The appropriate Attribute MSA for Automatic Visual Inspection Machine

[jefnik3201028]

Hello guys,

Just want to get comments from the cove experts about the appropriate MSA for the automatic visual inspection machine we have. The machine basically compares images based on the set accept-reject criteria so the operator's job is to feed the parts to be inspected and the machine will do the inspection by itself.

Now our calibration personnel is planning to apply an MSA study on these machines following the procedure defined in MSA (4th ed) section C, using 50 samples and 3 appraisers (operators).

Looking closely on the measurement system for this machine, the operators do not really have any effect on the outcome so the result of the study seems not to serve the purpose.

I was thinking, should the MSA be done between Inspection machines using the same samples ? please advise if this is the right approach.

Likewise, is the cross-tab/ kappa analysis limited to 3 appraisers or it can be applied to more than 3 ?

Regards,
Freddie

[NumberCruncher]

In Reply to Parent Post by jefnik3201028

Hello guys,

Just want to get comments from the cove experts about the appropriate MSA for the automatic visual inspection machine we have. The machine basically compares images based on the set accept-reject criteria so the operator's job is to feed the parts to be inspected and the machine will do the inspection by itself.

Now our calibration personnel is planning to apply an MSA study on these machines following the procedure defined in MSA (4th ed) section C, using 50 samples and 3 appraisers (operators).

Looking closely on the measurement system for this machine, the operators do not really have any effect on the outcome so the result of the study seems not to serve the purpose.

I was thinking, should the MSA be done between Inspection machines using the same samples ? please advise if this is the right approach.

Likewise, is the cross-tab/ kappa analysis limited to 3 appraisers or it can be applied to more than 3 ?

Regards,
Freddie

Hi Freddie

If I were to do this study, I would make it fully crossed. Different machines, different operators.

The big problem is doing the analysis afterwards. The MSA manual only gives the example of a 3x50 study. I don't have any spreadsheets for this type of study, at least not spreadsheets that I would be confident about posting here.

If you have Minitab, have a look at the GRR options, it may well be able to handle the more complex study.

The cross tab/kappa approach can be applied to any number of people. Once you have worked out how to replicate the tables in the MSA manual, there is no great intellectual leap to be made in extending this approach to any number of people. There is a lot of work however.

NC

[Bev D]

if the operator truly has no effect on the results (they have no influnce on positioning for example) then you wouldn't have to include operators. You definitely should include the different visual inspection machines.

[NumberCruncher]

In Reply to Parent Post by Bev D

if the operator truly has no effect on the results (they have no influnce on positioning for example) then you wouldn't have to include operators. You definitely should include the different visual inspection machines.

Hi Bev

I agree with you except...

I like to have evidence that there is no operator effect. If nothing else, it keeps the auditors quiet.

And who knows, you might just get a surprise one day.

NC