cheating the cpk

[rsm7400]

Hello. Thanks for any useful information in advance!
Background info: My questions are related to a thin part that is produced, but then gets a coating on it.
the parts are made to .040" +/- .0005" and then coated with a final spec of .042" +/- .001" and a cpk>1.33
the spec on the coating is .000600"-.001900"

I have 2 questions relating to cpk.
1) the .040+/-.0005 is only process spec that we determined to ensure we meet the final spec (might be useful to note the part grows a tiny bit when coated). Our cpk on our raw part is .68 at that tolerance. Because it is only a process spec we determined, if we say the spec is .040+/-.001 the cpk jumps to 1.39, is there any negative statistical implication to doing this? It is just to raise the raw parts cpk but not actually changing how it's made.

2)Our coating process has a cpk>1.33 but the raw part is at cpk=.68 (or 1.39 if top is valid). So if you combine two statistically capable processes does that mean the final will be? It doesn't seem like it would be, especially if we are cheating the first to be higher (remember it was only a process spec).

Thanks for the help, and dont be harsh if something is super wrong with this please!

Ryan

 

[dgallup]

I'm assuming you are only coating one surface.
Then using your initial values:
Stock 0.04000+/-0.00050
Coat 0.00130+/-0.00065

Final 0.04200+/-0.00100

Assuming normal distributions and your tolerance=3sigma then your finished Cp is only 1.22 and the cpk is only 0.366 because you are not mean centered.

Second case:
Stock 0.04000+/-0.00100
Coat 0.00130+/-0.00065

Final 0.04200+/-0.00100

Cp is 0.88 and Cpk is 0.25

This is just doing the root sum squares calcs on your spec, not your actual measured data. Does not look to me like you have a viable tolerance stack in either case.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.