Process capability or Process control - Which comes first?

[jrubio]

Well, the AIAG methods don't always line up with what the rest of industry does...so the use of Cpk and Ppk for Automotive and for everyone else is different. that cases a LOT of undue confusion.
But, a long long time ago I understood this so my memory might not be correct and I no longer have the AIAG manuals (thank what deity you believe in - or not) but I'll take a shot the Ppk for unstable processes: I THINK I REMEMBER that the comment you refer to is that if you have a process that is in fact in statistical control then it must have a Cpk of 1.33. If your process is NOT in statistical control then it must have a Ppk of 1.5 (or higher, I don't remember the exact numbers).
Please note that I may have remembered this incorrectly adn I certainly don't endorse it. (In past posts I have stated my belif that all capability indices should be eliminated.)

For me the different oabout of ppk and Cpk is logic.

The ppk at the begining trend a exponential curve when it reach the stable limit of capability you now can calculate the index cPK
-
-.................XXXXXXXXXX
-.............X
-.......X
-....x
--X
-x
----------------------------------

<---PPK----><----CPK---->

 

[Tim Folkerts]

I'm mostly an academic, so I tend to respond the way that seems logical and intuitive to me, rather than what a particular authority or book says. And I don't have a copy of any AIAG manuals, anyway, so I can't check what they have to say.

Mathematically, the only difference between Cpk and Ppk is how the standard deviation is calculated.

• Cpk: calculate the st dev for each of the subgroups, then combine the results.
• Ppk: combine the subgroups, then calculate the st dev. for the entire set of data.

If the process is stable and in control, then both methods should give the same value for the st dev and Cpk = Ppk.
If the process is stable in the short-term (within a subgroup) but changes in the long-term (variations when changing shifts, batches, tools, etc), then Cpk > Ppk.

If the process is not stable, then both Cpk and Ppk will vary in unpredictable ways. AIAG and Delphi are welcome to say they will accept Ppk values but not Cpk values for unstable processes, but I don't see a mathematical justification that would require such a decision.

Tim F

 

[JRKH]

Hi all,

I would like to understand that in case of new process , what comes first , process control or process capability?

Too much detail folks - - -

First you find out what the process is doing - capability
Then you begin eliminating special causes - Control
Then you check the process again - Capability
Improve some more - Control

Repeat as needed.

James

 

[jrubio]

I'm mostly an academic, so I tend to respond the way that seems logical and intuitive to me, rather than what a particular authority or book says. And I don't have a copy of any AIAG manuals, anyway, so I can't check what they have to say.

Tim F

Tim Thank you.
It is very outstanding your way of thinking "so I tend to respond the way that seems logical and intuitive to me, rather..." for me this is great,

Can you please provide some reference about Cpk and Ppk as you indicate?



Thanks you.

 

[Bill Ryan - 2007]

Too much detail folks - - -

First you find out what the process is doing - capability
Then you begin eliminating special causes - Control
Then you check the process again - Capability
Improve some more - Control

Reteat as needed.

James

While I've enjoyed the discussions, I think you've nailed the answer to the OP's question nicely

 

[jrubio]

I'm mostly an academic, so I tend to respond the way that seems logical and intuitive to me, rather than what a particular authority or book says. And I don't have a copy of any AIAG manuals, anyway, so I can't check what they have to say.

Mathematically, the only difference between Cpk and Ppk is how the standard deviation is calculated.

• Cpk: calculate the st dev for each of the subgroups, then combine the results.
• Ppk: combine the subgroups, then calculate the st dev. for the entire set of data.

If the process is stable and in control, then both methods should give the same value for the st dev and Cpk = Ppk.
If the process is stable in the short-term (within a subgroup) but changes in the long-term (variations when changing shifts, batches, tools, etc), then Cpk > Ppk.

If the process is not stable, then both Cpk and Ppk will vary in unpredictable ways. AIAG and Delphi are welcome to say they will accept Ppk values but not Cpk values for unstable processes, but I don't see a mathematical justification that would require such a decision.

Tim F

Unfortunately we are in Autobile Arena, and AIAG rules are requirements, if you consider that these rules are not valid, This is admirable, the best way is to discuss with AIAG and Delphi to change it, but we must obey untilm new revision.

 

[Bill Ryan - 2007]

Unfortunately we are in Autobile Arena, and AIAG rules are requirements, if you consider that these rules are not valid, This is admirable, the best way is to discuss with AIAG and Delphi to change it, but we must obey untilm new revision.

I'm sorry, I must have missed something. What rules are you referrring to and how have Bev's and Tim's answers "broken" those rules?

We supply Delphi also and when they ask for a Cpk or Ppk value - that's what they get. If the value doesn't meet their established criteria, it only means we need to employ some method to insure they get "Zero Defects" (let's not go there!) until such time that we have improved the process to meet their requirements.

 

[jrubio]

I'm sorry, I must have missed something. What rules are you referrring to and how have Bev's and Tim's answers "broken" those rules?

We supply Delphi also and when they ask for a Cpk or Ppk value - that's what they get. If the value doesn't meet their established criteria, it only means we need to employ some method to insure they get "Zero Defects" (let's not go there!) until such time that we have improved the process to meet their requirements.

Hi Bill

AIAG PPAP Manual is a Customer requirement, Unless for Ford, GM, Delphi, Visteon...

If not followed, them Customer requirements are not met.

When Customer ask for PPk according to the manual it means that the process is chronic unstable not stable, if you provide them the index stating that it represent a stable index, therefore the Customers requirements are not met, this is my point of view.

 

[Bill Ryan - 2007]

Hi Bill

AIAG PPAP Manual is a Customer requirement, Unless for Ford, GM, Delphi, Visteon...

If not followed, them Customer requirements are not met.

When Customer ask for PPk according to the manual it means that the process is chronic unstable not stable, if you provide them the index stating that it represent a stable index, therefore the Customers requirements are not met, this is my point of view.

My "issue" is that you specifically mentioned Delphi. Their requirement is an initial capability index must meet a Cpk of "X" (I'm not trying to bash Delphi - other automotive customers have the same requirement with, possibly, slightly different index thresholds. Wait a minute - does that mean I'm bashing all of them??? ). I am not a stats guru by any stretch of the imagination but Cpk is a long-term index and Ppk is a short-term index. How can Delphi ask for a long-term index for a PPAP submimssion when the process has probably not run "long-term" yet?

(Another bug-a-boo of mine - customers not knowing/understanding what their requirements mean - or worse, "customizing" their own definitions. No wonder there is a need for this type of forum )

Wow, is it time to branch this off? I think we may have meandered from the original question.

 

[jrubio]

My "issue" is that you specifically mentioned Delphi. Their requirement is an initial capability index must meet a Cpk of "X" (I'm not trying to bash Delphi - other automotive customers have the same requirement with, possibly, slightly different index thresholds. Wait a minute - does that mean I'm bashing all of them??? ). I am not a stats guru by any stretch of the imagination but Cpk is a long-term index and Ppk is a short-term index. How can Delphi ask for a long-term index for a PPAP submimssion when the process has probably not run "long-term" yet?

(Another bug-a-boo of mine - customers not knowing/understanding what their requirements mean - or worse, "customizing" their own definitions. No wonder there is a need for this type of forum )

Wow, is it time to branch this off? I think we may have meandered from the original question.

I agree :truce: