# GR&R (two dimensions on a shaft) with few parts and only 2 operators

W

#### winding

I have been asked to put together a GR&R for two dims on a shaft. One is the diameter (.5906/.5907) and the other is run out (.0001)

We will have a max of 5 parts available to measure and only two operators. We want to perform the GR&R on two pieces of equipment: a laser mic and a CMM

Can anyone give me some suggestions on how to do this with such a small number of parts and only two operators?

#### normzone

Trusted Information Resource
Wow ... I'm only a tyro at dimensional stuff ... but wouldn't that .0001 diameter tolerance allow a .0002 runout ?

My second question, "who the hell designed that thing", will be saved for another time.

N

#### ncwalker

normzone: Not necessarily. You can have a runout less than diameter tolerance. In fact, that's WHY you have a runout tolerance. It means that the diameter can be anywhere in the 0.0002" range but wherever it is in this range, the runout must be less than 0.0001"

Also, I agree, who designed that? I work in a pretty tight tolerance industry with rotating things and we are pretty high tech. This is a tight tolerance.

winding: You just fill in a Gage R&R with what you have and you will get a result. There is not "special way" to calculate it when you have LESS than appropriate number of samples.

Here comes my rant:
The number of samples, regardless of how it is configured, affects the confidence interval of the mean and the standard deviation. What this means is the results of a Gage R&R which are ALWAYS reported as ONE number, really should be reported as a range. Yet we do not do this.

Everyone considers the hurdle to be 10%. But if I did a study with 3 parts measured twice by one operator and got a 9.5% OR I did a study with 25 parts measured 5 times by 3 operators and got another 9.5% we would just accept them as both passes. When CLEARLY we should be really concerned with the first one. Good Lord world, you have the study in Excel or Minitab. All the required stuff to calculate the confidence intervals is THERE. And yet we do not do this ... myself included.