Is it possible to come up with a CPK for a part with a single control limit. We take readings on the part and the minimum reading allowed (the single control limit) is 25, however it can be above 25, just not below 25. Therefore we have readings anywhere from 25 to 55. How would I go about calculating a CPK with the single Control Limit of 25? I'm sure this is very simple for most everyone, but I'm very new to the Quality Assurance program. Any help would be greatly appreciated. Thank you all.

Welcome Bob in KY!

I think you are confusing between Control Limit and Specification Limit. In this case 25 appears to be the Lower Spec Limit. (Is the upper spec 55? If yes, then you clearly have two sided specs.)

This topic has been discussed before:
Calculating Cpk on a single sided tolerance (http://elsmar.com/Forums/showthread.php?t=6552)

CP and CPK for product w/ only LSL (http://elsmar.com/Forums/showthread.php?t=6410)

Thank you Atul Khandekar, I now see that I cannot get a Cpk and must use a Cpl. Thank you again.

Hi Bob,

I have several parts in which there is only one spec limit. CPK is calculated by "x(bar) - LSL / 3s".

Good luck,
-Fox

Cleverfox, pardon me for being ignorant on the subject but what is the "s" in the "3s", of x(bar) - LSL / 3s. Thanks.

would the "s" in "3s" be Standard Deviation?

Bob,

The "s" is for standard deviation; calculated as s = R(bar)/ d2. The "d2" is a constant and its value is found on a chart in the AIAG SPC reference (or other SPC guides) manual.

-Fox

Your 25-55 should be the sampling amount,right?.
Actually,it would be stlll OK if over 55. The reason why we have to collect 25 above is that the sampling distribution of X-bar will be approaching to Normal distibution. This is derived from "Law of large number" in statistics.

I still felt that cpmk could be better estimate for one sided specs. I added a spreadsheet with an example of "with larger variation on target or less variation far away from target.

by the way

This is derived from "Law of large number" in statistics.


is the central limit teorem?....

I am switching from a sample size of 5 to 2 pcs. I have a history of spc on 5 piece chart. I had calculated economic control limits( infinity qs) on the 5 pc charts. The two pc sample size would be before the operator dumps the parts into a bin. I am usign visual spc software
I would like to hear feedback on anyone using other control limits when the CP and Cpk is large and also any feedback on the type of chart for a sample size of 2.

feedback on the type of chart for a sample size of 2

The use of sample size of 2 is Individual and moving range charts or IX-MR, Donald Wheeler described it in "Advanced topics in Statistical Process Control" as one of the more flexible charts (even he used it for attribute data).

IMO, IX-MR chart is powerfull and usefull for batch processing.
:agree1:

You can still use an X-bar & R chart with a sample size of two. We frequently use sample sizes of two or three for multiple cavity tools (dies) or multi-fixtured machining centers. Your calculated Control Limits for a smaller sample size may or may not be more "economical".

Bill

Thank you for your responses. I want to let you know this is a stamping operation , variable data being collected. The economic I refered to was in calculating control limits. This is similar to the modified control limits, I beleive this was in a book by AT&T.
It allows the average to move depending on the material , set up and tool wear while maintaining customer requirement for Cpk. I thought batch processing was with chemical, does it also refer to sampling?
Is anyone out there using control limits calculated this way?

Hi Let me know what kind of Character u r monitoring for a deep draw part

It allows the average to move depending on the material , set up and tool wear while maintaining customer requirement for Cpk

For different means, you can use Target IX-MR or Short run IX-MR (for individuals) and Target X_Avg-R or Short Run X_Avg-R (for subgroup size of 2), depending on the equality of the standard deviation (are them equal for all products?).

I thought batch processing was with chemical, does it also refer to sampling?

The batch processing is related to chemical processing, the values of the samples are too similar in short term (homogeneous solutions, big volumes, so small changes), so the use of X_Avg-R, is doom. This happend specially with systems that take automatic meassures and the time between samples is reduced.

I tink is not your case, looks more like you permit tool wear until a limit (for example in glass bottle making), there are many against precontrol but It could work in this case, or use CUSUM or EWMA to detect changes on the mean (if you don't spect any big changes for the mean).

:bonk:

We are measuring an ID with a bore gage and some with the best fit plug in .0002 increments until we can chnage over. The height under a large diameter, lookign for the lowest height under thta flange and the foot diameter which is done with a micrometer. On the foot dia. small we are lookign for the highest and lowest dimension. The latter measurment we are looking to improve the method of measurment.
I wouldn't say it was pre-control. I am able to relate to actual formulas noted. In soem cases I cant have my target at nomina because of subsequent operations. Has anyone knowledge of these economic formuals?