Imaging with sound
Bruce - Well the ultrasonic waves, similar form to acoustic waves, so we know they're travelling in air, but they also travel in liquids and solids. And so, dealing with bits of metal then ultrasonic waves travel really very nicely in those bits of metal. So, we have a series of sources which emit ultrasound into the metal structure and they're reflected from all of the interior structure which obviously includes cracks. We can then pickup with our microphones, ultrasonic microphones, actually they're small electric elements but doesn't really matter there. We can pick up that reflected energy, process it and from that, produce an image of the interior of the component.
Ben - A lot of the materials that we might be looking at to try and find cracks are actually very highly, very precisely made materials in the first place. Often, they can be a single metallic crystal. Does that actually affect the way that the ultrasound would travel through?
Bruce - Yes. The crystal structure is utterly paramount to how waves travel in so that the single crystals are an isotropic in their material properties for which as far as ultrasound is concerned means that waves travelling in different directions go at different speeds, and you can imagine that if we're not very careful, that could create all sorts of confusions we have to compensate and take account of the fact that the velocities are different and the speeds of sounds are different in different directions. And if we do that then we can produce images as if the velocity was uniform.
Ben - So essentially, you use some kind of algorithm or some kind of computer modelling to compensate for how you know the sound would travel in a pristine example and that therefore means that you can compare your example to pristine, and easily see if there's something wrong.
Bruce - Yes, we need to have a model of the structure we're trying to test in order to correctly image it, and one of the big challenges is when you go to a component and try to make a measurement where there's uncertainty in that model. And so, one of the things we have been working on is what we call autofocus techniques where your camera automatically adjusts to put things in focus whereas we have been working on ways of trying to automatically extract this picture of what the structure is like. So this model without knowing it beforehand or say, it's the particular principle orientation direction is unknown. But we can extract that from the data itself if we know something about the structure like, for example, how thick it is.
In terms of detecting cracks, the real challenge is, somewhat different to the medical challenge where you've got this very rich image. The interior of the piece of metal is you'd have thought relatively simple and it is, but there's a lot of scattering from other things within the material like grain boundaries, although they don't occur in the single crystal example that you just talked about but in most metals, there are many grain boundaries, all reflecting ultrasound and there may or may not be cracks and so, the challenge is to achieve that high resolution imaging performance of the crack surrounded by these other scatterers which basically set the signal to noise ratio of your measurement or the best possible measurement that you can make.
Ben - And I assume that signal to noise is ultimately what will limit the resolution with which we can view this and the actual minimum size at which we can first detect that crack.
Bruce - Yes, and that's always the challenge, to detect the smallest possible crack and that's where we're always trying to push the limits of what's possible there and that's what I'm interested in as a researcher.