Evaluating Process Stability - Unstable Processes - Cpk/Ppk?

[niotusen]

Is there any ideas how to see if our processes are stabel or instabel?
Evaluating Cpk/Ppk is one way?
You can evaluate on a Xbar R diagram over the time the process runs?
Is there any other ways/hints to evaluate the stability of a process?
Thanks!
/Niotusen

[Narfeldt]

Hi
Search for pre control in this forum.
Instead of Xbar and R you can use other "XY charts" as Individuals with moving average.
Depending on your process (administration or manufacturing) you can use charts over defects/not good ie PPM, dpmo.

[Steve Prevette]

Originally Posted by niotusen

Is there any ideas how to see if our processes are stabel or instabel?
Evaluating Cpk/Ppk is one way?
You can evaluate on a Xbar R diagram over the time the process runs?
Is there any other ways/hints to evaluate the stability of a process?
Thanks!
/Niotusen

Cpk and Ppk will NOT tell you if a process is stable. In fact, in order for Cpk or Ppk to be meaningful at all, the process must be stable prior to making those calculations.

One way (and probably the most common way) to tell if the process is stable is to use Statistical Process Control (control charts). SPC includes using rules that will tell you if the data (and by inference the process) are stable. It is best if you have at least 25 data points, but you can tell if the process is not stable well before 25 points many times.

You may find one way to do SPC at my company's website at http://www.hanford.gov/safety/vpp/trend.htm. This is the procedure we use. There are variations out there.

[Darius]

It's not a common way to put stability but...

Cpk/Ppk or Cp/pp is known as inestability index and is nothing else but Total_variation / Variation_within, and altho it it self doesn't show a process stability, it's true that the nearer to 1 the total_variation and within_variation estimates are more or less equal and if variation is equal (and you take for granted that the mean of the process hasn't change) desto the process is stable.

beware of the taken something for granted, but for variability is a nice measure (altho no magical number for stability in such index exist, as far as I know).

[The Taz!]

Steve,

You are right on the money! Where is Shewhart when you need him! I think that the old Western Electric rules would help here. . .at least as a staring point.

[ben]

Niotusen,

May I recommend that you do some studying on process control and capability. There are many books and websites that can be of immediate help. I personaly like the approach to process control taken by Donald Wheeler.

Keep asking questions here at the Cove; there are some very talented and experienced individuals here who will help with great patience.

You are starting a fastinating journey; I wish you well.

Ben

[niotusen]

Thanks a lot"
I ´ve started my journey!
I´ll be back!
Thanks a lot!
/Niotusen