Detecting flaws in materials using sound waves.......

16 Feb 2010

So, would anyone have some advice on how to get started using the mbed to build a device that can detect tiny airbubbles, cracks, or other flaws in materials like carbon fiber structures, etc...?

I'm guessing some sort of ultrasonic transducer, some fancy coding to send a pulse it and produce some levels based on the density of the resin, etc...  Anyone done this before?



19 Nov 2011

I think it's not as simple as just using a ultrasonic sensor. By searching a lot on the internet about this subject i've found that all systems for finding cracks in metal pipes, medical systems etc... are using several ultrasonic sensors steered sequencially in a phase array-system. Look at wikipedia about phase-array and you'll find a lot of info.

19 Nov 2011

One of my friends was doing a phd on this sort of thing; imaging with cheap sensor arrays. Can ask if this is something you really want to get in to :)

19 Nov 2011

I used to work for a company that detected flaws/thinning in steel plate/tube, the equipment is very specilised and rediculsly expensive !!

But basicaly it boild down to an oscilloscope (special) and an ultrasonic trsnsmitter/reciver,

and most importantly a VERY experianced operator.

sorry to put your fire out


19 Nov 2011

I think you should use Image processing. You have to connect a CAM to Mbed and take a picture with a CAM and then read it with a Mbed and process it Maybe you have to use DSP

24 Nov 2011

Wow! Thanks for the replies guys! I actually am really interested in doing this... I'm working on a project that this would benefit greatly... Plus, how cool would that be? :)

Rudy, I'm sure it's complicated... and I've done some searching around.. I think they use stuff like this to do annuals on airliners, etc as well for metal fatigue... I might be wrong about that.. ;)

Simon, heck yea! I'd love to hear about your friends work! Thanks!

Ceri, I understand... But I also tend to try harder when someone tells me it's too difficult or not possible.... ;)

Ali Reza, I think the biggest problem (unless it's in another spectra of light) would be the depth that imaging can provide... I'm looking at trying to find bubbles that may be several layers of fiber below the surface...

I was thinking something like a sending unit on one side, with the receiver on the other in some sort of rig to keep them aligned, etc. that would allow you to pass the device across the area under test and output either via ethernet (or wireless) to a remote terminal that might be used to do the secondary signal processing/analysis, etc...

Anyways, anything on this would be great to hear!! Thanks!

24 Nov 2011

Hi all, I presume I'm the one to whom Simon was referring?

My experience is pretty much exclusively in the medical ultrasound area rather than NDT. That said, the physics and basic principles are the same and I'm interested in what more can be in the NDT world.

The expensive (and its not *that* expensive) bit is the transducer. Typically, the frequencies you're dealing with are pretty low (in the MHz range and lower), so its not like hard core RF (though its not exactly audio either!). Blatek ( ) are pretty prevalent in the industry and they make NDT style transducers (though again, my knowledge is medical ultrasound related).

I'm not aware of ultrasound being used for plastic NDT, though that's more likely down to my ignorance of the field. The subtlety is just going to be down to coupling the transducer to the material. If anything, I expect plastics will allow more freedom in how you can couple to them (though that's just a feeling).

I'd be really interested in chatting about this as it's pretty closely aligned with my interests and what I'm up to at the moment. I presume you have an interesting application in mind? Where are you based?

24 Nov 2011

Another point, I actually have a good chunk of the bits to mash together a basic medical ultrasound machine (transducer array, rf switches, a/d, d/a). With a bit of hacking and some impedance matching, I'm sure it could be repurposed for an NDT project (it also needs to be built, based on my slightly speculative design!). (the parts are superfluous to current activities)

26 Nov 2011

Hi Henry, thank you for the reply!

I looked around a little, and found products like these.... I also found a nice overview.. (I'm sure Henry this is very basic, but if anyone was reading with interest..I figured I'd throw in some background..)

Fun stuff? I think so... And there's a need.. and if we can make one, and on the cheap.... All the better!! And what better platform than an mBed! =)

So, my first question is exactly what you mentioned... Coupling to the material... It looks like (from that last application note) they use a probe that has two tips..tx and rx... I'm betting there's probably optimal configurations for every different application .. like trying to find delamination near the surface vs. 5 layers down... or bubbles, etc.. I don't think it'd be much fun to use jelly on these surfaces like hospital ultrasounds do.. ;)

My application is essentially the same as aircraft composites.. Mostly foam core, kevlar and carbon fiber with epoxy resins... I would love to develop a device based on the mbed that can detect flaws and generate a map of the weak spots or even better areas of delamination vs. cracks, etc.. Eventually something where the mBed samples the data and pushes the data though ethernet or wireless to a pc to display and process the data..

And really long term stuff would be a realtime solution would be less scanning for specific flaws... where the mbed has a set of transducers or microphones spread across an airframe and detects a crack or stress and reports it... It even occurred to me that it might be possible to use light (an LED through a photodetector) through the resin to do this..

I'd be very interested in talking some more about this as well.. I'm based in the US (Arizona)...

26 Nov 2011

That little device looks pretty cool.

So, the big problem you're going to have with foam cored carbon fibre is the fact that foam is mostly air, and so will look like a big acoustic impedance mismatch with the CFRP. The result will be exceptionally poor acoustic coupling to anything but the top layer.

Now, I actually had a meeting yesterday to discuss the use of ultrasound in an area that might have some serious overlap with the sort of thing you're discussing. Its the kind of project that might well yield the sort of kit that is perfect for what you're describing (assuming the physics holds up!). I'm aware of the serious issue and current poor state of NDT testing of CFRP airframes. Is this an area you're in commercially?

The other thing is that *probably* the mbed is underpowered for being the core of this kind of application. It'd be worth playing around with something like the Beagle Board (or more recent Panda Board), which provides some serious grunt to get things done of the kind that you're likely to need.

28 Nov 2011

Henry, I think that might be.. At least for my purposes...which are mainly to tell if the epoxy has delaminated from the foam core... As the multiple layers of fiber are vacuum bagged and the resin infused into all layers at once, there's not a huge chance for them to delaminate from one another...mainly the foam.. Although, there is the chance for air pockets, or dry spots (which is another part I'd like to find)... I'd be more than happy to scan both sides without being able to penetrate from surface to surface.. Although I am sure the foam still poses a problem in the sense that it probably absorbs any acoustic energy without reflecting much...??

Sounds like that might have been an interesting meeting! I'm not in the industry commercially, just a 'hobby' for the moment... ;)

I might agree about the mbed being a little short on power for this..and I like the idea of the beagleboard...Maybe a middle ground could be something like the beaglebone?? ... Although I'm still wondering if the mbed might be a good fit since it's role in my mind wasn't so much the signal processing end as much as driving the transducer, etc.. and sending that data back to something like a beagleboard, etc. to actually crunch the numbers....