Hello everyone

Cp is a measure of process variation. So if a process is in control and you have a high value for Cp (i.e. low process sigma), then you don't have to perform an MSA? At least not if you are interested in R&R as percent of total tolerance.

Is this correct or am I missing something... :confused:

There's another thread containing the answer of your question. See

Which should be decided first : Cpk or all equipment/technician GRR (http://elsmar.com/Forums/showthread.php?t=10069)

BTW: How high is your Cp? And what's with your Cpk?

Barbara

If nothing else, you would want to do your MSA to validate the measurement results that you are getting.

No point in having a Cp or Cpk of 5+, when your R&R is >75%.

If nothing else, you would want to do your MSA to validate the measurement results that you are getting.

No point in having a Cp or Cpk of 5+, when your R&R is >75%.

I agree with you, but started thinking about this and did some calculations.
Let's say most of the process variation comes from the measurement system. In that case we can put

process sigma=sigma GRR.

Then Cp=(UTL-LTL)/6 sigma GRR

And GRR%TOL=6 sigma GRR/(UTL-LTL)

combining above gives GRR%TOL=1/Cp

I admit that you need to have a high Cp to say that the measurement system is good enough (Cp>10) for a 10% acceptance criteria, but this might be useful in some situations.
So with Cp>5 you have GRR%TOL<20%.

Please tell me if I made a mistake somewhere.

I agree with you, but started thinking about this and did some calculations.
Let's say most of the process variation comes from the measurement system. In that case we can put

process sigma=sigma GRR.

Then Cp=(UTL-LTL)/6 sigma GRR

And GRR%TOL=6 sigma GRR/(UTL-LTL)

combining above gives GRR%TOL=1/Cp

I admit that you need to have a high Cp to say that the measurement system is good enough (Cp>10) for a 10% acceptance criteria, but this might be useful in some situations.
So with Cp>5 you have GRR%TOL<20%.

Please tell me if I made a mistake somewhere.

since your measurment resolution should be 10% of the process limits can you explain how you can get process sigma = sigma grr? the remaining equations are moot point since your very first statement is improssible.

note: if the measuring resolution is greater than 10% of the process limits then you had an inadequate measuring system to start with.....no GRR need!!

Let's say most of the process variation comes from the measurement system.
And where does this certainty come from, if not MSA?

since your measurment resolution should be 10% of the process limits can you explain how you can get process sigma = sigma grr? the remaining equations are moot point since your very first statement is improssible.

note: if the measuring resolution is greater than 10% of the process limits then you had an inadequate measuring system to start with.....no GRR need!!

Hi qualeety

By the above I make the assumption that you didn't read my first post in this thread? I am talking about the use of a gage for inspection, and not for process improvement. Big difference between R&R as % of total tolerance and as % of total variation. If a process is as capable as I stated above (Cp>5), it wouldn't be my top priority to redesign the measurement system in order to make further process improvement, centering is enough. :)

And where does this certainty come from, if not MSA?

Hi JSW05

Well, by this I mean that the probability, that the variation caused by the measurement system is larger than total process variation, is small. And therefore the worst case scenario would be that they are of equal size.
I would not bet my life that this assumption is correct though, two sources of variation could eliminate each other. :tg:

It frustrates me when people confuse Cp with variation. Cp and Cpk are measures of how good your SPECIFICATIONS are versus the inherent variability of the process. You can make Cp any number you want by changing your specifications.

MSA and other variance components analysis determine how much of the variability is attributable to the different sources. Therefore, a MSA should be done regardless of Cp. You should also know how much of your process variability is attributed to the measurement system.

Theoretically, when your Cp and Cpk gets large enough, you tighten the specifications. When you Cp and Cpk are too low, you improve the variability.

It frustrates me when people confuse Cp with variation. Cp and Cpk are measures of how good your SPECIFICATIONS are versus the inherent variability of the process. You can make Cp any number you want by changing your specifications.

No one is confusing Cp with variation. Here is the full story. I work as a manufacturing engineer in the aerospace industry. The parts we manufacture are complex and tolerances are small. We are not in the position to change any specifications. So to me it's of little interest if specs are good or not, my job is to see to it that process variation is small enough i.e. Cp and Cpk are high. In other words low Cp, to me means that variation is to big and nothing else. Also the consequenses of failing to meet specs could result in accidents. This is why we want to make sure our inspection process is ok. Since we have a lot of different gages, reducing the need for MSA for processes with small variation would free resources to improve other processes, with low values for Cp and Cpk.

well I'll wade in a here a bit.

IF you have a controlled process and your process variation (Pp & Ppk) is small compared to to your tolerances, THEN the measurement error cannot be very large. Remember that the OBSERVED variation is the square root of the sum of the actual variation and the measurement error, hence a small observed variatoin (relative to the tolerances as stated above) = a small enough measurement error.

As for 2 sources of variation eliminating each other it does happen but not on every measurement: for the observed variation to be small compared to teh tolerances, the measruement error and the actual variation would have to be almost always opposite of each other for the OBSERVED variation to be small....

Now if it were me, there are several different scenarios for performing an MSA. If I had a stabel process and was fairly confident in my ability to detect any negative trending (SPC, etc.) and catastrohic shifts were unlikely AND the severity of the defect escaping was very low OR catastrohic shifts were catchable via some poke yoke device THEN I wouldn't perform an MSA. Otherwise if the severity was high enough and catastrophic shifts were possible I would do the MSA to determine if the gage would detect parts at the spec limit and then I would also most likely guardband.

Certainly I wouldn't put this process on the top of my problem solving priority list if there wre others with worse process performance.

As for teh comment concernign reducing tolerances for very good processes - *I* would think twice about it. If there were evidence that the specs were too loose (field failures that werein spec) I would change them - regardless of the process variation actually. If there were no evidence of 'too loose' specs and their was engineering logic or data supporting their validity then I woudl not change the limits. To me tolerances and processes performance are - and should be - separate things.

There are some who feel that if your performance gets better, you should automatically decrease your tolerance - but if decreasing the tolerance has no value add to the performance (and to the customer - why do it? If we automatically decreased the tolerances on well perfromign processes, we get in the endless loop of having to now improve our Performance to get the desired Ppk - which has tightened up over the years - then we tighten our tolerances then we improve our performance and so on as we fly in ever decreasing concentric circles until we fly up our own...self.

No one is confusing Cp with variation.

my job is to see to it that process variation is small enough i.e. Cp and Cpk are high.

The statements above seem to be contradictory.

Since we have a lot of different gages, reducing the need for MSA for processes with small variation would free resources to improve other processes, with low values for Cp and Cpk.
In principle there's nothing wrong with this idea, but the fact remains that without empirical data, you're guessing about the contribution of gage error. The best way to a reasonable shortcut is use of surrogate data. If you do MSA on a particular feature with a particular type of gage, the results may be used to satisfy MSA requirements for other parts with analogous features measured with the same type of gage.

No one is confusing Cp with variation. Here is the full story. I work as a manufacturing engineer in the aerospace industry. The parts we manufacture are complex and tolerances are small. We are not in the position to change any specifications. So to me it's of little interest if specs are good or not, my job is to see to it that process variation is small enough i.e. Cp and Cpk are high. In other words low Cp, to me means that variation is to big and nothing else.

Dear JSW05

You have to excuse me, I might be stupid but I don't see any contradictions in the above statement. :) If process variation is small enough Cp is high, or do you have another opinion. This does not mean Cp=variation...

About the second part of your post I do agree with you that you can't be sure of the contribution of of gage error. So I guess that answers my question.
Thank you for discussing this subject with me, I appreciate it.

Hi Bev D and thank you for your thaughts on this subject.

Dear JSW05

You have to excuse me, I might be stupid but I don't see any contradictions in the above statement. :) If process variation is small enough Cp is high, or do you have another opinion. This does not mean Cp=variation...

About the second part of your post I do agree with you that you can't be sure of the contribution of of gage error. So I guess that answers my question.
Thank you for discussing this subject with me, I appreciate it.

Hi Bev D and thank you for your thaughts on this subject.

Technically, Cp is not equal to variation.......but Cp is inversely proportional to variation for a given tolerance..........so when one changes the other changes as well and visa versa....

so, technically JSW05 is correct, stating that Cp is not equal to variation!!!

Technically, Cp is not equal to variation.......but Cp is inversely proportional to variation for a given tolerance..........so when one changes the other changes as well and visa versa....

so, technically JSW05 is correct, stating that Cp is not equal to variation!!!
The point was originally made by Statistical Steven. I understand Qaware's point though, in saying that if Cp is high, then it usually follows that variation is relatively low. You can't talk about Cp or Cpk without reference to spec limits, so variation in this regard is necessarily a relative concept.

Technically, Cp is not equal to variation.......but Cp is inversely proportional to variation for a given tolerance..........so when one changes the other changes as well and visa versa....

so, technically JSW05 is correct, stating that Cp is not equal to variation!!!


:frust: And that's just exactly what I said to, you quoted it yourself!!!

:frust: And that's just exactly what I said to, you quoted it yourself!!!

oops... :o ..you are right !!!! :agree1:

So conclusion from this discussion would be that if Cp is high enough you probably have a low value for GRR%Tol. This is not a completely solid statement though. You still have not proved that there is no interaction between two or more sources of variation and the measurement system. However in a situation where you have many processes to analyze, you can use Cp as one of your priority factors for where to start your work.

(No hard feelings qualeety ;) )