Minor Nonconformity - Ruler Attribute Study

[dwperron]

Over where I work we have come up with this simple rule for using rulers for measurements:
If what you are measuring has a tolerance, you must use a calibrated ruler.
If there is no tolerance on the measurement, then consider the ruler as an aid, not a measuring device.

Also, like any measurement device, the resolution of the ruler must be adequate for the measurement accuracy.

 

[Mordacious Enigma]

Thank you for the replies, everyone. I had a three day weekend and didn't feel comfortable talking about work related matters outside of work hours. I am confident in moving forward with a variable study. However, as mentioned, +/-1mm is really tight, which would make the data look weird, and possibly irrelevant to an auditor. I wanted to create an mock Excel where I had two operators and 30 parts, but all I ended up getting was 0mm/1mm in my head (since those were the ranges).

It just doesn't make sense to me to present what is essentially binary code, but I guess it is better than no data at all, hence the finding. As far as the fixture idea, I can pass it along since we have numerous 3D printed fixtures we make in-house that I "calibrate". I would say I am surprised that this whole thing got to three pages, but prior to creating my account, I ran across 20+ page thread about whether or not rulers need to be calibrated. I'll move forward with the Excel to address the finding, and talk to the Production Manager and my boss tomorrow about removing the rulers altogether for either something like a go/no go gage, or nothing at all. All our control plans are changing at some point this year.

 

[Bev D]

Well as others have said ‘nothing at all’ would likely be not acceptable at all given that this tight of a tolerance is most likely a critical or at least major characteristic which would require some kind of acceptance testing. You would definitely need to confirm that this kind of characteristic doesn’t’ require a control plan…it could lead to a major.

That said you are really overthinking this and are just paralyzed by the yabbits. (Yabbit = yeah but what if something bad happens). In all reality you will probably have to default to a real attribute study (this is a Kappa test and NOT % agreement) but we can only know if you do the test. Do the test - no one will die. You will learn something valuable. Do the test. Just do the test - save your thinking for the actual results…

Remember that you will need (marginally, no more than 1.5mm out of spec) failing units and your sample size should be at least 20 units, preferably 30+

 

[Miner]

(Yabbit = yeah but what if something bad happens).

I love this. Never heard it before, but it is SO appropriate.