Gage R&R on In-Process Automated Inspection Equipment

C

Curtis317

Looking for perspectives and insight on doing a gage r and r on in process automated inspection equipment. The parts are on moving steel pallet are presented to the machine which does the measurement. We are going to run the same part 3 times through the measurement process and analyze the differences.

Your thoughts?
 

harry

Trusted Information Resource
There are two related threads in the similar discussion threads box below (just scroll down the page). I believe you can find your answers there.
 
C

Curtis317

Ajit, Thanks for the links. However the measuring device measures the parts to a matching part. So if it measures a 3 then a 3 is assembled to it. If it measures a 2 a 2 is assembled to it. What we did was run 10 parts through 3 times. That showed repeatability. Which was not too at all. I was just looking for maybe a more statistical method.
 

Bev D

Heretical Statistician
Leader
Super Moderator
are you looking at categorical data? (pass/fail) or continuous data? teh approach for these is very different. if pass/fail, then measurign only 10 parts 3 time is not a good approach.

can you explain what you mean by "if it measures a 3 it is assembled to it"
 
C

Curtis317

The machine measures a part and provides a number related to the diameter. For the study the actual diameter was used. The diameter results in a ranking. 1 through 5. The part assembled to it is also ranked the same. So once the measurement is done a ranking is provided where the part assembled to it is of the same rank. In doing the study of 10 parts run 3 times we compared the 3 results. I know of no other way of doing this. That is why I am asking the question. However I think this is actually the only way.
 
S

strange1

I have a related question: I have an Atomic Absorption Spectrometer, the SensAA. It tests the content of water relative to Nickle and other metals. I need ot conduct an R&R on it. It requires that the machine be "standardized" with a liquid reference. Then samples are run through it presenting the results.

SO the operators have no influence, aside from the "standardizing" , which is automated.

Regarding an R&R, how many operators? How many samples?

One operator...30 samples?

Three operators, 10 samples?

TIA
 

Bev D

Heretical Statistician
Leader
Super Moderator
one operator 30 'samples' measured twice each will provide far better resutls than 3 operators, 10 samples, measured 3 times each.

even if operator was a contributor (I have seen this on 'hands-off' equipment like you describe when sample prep is involved? you are far better off if you use 30 samples measured twice across multipel operators...

you are trying to estimate the measurement error standard deviation and standard deviation estiamtes are far more imprecise with small sample sizes than an estimate of the mean. For one thing they are heavily influenced by even one extreme result.
 
S

strange1

Thanks...I was leaning that way. The samples area pain to draw, but I do need the accuracy...Sucks being them I guess.
 
Top Bottom