix properties progressively from affected layers. If fiber strengths are exceeded then they can also be removed, but often fiber failure is taken to be laminate failure.

MD Nastran is somewhat separate from MSC Nastran. I forget when they parted ways (about 10 or even 15 years ago?). They should have about the same capabilities, though MSC has had a lot of other packages added (the 700-series solution sequences are what used to be Marc; I forget what the 600-series ones were).

But the usual basic Nastran outputs can be used reasonably effectively. The sign of the 1- and 2-direction stresses plus the level of the shear stress (the 12-direction) dictate which material is considered to have failed and so what properties are to be removed from a layer.

I notice in the Nastran quick reference guide there's mention of the Puck failure criterion with more sophisticated solution sequences. It's possible that these will enable a better approach if a modern solution sequence is used. I'm out of date with MSC Nastran's more recent nonlinear options.

 

[RPstress]

Oops. I was thinking of NX Nastran. MD Nastran is still MSC. It will not be any different from the point of view of 'progressive failure.'

 

[sonia518]

My model is a filament wound composite cylinder (9 layers (0/0/0/55/-55/55/-55/55/-55) resine vinylester/verre E. I'm modeling a tensile test of the tube and i would like to obtain the behavior of the tube after failure.

 

[RPstress]

So almost quasi-isotropic but the asymmetry makes that questionable.

You should be able to predict initial failure and fiber failure passably accurately (it depends a lot on what material strength data is available). Simple (and therefore inaccurate) estimates can be made of the laminate stiffness after resin failure just by removing the resin-dominated properties and observing the predicted extension. When a fiber failure occurs stiffness will drop markedly and strength will drop a similar amount. In practice with one third 0° plies laminate failure will follow the first fiber failure almost instantaneously.

Modeling laminate behavior after failures is questionable with a simple solver. Nastran's more sophisticated solution sequences might help but usually, even for uniaxial loading as you describe, something more like explicit analysis would be in order.

For uniaxial tension in line with the cylinder axis failure will be predicted all around the circumference. In practice it will occur at one position, depending a lot on things like porosity, and then spread. That will be essentially unpredictable, even with very sophisticated analysis.

Usually the cylinder ends will cause failure near the start of curvature. This will at least localize failure along the tube.

Any damage will help isolate failure position but will also complicate matters considerably. Modeling damage is very hard unless it's a simple hole or similar.

It depends what the post-initial failure analysis is for. Anything other simple predictions is questionable in my opinion and even those simple predictions are subject to considerable uncertainty.

Note that even with material failure data for the laminate, failure will be subject to considerable statistical variation. You can predict with confidence when it won't break but exactly when it will is hard!