Using Cpk for the purpose of determining sample size/frequencies

[thebigbear]

Hi,
I'm seeking some help with the above topic.

I'm researching the possibility of using Cpk for the purpose of determining sample size/frequencies.

I'm of the opinion that if a process has proven to be stable and common sources of variation eliminated, and it has held this state of control over a period of time, say 2 or 3 months then couldn't one say it's time to review the current sampling scheme.

If this can be done how is it done?

Is it possible to convert the Cpk value into something that relates to a sampling plan AQL?

Can someone please point me to some papers, journals or text that explains how this can be acheived?

Thanks in advanced,
Nick Viney

 

[Marcel Boere]

Nick,

What I recommend is to switch to x-s or x-R charts to monitor the processes average and spread (spc). For any continuous process this is an acceptable method with alert and action limits on e.g. 3 and 4 sigma.

Marcel

 

[AviGeva]

Hi Nick
Sample size and sampling frequency must be derived from the nature of your process. Nobody knows better than yourself.
Yes it is the time to review the process. Do you have any customer` (internal or external) complaint? did they increased or decreased?
If you do not have complaints, it is better to leavr it alone. it is a good one.
if not, you must analyze it again by the same methods you first used and recalculate UCL and LCL as well as CP and CPk.
CPk has nothing to do with AQL since, if greater, equel 1.3 it shows that the process is well positioned in the secification limits and promisses about 30 or something PPM.
AQL, even for zero acceptance plan allows much more deffectives by nature. Just remember the LTPD or LQ.

There is very good stuff in this site in the PDF files go and look for it.
Regards
Avi