Scaling a Gage R&R - 2 holding fixtures each with 10 nests

[RooBoy]

Background:
I'm modeling a Gage R&R.
I have 2 holding fixtures each with 10 nests to hold a singular component.
Even though my holding fixtures have multiple nests I am considering them as individuals as they are not truly identical.

Issue:
10 nests X each component 3 times in each nest X 2 operators X 3 Trials = Approximately 17 hours (inclusive of fixture loading, measurement, unloading). Add the 2nd holding fixture and I'm up to 34 hours of time to complete the GR&R. Not to mention taking the equipment out of service to prevent set-up tear down which would add another level of variation.

My questions:
Can I scale this model and still have a valid MSA?
Do I need to consider each nest an individual?

 

[ncwalker]

2 fixtures holding ten nests ... these are all MEASURING fixtures in some mass measuring machine? Or are these work holding fixtures that are holding parts to receive some operation? (I'm asking because to me, 20 measuring fixtures sounds like a lot ...)

 

[RooBoy]

These are not measuring fixtures. These nests are holding a component in order to be measured by a CMM.

 

[Miner]

Do you establish your coordinate system from the part itself, or from the fixture?

 

[Ninja]

Is the nest a functional part of the test? Or just a place to keep it on the CMM?
I'm guessing it's just a holding fixture.

If so, I would use the ten nests as my 10 samples.
2 operators
10 parts (in the ten nests)
Trial #1 on fixture #1
Trial #2 on fixture #2 (with parts in a different order in the nests)
Trial #3 on fixture #1 (with parts in yet another order in the nests)

When I did GRR 3x3x10 on small parts on the CMM stage...all 10 parts were on the stage at the same time...just not in a fixture since it wasn't needed.

Good call from Miner while I was typing...which specific coordinate system is used for the data offload...
When we did use a fixture, we set coordinates for the fixture, but then reestablished a new system for each part too...

 

[ncwalker]

Remember that a Gage R&R is a designed experiment. And it's OK to have a substitutional similarity assumption. What I mean to say is ... if one nest passes, they all "should" pass. Now, of course people will want you to demonstrate this, which is why I have "should" in parenthesis.

If you're hell bent on doing a 3x3x10 form instead of the more scalable ANOVA approach, I would use 10 parts, 3 nests, 3 times. BUT ... you're going to have to define your "worst" nests. Then, if those pass, they all should.

To identify the worst nests, you could measure a part twice on every nest and pick the three nests with the biggest delta, measurement 1 to 2. OR you could query the nests from some feature that made sense. Example: If you have male pins on the nests going into female locating holes in the part, then the nests with the smallest locating pins will have the most noise. (Or maybe those with the most tilt to the locating pins, I don't know your nest design).

In any case, I'd buy into a scheme like that.

 

[Miner]

The three operators are intended to be a random sample of the entire pool of potential operators. There is no reason why the nests cannot be a random sample (subset) of all the nests.

 

[ncwalker]

Good point, Miner.