Measurement System Analysis - Using Attribute (pass/fail) Data

[@user-rz6jt4km6c]

Thank you sir! has usual and common to all your videos, great detail and simplification.

[@Thriumphes]

Simple and effective, keep rolling mate.

[@paulallen5321]

AJIN - thanks for the encouraging comment, I like to keep it simple. Check out my books on LULU.COM. Drink tea and read the paper is a good all round quality book...

[@keithboone4800]

Another treasure.

[@paulallen5321]

Cheers..

[@jamesriepe]

Sorry I do agree with this method. It is based on speculation and you also have the human factors in your process. What happens if you deciding expert dies? Do you stop the production.

[@paulallen5321]

Hi J - No but if you pass/fail inspection criteria, someone has to decide and set the rules....no noe has to die, someone could just leave. It's a common problem, how would expect cosmetic criteria to be asessed?

[@jamesriepe]

@@paulallen5321 Operator 1 may be 90% effektiv today but only 50% tomorrow. You cannot calculate the human factor . You can’t put a number on it. I have been working with optical releases in the automobile industry for 25 years now.

[@paulallen5321]

What you're saying J is the result we get is an estimate, that's why we design the test carefully to get the best estimate. To be fair, if what you're talking about is the ability of an inspector to look at a full body shell of a car and see every blemish then yes that is a much more difficult test to run....but it doesn't mean you can't do that test...and learn soemthing

[@jamesriepe]

@@paulallen5321 No, what I am saying is the a human being cannot be considered as a measurement system. The chance of error is too high and that chance of error changes day for day.

[@paulallen5321]

J - I'm afraid the human element is in lots of measurement systems, as it is in lots of manufacturing systems. If you don't know how to reduce variability from people in all elements of our proecesses then we are in trouble...it would be great if we could automate everything but that's just not possible...

[@emilaaganovic9672]

Thank you so much for the video! Is there a minimum sample size requirement defined for visual attribute data?

[@paulallen5321]

Hi Emila - i usually use 20. 10 pass 10 fails...

[@hasangull4550]

Thank you for the video, Paul! I'm seeking guidance on selecting the appropriate MSA for an in-line vision system used for inspection. This system generates pass/fail outcomes without any operator interaction. While I understand how to assess repeatability by conducting multiple trials on identical parts (good and bad) passing through the vision system, I'm uncertain about addressing reproducibility. Could you demonstrate a scenario like this using Minitab? Your advice would be greatly appreciated. Thank you!

[@paulallen5321]

Hi Hasan - I made a video to answer your question... kzbin.info/www/bejne/bXibmIt7hcx6fLc

[@isaccobettin8321]

Ciao Paul, thanks for the video. Question for you: How do you evaluate repeatability if you run only one inspection of each sample ?

[@paulallen5321]

Isacco are you saying that your test is destructive?

[@isaccobettin8321]

@@paulallen5321 No, I'm saying that if I ask to my operators to inspect the 20 pieces only once, I don't see the repeatability. I would prefer to perform multiple inspections (i.e. 20 pieces evaluated 2 or 3 times by each operator) to understand if it is repeatable or not. What do you think?

[@paulallen5321]

Ok see your point - the reason is that you have a standard that they have to match. That is the effectively doing a repeatablity check. There is no standard in the R&R study only consistency.But let's put it this way, there is no harm in doing the extra repetition...you can certainly set the template up to do that in SPC XL...

[@bobmorkitchen854]

Hello, could you please specify where its decribed that its recomended to select 10 samples accepted and the other 10 rejected.Thanks

[@paulallen5321]

Doesn't have to be 10 and 10, but it does have to be the same. 15/15 or 9/9 etc....

[@aru-kun_00]

Hey Sir, How do you calculate Process Capability for C charts? (Attribute data)

[@paulallen5321]

Ardan - good to hear from you. The answer is that you don't. C charts count flaws on a product. If you ask a customer what is the accpetable no of flaws on their new car?? They'll say zero....