Attribute Gage R&R study - Automatic machine that tests for a particular feature

D

DJN

need to conduct an attribute R & & study, but I am unsure if I have the requirements correct. We have an automatic machine that tests for a particular feature. The only operator intervention is to place the assembled component on the test rig. The test rig has been foolproofed so that the operator cannot place the assembly incorrectly on the jig, so I guess no reproducibility(AV) issues. I have suggested that an attribute study is appropriate, as we do not need to know the actual dimension, only whether it has passed or failed the test. My thinking is to have 50 assemblies, with 20 good, 20 bad and 10 borderline. I would then conduct 4 trials per assembly to simulate two operators and two trials. Having said that, as the machine produces continuous data, would it be better to run a normal R & R, taking 10 parts and running 4 trials. The only problem I see is that I would have no different operators as the machine is effectively the only operator? Is my thinking sound in these circumstances? Any guidance appreciated.
 

Howard Atkins

Forum Administrator
Leader
Admin
Despite the fact that it is an automatic machine it can vary slightly in the same manner that a hand can change, equally the part can change so the interactions are different.
I would do the study as if there were 2 operators
 
D

DJN

Thanks Howard. I think that your suggestion is a good one. I will treat the study like a normal R & R with two operators. Many thanks for your input.
 
C

CanuckTester

DJN said:
Thanks Howard. I think that your suggestion is a good one. I will treat the study like a normal R & R with two operators. Many thanks for your input.
We have a similar question at our company - ie: an automatic machine that tests several features on one part. There are multiple test stations, and sensors. Some of these features are variable, and some of the features are attribute data. My feeling is that a variable Gauge R&R should be completed on each variable sensor, and attribute studies should be performed on the attribute gauges in the system, instead of an attribute study on the whole machine. I do not think that an attribute study provides very much useful data for troubleshooting or improving the machine's true acceptance rate.

Any thoughts/comments?
 
Top Bottom