Calculating correct Cpk - Getting different answers

[Mitzdawg]

Following my textbook and I'm getting answers I don't understand...
#1)
My dataset is ten readings: .631; .631; .631; .629; .633; .632; .630; .631; .631; .634.

Average: .6313; STDEV: .0014181
LSL: .605; USL: .625

Cp: 2.35 = (.625 -.605)/(6)(.0014181) Good process....

CpK: 1.48 = (.6313 - .625)/(3)(.0014181) (huh???)

How can the process be "capable" when none of the product it produced was in spec?

#2)
Dataset of ten: .055; .055; .052; .055; .055; .055; .058; .053; .055; .055

Average: .0548 STDEV: .0015492
LSL: .061; USL: .071

Cp: 1.08 = (.071 - .061)/(6)(.0015492)
CpK: 1.33 = (.0548 - .061)/(3)(.0015492)

Again, a capable CpK where nothing was in spec? And, a CpK that's bigger than the Cp?

What's going on? (Please be gentle...)

 

[bobdoering]

Re: Noobie Questions

Following my textbook and I'm getting answers I don't understand...
#1)
My dataset is ten readings: .631; .631; .631; .629; .633; .632; .630; .631; .631; .634.

Average: .6313; STDEV: .0014181
LSL: .605; USL: .625

Cp: 2.35 = (.625 -.605)/(6)(.0014181) Good process.......

Variation is acceptable, just not centered.

CpK: 1.48 = (.6313 - .625)/(3)(.0014181) (huh???)

Miscalc - it is: CpK: -1.48 = (.625-.6313)/(3)(.0014181)

How can the process be "capable" when none of the product it produced was in spec?

It's not capable.

#2)
Dataset of ten: .055; .055; .052; .055; .055; .055; .058; .053; .055; .055

Average: .0548 STDEV: .0015492
LSL: .061; USL: .071

Cp: 1.08 = (.071 - .061)/(6)(.0015492)
CpK: 1.33 = (.0548 - .061)/(3)(.0015492)

Miscalc - it is: CpK: -1.33 = (.0548 - .061)/(3)(.0015492)


It's not capable.

 

[Mitzdawg]

Re: Noobie Questions

Thanks for the answer.

The guide I'm using may be incorrect...
The CpK formula is stated as:
CpK= (mean - Closest Spec Limit)/3(stdev)
and goes on to say: "This could be a negative number, but we are interested only in the absolute value."

 

[bobdoering]

Re: Noobie Questions

Thanks for the answer.

The guide I'm using may be incorrect...
The CpK formula is stated as:
CpK= (mean - Closest Spec Limit)/3(stdev)
and goes on to say: "This could be a negative number, but we are interested only in the absolute value."

Oof! So wrong! Negative number has meaning!

CpK= lesser value of:

Cpl=(mean - LSL)/3(stdev)

or

Cpu=(USL-mean)/3(stdev)