Cracks, pinholes, or dimensional data?

[paulengr 44]

I work at a foundry that makes ductile iron pipe. The product is 20 feet long and between about 8" and 30" in diameter. Right now, they visually inspect everything. The majority of the defects are easy to spot. It's laminations, chunks of slag, and such. They also do measurements for the various dimensional details. So far, I can trivially figure out how to automate getting these done. Much harder is cracks, especially hairline ones, and pinholes. The feature size is not very big so it probably takes some work to do it. This is what I know so far: Eddy current sensing apparently works. One of our competitors is using it. I have no experience however with this equipment. It detects cracks but clearly based on their product, it doesn't do a good job on pinholes. From what I know from an electrician that used to work on it, apparently the machine wasn't terribly reliable either. Keyence makes distance scanners that could easily have enough accuracy to detect pinholes and cracks, but the problem is that I'd be scanning at a resolution of a few mm at most. With a 20 foot long piece of product, at maximum scanning rates, to maintain the current production rates of roughly one piece every 20 seconds, I'd be looking at rotating it at around 600 RPM to do this cost effectively, never mind the obscene data rates....basically, it becomes unrealistic. Anyways, anyone with some experiences doing automated inspections on castings?

[gravitar 3]

Personally I have zero experiencewith non-destructove testing, but are you familiar with this process: http://www.magnaflux.com/index.asp This is what engine rebuilders use to detect cracks in engine block cores, which is how I know about the process. Not sure how well it would work in your application

[BobLfoot 302]

Gravitar provides an excellent link. The Eddy Current Test sense a change in flux of the metal and with limits trips an alarm. Building on ideas already laid down. Think this thru. 8" dia is 25" circumference and 30" dia is 95" circumference. Using an eddy current or magnaflux ring sensor you'd need rings in standard diameters {25, 35, 45, 55, 65, 75, 85 & 95} and a feed rate of 1 FPS or 60 FPM. Combine this with a paint gun to spot mark the variation location and then divert the suspect unit into a more detailed scan station. Here "offline" you could use your keyence or other technology including visual to do an inspection of the highlighted suspect areas. With your skill I'm sure you can build this idea into a working system of inspections to put your competition to shame. Goo luck Paul.

[BudMan 1]

What about air pressure leak testing??

[BobLfoot 302]

Not to put words in Paul's mouth, but from my forge shop experience these parts are probably still very warm when he needs to inspect them and the chill of compressed air would adversly affect tempering.

[TimWilborne 108]

We toyed with the idea of of using something similar to test the splices of steel where I once worked. The variations in height and thinkness of the splices shot the idea down. Would casting variations have the same effect?

[paulengr 44]

Cracks are a structural concern. Whether they currently penetrate or not is not really the issue. Any crack will soon make it all the way through. Pinholes CAN theoretically be detected via pressure testing. AWWA specs require EVERY piece of pipe to be hydrostatically (water pressure) tested to hold at least 500 PSI for 10 seconds. That's the theory. The reality is that at a production rate in excess of 120 pieces per hour even in a "slow" shop (high speed ones are closer to 200 pieces per hour), the limiting factor is being able to pump that much water that quickly into it AND discharge all the air (air=compressible fluid), AND make a good seal on the ends. The seal is the whole problem. I can't figure out a good mechanical way to make a seal which can adapt to several different sizes other than the current method (push the pipe hard against a flat surface that is SOMEWHAT pliable to conform to the pipe). All the good ones I've seen have some serious hydraulic cylinders on one end that simply push the pipe solid against the backstop. In both of the ones we have, the machine is held together with a 3" solid gussetted plate on one end with four 4" shaft steel "tie rods" to hold the machine together. The hydraulic cylinders are 4" bore with 3000 PSI/1200 PSI feeds. The lower pressure is for high speed. One of them requires a jacking cylinder (a smaller bore cylinder inside the main one) for even higher "speed". The water system uses a similar setup. There are two 6x4 centrifugal pumps to fill the thing up (several hundred GPM flow rate). There's a little trick while doing it where we intentionally back off the hydraulic pressure close to the end of the low pressure cycle just enough to slide the pipe back about 1" by water pressure alone and then re-engage the cylinder. This causes the remaining compressed air to burp out all at once. Then a high pressure jet pump is engaged to crank it up to about 600-700 PSI max (40 GPM max). The practical reality is that this is a grey iron test. If the ductile process failed and QC missed it, then the pipe explodes and tears up the hydrostatic test press when the pressure is applied. There is so much pressure that if the settings aren't correct (size detection fails), it will pretty easily mushroom the end of the pipe. The seal is almost always worse than the pinholes so there is more water leaking from the seal than there is from any pinholes. If we could solve the seal problem, things would be much different. We could run the pipe up to pressure and shut the pump off, or even leave it running with flow meters inline. Then by monitoring either pressure or flow rate or both, we could pretty trivially measure the flow rate through a small pinhole...maybe. From experience with trying to do this on a much larger scale (10-50 mile long pipelines) where a pinhole means that we have an environmental problem, I have almost 0.0% confidence that pressure testing is ever going to work from a practical point of view. The trouble is usually that relative to the fluid volume, a pinhole is such a small feature that it is very difficult to detect much of a change in pressure or flow rate. We'd be better off trying to say use an ultrasonic leak detector scanning the surface during the existing pressure test equipment than using pressure/flow to do the job.

[BudMan 1]

Could you use inflateable seals on the inside or outside diameter? Something like.. http://www.ramerproducts.com/ http://www.fastestinc.com/generalindustry/...log.asp?PCA=299 http://pump.tuthill.com/Products/QuickSeal/Quickseal.cfm But bigger?

[Bob O 8]

Just another idea...How about pushing the pipe up against a cone to seal the end? Pipe_Seal.pdf

[robh 12]

X-Ray?

[paulengr 44]

I guess my question was really if anyone had actually done eddy current testing so I had a contact name. I hadn't intended to open a huge design idea thread. Many of the ideas are very good and we've kicked almost every one of them around already. I've even made several phone calls and a few experiments to see what will and won't work. I've attached responses below to the various dead ends that I already ran into. Maybe. The problem appears to be SNR (signal-to-noise ratio). To get decent signals, you need a lot of power. A simple X-ray source isn't going to do it. So instead it has to be a nice piece of hot radioactive material. The first problem here is of course demonstrating that nobody is going to be growing a third eye any time soon just because you have a hot source in the plant. The second problem is shielding it from whichever idiot wants to "play" with it. The third problem is that the size of the nuclear source needed to get decent penetration is so large that I might have to pay somebody from the NRC (Nuclear Regulatory Commission) to sit and drink coffee at the plant 24/7. There is already a lot of data on this approach because some steel mills use this exact approach for automated slab inspection, so it is pretty easy to find out the amount of power needed and get into contact with existing manufacturers of similar equipment. So I ran into a dead end on this one. EMAT. Google it. Kind of cool. You bring a high frequency magnet within a couple inches of the surface. This magnetically induces ultrasound into the pipe. Then use a laser to measure vibration of the surface...this gives you noncontact ultrasonic detection. It will not work close to the Curie temperature in the hot end of the process because the pipe is not magnetic at that point. But it will work at the cold end. This one is still on the table in my mind. At least one manufacturer is still very interested in trying it. It helps when one of our foundries and their headquarters are in the same town. A variation of EMAT uses all lasers. A high power pulse laser induces ultrasound. Same reading laser is used. This has the advantage that it works at high temperature and that you can get the equipment away from the radiant heat. The company's name is Lasson. I'm very intrigued but I get the feeling that this is more or less an R&D instrument company. I'd classify this as extremely far out on the bleeding edge. You supply all the legwork to make it work. Eddy current testing. Apply a magnetic field. The field "pops out" of the surface at the discontinuities from the field lines arcing out around the discontinuity where it is very easy to detect. The advantage here is that it's "tried and true"...there are already operating machines in the industry. The disadvantage is that results interpretation is a little more difficult.

[TConnolly 13]

We use a device called a scanning ultrasonic microscope for looking for cracks, small voides, and delaminations in metal/carbon composites at high speed. Many of these devices are highly customized. Your pipe would need to be able to travel through a tank of water. Sensors might need to be arranged in a submerged ring through which the pipe travels. In a SUM an emitter bounces an ultrasonic wave through the target. A receiver picks up the reflected wave and feeds it to an oscilloscope card in a PC. The software in the PC looks for peaks from the o-scope card. Sound reflections will be measured from the surface of the part (outside of pipe in your case), any inclusions, cracks, or delamainations inside the part, and from the back of the part (inside wall of pipe in your case). The software processing the signal from the oscilloscope card ignores the first and last peaks and any peak in between above a predetermined amplitude indicates a flaw in the part. There are a number of companies that make these, and since most are custom solutions then they should be able to adapt something that will work for you. I have no idea if its feasible for your application, but it won’t hurt to investigate it.

[paulengr 44]

It looks interesting but a quick google search didn't turn up any names.

[forqnc 2]

20 Years ago, when I worked for and Automtive company, we had to check suspension arms for cracks. This was done on a production conveyor, the part was washed down with some kind of iron water, we then ran the parts throught a magnetizer and an operator would use a black light since the iron in the water would stick in the cracks. after that they where de-magnetized and washed down again. I am not too good on all the details as is was 20 years ago

[BudMan 1]

I think this would be the Magna Flux style test as mentioned above by gravitar.