Measuring the same part on 2 CMMs, Different Results

What PCD SHOULD vs what it ACTUALLY does (in my experience anyway) can vary quite a bit. Thats why I keep the touch speed consistent from one machine to the next

Matthew D. Hoedeman parts are about 48 lbs. valve bodies.

This is the breakdown of the reports on both machines. We measured 5 dimensions total. Same exact part on both machines run 5 times. Using the same star build

1 ID that is constructed with 11 vector points (Tips 2-5)

1 auto cylinder ID, used for (xy orgin) (Tip 1)

1 auto circle OD (Tip 3)

1 auto circle ID (Tip 5)

and an angle dimension

Average Deviation Comparison

Constructed ID on A 0.0014, B 0.0008
auto Cylinder ID on A -0.0004, B -0.001
Auto Circle OD on A 0.006, B 0.002
Auto circle ID on A 0.0000, B 0.000

Again, CMM A matches the mic measurements closer than CMM B.

How would I begin to troubleshoot CMM B for accuracy?

Thank you all for the info.

Can you post the code ?
Do you repeat in dcc the manual alignment before doing a accurate alignment ?
Can you post average values AND standard deviations values ? (or each individual values)
Can you post calibration values ?
Did you check calibration spheres values ?

No answer, only questions, sorry, but it could help understanding...

...snip...

Did you check calibration spheres values ?

...

Something to try: use the same calibration sphere for both machines.

CMM A V2106 original program

CMM B V2017 copied program from CMM A

possible suggestions

1) Re-write program on CMM B in case something didn't migrate properly from CMM A.

2) Try star probe from CMM A and qual it on CMM B and then compare results.

Good luck.

I don't think I'll be able to submit anything else today.

I'm in south texas and under hurricane warning. We'll be closed tomorrow and possibly Monday.

Gonna spend the rest of the day backing up cmm stuff.

The part weighs 48 lbs, I guess it's aluminium, so it gives 8 dm3 of alu without holes, flanges...
It means that the dimensions of the part can be close to the cmm measurement dimensions.
Maybe the calibration of the cmm is not as good as it should be to the edges of the granit.
I don't find the MPEe of the454sf, can you calculate it for the dimensions that you talk about ?
And give us the part dimensions.....

We have 2 4.5.4 SFs.

checked a master ring to both CMMs. CMM A was measuring -0.0003 under nominal and CMM B was measuring 0.0001" over nominal. This this be a cause for concern on a individual cmm basis?

The 454 SF is an inexpensive shop-floor workhorse CMM, not a high-precision lab-grade CMM. The published first term of uncertainty is 3.1 microns, which is 0.000122”. The Rule of Tens applies: always use a measurement device whose uncertainty is a tenth or less of the total part tolerance. I would not choose a 454 SF to chase tolerance levels under .0012”.

I would like to know what the actual calibrated uncertainty is from these two machines’ latest Hexagon Service calibration is (and how many crashes ago were the calibrations done, hmmm?)
In any case, we can add to that the probe calibration Standard deviation, then we can add a few more microns for differences in touch-probe module health (any leaking oil?), add a few more and finally there may be a few more rounding errors in Qual Sphere size definition. Also long styli add more microns, plus changing wrist angles adds a lot more microns and those get multiplied by long styli.
That will get us close to or even over the 4 to 6 ten-thousanths inch differences we see here, but I agree there may be room for improvement to get the machines to match better.

I would run some MM-units tests on a certified ring gauge, not a part(!), swapping this equipment between machines:

• Qual sphere (created anew in the MM-units test program which needs to have 6 decimal places set)
  
• Probe body
  
• brand-new Light Force probe module
  
• brand-new large-dia short-shaft Stylus tip, like a 5x20
  
• probe qual done with 25 points over 4 levels, touchspeed set to 5
  
• Ring gage measurement with touchspeed set to 5.
  

Please report back with this data. We may be able to arrange having our Service Engineers get your twins to correlate a little better.


FYI… if you need to hold 0.0001” with long styli and changing wrist angles, then your employer needs to spend waaayyyy more money on high-accuracy air-bearing CMMs with analog wrist probes or even fixed-head analog probes.

Josh Carpenter , i must agree with you and all what you write above, but I just disagree with the word"uncertainty" that you use in MPEe first term. The MPE is a tolerance, not an uncertainty. It's only the deviation value that you mustn't exceed on a calibrated standard measurement.

Josh Carpenter , i must agree with you and all what you write above, but I just disagree with the word"uncertainty" that you use in MPEe first term. The MPE is a tolerance, not an uncertainty. It's only the deviation value that you mustn't exceed on a calibrated standard measurement.

Jeffman, you are entirely correct.
I employ the first term as a means to introduce the idea that the measurement device has a measurable amount of uncertainty in it's results.
As I pointed out the effective uncertainty comes from many other factors like long styli, etc, but we begin with the tolerance the machine is supposed to hit when making beeps on a standard.

I don't think I'll be able to submit anything else today.

I'm in south texas and under hurricane warning. We'll be closed tomorrow and possibly Monday.

Gonna spend the rest of the day backing up cmm stuff.

Hey you're in my territory!

Yep, looks like that "tropical depression" is picking up speed... good luck!

Hit me up if you need to get this sorted out, I work with the Service guys all the time.