Receive a mail from my Manufacturing Engineer, he want me to review a process performance as the Cpk value reported is “Out of Control”! I m shocked while reading the mail, I scratch my head and try to figure out what he is trying to tell me. I quickly flip through the report I send to him, notice that the Cpk value reported is 1.8, which is very good. Hmm, just don’t understand why he keeps saying it is “Out of Control”.
The best thing I can do is call him for clarification. Over the phone he told me the process Cpk control at 1.67 +/- 0.05. 1.8 is out of the UCL range, it is not acceptable and I have to provide him an improvement plan.
Now I realize, he seem like using a SPC chart to monitor the Cpk value. Can we do that?
Anybody mind to share your opinion?
Wednesday, September 2, 2009
SPC Chart To Monitor Cpk Value ?
Posted by I Talk - You Do Labels: Six Sigma
Subscribe to:
Post Comments (Atom)
1 comments:
Not really a good application of control charts. Technically, he could plot it as an ImR. But in this case he went OOC above the UCL. This is generally a good thing. Now he has to figure out why his process improved and how does he keep it there?
Cpk is a function of average, std. deviation and specification limits. In an IMR chart the specification limits are really constants so were not really violating the rule of putting spec limits on a control chart.
I would like to see the underlying processes controlled by something like a Xbar-R chart. Since this is a plot of average and range (which can be used to estimate Std Deviation (Rbar/d2)) it contains all the same elements as Cpk. An Xbar-R would give him a better indication regarding the source of his process shift since his control rules would indicate whether average or range went out of control.
Even if he chose to maintain Cpk on an ImR I would suggest he consider the following:
(1) are his control limits of +/- 0.05 practical and actionable.
(2) how often is he plotting Cpk? Most of the run plots I have seen using Cpk are monthly. If so, 30 days of aggregate data can hide a lot of issues. Unless the underlying processes are also controlled his plots as control charts may not be very practical.
(3) Part of control charting is to consider your frequency and sample size. We normally chose these based on knowledge of the process and a management decision regarding acceptable ARL (average run length) or ATS (average time to signal) process shifts. Often we are sampling several times per day. In that case ImR of a Cpk is not a very realistic tool. Again, better to use an Xbar-R.
Post a Comment