Gage R&R

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 0.16740 11.26
Repeatability 0.09878 6.64
Reproducibility 0.06863 4.62
OPERATOR ID 0.00000 0.00
OPERATOR ID*BSN 0.06863 4.62
Part-To-Part 1.31940 88.74
Total Variation 1.48680 100.00


Study Var %Study Var %Tolerance
Source StdDev (SD) (5.15 * SD) (%SV) (SV/Toler)
Total Gage R&R 0.40915 2.10712 33.55 13.17
Repeatability 0.31429 1.61859 25.78 10.12
Reproducibility 0.26196 1.34912 21.48 8.43
OPERATOR ID 0.00000 0.00000 0.00 0.00
OPERATOR ID*BSN 0.26196 1.34912 21.48 8.43
Part-To-Part 1.14865 5.91554 94.20 36.97
Total Variation 1.21934 6.27962 100.00 39.25


Number of Distinct Categories = 3

See the above results of an actual gage R R. Anyone can suggest if this gage R R is acceptable or not?
Notice the (%SV) vs. (SV/Toler)

What do you think? Any suggestions.

Note: I judge the acceptance criteria by looking @ all indices, inculding the # of distinct categories

Thanks in advance

Any gage R&R experts interested in looking at this?

Gage R&R

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 0.16740 11.26
Repeatability 0.09878 6.64
Reproducibility 0.06863 4.62
OPERATOR ID 0.00000 0.00
OPERATOR ID*BSN 0.06863 4.62
Part-To-Part 1.31940 88.74
Total Variation 1.48680 100.00


Study Var %Study Var %Tolerance
Source StdDev (SD) (5.15 * SD) (%SV) (SV/Toler)
Total Gage R&R 0.40915 2.10712 33.55 13.17
Repeatability 0.31429 1.61859 25.78 10.12
Reproducibility 0.26196 1.34912 21.48 8.43
OPERATOR ID 0.00000 0.00000 0.00 0.00
OPERATOR ID*BSN 0.26196 1.34912 21.48 8.43
Part-To-Part 1.14865 5.91554 94.20 36.97
Total Variation 1.21934 6.27962 100.00 39.25


Number of Distinct Categories = 3

See the above results of an actual gage R R. Anyone can suggest if this gage R R is acceptable or not?
Notice the (%SV) vs. (SV/Toler)

What do you think? Any suggestions.

Note: I judge the acceptance criteria by looking @ all indices, inculding the # of distinct categories

Thanks in advance
My initial take is that the GR&R takes up too much of the tolerance (~13%). But what I concentrate on is the repeatability to reproducibility "ratio". That is I would expect that my reproducibility would be larger than my repeatability. Given that is not the case for you, I would try to understand why repeat measures on the same part is so large. Just my $0.02.

Any gage R&R experts interested in looking at this?
Not enough information; in order to do a thorough evaluation we'd need to see how much tolerance the process is using (the 13% GR&R value may or may not be acceptable, based on this factor). If this is an average-and-range study, and it appears to be, we should also be able to see statistics such as the UCLr value, and the individual ranges of the operators. BTW, I've never thought much of within-study variation as a deciding factor. For practical purposes I'm usually more concerned with process variation, although within-study variation is useful for helping to judge the integrity of the study (i.e., if there's a significant difference between within-study and process variation, something's wrong somewhere).

Observed Study variation is about 40% of tolerance. I am not sure if the parts chosen are truly representative of spec - or even the process as JSW said as no info is given about process spread. Are we looking at a fairly capable process? If this is so, or if the parts do not cover the prodess spread, you may see low nDC and a large GRR% of Study. (33.55% here, also contribution of GRR is over 10%). What is the instrument being used for - process improvement or inspection?

Thanks to all for replying/comments:
Here is the deal:
This is an end of line tester. I came in late in the game; I would have performed destructive/nested gage RR rather than the traditional gage RR.
Process tolerance = 16

So the gage RR here is 10 parts 3 operators, each meaured each part 3 times.
Looking forward to your comments.

Thanks,

Master BB

It is a good idea to always look at # of distinct categories in R & R output which tells you physical interpretation. Here is an explanation.

We always say that a variable data is better than attribute data. But as gage R & R get poor, the gap between variable data and attribute data reduces.

Gage R & R-------Measurement system ability to discriminate
10 %------------ 14
20 %------------- 7
30 %-------------5
50 %----------- 2

You can see that at 50 % Gage R & R, the measurement system can discriminate only in 2 categories which is like attribute data. 4 is bare minimum and 5 or more are desirable.

In your case, # of distinct categories is 3 which indicates that measurement system needs improvement.
Trust this clarifies.

Arvind

Thanks for your reply.
This is a unique case of Gage RR.
Once agian I look at all of the indices of Gage RR.
In Minitab you get % Contribution (of VarComp), % Study Variation, & % Tolerance (SV/Toler) as well as number of distinct categories.

If any one of these indices fail the criteria, I would not accept that gage. My customer want to only go by the % Tolerance (SV/Tolerance) as an acceptable criteria (anything less than 30% is acceptable "Customer"). And I am disagreeing with my customer. The MSA AIAG manual is not that quite clear when it comes to all of the indices.
Who is right here?
Can you refer me to a published paper regarding the subject & possibly from a Credited Journal.

Any comments, replies regarding the concern greatly appreciated.

Regards,

Master BB

My customer want to only go by the % Tolerance (SV/Tolerance) as an acceptable criteria (anything less than 30% is acceptable "Customer"). And I am disagreeing with my customer.
There are no universally accepted pass/fail criteria for GR&R, because it's a type of experiment. The results are what they are, and it's up to you to decide how to act on them. If the measurement system is being used to evaluate the customer's material, and they're concerned only with one output statistic, and they have been informed of the risks inherent in that choice, and are willing to accept nonconforming material that results from measurement error, then God bless the customer.

There are no universally accepted pass/fail criteria for GR&R, because it's a type of experiment. The results are what they are, and it's up to you to decide how to act on them. If the measurement system is being used to evaluate the customer's material, and they're concerned only with one output statistic, and they have been informed of the risks inherent in that choice, and are willing to accept nonconforming material that results from measurement error, then God bless the customer.
Let's understand something. Most customers do not care about measurement error. How often do you buy something and ask what is the measuerment error on the deli scale? I give the customer alot of respect for asking how much of the tolerance does the measurement system take.

Let's understand something. Most customers do not care about measurement error. How often do you buy something and ask what is the measuerment error on the deli scale? I give the customer alot of respect for asking how much of the tolerance does the measurement system take.
There are two sides to this, as with most stories. I think most customers care a great deal about meaurement error, or anything else that might negatively affect product quality; it's just that they don't understand it. Often the level of concern only heightens when something bad happens. The OP is doing his best, it seems, to prevent bad things from happening, and to protect his customer from his own ignorance. The deli scale analogy is apt, because we assume that there protections in place against inaccurate scales, so unless a person is very cynical, it's not a big concern. In manufacturing the same principle applies: our customers have a right to believe that we have protections in place, and should be able to assume that gages are regularly calibrated and that we have the expertise to assure that that an efficacious measurement system is in place.
Nonetheless, if the customer has his gun pointed at his foot and we repeatedly tell him that he should point it elsewhere, and in the end he shoots himself in the foot, then he has to be ready to accept the responsibility for it.

Of the 5 indices you mentioned, there are only two independent and rest are identical.

1) R & R
a) Gage R & R is most frequently used.
b)Square of gage R & R is Variance
c) # of distinctive categories is also calculated from gage R & R.

2) Precision by tolerance-
To calculate this index, you need to know tolerance specification. Tolerance is not needed for Gage R & R.

R & R is better than P/T ratio because when process capability improves, gage R & R deteriorates which requires you to look for better measurement system.
Precision/Tolerance remains the same even when process capability improves so in that sense, it gives you a false security that measurement system is OK.

As I posted earlier: The number of distinct categories is below 5, % Study Var is over 30%; Only the % Tolerance is under 30%, Over 15%.

Now the customer wants to tighten the spec based on this results. This is extremely amazing; how can you revise the spec based on % Tolerance & not other methods including but not limited to Cp, Cpk, Pp, & Ppk.

Thanks in advance.

As I posted earlier: The number of distinct categories is below 5, % Study Var is over 30%; Only the % Tolerance is under 30%, Over 15%.

Now the customer wants to tighten the spec based on this results. This is extremely amazing; how can you revise the spec based on % Tolerance & not other methods including but not limited to Cp, Cpk, Pp, & Ppk.

Thanks in advance.
I still don't understand. Your customer wants to tighten a spec, based on the fact that the measurement system is questionable? Makes no sense, on any level. Does the customer realize that GR&R results will only be worse if the spec is tightened?