Dear All,
Recently my customer required my company to perform all equipment/technician GRR first in order to ensure whole measurement system is stable as well as their products.In my opinion,once if Cpk is high enough we could tell gaging variation is far from affecting total variation and their products won't be misjudged. I am not very sure if I am right or not.
If I am right,how do I persuade my customer?It seems they don't accept at all but can't raise any theoretical viewpoints.
If I am wrong,please kindly advise me your correction.

While I understand your point of view, please keep in mind that process variation and gage variation are two different things. Cpk measures how stable the process is, while a gage R&R measures how reliable the gage is throwing in the variable of using different operators. IMO, It is quite possible to have a great Cpk, while at the same time having a lousy gage R & R, say perhaps because of a poor gage design. I think that what your customer is asking for is reasonable.

What does everyone else think?

Jonell

Hi Jonell
Appreciate a lot for your response.
I agree with most of your of viewpoints,except gage variation and process variation are two different things. As far as I was concerned,the latter contains the former.
Maybe I should correct my title : If time is limitted(actually always is) , could we use Cpk as a priority screen to decide which equipment should implement GRR first?

Hi Jonell
I agree with most of your of viewpoints,except gage variation and process variation are two different things. As far as I was concerned,the latter contains the former.
Maybe I should correct my title : If time is limitted(actually always is) , could we use Cpk as a priority screen to decide which equipment should implement GRR first?

I agree with Jonell regarding his comments that gage and process variation are different.

Keep in mind that the purpose of a Gage R&R is twofold, to determine the variability of the gage itself or the manner in which the gage is being used to take a specific measurement, and the variability due to the operators technique, or lack of technique in some cases.

I have always been of the opinion that unless the measurement method has been validated, than the data being received is suspect. Your first post you made the comment, "if Cpk is high enough we could tell gaging variation is far from affecting total variation and their products won't be misjudged." One possibility is that the Cpk is high because of measurement error. There are a couple of reasons that this belief is untrustworthy:

1) There isn't enough resolution in the measuring device so that measurement itself tends to group the data (i.e., if my range is 0.010 to to 0.030 and my device only reads two decimal places, I can end up with a lot of 0.020's when in reality I should be getting 0.014 or 0.016).

2) The measuring device isn't reading accurately over its entire scale. In this case I'm reading 0.020 on the device, but in reality I'm getting 0.012. Perhaps the gage has fallen out of calibration, perhaps an internal component is wearing out, etc.

3) The operator is consistent, they just aren't very accurate. Similar to item #2, its just that the operator is consistently producing readings which are 0.010 from what the actual measurement should be.

Gage R&R is also a method to determine and verify the ability of your operators, not just the method or device. Perhaps this is why your customer is requesting the studies. They want to be sure that the people producing their product are adequately trained.

What I would probably look at is which products/lines have the highest rates of defects. Because your customer is demanding this, it may also be of benefit to sit down with them and both of you come up with a priority list. Also be sure to identify which method (x-bar/r or ANOVA) to use. In this manner you satisfy their greatest concerns first, and they see you as being proactive.

That's my 2 cents worth......

Regards,

Wayne

Welcome to the Cove ysdennis and Wayne :bigwave:

Wayne: Nice response. I agree that without knowing the R&R of a gage (along with linearity and bias) there shouldn't be much faith put in the capability study results.

ysdennis: Do you have Minitab or some other type of software that will calculate your "number of distinct categories"? AIAG recommends greater than 5 categories. That number will, basically, tell you whether you have enough resolution.

Dear All,
Thanks for all your support and instructions.
I learned a lot form you all. Now I know how to discuss with my customers and give them more understandings.

My understanding is that you have to qualify your gages before you use them. This is to ensure that they are telling you what you need to know consistantly and accurately. If you do not know your gages are accurrate, repeatable and right with reproducible results how can you know that your process is really running where your gage says it is. it could be throwing skew or any number of other issues into your system and you would be blind to it. Calibration and GR&R are the ways you protect against these issues. You have to qualify the gages and operators before you use them to track capability.
What types of gages are you using and for what type of product?

Mark

Wayne, good explaination.

Thanks for your comments. Actually, I read what indicated in an article here in my country. I have ever used Minitab to implement GRR. But I didn’t really understand how poor resolution and device inaccuracy made good Cpk and poor GRR ,until Wayne’s detailed explanation. About gage types ,some are almost “automatic” except load/unload steps. This reminds me another question that “If measuring device is automatic, does it mean there’s no need to evaluate reproducibility?”

If there is any human intervention anywhere in the measurement process, then you need to evaluate repeatability and reproducability. Can the part be loaded incorrectly? Can the measured value to incorrectly interpreted or recorded? These are just two things that are part of your MSA.

And be sure to have the measurements done by people who will be normally doing it. One of the fallacies that I have seen over the years is people who insist on performing GR&R studies in the lab, using trained technicians. After obtaining an acceptable score, they are suprised when the production team can't match their results.