Is it possible to have a process that is too good. The reason I ask, is that I have conducted a stability study on a gauge. The Xbar chart shows points out of control. The mean is 36.007. The UCL is 36.014 and the LCL is 36.001 The rbar is .011 and the specification limits are 36.000±.2 I also get a CPk figure of 14.63!! Which I know is invalid if the process is not in control. The range chart is in control. I am mystified as to why the chart would indicate that the process is not stable. Perhaps someone can enlighten me.

I am mystified as to why the chart would indicate that the process is not stable.

We have no way of knowing why you have points beyond the control limits. What are you measuring? Do you see similar problems in the process data?

Yes - We cut wire to length using an automatic machine. The machine was very repeatable and having a product spec of +/- 1/2" didn't hurt either. We had CPK's in the 70 - 80 range. After a few years we backed off on our frequncy of measurements. Never had a problem with this number from the customer or the auditor (once they understood the process)

Is it possible to have a process that is too good. The reason I ask, is that I have conducted a stability study on a gauge. The Xbar chart shows points out of control. The mean is 36.007. The UCL is 36.014 and the LCL is 36.001 The rbar is .011 and the specification limits are 36.000±.2 I also get a CPk figure of 14.63!! Which I know is invalid if the process is not in control. The range chart is in control. I am mystified as to why the chart would indicate that the process is not stable. Perhaps someone can enlighten me.

I don't know about effecting stability but it does effect GR&Rpv.

For example I recently improved a process very significantly 1.8 to 6+. GR&Rpv on new process failed because I cant improve the gauge enough for the drop in the process range value, needless to say I applied common sense MSA Page 23, and told my customer regarding cost effectiveness of improvements MSA Page 46

Martin
Older not wiser

The problem i see is in your measurement system. Instead of measuring to 1/1000th of an inch, try to round off to 1/100th of an inch. Your chart will be more stable and in-control.

Remember, out of control condition includes other unusual patterns (runs, cycles, etc), not just points beyond the control limits.

Did you do a gauge R&R study and plot the numbers on the control chart? Such that the range data is the range of the repeated measures of the same unit and the average values are the averages of the repeated measurements of each part?

If so - this would give you in control range values and a lot of 'out of control' averages...because you are calcualting the control limits from the average range which would be very tight if your gauge is reasonable. Then the averages are from PARTS wich vary greatly from each other - in comparison to the difference between repeated measures of the same parts...

Also, depending on your software or formula choice - if you calculated the traditional Cpk value (as opposed to the Ppk value) you would have used the within subgroup standard deviation which comes from the range of the repeated measures...again this woulg give you a very small standard deviation compared to your spec limits...

The problem i see is in your measurement system.
I have to agree, it's like you are trying to measure a klick using a meter stick while gauging errors in centimeters. Reduce the accuracy of the gage or tighten up the UCL & LCL.

Is it possible to have a process that is too good? Cpk is over 14

I suppose it is possible: Let's assume you buy top end CNC equipment, and then use it for producing a low spec product you could easily make with an old lathe from 1870 (An extreme case, I know, but just to make an example)?

/Claes

It's posible, depends on specs..., if you have an american football court for specs and a football payer when a place-kick goes through the goalpost as a process. :lmao:

In a gage R&R, you want the range chart (test-restest variation) to be in control and the X-bar chart (part-to-part variation) to be out of control. The X-bar out of control means your gage can tell a difference between the parts. I'm confused on why you're calculating Cpk on a gage R&R study?

In a gage R&R, you want the range chart (test-restest variation) to be in control and the X-bar chart (part-to-part variation) to be out of control. The X-bar out of control means your gage can tell a difference between the parts. I'm confused on why you're calculating Cpk on a gage R&R study?

The OP said he was doing stability, not GR&R. Evaluation of the readings in relationship to the control limits is the same in stability studies as in production Shewhart charting. See the AIAG MSA manual, 3rd Edition pages 83-84 for details. Nonetheless I'm also not sure that Cpk is relevant, which is one reason that I initially asked for more information.

That's right. In a stability study, we are measuring the "same" part repeatedly at a set time intervals. This does not represent the mfg process. The idea behind control chart is to detect changes in the measurement system over time and not in the process of manufacturing the parts. IMO, Cpk is not relevant.