We are doing some R&R studies on some gages we have prioritized. I've looked on page 131 of the MSA manual in reference to the attribute study. We did a study on the use of a go no go gage from 3 operators, 20 samples each 3 x. Would the attached Attribute study satisfy IATF requirements? Looking for any input. Thanks.
IATF & Attribute Study requirements
- Thread starter KyleB
- Start date
You're welcome. You should be aware that although you think it makes sense and I think it makes sense, you might have benighted customers who don't.
Ooi Yew Jin
Starting to get Involved
For me, most of them want to get pass in the study for audit in purpose and may forget the room for the improvement in the study results.
We can further check with appraiser 1 whether there is any special cause when checking the trial number#3.
This information can be turned to the knowledge or input on further improvement.
We can further check with appraiser 1 whether there is any special cause when checking the trial number#3.
This information can be turned to the knowledge or input on further improvement.
John Predmore
Trusted Information Resource
If I was reviewing your study, I would ask how were the 20 parts selected. The legitimacy of the result doesn't matter whether calculations were done correctly if the samples were not chosen appropriately.
AIAG (2nd edition section 6) says in a short study of 20 parts, some of the parts should be slightly below and some slightly above acceptance limits. You have not explained whether both directions are represented in this study, and how far out of limits are the NoGo parts. It is recommended that this determination (how far out of limits) be made using variable measurements.
Making Sense of Attribute Gage R&R Calculations - iSixSigma
"Practically speaking, if 'clearly good' parts and 'clearly bad' parts are chosen, the ability of the measurement system to accurately categorize the ones in between will not be tested."
A better approach (the long method study) involves 8 sample parts which are known to reflect the full range of variation, equally spaced. The 8 parts are measured 20 times, and what you want to see are one part accepted by the attribute gauge 20 times out of 20 and one part rejected 20 out of 20. The other six parts should be accepted 1 < a < 19.
The Minitab support website has good explanations of how the long or default study methods are conducted. You don't have to be a user of the software to learn from their website.
AIAG (2nd edition section 6) says in a short study of 20 parts, some of the parts should be slightly below and some slightly above acceptance limits. You have not explained whether both directions are represented in this study, and how far out of limits are the NoGo parts. It is recommended that this determination (how far out of limits) be made using variable measurements.
Making Sense of Attribute Gage R&R Calculations - iSixSigma
"Practically speaking, if 'clearly good' parts and 'clearly bad' parts are chosen, the ability of the measurement system to accurately categorize the ones in between will not be tested."
A better approach (the long method study) involves 8 sample parts which are known to reflect the full range of variation, equally spaced. The 8 parts are measured 20 times, and what you want to see are one part accepted by the attribute gauge 20 times out of 20 and one part rejected 20 out of 20. The other six parts should be accepted 1 < a < 19.
The Minitab support website has good explanations of how the long or default study methods are conducted. You don't have to be a user of the software to learn from their website.
One of the problems with attribute studies is that it's sometimes very difficult to come up with "bad" parts, and asking for ones that are "bad" on both ends of a bilateral tolerance can be too much to ask. Sometimes the efficacy of a gage can be deduced simply by measuring it, and other times it's not so easy, such as in the case of gages used for verifying GD&T callouts. Let's not forget that the OP was asking about the format of his template and not a tutorial on attributes gage studies.
Yes, Both above the upper limit & below the lower limit were used.
I vote for continuous or variable data vs discrete Go/No - Go events...Why? If one only collects binary events, how do you know if the process if shifting out of spec? When you do its already too late you're SOL....rejecting 10 parts out of say 40, its already too late, you have lunched 10 parts or at least have to rework the same. Be proactive rather than reactive....my 5Euros worth
Similar threads
