Machine Differences

[[Mi...]]

Hi,
We are utilizing three CMMs in the quality department. One is a Contura G3 and the other two are Contura G2’s. The Contura G3 was bought in 2017 and the G2’s were bought in 2009 and 2011. All three machines are utilizing the Vast XXT with the RDS-CAA head. Our calibration intervals are 1 complete and 2 enhanced every year, so our machines are getting the complete calibration every 3 years. We are currently looking into doing a complete on all three yearly.
Recently, our machinist have been complaining about the CMMs giving out difference results with the same part, program and setup. Our machinist currently run their own parts on the CMM. Up until early 2021 we had a CMM operator running all the parts. When looking into this, I noticed the same part and program run on two different machines were reporting 2 difference results and this became quite alarming. One machine was reading all green while other was reporting a few red results, mostly diameter and roundness. This has seem to cause quite the backup in the quality department because machinist only want to run their parts on the machine that gives them the best results.

Furthermore, I decided to take a deep dive into this issue this week. I utilized a Glastonbury Gage and a program to measure the gage utilizing 1 stylus at 3 different orientations. I did, X-Y, Y-Z and Z-X. This is shown in the attachments below. As you can see from these results,2022-01-28 - MIN-MAX Report X-Y, Y-Z, X-Z.pdf
the same program on all three machines are reporting different results, some much bigger than the allowable specifications. All these programs were run after qualifying the stylus used with the gage.

I am posting here to see if there are any other shops that utilize 3 or more of the same CMMs and seeing something similar. Is there something that can be done?

Regards,

QUICKCHECK MASTER Z-X.jpgQUICKCHECK MASTER Y-Z.jpgQUICKCHECK MASTER X-Y.jpg

[[Da...]]

Have you checked the styli for any damage? I had a similar situation with 2 CMMs one using a TP6 and the other uses an XXT, I had to use a 60X eye loop to see there were cracks inside the ruby at the shaft. After changing the stylus the machines were in agreement again.

[[Mi...]]

Please sign in to view this quote.

Yes, this was the first thing we looked at. We do a lot of scanning so I looked for material build up and any flats that may have developed. We have placed the the stylii on the comparator and is straight as straight can get.
We also took this a step further and just replaced the stylii to be sure it wasn't a stylii issue.

[[Ri...]]

Oh... The old tale of This Machine Reads Good and This Machine Reads Bad. I heard this tale for about 10+ years at my previous place. We had 5 CMMs in the lab, some G2s, some 2014s, and an Accura. 99.9% of the time the difference between Good/Bad were either A) within machine specification or B) within GR&R allowance.

I've been down that road with the GSG QuickChek. I feel like there are way too many factors that come into play when it comes to trying to correlate measurements from one machine to another. Temperature, cleanliness, workholding, stylus qualification, point density, filter/outlier, fitting algorithm.

Everyone focuses on MPE (EL) too much, and that number only applies from measuring from one point to another linearly. There is also MPE THP which is your form measurement allowance.

Then there is the other thing which is which machine is correct? You are comparing your CMM results to the cert from GSG, but GSG doesn't use a CMM to certify the QuickChek - they use a Federal 136B-3 Comparator, so the results you get off the CMM, and the results you get from the cert are never going to correlate.

My suggestion is to have the CMMs calibrated, and immediately following calibration run this measurement routine on your QuickChek. I would suggest running each position 3-5 times to track the average and range. Now you have baseline measurements directly following a manufacturer calibration. Whenever there are questions, throw it back up there and see if there are any changes.

At the end of the day, a machinist never makes bad parts, so their job is to make your job as difficult as possible. Just smile and nod.

[[Ke...]]

I would use the exact same stylus on each machine, and compare probe qualification results for each machine.
What is the temp range of the lab? You may need to verify the quickcheck "morning, noon, and night" to capture the (observable) uncertainty of the measurements, due to temperature variation (although quialfying the stylus immediately before inpection would minimize this, I'm not sure how tightly you are trying to keep the results)

[[Mi...]]

Outside of calibration issues or damage the results should be comparable, within the error of the machine.

Don't assume people are cleaning the parts well either. Every time I'm presented with this issue ( we have 33 CMMs) I take their part, clean it myself, clean the probes myself, and recreate the test they said was bad. Often enough the problem has vanished just with proper cleaning and it's a teachable moment.

But also, are you looping your Base Alignments?

I've just gone through the process of doing that on a production program that we were having issues with quality of the measurements and repeatability, especially where CNC machinists were loading the part and not necessarily taking enough care while doing so.

Adding a looped Base Alignment, made the repeatability improve so much that I could take a part from our floor Duramax with XXT, and run it on our Contura with XT Gold and then compare it to the customer's report ran on a Prismo and get VERY consistent results, where before we couldn't even come close consistently. Most importantly anybody could take a part and run it on any CMM and they get the same result.... they kept trying after I made this change and were surprised that every time they rechecked a bad part they got the same numbers within a few microns.

If only you ever run the parts, and you're fixturing is bang on, a looped base alignment isn't likely necessary, but if you're letting other people do it, you really should be doing it. Nothing is worse than the CMM giving two different results for the same part.

[[Mi...]]

Please sign in to view this quote.

We hold and maintain 68.5 to 69.5° year around. humidity between 40-45. What caught our attention was 1 part specifically. it was .0007-.0008 out of round on one CMM and .0002 of the other.

[[Mi...]]

Please sign in to view this quote.

We do throw the Glastonbury gage on the CMM following CMM calibration or service. We usually only do that with one orientation and call it a day. Going forward I'm going to make it a priority immediately after calibration to run all the orientations on a loop for the machine to get a baseline. I just wish there was something we could come up with now. Seeing as we just had service tech on site in September to do a complete on one of the CMM's and enhanced on the others. This would be quite the expense to have them out 3 months later to do a complete on all three.

I just nod, no smile 😐

Thank You!!

[[Mi...]]

Please sign in to view this quote.

As far as base alignments go we have a generic base alignment that we use. Vise, Chuck, Angle Plate. we just use those and the first characteristic is a alignment (translation) to were the part is located and build off of that translation. I guess we could loop the beginning primary alignments.

[[Me...]]

I've had a similar issue. I ended up rebuilding all the probes on the machine in Q.
That is over writing the probe data as if I had broken a probe.
Some how even using the passive scanning for qualification wasn't working.

[[Mi...]]

Please sign in to view this quote.

It's a hard pill to swallow but that's exactly the situation I just fought with.

Looping secondary alignments within the program doesn't work well.

You really need to build the Base Alignment off your part (you could use a Start Alignment off your fixturing if you wanted), and then loop the Base Alignment. Because you're fixturing isn't what's changing run to run, the part is so that's what you need to iterate.

It's hard to explain or convince anybody, but you try it and watch how the CMM corrects on subsequent loops if your fixturing isn't perfect and you'll be sold.

[[Da...]]

We have 3 Micuras right now, with a 4th waiting to be assembled. We check hydraulic components ( cages, valve bodies and so on...). We almost never need a distance from one cylinder to the next, but we do need super accurate diameter sizes and form checks.

We developed this process along with Zeiss customer service. this is a double check to run with the regular calibration. We use a XXX master ring gauge that is more or less the average size of all the parts we check. Our lab is temp controlled to 68 +/- 1. We also have a temp probe.

Master probe is run and then the smaller of the two probes we will use is qualified. The ring is checked with 3 circles . The alignment uses the top and bottom diameters recalled into a 3D line for spatial and X/Y alignment, Z is a single point at the plate. The alignment is looped. The ring is checked and the middle diameter is used. Without touching the part the probe is changed to the master and rechecked. Probe is calibrated and the process starts over. 5 diameters are checked and recorded. The entire process is repeated with a different sized probe.
Each machine is run, producing 5 diameters per each of 2 probe sizes. Each set is averaged and recorded.
Seems long winded but it works.

Calibrating a probe only sets the diameter of the probe ball and its location in respect to the other probes. It does NOT correct an error of distance. So diameters will be corrected but distance between 2 diameter features will not. That's what recalibration does.

[[Je...]]

.

Have Zeiss remote into your CMMs and provide you with collision data over three months. This excel file will tell you the speed upon contact, stylus displacement, whether the machine was in cnc mode or manual, and other useful data. I've also heard a rumor from a little bird that a third party application exists by a talented programmer (cough, ahem, Eric Moberg) that can pride collision data.

A collision doesn't necessarily mean mechanical damage to the sensor or stylus, but it can move the geometry of the stylus system enough to throw it out of calibration. One example would be if someone happens to not lock a clamp in the cmm fixture, which causes a crash. Instead of recalibrating the probe, the operator/machinist just continues with the measurement.

Just another variable to consider. You're goal is to align the measurement process of all three machines.


Jeff Frodermann
Meier Tool & Engineering
Anoka, Minnesota


.