Measurement comparison

[[Cl...]]

We do a measurement comparison using a different measurement method, whenever we write a new program using
the CMM/Keyence/Zone3, etc. I've been asked to write a work instruction outlining this procedure. What I'd like your
input on is what tolerance range to apply to the results of the comparison. Currently we are requiring (unofficially) that the comparison result be within 10% of the program result.

[[Da...]]

Hello Clarke,
We are a contract manufacture working mainly in the medical and robotics industry and 10% is the number we try to stay under when correlating results from one piece of equipment to another.

[[Cl...]]

Thank Dakota!

[[Je...]]

My standard has always been a Precision-to-Tolerance review of both machines using a total of 10% of tolerance at most. Preferably conduct a GR&R on both machines and combine the data under a single analysis.

Correlation is a VERY difficult topic to troubleshoot and implement. A value correlation of 10% of tolerance adds significant measurement uncertainty if your original equipment standard was already 10% of tolerance for program approval (this is pretty standard unless a GRR is conducted). You have built in a potential error stackup of 20% attributed just to your measurement method.

[[An...]]

I personally find a correlation of 10% to be a lot.
First of all, you have to ask yourself whether different measuring devices (tactile, optical) are at all comparable with each other.
Each manufacturer has its own guidelines of calculation methods, the design of measuring devices is different and thus the achievable degree of accuracy.
The reference value for us is always the tactile measurement on the CMM.
For new parts, we always make comparative measurements on other devices and internally give ourselves a comparison tolerance of a maximum of 5% of the specified dimension on another measuring device.
If the correlation is exceeded on an optical measuring device, for example, the confidence remains with the tactile measurement and is measured again tactilely for further control.

[[Ma...]]

We did some cross measuring with customer - stay at 10%.
It was a same part measured by their and our program - but strategy wasn't the same and part was forged and sandblasted - so 10% with no same scan path can be relatively small deviation.

[[An...]]

ok, 10% is tolerable for something that is forged.
I thought it was about a part that was turned / milled manufactured.

[[Ri...]]

I use this MSA spreadsheet, it has proven to be very good to view your results.
There is a sheet showing some data from a 20 pc test.
I have another one that is for 35 pc's, just cant find it.

Copy/Paste your data to each relative column
Add the Nominal and tolerances
and it calculates the rest at the far right by percent.

#Master-Correlation or Capability Study 20 pc.xlsx

[[Cl...]]

Awesome, thank you Rick!