GR&R for a Test Fixture

[lornzky]

Hi,
In our test fixture qualification procedure, it was mentioned that Pp/Ppk is to be used as a measurement to qualify a test fixture.
But I disagree to this approach, because to me GR&R is more appropriate due to the following.

- Test fixture/equipment only tells us if the board passes or fails based from the pre-defined test parameters and limits provided by our customer
- Once the customer defines those test limits, they won?t allow these limits to be changed just to have a good ppk/pp result because what they care about is just for the assembly to pass for its intended application.
- The test fixture/equipment does not have any control on the input(PCB assembly) to get a good output.

Any thoughts is much appreciated... Thanks

 

[Stijloor]

A Quick Bump!

Can someone help?

Thank you very much!!

 

[Watermelonsg]

totally agree with you.
When it comes to fixture qualification for a measurement tool, GR&R is more relevant as it will tell us the variance from the measurement system (gage and appraisers) in relation to the total variation. Based on the result, we will know whether or not it suits the requirement as a tool to measure the output from a certain process station.

When GR&R passed the requirement, lets say the GR&R STD% is less than 10, the next job for the team will be to check if the process is capable and stable. Now Cpk/Ppk comes into play. In the event Cpk fails the min requirement, then the focus is on the process rather than the measurement.

It appears that the way you mentioned skipped GR&R.

 

[Semoi]

The key argument why Pp and Ppk are not relevant is that they describe the process and not only the gage. Thus, in order to qualify the gage we have to "remove" the process from the gage and consider only the later. However, note that there are two different gage RR studies: One focuses on the product specification and thus answers the question whether or not the product is within specification. The second focuses on the process optimization and answers the question whether or not we are able to use the measurements to optimise the variation of the process. The later is usually more demanding.

PS: Could anybody tell me how I am able to insert equations. Is this environment able to interpret MathJax? Usually it is much simpler to describe things using formulas.

 

[Watermelonsg]

The key argument why Pp and Ppk are not relevant is that they describe the process and not only the gage. Thus, in order to qualify the gage we have to "remove" the process from the gage and consider only the later. However, note that there are two different gage RR studies: One focuses on the product specification and thus answers the question whether or not the product is within specification. The second focuses on the process optimization and answers the question whether or not we are able to use the measurements to optimise the variation of the process. The later is usually more demanding.

PS: Could anybody tell me how I am able to insert equations. Is this environment able to interpret MathJax? Usually it is much simpler to describe things using formulas.

That's good that you further detailed two types of Gage R&R people can choose to use depending on their intention.

I would suggest to use % Tolerance 1st to qualify the gage system for the process to produce parts meeting product specification --- capable. (I would say most of the time it would pass as the process variation should be less than the tolerance range. So personally I am opposed to use %Tor. Probably this is the reason you mentioned the later is usually more demanding?)

In order to control the process making it stable, we need to look at the % Gage std. This will tell us if the Gage system is sensitive enough to signal process variation rather the noise from the gage system. And we should do % Gage std regularly.

 

[Semoi]

I'm neither a statistician nor a quality person, but I am certainly closer to the former. Since I have seen to many different definitions for each terminology I do not understand what you mean by

I would suggest to use % Tolerance

In this forum I have seen that many use Minitab. Thus, my best guess is that

where L is one limit, and StudyVariationis the amount of variation caused by the measurement system and by the differences between parts (see Minitab homepage). StudyVariationis is calculated as 6 * the standard deviations of each source of variation (see Minitab homepage). Hence, this is the inverse Cp value, right? (omitting the factor 100 for the conversion to percentage, and assuming that xbar is centred between the limits)

However, in your first post you agreed that a Pp and Ppk value does not make sense to qualify a gage. Hence, I reckon that you are using a different program and that the definitions differ from Mintab.

PS: As I do not have enough reputation to add the links, I had to remove them.

 

[Watermelonsg]

I'm neither a statistician nor a quality person, but I am certainly closer to the former. Since I have seen to many different definitions for each terminology I do not understand what you mean by

In this forum I have seen that many use Minitab. Thus, my best guess is that
View attachment 28569
where L is one limit, and StudyVariationis the amount of variation caused by the measurement system and by the differences between parts (see Minitab homepage). StudyVariationis is calculated as 6 * the standard deviations of each source of variation (see Minitab homepage). Hence, this is the inverse Cp value, right? (omitting the factor 100 for the conversion to percentage, and assuming that xbar is centred between the limits)

However, in your first post you agreed that a Pp and Ppk value does not make sense to qualify a gage. Hence, I reckon that you are using a different program and that the definitions differ from Mintab.

PS: As I do not have enough reputation to add the links, I had to remove them.

Hi Semoi, good morning to you, i assume you are from Europe or European time zone.

You are right, %Tor is what the formula you pasted representing. This is for one sided specifications. In the example I assume L refers to Lower specification.

When quick glance at the formula may lead ppl to the quick impression that it is inverse Cp. you might change your mind if you take a 2nd look, this is because the studyvariation refers to gage system which does not include the part to part variation. Agree?

So %Tor is not related to pp/ppk or cp/cpk. And we will compute them after the process and gage system have been validated, and probably doing regular MSA calculating % Gage R&R instead of %Tor? Thoughts?

 

[Johnny Quality]

Depending on the "test fixture" it might make more sense to use an attribute study, for example if the output is either a pass or fail with no continuous data.

In regards to the process indexes I have this cutout in my office to remind myself the relations between them. It helps when doing SPC studies and deciding where to focus your efforts.



This picture is from an article by Donald J.Wheeler available here.

 

[Semoi]

Hi Watermelonsg,

yes, I am in Europe. Thus, I am several hours ahead of you

Regarding your comments, I am sorry, but I have to disagree.

2nd look, this is because the studyvariation refers to gage system which does not include the part to part variation. Agree?

As I said, when you start reading about Six Sigma you will find that the authors use different definitions. Thus, it is important to define which definition you are using. As I wrote the StudyVariationis in Minitab takes both contributions into account. The full text is as follows:

What is study variation?
In gage R&R, study variation (Study Var) is the amount of variation caused by the measurement system and by the differences between parts. It is calculated as 6 * the standard deviations of each source of variation.

You use many abbreviations and it is hard to understand them unless you tell us which reference you are using. The manual of Minitab is available online. Are you using Minitab? If the reference is not available online (for free), it would make sense to add the formulas for each of your "abbreviation", so that they become unambitious. E.g. I don't know what "% Tor" means to you. The same is true for "% Gage R&R", where some authors use the components of the standard deviation and others use the variance components. Of course you might say that this really does not matter, however, it is tiresome to discuss things, which are not well-defined. It just adds a layer of uncertainty, which is unnecessary.

Looking forward to your reply.

 

[Watermelonsg]

Hi Watermelonsg,

yes, I am in Europe. Thus, I am several hours ahead of you

Regarding your comments, I am sorry, but I have to disagree.

As I said, when you start reading about Six Sigma you will find that the authors use different definitions. Thus, it is important to define which definition you are using. As I wrote the StudyVariationis in Minitab takes both contributions into account. The full text is as follows:

Hey Semoi, good to know you here.

After I re-read my statement "studyvariation refers to gage system which does not include the part to part variation." and help page of Minitab. I found it not right technically. I should have said each source of variation can be expressed by StudyVariation(6*SD) where SD is StdDev. And in relation to the process tolerance, the weight(technically it should be %Tolerance (SV/Toler)) of the Gage R&R can be calculated as

studyvariation(of Gage R&R)/Tolerance*100%.

This is what I meant by %Tolerance(and by default when I call it this name, i meant the ratio for Gage R&R to the tolerance, I should have been more precise), and its referred to when deciding if we should release the gage.

I have included a table to results using Minitab below:


If we calculate the %Tolerance for Total Variation, using the same formula, it is


studyvariation(of Total variation)/Tolerance*100%.

it is the reverse cp as what you commented. Thats why we have different interpretation I guess.

Speaking of % Gage R&R or % Gage std, I meant %Study Var (%SV). I guess I need to change the term from now on to avoid confusion.

As an expansion of the discussion, did you find any guideline for %Tolerance for gage study of destructive test? In AIAG MSA 4th edition, pg78, it says <10 - acceptable, 10-30 maybe acceptable. Does it imply it applies to destructive test as well?

I feel the requirement should be more stringent for destructive test as it does not take reproducibility into account(you can find the above snapshot in which reproducibility is 0). What is your opinion?