Cpk for Geometric Tolerance

[WardHollowayPE]

Customer spec. calls for .005" T.I.R. on a round valve stem. Also customer wants a Cpk of 1.67. How can I calculate this? Should my target be set to zero? What should my upper and lower limits be? Am I missing something here?!?

 

[PSE]

Your nominal value should be 0. You will need to conduct a series of measurements to determine your control limits. Their are numerous formulas that you can work from there based upon sample size etc. that you can use to calculate Cpk. Make sure that your measurement equipment is calibrated and has an acceptable gage R&R otherwise your data may end up meaningless.

http://www.aiag.org/

used to have a good publication on Fundamental Statistical Process Control that you would find helpful in setting up your evaluation. Basic SPC software can get you going as well.

Regards

 

[WardHollowayPE]

If my nominal value is set to zero, whats my lower limit? Also all values of TIR are positive numbers. This will lower my Cpk if nominal is zero.
I don't think you can get a Cpk value for TIR. My question should have been, "Is there a Quality measurment for TIR that is equivalent to Cpk?"

 

[PSE]

Zero would also be your lower limit. If you keep in mind that you will rarely if ever come across a measurement of 0 TIR you still should be OK. Your average values will end up being your natural centerline and it will be the range between the samples that will have the largest effect on Cpk. I will continue to do a bit more looking into "one sided" tolerancing.

Regards

 

[gmorin1]

I have come up against this problem in the past. The basic problem is that none of the software programs are capable of calculating a "single tailed distribution" correctly. I have had experience with a program that Boeing distributed to suppliers on the 777 program 10 or so years ago. That program performed the calculations correctly and gave a CpK value that was "real".
The company that I worked for at the time is long gone and I now find myself with the same problem again. I still cannot find suitable software for the single tailed distribution.
If anyone else has an answer, help is very much appreciated.

 

[DennisP]

We use Minitab for statistical analysis of many one sided tolerances. It seems to work just fine.

 

[gmorin1]

DennisP - thank you for the sugestion. I just happened to have a demo disc of MINITAB that I had not tried yet. Just what the Doctor ordered.
Thanks again.
Gary