Gage Repeatability for an Automated Leak Tester

TBoneSupreme

Registered
Hi,

I am hoping someone here can help me with this...

I am looking to do a gage r to prove out a measurement system. It is an automated leak tester. We only have one. It is not operator dependent, so reproducibility is not needed.

The MSA manual appendix D lists a repeatability study, but it states that it is for preliminary purposes only, and cannot be used for gage acceptance.

I need to do a long term gage R that will be approved by our customer for PPAP.

Is there a standard for this? Would it be acceptable to repeat the 1 part 10 trials test on multiple parts?

Any help would be greatly appreciated.

Thank you
 

Howard Atkins

Forum Administrator
Leader
Admin
If you look at the bottom of the page you will see references to similar discussions.

When you say the test is automatic, is the part loaded automatically?

Even if it is the position in the fixture will never be 100% the same each time so there will be differences.
 
O

OmarEn

I have a very similar question on my leak tester.
the installation is not automated, actually is manually made by an operator.

Somebody told me that a Nested study would work well for this kind of applications.

do you guys have any idea of which method would can we use?

thanks in advance!
 

Johnnymo62

Haste Makes Waste
Ford has us use only crossed, not nested.

We did use ten samples sent through our automatic tester three times by three different Operators.
 
O

OmarEn

Actually, my customer is GM. and I do not think a crossed GR&R is useful for this application since you are considering the operator contribution. In this case, the operator does not contribute to the study as this is an automatic process.
 

Bev D

Heretical Statistician
Leader
Super Moderator
Actually, my customer is GM. and I do not think a crossed GR&R is useful for this application since you are considering the operator contribution. In this case, the operator does not contribute to the study as this is an automatic process.

Actually the nested study is not for operator contribution. A nested study is used when the test method is ‘destructive’ and a part cannot be physically tested twice.

If your device can be tested twice a crossed study and a nested study will yield the same result. OmarEn - your leak testing can be done twice - or more - on the same device.
 
O

Omar.

Hello Bev D,

Yes, the part can be tested twice if you would, but values are very different always.
When a part is bad (with leak), always is bad no matter how many times you tested, and the same situation when a part is good.
 

Bev D

Heretical Statistician
Leader
Super Moderator
Omar - as I stated in the other thread where you posted the same question, all parts will have different values each time you test them. What matters is how much variation in measurements there is in relationship to the specifications and/or the actual part variation. We must determine if the measurement system is acceptable for use or if we must improve it to make it useful for testing. This is why we require a measurement study. I also gave several examples of how this works with functional or non-destructive testing in the document I referenced. Did you read it?

You have asked twice what you should do. You have gotten an answer both times - the answer won’t change...I am confused as to what you really want. Why wouldn’t you just do the test as outlined? It is easy and simple. We can help you understand the results...


If you really believe that bad parts are always bad and good parts are always good then your study will demonstrate that.
 
O

OmarEn

Hello Bev D,

Thanks again. my actual question was.. what type of study should I use?
I was trying to use a nested one but you said in a last post that It was not the one I need.

Thanks again.
 

Bev D

Heretical Statistician
Leader
Super Moderator
Omar: You need a crossed study. Take 30 tubes and measure each one randomly twice. You do not need to measure them more than twice. Do NOT perform the second measurement immediately after the first measurement. You must have a second independent measurement of each part. If you read teh document on functional testing that I provided you will see how to analyze the results.
 
Top Bottom