Stamping application

I'm looking to measure the deviation of the coined thickness of a part being produced in a Minster 45 ton press running some progressive die tooling. The rate of production is 300 parts(strokes) per minute. I will mount an encoder on a jack-shaft for stroke position indication. I should have about 50-100ms to make the actual measurement so the timing is not critical. The thing I wanted to bounce off you guys is the tolerance. The customer wants to measure a deviation of +-0.0002" (yes that's two ten thousands of an inch). This doesn't sound like it's possible to do with an LVDT in a high vibration area but, Iv'e been told it may be possible with a load monitoring strain gauge arrangement sandwitched in the base of the coining station. What do you guys think? Any reasonable opinions would be helpful.

Sorry guys, I realise that this may not be the proper forum to ask this sort of question but, I don't know of any other decent applicable forum. I'll take alternate forum recommendations if you've got them.

Not in my field so can't really say.

This forum splits into more specialist fields, not sure of the response time though.

I wou;ldn't give up on someone here responding though, there are people from a wide range of experience here. Not everyone reads to forum every hour, some may read it once a day or less, so you may still have some decent responses from here.

Whoops! I mis-spoke (to borrow a phrase). The rate is 180 parts(strokes)/minute - 3 parts/second - 0.333 seconds per part. I'm assuming the bottom dwell is about 1/3 of the total cycle which is where I will trigger the measurement.

MCAFONE

Regardless of all else any quality checking process is not value added - it is overhead. Definitely needed though - no argument - dies wear as they are used. Doing it automatically does away with hand measurement which is very time intensive. Even with automated do you really need to measure each part that is produced? Would you be able to maintain quality at a lower sampling rate? How about one in ten or one in 100? This would possibly get your QC measurement away from the jarring of a press and would allow more time to do the measurement. The longer time frame may allow another measuring method.

Dan Bentler

The current method of Q.C. is to take a representative sample as you stated. The problem here is that this is a safety oriented device (a burst disk for an automotive air bag assembly) and there was a test failure where a disk fractured instead of blooming outwards. So, it would seem that 100% inspection is necessary. I originally thought of maybe doing some kind of eddy current test for overrall thickness but, this won't tell me the remaining thickness of the coined area which is a square having sides of about 0.0035". The coined pattern is a cruciform or cross shape. Previously, I was able to do this with a carbide needle mounted on an LVDT but this only gave me +-0.003" in real operation. Yea... I know this is an ugly one - I never seem to get the easy ones (Weeeeee!).

Sorry, I looked at my previous comment and realised that I did not explain things too well. The part is a 0.75" round disk, made from 0.025" Inconell 600 flat stock. We intend to coin a cross pattern in the center of this material before clipping out the disk. At the center of the cross pattern is a flat square land about 0.0035" on each side. The resulting coined thickness will be about 0.015". In a similar application with larger tolerances, I have previously used a carbide needle attached to an LVDT mounted in a dummy punch to probe this square which gave me +-0.0035" accuracy. I am now looking to detect variations on the order of +-0.0002".

This will be tuff. Good luck.

I did a machine once using an LVDT and carbide needle measuring a paint nozzle for airless paint guns. It is very hard to present the part correctly for the test and get a good test everytime. Doing it in process is even worse.

I think your best bet is a strain gauge load cell in the press tooling to monitor the tonnage. The hard part about this will be speed. Most of your stamping press controls that have tonnage monitors will be way to slow for your needs. They will also not have the resolution you need. I suggest doing some math and see just how fast you will need to be to get a peak tonnage reading.

.0002 inches is easy.

If you were using an LVDT then you might just be able to use one of these 1/2 mircon (ten times what you needed) resolution probes. The probe is a linear optical encoder, much faster than an LVDT because you don't need to wait on an A/D conversion. You can also send a signal to latch in the reading and then read the result at your leisure.

Charles - Part positioning is the least of my worries since everything is being held via pilot holes punched in the strip in a previous operation of the progressive die. I should be within 0.0005" for part positioning repeatability.
Alaric - I see what you're getting at but, the vibration from the press alone will throw the reading off at least 0.001". Also, I believe Heidenhain makes a similar product.
I spoke with a tech guy from Helm Instruments (www.helminstrument.com). They have a device that will evaluate the base signal characteristic of a load cell mounted in the die shoe and then indicate any major deviations in that signal. It sounds a little iffy because it's not a direct measurement of the part so I'll probably have to prototype this to see if it will work...

mcafone said:

The current method of Q.C. is to take a representative sample as you stated. The problem here is that this is a safety oriented device (a burst disk for an automotive air bag assembly) and there was a test failure where a disk fractured instead of blooming outwards. So, it would seem that 100% inspection is necessary. I originally thought of maybe doing some kind of eddy current test for overrall thickness but, this won't tell me the remaining thickness of the coined area which is a square having sides of about 0.0035". The coined pattern is a cruciform or cross shape. Previously, I was able to do this with a carbide needle mounted on an LVDT but this only gave me +-0.003" in real operation. Yea... I know this is an ugly one - I never seem to get the easy ones (Weeeeee!).

Click to expand...

I assume the client has done a thorough job on the fault tree analysis and the proposed solution. Wouldn't some sort of iradiation/absorbtion test (xray) or such give you a better idea of the part? a Fracture could be an inherent material failure as well.

So if your going to probe these parts how big is the bottom of the cross? How big of a point on the end of the probe? I was thinking the cross would be a wedge shape in the bottom.

They have a device that will evaluate the base signal characteristic of a load cell mounted in the die shoe and then indicate any major deviations in that signal. It sounds a little iffy because it's not a direct measurement of the part so I'll probably have to prototype this to see if it will work...

Click to expand...

Is the tonnage on the part proportional to the thickness of the part? I am assuming this is a mechanical press and not hydraulic.

It is a pretty wild guess, but Keyence advertises (http://www.keyence.com/products/vision/laser/lkg/lkg.php) a laser displacement sensor with measurement resolution of 0.01 um. Which is 50 times better than your spec (0.0002" = 0.5um).

Might be worth a look. Sure not cheap.

CharlesM
In a former life I did a 9 years tour in a stamping shop. Sorry no exposure to "Inconell" is it a uniformal material draw. In steel your Hardness can vary up to 3 points in a 36" lenght. With that being said I tried to use tonnage monitors (Data Instrumants, Link, Helm, Tolledo, ect.) to do setups, and monitor for wear in the tooling. The variation in hardness will play hell with your tonnage. The stainless that was fully annealed did not have any problems. What type of press did you said a Minster and maybe I missed it (conventional crankpress with no dwell at BDC). We did some coining and drawing in crank style. We finally just used the tonnage monitors for tooling breakage (if that is a word). The punch monitors work OK from what I can remember. Just trying to point out my mistakes hoping you can avoid them.

We once made press monitors for Minster, stolley and Bruder Presses

These were mechanical crank presses that stroked at 600 ppm and made beer can ends with the pull tabs and all. This was important work There were 8 different dies or strikes required to make one can end. This system was similar to the Helm Instruments plan of attack.

0.0002 inches will not be possible in the press. You should look at measuring thickness later. The problem is that the press warms up and metal expands. This in turn causes the load cell forces to go up. We monitored the forces on all the load cell on the press ( 24-32 ) each stroke. We calculated the min, max and average for 15 second intervals. We also had a running standard deviation. Errors were caused by strikes that were about 6 to 7 standard deviations from the norm. We could easily tell when bearings were going to go out, we could see dirt in the dies or dirt on the ends, fractures and other defects. We used a Radisys VME computer for this so we have plenty of speed even though we were using a 386. This was 15 years ago. The point is that I know about this topic.

You will see 0.0002 inch differences in thickness by the differences in pressure but it will not be exact. Metal compresses when a force is applied. See Young's Modulus or the modulus of steel. The CHANGE in force will be roughly proportional to the CHANGE in thickness. The problem is that the average force changes as the press expands so everything is done by differences in terms of standard deviations.

The laser technique is good if the update rate is fast enough. I would prefer a non contact measuring device. It should be just down stream of the press so little product is wasted when a defect is found. A lot could also depend on how the surface reflects the laser.

A sonic means may work too. A good part will vibrate at certain frequencies but not others a FFT could find these frequencies and if the FFT plot for the part doesn't match the standard the part is rejected.








I would look at