The probe emits a magnetic field, changing with time, that attempts to penetrate into the sample. The probe detects the magnetic field that returns from the sample and reports it for analysis. Most non-metals do not affect the magnetic field, making this technique valuable where there is a covering (such as paint) that hides the features of interest.
The magnetic field is sucked up by most steels and iron alloys, and will tend to sneak out when there is a small gap or hole between two bits of steel. The concept of measuring that exposed field to look for gaps and holes is called magnetic flux leakage. Even though it can be done with the same hardware, this is not what the eddy current approach aims to do.
When the magnetic field tries to move through a conductive metal, such as aluminum or titanium, electrical currents are generated that try to stop it moving. Those currents cannot stop the magnetic field for long, but metals that conduct electricity better (such as aluminum) can slow it down more effectively than other metals. If the probe is trying to push the field in (and suck it out) fast enough, the slowing will be enough to be visible to the probe measurement so we can detect that there was metal there.
Of course, when looking for corrosion and cracks, we normally do the opposite. We become suspicious when the magnetic field was able to penetrate into the material faster and more easily than we expect. This means that those electrical currents were unable to flow for some reason, and we then have to decide what it was that stopped them. Corrosion reduces the amount of metal, so there is less room for the currents to flow and cracking will make a tiny gap in the metal that the currents cannot cross.
It is important to make sure that the currents flow in a direction where the flaw we are looking for will get in the way. For example, if we are looking for cracks at a fastener, the current that flows in the same direction as the crack is pointing will not be affected because it doesn't need to cross from one side of the crack to the other side.
Cracks, for example, are especially dangerous when they connect from one fastener hole to the next adjacent hole because this greatly increases the stresses on the material on either side of the two fasteners. This is also the direction that cracks are most likely to grow if the joint is in ordinary compression or tension. This means that we want the current to flow between the fasteners.
As a general rule, if the sample surface is mostly flat, the pattern of the currents inside the metal is going to be very similar to the pattern of wire that is used in the coil that is generating the magnetic field. For that crack problem, for example, we want a flat coil that has lots of wires going at right angles to the line of fasteners. Since the wires have to be connected together, the full transmitting coil looks like two spirals next to each other with one going clockwise and the other going anticlockwise.
Remember, each flaw has a specific current direction that makes it visible. If your probe doesn't generate a current in that direction, you cannot see the flaw. Make sure you choose (or build yourself) the right probe coil for the job you're trying to do.
It's easy to say that the flaw will make the current take a different route. We have to determine what magnetic field change will be generated so that we can place the detector in a location that has a chance of seeing the difference.
With electric currents, we can generate complicated answers by adding together lots of separate easy cases that we know how to solve. In this case, we say that the detected signal when a flaw is present can be computed by adding together the signal that we normally get with that from a special current pattern. That special current pattern is simply the one which generates an exactly equal and opposite current in the flaw (so that there is no net current, in real sample) and somehow manages to get from one end of the flaw to the other end of the flaw in a loop. In most cases, that current flow looks like two little circles next to each other, that touch each other and whose currents go round in opposite directions.
When you have two little circles of current like that, the magnetic field forms a circle that sticks out of the material and goes through the two circles. The place where the probe is most likely to be able to measure that flaw signature is when that magnetic field loop is highest above the surface, is parallel to the surface and at right angles to the main current pattern in the material sheet. Alternatively, the magnetic field can be seen on either side of that highest point, when it is at an angle, partly in the same direction and partly sticking into or out of the surface of the material.
It is tempting to measure the obvious magnetic field, when it is highest above the surface. However, the magnetic field caused by the normal current pattern in the material (where there is no flaw) also goes in this direction, so we would be looking for a tiny change in a big signal. It is a lot easier to see a slightly smaller change but in a small signal because the contrast is better ... there is less 'glare' to deal with. Therefore, the best field direction to measure is the one that is at right angles to the surface and this makes the sensor stick up from the probe plate that contains the coil itself.
Even if your probe generates a current that tries to pass through the flaw, if the magnetic field generated by the flaw doesn't go where your detector will be able to measure it, you cannot see the flaw. Again, always choose (or borrow from a friend) the right probe for the job.
Clearly, the key to all this is in the design of the probe. These are always going to cost money because they are specialist pieces of equipment that someone has to build in small quantities. However, we can try to make it easy for groups of people to build, buy, modify and share a set of probes between them. The Open Source concept, which encourages design sharing and improvement, is ideal for this. In this case, it doesn't even stifle manufacturing because these things still are going to cost money to make. If the community comes up with a design that they all want to use, someone can make money selling them. By having a shared design, the purchasers have enough information to modify these purchased probes and come up with improvements and new uses, which the manufacturer can incorporate into the next build run.
A preprint of a scientific paper that discusses this fastener crack detection method in more detail is available in PDF and HTML formats. Because that probe can both detect corrosion and cracks, we are planning to make this our first prototype on this project.
If you would like to suggest other comments that should be here, please write and send them in.