From: Stan Hilliard
Date: 29 Aug 1999
Time: 19:14:08
PREFACE: Mark's message at the top of this thread is from an eight-message thread on Deja.com at misc.industry.quality. That thread is no longer amendable. He has allowed me to paste his question here to continue the conversation.
Greetings Mark,
Its not a silly question. Your customer wants you to insure that 99.99% of the parts in his 200K order are in tolerance. My two questions would be these:
(QUESTION Q1) Is this process capable of ever meeting that specification?
(Question Q2) How can you verify with data that this particular order of 200K parts meets the specification?
PROCESS CAPABILITY: In order for 99.99% of the parts to be between 5.475 and 5.525, and with STD=0.0112, the mean of the population of 200k parts must be:
As high as L+Z*Std:
5.475+3.891*0.0112=5.4750+0.0436=5.5186
and as low as U-Z*Std:
5.525-3.891*0.0112=5.5250-0.0436=5.4814
(Note: I used the two-sided Z=3.891 that enclose 99.99% of the parts, assuming the process is normal and centered.)
Of course the population mean cannot be greater than 5.5186 and simultaneously be less than 5.4814.
The answer to Q1 is no -- this process cannot meet that specification. This conclusion is the same as you would get with the Cp method of analysis.
The next question is what are the ways to solve this. I, and hopefully others, will post our ideas.
Sincerely, Stan Hilliard, CQE,CRE,CQA,PE shilliard@samplingplans.com