Sorry, what do you mean by "call it quits" ?
The Gaussian curve - or normal curve - is a function whose values extend to +/- infinity. So, in order to deal with that they artificially consider +/- 3 sigma the "endpoints" of the function (or "call it quits" at those points - an American colloquial term).
How to tell which gage has less than 10% for us to choose before we use for the measurement/study?
You have touched on one of the problems with GR&R. To use another American colloquialism - chicken and the egg. You cannot tell ahead of time if a gage is good enough to measure the process variation you need to supply to the GR&R process. But, since that is not a precise process anyway, use the best gage you have available - best being judged as least potential for gage and measurement error (High resolution, little operator intervention, measured at the same spot on the part)
With reference to first version of AIAG SPC Manual:
1) Page 126 says "If the process variation is known....", how do you know or how can we tell?
My guess is a study of the Total Variation Equation - but it is unlikely they were thinking of anything that sophisticated. Most likely they are thinking of a capability study.
With reference to first version of AIAG SPC Manual:
2) What is this "process variation"?
It
should be the
portion of the
total process variation specifically attributable to the process alone - that is, stripped of non-process variation, such as of operator, gage, measurement and lot variations, as examples.
With reference to first version of AIAG SPC Manual:
3) Page 119 says to assess a measurement system prior to engaging in spc, if it is, we will have to wait until the production run is completed in order to select parts that represent the range of the process. Is it correct? Can we do GRR and SPC at the same time? especially when we run a new part?
Again, this is a problem with the concept - chicken and egg, again. How do you know that the calculations you used for your capability and SPC control limits were not incorrect due to "chunky" data of statistically insignificant gage resolution? You best hope is to use a gage that had satisfactory GR&R on similar processes. Also take care not to introduce measurement error - that is different from gage error. That is using a perfectly good gage incorrectly. That allows a gage that passes GR&R to generate inaccurate data.
With reference to first version of AIAG SPC Manual:
4) Page 126 of MSA says "The specific gage the team using has a % GRR = 25% of the tolerance....". How do we know? This section is about Attribute gage, how to establish %GRR for attribute gage system? , if we have not conducted a GRR for this gage using the parts understudy in variable manner, how to arrive at %GRR?
Not sure of this reference - I only have 4th edition handy.
With reference to first version of AIAG SPC Manual:
5) In your earlier post, you said to use gage that is not more than 10% of the tolerance (part or process?), how to tell this % before we do GRR? and the system has not been used on this new part yet?
The gage has a resolution - the smallest readable increment. That should be no less than 10% of the tolerance - and even that yields very low ndc, but it is a common starting point
