Does Control of Cp-Cpk of Mass-production is considered as SPC?

[Fadhilah Cholish Azhari]

So, at the company i work for, the control mechanism for Mass-Production SPC is judged by the value of Cp-Cpk.

Does can be considered as a "proper" SPC or there should be pattern/trend or special cause criteria?

 

[Johnny Quality]

No,

How can you "control" production with a Cp-Cpk value? It's an end of process value and tells you very little what to do if someone doesn't like the value.

Consider control charts to determine common/special causes of variation so your organization knows when to leave a process running and when to investigate and adjust; there is no better tool.

 

[Bev D]

So, at the company i work for, the control mechanism for Mass-Production SPC is judged by the value of Cp-Cpk.

Does can be considered as a "proper" SPC or there should be pattern/trend or special cause criteria?

NO. NO. NO NO.
Cp/Cpk hides the data needed to understand and control your process. You need the time series data.
Cp/Cpk has NOTHING to do with statistical control. NOTHING.
Depending on the sample size and the frequency of sampling it can tell you something about the process’s relationship to the specification. But only if the Cpk is greater than about 1.2. There is NO real world relationship of the Cpk value to the number of parts that might be out of spec. (Theoretically there is but the theory doesn't actually exist in real world processes.).

Your organization has bought into snake oil mythology…please do the work and start reading Donald Wheeler. (SPCPRESS.com has many, many free articles.). See my essential references list as well for other free articles.

 

[Golfman25]

There is NO real world relationship of the Cpk value to the number of parts that might be out of spec. (Theoretically there is but the theory doesn't actually exist in real world processes.).

If this is true, and I don't doubt that it is, then what is the point of us knocking our heads against the wall trying to achieve a certain CPK?

 

[Bev D]

The automotive and other industries bought this crap hook line and sinker in teh eighties/nineties. Because it was easy fo teh SQE and required no thought. Just that bright line of 1.33. Laziness - both intellectually and analytically and physically. A little thought will show what a crock this abomination is: how is substituting a single dimensionless number better than plotting your data in time series? A single number to describe variation? You have to turn off your brain to accept this*.


Now the original use of Cpk didn’t have any translation to defect rate it was a simple ratio of the process variation to the tolerances. Some usefulness but never intended to be used in isolation to actually plotting your data and thinking about it.

 

[Fadhilah Cholish Azhari]

I see, thank you for the answer.

The reason i open this forum is to validate our perspective if our current SPC system is already proper or not.

As a background, our SPC system may be *roughly describe as:
Operator measures the part -> Operator input into the measurement format -> Operator input/plot the measurement value to SPC plot format-> Check whether the Cp-Cpk value is acheiving the target or not.

Well in fact, we do plot our measurement as a time period in our SPC format. But, our current work instruction still just mentioning to check the Cp-Cpk value. No criteria of abnormality pattern yet.

From my knowledge is, the "proper" SPC should be as time period and also be plotted as "real-time" as possible.

But our obstacle to conduct a SPC is in the inputting process of the measurement data. Due to the large amounts of the measurement data should be inputed every day, the measurement data may took days to be plotted.

 

[Bev D]

Just stop comparing to a Cpk value - the way you are doing it is silly and useless if I’m interpreting you correctly.

Things to focus on:
What sample sizes are you using?
What kind of charts are you using? I, MR. Xbar, R ?
Why does it take so long to plot? Are you using paper charts? Charting software? Is someone other than the operator plotting the chart? Maybe printing It?
I don’t understand why a chart can’t be plotted as soon as a subgroup is entered?

I think we are still missing some very valuable information to help you.