Dear all,

Recently we performed an R&R study and the problem we encounter is the following:

For all three variables the number of distinctive categories is not enough. For five different specifications the distictive categorie number was three times higher than 5 and twelve timer (much) lower.

Now, I know (at least I think so) what this means, however it is impossible for us to perform more accurate measurements (add a significant decimal, for example) because there is no equipment available.

When I look at the measuring results and compare the results of the different operators I can simply draw the conclusion that every operator is capable of measuring. All results are very much alike.

The big question for me is though: will an auditor take no for an answer? Strictly by the book our equipment does not meet the requirements.

How do you people think about this?

Thanks in advance,

Patric Wessels
The Netherlands

As you say, one aspect is discrimination of the device.

However, we found that the process spread of the parts measured also has a big impact on ndc.

Did the parts you measure cover the entire expected process spread (6 sigma?)

If they are all very close in size, you will get few distinct categories and poor R&R results.

If the process spread is not known, you could use the tolerance as a guide.

We made the mistake of taking parts all produced back to back, with almost no variation between them. Results were poor.

We ended up collecting GR&R study parts over several weeks and several diferent set ups to get closer to the process spread. This fixed our ndc probelm and gave acceptable R&R results.

Some people also "make" a set of study parts that span the tolerance range.

Do you have the AIAG MSA book? I recommend it. This is well covered in the book.

Page 46 says ...for a stable, highly capable process using a stable "best in class" gage at the practical limit of technology ...effective resolution may be inadequate and improvement of the gage becomes impractical....

Your auditor would have to accept this if it applies in your case.

Good luck

Yeah,

I don't know where the thread is now, but we were having a similar problem and Caster responded to it as well.

Its not very funny, but if you meet ideal conditions for quality control of having no variation at all, it can ruin your GRR since people and gauges are not so perfect. I'm gonna posts the consensus of what I developed concerning MSA GRRs below to see if it helps you.

Suggestions for MSA GR&R’s:
(Meeting the UCL, %GRR, and ndc requirements)

1. Use the GR&R report that calculates Total Variation (TV), not Tolerance (Tol).
There are two GR&R report formats included with the software package for PPAP’s. The one that calculates Tol or Tolerance is only to be used when the process variation is known. Since the process variation is not known, the report that calculates TV or Total Variation should be used.

2. Use the 10:1 Rule of Thumb when measuring a parameter.
If the parameter to be measured has a tolerance of 0.001, use a gage which is accurate to (not just able to measure to) 0.0001. This will ensure that enough discrimination exist so that measured values will not fall out of the UCL value, making the measurements “Out of Control”. If any values do fall outside of the UCL value, a new value entered in will not greatly affect the recalculated UCL value.

3. Try to measure parts that will have some significant range between part parameters.
If all the parts measured had similar values across the page, the calculated “Rp” or Part Average Range would be low, which in turn can cause a low “ndc” value and a high “%GRR” value. If the variation between parts is low and the ndc or %GRR is not being met, try to measure parts that have more variation, still falling within tolerance of course.

4. Make sure that measurements taken on the same part between different appraisers do not have significant differences (Reproducibility).
If Appraiser A and Appraiser B or C show significant differences in measurement of the same part with respect to the accuracy of the gage, have the appraisers re-measure the same part.

5. Make sure that measurements taken on the same part by the same appraiser do not have significant differences (Repeatability).
If Appraiser A, B, or C re-measures the same part 3 times and shows significant differences in his/her measurements, have them re-measure the part.

6. When form and measurements are completed, make sure cells are formatted to show the degree of discrimination.
If a parameter is measured to the accuracy of 0.0001, it should show a value with 4 decimal places. If a parameter is measured to the accuracy of 0.001, it should show a value with 3 decimal places.

Thanks for your reply Geese.

Referring to the suggestions you give please let me elaborate on those:

1)I did use the TV calculation, however because I had more than 3 operators that performed the measurements (everybody at our company in manufacturing has to be able to measure) I used a quality Engineering Handbook to make a spredsheet with al the calculations and factors. Using an example from this book I checked the spreadsheet.

2)I can be short about this; our gage is able to measure with an accuracy of 0.1° (angle) but the variance in our production is very low, about 0.2-0.3°.

3)The used parts have a range of about 6 sigma (because our product is a web, like paper production, this is the spread across the web. This spread is the maximum in our production)

4)This is OK

5)This is OK

6)This is OK

Summarizing I think our problem is that there is no suitable equipment at this moment that can measure with the required accuracy and make enough discrimination between our parts.

Please let me know how you people think about this.

regards, patric

The only other question I have, is what is the application for your MSA? Are you submitting it for a PPAP or some other customer submission format? There might be waivers they can make if you are upfront with them on the issue of your gauge accuracy. This can be tougher to do if there are distributors between you and your end customer requesting the MSA. Then you have to pass the idea down the chain and back up again. I suggest an explanation in an email that your distributor can forward to their end customer if that is the case. With PPAP submissions, the customer has the final say on everything; the costs of doing business but sometimes a helping hand.

I know it seems silly that if you're within tolerance, your MSA can be bad, but statistics and all bell curves must be satisfied :) . The only other thing I can suggest is browse the market for a more accurate gauge that is accurate to 1/10th the tolerance called for on your drawings or specs.

One other thing. Does this web you mentioned have a performance role or is it purely for appearance. You might be able to classify it as such if it serves only for looks and not function. For instance, on a PPAP (sumbission format for automotive industries), an Appearance Approval Report form is used on items that serve for appearance only and would probably not need to be measured. However, if your Control Plan calls for this measurement and specifies the gauge, you must have an MSA for it.

Not trying to raise more questions. Just trying to get an idea of what your MSA is used for.

...try to measure parts that have more variation, still falling within tolerance of course.
Why do you say they must be in tolerance?

I say they must be in tolerance because a customer who sees that and doesn't know what the purpose of the MSA is will question it. Its kindof a common sense issue I guess. Also, on the suggested format, the part tolerance of the dimension being measured is filled in at the top of the form we use for GRR.

Its a safeguard to not get a MSA submission sent back to you. You could do the MSA and show that measurement which is out of tolerance and all of your numbers will be good, but your distributor or end customer might send it back and tell you to resubmit. Its an attempt to cut down on bouncing submissions around, not an attempt to undermine the integrity of the MSA.

If you have a lot of parts that are out of tolerance and not meeting spec, thats a quality department issue and should be caught and corrected. You don't want to be submitting bad product, but at the same time you dont want to be submitting documentation you know will not pass. I'd rather correct the problem and then submit information on good product.

I've mentioned the (in tolerance) line in some of my other posts. If its horribly wrong and can potentially burn someone, let me know and I'll stop putting it in and try to find the other posts I've made with it.

At this moment it is like this:

We started about 5 years ago as a small company (called Dejima Optical Films, 25 employees) and I am the one "chosen" to do the quality. We only deliver our product (a polymer coated substrate for the LCD industry named Twistar™) to one company that is owned by the people who made use "exist" (the finance). This company laminates our product with a polarizer and delivers it to so called LCD manufacturers.

The MSA is only for internal use at the moment. I just want to know where we stand regarding measurement of our product.

This is in very short our situation. If it isn't clear to you please let me know and I will be happy to tell more.

Best regards,

patric

I say they must be in tolerance because a customer who sees that and doesn't know what the purpose of the MSA is will question it. Its kindof a common sense issue I guess. Also, on the suggested format, the part tolerance of the dimension being measured is filled in at the top of the form we use for GRR.

Its a safeguard to not get a MSA submission sent back to you. You could do the MSA and show that measurement which is out of tolerance and all of your numbers will be good, but your distributor or end customer might send it back and tell you to resubmit. Its an attempt to cut down on bouncing submissions around, not an attempt to undermine the integrity of the MSA.

If you have a lot of parts that are out of tolerance and not meeting spec, thats a quality department issue and should be caught and corrected. You don't want to be submitting bad product, but at the same time you dont want to be submitting documentation you know will not pass. I'd rather correct the problem and then submit information on good product.

I've mentioned the (in tolerance) line in some of my other posts. If its horribly wrong and can potentially burn someone, let me know and I'll stop putting it in and try to find the other posts I've made with it.
I have used out of specification parts in my variable Gage R&Rs quite often. My feeling is that it's just as (if not more) important to be confident in "out of spec." measurements as I am with the "in spec." measurements. I've never had a customer question the "out of spec." measurements.

I'm no "expert" regarding MSA so unless someone more knowledgeable chimes in, I wouldn't go deleting that from you other posts.

Patric
You are using MSA internally but, apparently, don't have a requirement to submit studies to your customer. I guess I'm not clear on your question of "where we stand...". Are your MSA studies agreeable with the guidelines put forth in the MSA booklet?

I'm also curious Patrick if you have outside audits on what you do internally. Its good to do an MSA if you are doing it for your own benefit, but if you know that you don't need to (not submitting MSA's to customers), meeting the ndc numbers should not be a big issue. It should be understood why you don't meet it though so youll know in the future what is required.

In reply to Bill, if we were supplying one customer or just a few customers, I would more than agree with you, but we deal with distributors and those distributors sometimes have distributors which means the probability of everyone understanding the MSA's in the chain from supplier to end customer is somewhat remote. So we've made the effort to make sure small things will not send submissions back up to us. I agree that a number out of tolerance is valuable on an MSA, its just the practicality of human error in the submission process that leads to problems and we dont' always have direct access to the customer to give and get information.

In other words, there are people out there that have a checklist they checkoff on submissions even though they don't know the significance of the things they check off. If they see what they think is a discrepancy, it can slow down the process considerably.

We do not really have to submit MSA to anyone, but because of the fact that the past two years our operator group was extended from about 6 people to 14 and we started in a five shift production the need for measurement (inspection) by all operators was born. I then just wanted to get confirmed that there are no significant differences between the operators concerning the measurements.

And then, because they all had to perform a lot of measurements, I grabbed the oppertunity to do the MSA. Strictly by the book we can not follow the rules for the MSA. For example, we can not take random samples because our product is on a roll. We only have sample from the start and end of a roll (i.e. production lot). There are lots of (complicating) details that I shall not mention here.

I think, if I understand all reactions to my question, that we do not have a problem (yet). Perhaps things change when we want to achieve QS9000 (or ISO) certification, but I am not really familiar with the subject yet. I have to learn a lot, my "quality carreer" is just beginning.

So, thanks for everybody who replied to my question.

patric

In other words, there are people out there that have a checklist they checkoff on submissions even though they don't know the significance of the things they check off. If they see what they think is a discrepancy, it can slow down the process considerably.
Gotcha!!