[thedoc (OP)]

Paul asked the Naked Scientists:
   
If you send a vibration along a piece of metal which is being stressed until it breaks, is there a way of predicting when it will break?

What do you think?

[Colin2B]

It would be worth trying to measure the resonance of the sample as I would expect the frequency to change as it reaches the yield point and the crystalline structure changes.
You could also try firing pulses down the sample, they should reflect off the end and you might get a second closer echo at or near yield.
Beware brittle metals fail quickly without much deformation and might not give much warning

[chiralSPO]

When you say "predicting when it will break" are you asking how to predict the frequency of a vibration that would break the structure, the amplitude (intensity) of the vibration that would break the structure, or the time (actually "when") a structure will break given a specific vibration?

[alancalverd]

Yes and no!

You can use correlated fourier analysis to extract the resonant spectrum of a structure from repeated tapping. There are a number of machines around that use this technique to detect flaws in concrete structures including tall buildings and bridges, and doing it on a single wire or bar is vastly simpler.

The problem is that whilst you can usually cut out and patch, or fix a support over, a small defective chunk of a concrete structure, a change in the resonant spectrum of, say, a single stressed wire usually indicates that it has already undergone the first phase of irreversible failure. Some metals actually get tougher before they reach breaking stress, but you have to be pretty slick to stop incrementing the stress at that point, and in nearly all cases (apart from a few "memory alloys") any change in resonant spectrum indicates an irreversible change in crystal structure.