Gage R & R for large amount of inspections

#1
We currently have a large amount of both variable and attribute inspections on our control plans. In the past customers have only asked for GRR on critical checks. I am curious how others are meeting this requirement- Through annual GRR? Grouping SME into families? How are attribute checks handled? How about camera systems?
Any input is appreciate.
7.1.5.1.1
Statistical studies shall be conducted to analyze the variation present in the results of each type of inspection, measurement, and test equipment system identified in the control plan. The analytical methods and acceptance criteria used shall conform to those in reference manuals on measurement systems analysis. Other analytical methods and acceptance criteria may be used if approved by the customer.
Records of customer acceptance of alternative methods shall be retained along with results from alternative measurement systems analysis (see Section 9.1.1.1).
NOTE Prioritization of MSA studies should focus on critical or special product or process characteristics.
 

Miner

Forum Moderator
Staff member
Admin
#2
An R&R study need only be performed one time unless there has been a fundamental change in the measurement system. There are a considerable number of searchable threads that discuss this topic. Grouping gauges into families is an excellent approach to minimizing the volume of studies required.
 

BMcCalla

Starting to get Involved
#3
I disagree with the one time only statement above. One R of R&R is for Repeatability, which focuses on the equipment, which is what you get when you pay extra for the R&R from the gage vendor, but the other R is Reproducibility, which includes the human element of error, so you don't know the true impact when you constantly change inspectors. Second, although a frequency isn't specified, some customers state that they do not want any data in a PPAP submission that is more than a year old, so the one and done for Gage R&R may not work for new PPAPs. Third, attribute studies focus entirely on the human error in the measurement, so again, my earlier point is that the risk is greater with an influx of new inspectors. Miner's point about family of gages would save a lot of time, but if you are a Ford supplier, the STA has to OK this approach. Also in IATF FAQ#6, they have taken a very narrow interpretation on what constitutes a "family of gauges". Quote "A complete statistical study on each single piece of equipment is not required. Instruments with the same characteristics (e.g. measurement range, resolution, repeatability, etc.) can be grouped and a sample instrument (representative of the gauge family) can be used for the statistical study."
 
Last edited:

Miner

Forum Moderator
Staff member
Admin
#4
I disagree with the one time only statement above. One R of R&R is for Repeatability, which focuses on the equipment, which is what you get when you pay extra for the R&R from the gage vendor, but the other R is Reproducibility, which includes the human element of error, so you don't know the true impact when you constantly change inspectors.
The type of ANOVA analysis performed in an R&R study assumes that the operators were randomly selected and are representative of all operators. Therefore, it will not change if you change operators unless you have stacked the deck by selecting your best operators.

Regarding gage families, the linked resource I provided in post #2 provides guidance on how to create the families.
 

BMcCalla

Starting to get Involved
#5
Miner, your guidance on family of gauge choices is a nice tool. If the study sample assumed all operators would measure alike, would this not tend to dismiss the risk of an operator not using the measuring device correctly? It is easier to make the point with attribute tests. Let's say that I am to give a Farnsworth Munsell Hue test to my color inspectors or give a visual inspection attribute test with KAPPA results to my injection molding inspectors. Next year I have a new crop of inspectors. I would likely retest all the inspectors to minimize risk that one of my inspectors doesn't have sufficient competency to judge parts, creating higher risk of Type 1 and Type 2 errors. The assumption all inspectors were properly trained and are knowledgeable of the inspection criteria is now too high a risk for medical devices or even automotive parts manufacturers to forego ongoing MSA studies.


1552398380285.png
 

Miner

Forum Moderator
Staff member
Admin
#6
Miner, your guidance on family of gauge choices is a nice tool. If the study sample assumed all operators would measure alike, would this not tend to dismiss the risk of an operator not using the measuring device correctly?
It doesn't assume that the operators measure alike. It assumes that the operators not selected will measure like the operators that were selected, specifically with the same amount of variability between operators.

It is easier to make the point with attribute tests. Let's say that I am to give a Farnsworth Munsell Hue test to my color inspectors or give a visual inspection attribute test with KAPPA results to my injection molding inspectors. Next year I have a new crop of inspectors. I would likely retest all the inspectors to minimize risk that one of my inspectors doesn't have sufficient competency to judge parts, creating higher risk of Type 1 and Type 2 errors. The assumption all inspectors were properly trained and are knowledgeable of the inspection criteria is now too high a risk for medical devices or even automotive parts manufacturers to forego ongoing MSA studies.
This is quite different because in this scenario the operators are the gage. The Farnsworth Munsell Hue test is essentially a calibration test of the operator/gage. As calibration is required for all gages, this test is required for all operators. In this case you would have greater variability between operator/gages than you would between multiple micrometers.
 

BMcCalla

Starting to get Involved
#7
I agree the operators are the focus of the "calibrating" activity in the hue test I mentioned, but my point is that operators also factor in variable studies, even if it is to a lesser extent.
 

Jim Wynne

Super Moderator
#8
I disagree with the one time only statement above. One R of R&R is for Repeatability, which focuses on the equipment, which is what you get when you pay extra for the R&R from the gage vendor, but the other R is Reproducibility, which includes the human element of error, so you don't know the true impact when you constantly change inspectors.
I've never heard of getting R&R data from a gage vendor, and I don't know what purpose it would serve given that GR&R generally requires measuring production parts. You will never know the "true impact" of human error, and getting the knowledge you seem to think is useful would require all operators to participate in gage studies. The idea of selecting operators that are representative of the population assumes that people have the requisite competence in using the equipment. Part of the purpose of doing gage studies is to verify, as far as possible, that the instrument in question is appropriate for a specific application.

Second, although a frequency isn't specified, some customers state that they do not want any data in a PPAP submission that is more than a year old, so the one and done for Gage R&R may not work for new PPAPs.
The existence of ignorant customers is something we all have had to deal with. I've found that explaining why new studies aren't necessary or useful will often solve the problem, but when the customer insists, you have to do what they want. An ignorant customer asking for something useless doesn't make the object of the request technically useful.

Third, attribute studies focus entirely on the human error in the measurement, so again, my earlier point is that the risk is greater with an influx of new inspectors.
Attributes studies do not focus entirely on human error in evaluation. An important thing to be learned from attributes studies--especially visual evaluation-- is whether or not the expectations are reasonable. For example, if it's found that an inordinate number of operators disagree with one another, there's probably a problem with what's expected of them for one reason or another. Not being able to discern the defects isn't necessarily due to operator error.
Miner's point about family of gages would save a lot of time, but if you are a Ford supplier, the STA has to OK this approach. Also in IATF FAQ#6, they have taken a very narrow interpretation on what constitutes a "family of gauges". Quote "A complete statistical study on each single piece of equipment is not required. Instruments with the same characteristics (e.g. measurement range, resolution, repeatability, etc.) can be grouped and a sample instrument (representative of the gauge family) can be used for the statistical study."
I don't see that as being a "narrow interpretation." It describes the criteria for creating families.
 

BMcCalla

Starting to get Involved
#9
Per Mr. Wynne's point. I agree that getting studies from the vendor of the gage does not provide a lot of useful information, but I have seen it commonly done. Even on location, I have seen substitutions for measured parts as samples (often gage blocks) used for R&R. Per the customer viewpoint and IATF's definition, a family would mean that a 1" mic and a 2" mic can not be grouped in the same family, nor can a 6" caliper and a 12" caliper, due to different measurement ranges, where a single caliper or a micrometer study would have sufficed in the past. I agree with your assessment of customer expectations and beliefs, where one IATF subscribing customer currently requires all gauges that touch their product require R&R unless a waiver is granted. I believe in a risk-based model for gauge R&R, with special characteristics and those individuals that measure the SCs taking precedence, but some of us have to follow the customer's wishes for a blanket approach, namely all measuring devices cited in control plans.
 
Last edited:
#10
You have to be VERY careful with gage vendor R&Rs.

For a simple example, if I am selling you a digital caliper and you want an R&R to certify the caliper is good, I'm going to go get gage pins and give you a fantastic R&R and say "My gage is good, so pay me."

We say "Gage R&R" but the AIAG term "Measurement System Analysis (MSA)" is a better term. First, because the actual R&R is a tool in the kit. But more importantly, it is a measurement system - and the parts themselves are part of that system.

So you will pay me for my calipers which I R&R'd for you on gage pins, but you make rubber bands. Suddenly, my calipers won't be so good. If you're going to use the calipers to check rubber bands, then you R&R the calipers ON rubber bands, and check the SYSTEM.

I have seen MANY suppliers get burned by this. They buy an expensive automatic gage and the gage supplier vets it on master parts instead of production parts, then it doesn't work in production. Ooops. Already paid the gage manufacturer, so he's gone. And the supplier burns money to get the gage working.

In your contracts with gage vendors, a payment milestone needs to be tied to passing an R&R on production parts the gage was designed to measure. You have been warned. :)
 


Top Bottom