Minumim number of parts to conduct the Process Capability Study

[Eduardo - 2003]

Is there a minumim of part to conduct the Process capability studies? I did 50 pieces, is this acceptables?

 

[Al Dyer]

Thanks,

The first thing I would ask is who is the customer?

 

[Eduardo - 2003]

The customer is Sharader Electronic....

 

[Al Dyer]

For what it's worth, my background is automotive. At the very least I would follow their guidelines and on a personal level would accept nothing less than 20 samples of 5 each. And that is just the start.

Capability needs to monitored over a time period, not just a sample size!

 

[Howard Atkins]

Originally posted by Al Dyer
For what it's worth, my background is automotive. At the very least I would follow their guidelines and on a personal level would accept nothing less than 20 samples of 5 each. And that is just the start.

Capability needs to monitored over a time period, not just a sample size!

This is the definition from the PPAP handbook......... but by my experience the customers ask for different amounts.
There is a difference between PPK for an initial run and CPK for a run over time.
What the customers are usually looking for is the width of the distribution. I have a number of experiences where the Std deviation is very low, but not centered and customers have changed tolerarances to give a good C/P PK.
There is a tendcy to sprinkle drawings with critical characteristics which are not really justified. The French have a system of ranking CC's from 1 to 4 and there can be 20 plus on a drawing.
The best way is to ask the customer what he wants in the PPAP, then there are no unpleasant surprises.

 

[A. Stuart Dyer]

Agreed Howard,

Even within the big 3 there are variations on what they want, and we give them what they want. But when it comes to rubbing your back against the tree there are certain numbers of samples and group sizes that are required and as long as the warrant is signed and there is evidence that you will improve on the process, there will be no problem.

To anybody, just think that because a B3 engineer sais he wants this or that, that an auditor will not disagree and issue a finding. Sometimes, even when you have it in writing it does not mean the auditor can't site a nonconformance.

 

[D.Scott]

A Stuart

To anybody, just think that because a B3 engineer sais he wants this or that, that an auditor will not disagree and issue a finding. Sometimes, even when you have it in writing it does not mean the auditor can't site a nonconformance.

While there is probably a bit of truth in what you say, the fact still remains - The requirements of the customer always overshadow the requirements of QS-9000. If the customer has asked for something in writing, the auditor should not be arguing or writing a nonconformance. Regardless what his personal feelings, he should move on to the next point.

Dave

 

[remsqa]

Sample Size

Hai


The better person is the customer.

But the process capability study is done for better result and better understanding of the process ablity and performance for continous improvement .

So i rembember for better understanding the histrogram should have minimum 13 to 25 average values ,with 10 to 15 clause intervales .

So it is sugested to go for the sample sizs of 25*5 ie 125 components with 25 Sub groups.Any improvements pl reply

REGARDS

R.L.SATTHISH KUMAR

 

[Steve Prevette]

The 25 sets of 5 (or 4) is a "correct" answer, but not for the reasons given. It is not related to a histogram, but to a control chart. Dr. Shewhart originally gave the suggestion that do not declare a process stable until you have 25 stable points from subsamples of 4 on a control chart. And process capability is meaningless unless the process is indeed stable.

 

[qualitygoddess - 2010]

It's all in the control chart

The 25 sets of 5 (or 4) is a "correct" answer, but not for the reasons given. It is not related to a histogram, but to a control chart. Dr. Shewhart originally gave the suggestion that do not declare a process stable until you have 25 stable points from subsamples of 4 on a control chart. And process capability is meaningless unless the process is indeed stable.

Steve makes a critical point. Before conducting a cap study, you have to know that the process is stable/in control. I advocate the creation of a control chart for the measured variable. Follow the basic rules for creating one of these (see Steve's post), and then you can calc process cap.