Gage R&R Study - Marking Parts, Number of Parts and Number of Operators

[Dawn]

I believe you both are correct, as the parts are to be checked at random, but also must be identified (marked). Mark the parts with a black marker in one spot on the part; number the back of the parts; then mix them up for each operator while someone else writes them down. They are still being checked at random, but they are also being checked in the exact spot. This will give you true readings.

 

[David Drue Stauffer]

Gage R&R Study

I recently embarked upon a journey to write a course (ppt pres.) to teach fledgling companies how to perform Gage R&R Studies.
In my endeavor to do so, I consulted the chat rooms of net. I pulled off a discussion that said,
"A typical Gage R&R study involves three operators taking two measurements of ten "identical" parts. The unlabeled parts should be passed to the operator in random order so that there is no bias when performing the measurements. It is important that the operator is not aware of which part they are measuring so that no attempt can be made to try to replicate the previous measurement.

After all of the measurements have been made, a series of calculations is performed to determine the % R&R. This value reflects how effective the measurement system is. A % R&R of less than 30% is considered to be acceptable by most industry standards."

When I learned to do R&R studies, I believe in accordance with AIAG, I understood that ten parts that were selected from a manufacturing process were to "labeled" 1-10, and a spot marked on each part as to the location the measurement was to take place in order to isolate gage system error from part process error!

It seems that if the parts were measured at random in random order from operator to operator, that this would introduce a sizable amount of variation into the equation that would make the grand average numbers quite high causing the resulting percentages to escalate! Also, 10% is the acceptable limit, not 30. 11-30 is acceptable ONLY based upon the application, in other words it has to be monitored closely. 31% and above would be unnacceptable under any conditions.

The above mentioned excerpt was from Ingersoll-Rand's Athens, PA plant that produces power tools.

If you have any comments or solution to either my or their understanding of how the R&R is to be conducted, please comment!

Dave S.

 

[Laura M]

I would add that the "measurement system" is being evaluated. If the parts can't be measured in the "exact same spot" when the gage is used in production, then you shouldn't specify where to measure for gage error. You may want to do that for diagnostic purposes...is it the gage or the process, but don't fool yourself. "Measurement System Analysis" isn't getting a gage to pass R&R.

 

[David Drue Stauffer]

Marc
I believe the idea here is to evaluate how much of "process error" is due to "measurement error". To isolate the measurement error from the process error for a particular characteristic such as overall length of the part, and a 0.001" resolution caliper is being used to measure the length with a tolerance of say +/- .005". Then an R&R study would be in order to discover if perhaps the gage being used is sufficient to achieve the required capability indexes. My point is that to reduce or isolate "part variation" as much as possible from the R&R result, that the "at random parts" should be measured at the same place for each trial AND for each operator. I could be wrong here, but that's the way I learned it some years ago.
Thanks for your response.
Dave.

 

[Batman]

Hi Dave!

I think you are thinking correctly, but perhaps not stating correctly.

The random thing:
The parts are numbered 1-10, and each part's measurement is recorded on the log as it's measurement. Part #1 readings ALWAYS gets entered in part #1's line. This is for the first reading for the first operator as well as the last reading for the last operator. The procedure as stated in the AIAG is to give each operator the numbered parts in random order. That way part #1 is not always the first part measured, part #2 is not always the second, etc.

The spot marking thing:
Since the Gage R&R is to determine the measurement system - operator and gage - error, it is necessary if your parts are in any way irregular. Later, when the measurement system is deemed acceptable, you can then determing just how irregular your parts are, and that is when it is appropriate to measure the irregularity. Adding part variation to the measurement system evaluation is not good practice. Unless your parts are or are equal to NIST standards, it certainly appropriate to mark a spot on the parts.

 

[Laura M]

I agree that part variation shouldn't be "included" in a gage r&r, but if the r&r is done in different circumstances than it is used on the floor, it may not provide an accurate measurement. I've been involved in situations where gage engineers were telling how we could get a gage to pass..."hold the part here," "it takes getting used to", etc. So my point was, don't "get a gage to pass", make it realistic because otherwise it can get into production and you may be falsely accepting or rejecting parts.

 

[Batman]

Yup, get an engineer into a discussion of whether a part is good or not and the first thing they ask is "How'd you measure that?" particularly if you did not consult them first when picking the gage.

So, get them into the discussion during the planning stage of the GR&R. Let them tell "how to hold" and "get used to..." Then do the R&R. If a gage is one of those "get used to..." types, I make sure the parts are marked so the only error is the operator not used to the gage. Then I go back and ask for a more user friendly gage. Sometimes I even get the engineers to participate. Graph that some time and discover how well your operators are measuring and who should not be measuring parts.

Seriously, you should not (to the extent possible) include any part deformity in the variation calculations when performing a GR&R. If there is part deformity it will be better detectable by a reliable measurement system. You may reject a good measurement system based on poor parts, when what you really want is a measurement system that will detect poor parts.

 

[David Drue Stauffer]

Batman, I totally agree, as I have done extensive search since I originally posted my topic, this clearly solidifies the process in my mind.
My first training class was on GR&R, & was a success, the customer left well pleased, so much so that a sister company of theirs has now booked for the same training plus a Metrology class as well.
Thanks to all of you who responded for your help. This is a great source of information and discussion of ideas. Dave