S. Udhayha Kumar said:
Could some one help me in understanding with simple example
a) how the gauge performance is to be assessed and the curve is to be drawn?
b) what is the influence of attribute GRR% as the Ppk varies?
You make good questions...
This post is in connection with GRR none attribute with Cp.
http://elsmar.com/Forums/showthread.php?t=17521
attribute GRR%, I think this has no sense, GRR attribute is Go /No Go, Good /No good and no %
But please I beg your pardon if I am wrong, I left the Autobile sector 4 months ago and many changes are on course.
Lets say better cpk instead of ppk due to if ppk your process is not stable and you can not submit parts until to obtain Cpk (Stable process, see PPAP 4 th Edition).
When the MSA is variable the formula is developed by MSA in the previous link, but if the MSA is attribute the relation is all or nothing and therefor how to establis a correlation between a none attibute var with an attribute one?
I mean:
Gage no OK. Simply you can not use this tooling, this is not mean that the parts are bad parts (Lets say Cpk<1.67 ppm>2.3ppm) the problem is that a formula could not be establish and nobody will know if the parts are bad due to the Gage have an error that nobody had previously established, simply you know that the error is higher than the Go/No go tolerance but it is not quantified.
That´s why it is better to make none attribute studies,
but le me try to make you the correlation. No previously made by IATF. So this is in my way and you now...
Lets say.
Gage Diameter.
Go: Diameter: 32mm+-0.5mm ->Tolerance = +0.5-0.5 -> 1mm
No Go Diameter 35mm+-0.5mm
Therefore if the part is not go this means that the Diameter is higher than:
>35mm-0.5mm -> 34.5mm
Relative Error= (32-34.5)/1mm -> Relative Error=2.5 -> 250%
Absolute error= 32-34.5 = 2.5mm
But you only know that the error is higher than 2.5mm not 2.5mm, this is the problem.
Therefore now is to establish a correlation between Cpk, Error.
cp= (Process variance + Error )/ 6Sigma
This relations are not enough exact, only to make you understand the correlation.
So it is better to simplify:
Criteria: Cpkobs similar to Cpact
Gage is OK -> You can use it and let approximate than Cpkact is similar to Cpkobs.
Gage is not OK -> Search other Gage.
Part.
Part do not Go -> Error higher than Tolerance Gage Go /No go characteristic - Refuse
Part Go -> Error less than Tolerance Gage -< Accept part.