Gage Reproducibility & Repeatability for POKA-YOKE (pass/fail) gage?

[ario80]

Hello,

This is my first post. My customer asked for R&R study for all measuring devices utilized to verify his product. On of them is a poka-yoke gage. It is meant to verify if a detail is too high (low). As avery POKA-YOKE the result is OK/NOK. Is it possible to make a study for this? How should it be done?

Thank you for your help in advance!

 

[Bev D]

Re: MSA (R&R) for POKA-YOKE ?

Yes. you treat it like any other pass/fail inspection or gage. if you search this forum for "kappa" or attribute R&R you will find many helpful threads.

if you google "When Quality is a matter of taste" by David Futrell you will find a free copy of an articel that details the math.

I also posted a spreadsheet that will help: http://elsmar.com/Forums/showthread.php?t=34889

basically you must identify some units that are passing and some that are failing by using a different method such as a continuous data gage that has good repeatability. It's best to start usign parts that are marginal to the spec limit. USe the 'truth' as the 'first' reading, then the second reading is from your poke-yoke gage. This establishes the zone - if any - where the gage and the user has trouble. You can use this approach to qualify new operators and or replicate gages...

If there is a zone where the gage has trouble making the correct call you may consider selecting a second sample set that has teh same distribution of parts as your process and then performing 2 repeated measurements of each unit to establish the actual 'escape' rate and false rejection rate given your true process. OR you can guardband the gage to over reject in order to not allow any real failures thru...

 

[ario80]

Hello Bev D,

Thank you for your detailed reply. My poka-yoke checks one dimension it is 110 +/- 0.5 mm so one gage is 110.5 and too high components will not pass. In that case should I pick let's say 30 pcs, measure them. Include in that 30 pcs. 2-3 NOK parts and check if every operator finds these pieces using poka-yoke?

After that how to communicate this to the customer. Just by saying that test was done like above and all NOK parts were found ?

Cheers!

 

[bobdoering]

Re: MSA (R&R) for POKA-YOKE ?

It's best to start using parts that are marginal to the spec limit.

This is true - and you should have these parts for your "rabbit" parts anyway for routine verification that the poka-yoke is still working. The worst thing that can happen is that you ignore it because it is checking 100% - only to find it did wear (and it will) and starts to let bad parts slip through.

 

[JaxQC]

When you write up you test results just call them PASS and FAIL. All operators should agree and catch all the FAIL parts. It’s not incorrect to see the word fail on an attribute gage study (provided it’s the correct ones failing). Short sample version below.

Also make sure to put the Red part/Rabbit part into the gage control system. "Calibration" would include 1) can mfg find it when it's time and 2) whatever makes it bad is still in place 3) if it does come up missing you have a record of what to make a replacement part to be the same as the original.


oper 123Part 1PASSPASSPASS2PASSPASSPASS3FAILFAILFAIL4PASSPASSPASS5PASSPASSPASS

 

[JaxQC]

Yea that cut & paste didn't work so well but hope you get the idea.

 

[coffee chang]

Hi all,

I have a jig to measure inner dimension of a part which I will perform attribute GR&R. Let say if I don't have Not OK part at hand.
Since the dimension is small, can i use a caliper instead? Set it to value which out of specification and try to slot in the jig.

Thanks...

 

[Citizen Kane]

But, if you have a poka-yoke, you shouldn't be allowed to have NOK as a position. Either works on the good way, or it doesn't - no monitoring needed.

 

[bobdoering]

I have a jig to measure inner dimension of a part which I will perform attribute GR&R. Let say if I don't have Not OK part at hand.
Since the dimension is small, can i use a caliper instead? Set it to value which out of specification and try to slot in the jig.

This approach sounds risky. One of the functions of a GR&R is to assure that any part configuration issues, such as cocking, can be accounted for. A caliper may not adequately duplicate the part feature.

I suggest fabricating a not OK part, or at least simulating that particular feature and the associated locations of where the part will rest if it is too expensive to duplicate the whole part.

 

[bobdoering]

But, if you have a poka-yoke, you shouldn't be allowed to have NOK as a position. Either works on the good way, or it doesn't - no monitoring needed.

I may be misreading your statement, but the device is to ensure all parts are OK, and it must stop all bad parts, which makes it a go gage. You MUST ensure that it will stop bad parts - every time - and the correct test is to present bad parts to it repeatedly and make sure 100% do not pass. Then, you need to make sure 100% of the good parts do pass. That is the point of using an attribute gage R&R on a poka-yoke.

In addition the 'rabbit" parts are used on an ongoing basis to ensure wear over time, etc. still does not let a bad part pass.