Is a Stable Process (within Control Limits) required for Calculating Pp, Ppk?

[sylwia]

Hello,
do you know if Pp, Ppk analisis required stable process (in control limit)? To calculate Cp, Cpk I must have a stable process, but what if I have make only Pp, Ppk analisis? Where can I find answer for my question in SPC manual?
thank you

 

[Bev D]

By definition a process that is not stable is not predictable.


If you have an unstable process you can calculate a Ppk number but that number has no predictive value; it only quantifies actual PAST performance. In this case, (usign the classical definitions of the indices) Cp, Cpk and Pp are all useless as they are concerned with short and long term capability whcih is predictive of the future...

The complexity with using control charts to assess stability is the non-homogeneity of most real world processes. Traditional Shewhart charts are based on a homogenous process stream. If your process is not homogenous and you don't select a rational subgrouping scheme your process will appear to be unstable because you will have "out-of-control" points on the chart.

 

[sylwia]

To calculate Cp, Cpk I must know Pp, Ppk. Auditor says: Processes capability for some characteristics has been approved (Pp, Ppk>1,67) despite the fact that graphical analysis shows that the process is not under control (out of control limits). I don't understood why this is nonconformity? Pp, Ppk analisis contain special causes.. What shall I do to auditor satisfaction? I know special causes and I know, they will be always in my process (screwing operation- torque moment)--> it is impossible to move special cases from my process (different components delivery)..

 

[Bev D]

In order to answer your question regarding why the auditor has issued a non-conformity we would have to see the data that the auditor is looking at. Can you post the data?

As I posted earlier, you may not have an unstable process, you may simply have a non-homogenous process which will appear unstable on the traditional charts for homogenous processes.

sources of non-homogeneity are quite common and often they are not economically 'removable' but they can be controlled. in some cases (like with tool wear) we simply use a different kind of chart to tell us when to change the tool and when some other special cause disrupts the process. in some cases it is a matter of selecting a different subgrouping scheme. and there are times when we can in fact remove special causes - we just need to think about it differently...

If you post your data and describe the sampling/subgrouping scheme we can probably help you develop an appropriate control chart...and answer your question regarding the auditor...

 

[sylwia]

thank you for your attantion
Results for analisis I collect for recording system (automatically) after operation (one month back) , process is automatically,
subroup size=1, IMR chart
In atachment results and graphs

 

[Darius]

despite the fact that graphical analysis shows that the process is not under control (out of control limits).

The presence of data points outside of control limits doesn't show that your process is not under control.

Extracted from http :// www. statit. com/support/quality_practice_tips/criteria_for_lack_of_control.shtml - OBSOLETE BROKEN 404 LINK(s) UNLINKED

Even without supplemental criteria, this one criterion will often give indications of lack of control (points outside the 3-sigma control limits) even though no special-cause variation exists. The probability of this happening is called the false-alarm risk or alpha risk. For example, with a normally distributed process that has a known mean and standard deviation, an Xbar chart with 24 subgroups has a false alarm risk of 0.063. This may be calculated by noting that the probability of any one point being inside the 3-sigma limits is 0.9973 so the probability of all 24 being inside is 0.997324 = 0.937. Therefore, the probability of one or more points outside the 3-sigma limits by pure chance is 0.063.

And you are very, very far away from specs limits.

note: Your process look OK to me, almost 1 per 100, not bad.

 

[sylwia]

Darius
I don't understood value with false alarm- how can I calculate this value?