What Do 'They' Want in a Gage R&R? Every gage on every feature on my control plan?

Gage R&R is made up of thee variations: Appraiser, part & Gage. When Someone asks for a gage R&R what do they want? I'm asking is that the goal is 10% of tolerance or less.
If my process is out of control my part variation will kill my R&R. Is there a way to extrapolate a gage R&R to a point where it can be defined that the gage should not be used below X tolerance? I'm trying to keep from doing an R&R on every gage on every feature on my control plan.
Well, Marc used the URL address and had no trouble.

[Roger Eastin]

I believe that if you set up your Analysis of Variance MSA study correctly, you will be able to see your gage variation regardless of part-to-part variation. You should be able to ratio your gage variance to your tolerance spread and determine the minimum tolerance that your gage can handle. (In fact, there are many software packages that will do all the work for you.) Check the MSA manual for specifics.

Yes, My Gage software will show just gage variation. Roger, I assume you feel Gage R&R is a total of gage, appraiser & Part variation. But, what do I put on my Dynamic Control plan? Total or just gage variation?
Interesting item Dyer gages advertises a line of gages with specfic gage R&Rs advertised. This must be only gage variation.
Would like more input.
tom

[Don Winton]

I am not QS savvy, but will try to address anyway.

Typically, I define three separate fields in my ANOVA report: Gage, operator and part. The part report is included to indicate variation of part-to-part, but strictly speaking, is not part of the gage R&R. I do not believe that part variation skews gage R&R if it is included in the ANOVA, therefore being known. The typical table I use is ten parts by three operators by one gage, but many combinations are available using ANOVA.

The report normally goes something like this:

Repeatability: The variation resulting from the inability of a measuring device to obtain the same result over and over.

Reproducibility: The variation among the persons doing the measurements or inspections using the same methods or equipment.

Inspection capability: Repeatability and reproducibility are quantified and combined to determine the inspection capability. Typically, in an ANOVA table, this interaction is the Gage x Operator f values. There are other methods (See "Total Quality Management Handbook," Hradesky).

Roger and Marc. You are probably more qualified to address this than I. Could not there be a statement in the control plan that states that gages used have <some value> capability and MSA studies on file to verify and validate the gage used? Just a thought.

Regards,
Don

With regard to the first post by thomash, part variation should not influence your gage study. If your parts are tapered, for example, measuring the same part 3 times at different locations will obviously inflate the variation. A gage study is to determine the variation of the measurement system, not part variation. Therefore operator and gage error determinations are what you want. I mark parts that are less than perfect, so each operator measures the same spot. Remember, repeatability is determined mostly by the range of readings from each part (in the classic R&R.) It is difficult to use historical data to determine gage error, since the (sigma of data)=((sigma of measurement)+(sigma of product)). In addition to the classic GR&R, I use a gage only determination for measurement systems that mostly have no operator influence, such as Laser mics and CMM. Measure the same part in the same place 10 to 15 times. Calculate the sigma, compare 6 x sigma to the tolerance. This takes out the (sigma of product) part of the calculation as indicated above, and makes the (sigma of data)=(sigma of measurement).

[This message has been edited by Batman (edited 01-03-99).]

[Marc]

Well, for my 2 cents, I'll stay shut up. Gage R&R isn't my forte. Roger, Don and Batman(?)(!) stated everything I would have said (in fact, more than I would have said 'cause it's not my forte). I will say there are three variables (I see this as an MSA question, in part):

1. Equipment Variation
2. Operator Variation
3. Part Variation

Ultimately each has to be considered. Your evaluation has to be as a system and as such you have to be ready to explain the part each plays in the system as a whole.

OK folks - one for the group. How about where you use one instrument for multiple measurements on each part and on 'other' parts as well?

[This message has been edited by Marc Smith (edited 12-31-98).]

My addition to my two cents. I may be mistaken, perhaps I am not understanding the words as printed, but one should not consider part variation [over time]as part of a classic R&R. Obviously one considers the variation seen with each of the parts that are used in the test, i.e. 10 parts, 3 measurements each, 3 operators, but the classic R&R is for determining Operator error, and Equipment error; it uses part measurement variation for the calculations. When one measures the parts for the gage study, the parts cannot change, this would add variation and muddy any analysis. In part, one can see if the overall error contains too much operator, too much gage, or how much of both. It is important in making decisions as to which or what to 'fix,' and if and when acceptable, how much confidence one has in the eventual measurement of the product. Here is where the part/product variation is detected.
Re Mark's question about one gage measuring multiple features or parts, this can be broad or simple. A CMM is one example, one with which I am familar. Do you have an example, or an idea?

[This message has been edited by Batman (edited 01-02-99).]

[Marc]

I list part variation as a consideration in the evaluation as a whole. Agreed not a part of classic R&R.

Re Mark's question about one gage measuring multiple features or parts, this can be broad or simple. A CMM is one example, one with which I am familar. Do you have an example, or an idea?

Nothing specific. Just fishing for comments.