About Process Capability Studies
Six Sigma – iSixSigma › Forums › Old Forums › General › About Process Capability Studies
 This topic has 9 replies, 8 voices, and was last updated 14 years ago by Chris Whittaker.

AuthorPosts

June 28, 2008 at 3:07 pm #50423
suresh kumarMember@sureshkumar Include @sureshkumar in your post and this person will
be notified via email.I would appreciate if anyone could help me in understanding what is process capability study means
0July 24, 2008 at 3:31 pm #174142Hai Suresh,
See it as a photo of the current situation of the data.
Input = Measurement on a parameter Y that you are interested in.Output = Picture + quality values Cp,Cpk,Pp,Ppk + (expected) reject ppm’s
Picture: Histogram compared to Spec (USL,LSL from customer).Cp and Pp = (USLLSL)/6Sigma is a quality number: how much room does your process use compared to the room your customer allows you. High value means good quality. (worldclass = Cp of 2)Cpk and Ppk take also into account if you are in the middle of the specroom (the interval [LSL,USL]) or not.Cp and Cpk use the shortterm Sigma in the formula; Pp and Ppk the longterm sigma. (Warning: “term” is not “time”)
You customer sees Ppk; your designers often make Cp.
For more info: follow a GB course.0July 25, 2008 at 7:01 am #174173
Michael MeadParticipant@MichaelMead Include @MichaelMead in your post and this person will
be notified via email.WOW. I am surprised nobody else answered this one.
Process capability is the inherent variation of the process. Thus, a capability study is measuring the inherent variation of the process. Usually process capability is described as 6 standard deviations when used for further calculations.
As Remi stated, most reports are given as a ratio of the process capability to specification limits. For example, Cp is (tolerance/6s); percent of tolerance used is (6s/tolerance)*100. The other formulae are also well known.
0July 26, 2008 at 11:02 pm #174199Cp, Cpk are used for stable processes (in control) whereas Pp, Ppk can be used to find capability of unstable processes.Thanks
0July 27, 2008 at 1:30 pm #174201
Paul F. JacksonParticipant@PaulF.Jackson Include @PaulF.Jackson in your post and this person will
be notified via email.I disagree AB!
The only significant difference between Cp and Pp, Cpk and Ppk is how the standard deviation is acquired. For Cp Cpk, it is estimated at any given interval from a cumulative sequential sample of a process parameter . For Pp Ppk it is estimated from a finite sample size of a process parameter.
Both types capability estimations Cp Cpk and Pp Ppk require evidence that the data used to derive them is predictable… random… “in control”… void of special cause variation!
Paul0July 27, 2008 at 2:44 pm #174202
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.WRONG!!! Control is a precondition for doing capability studies. Cp and Cpk use a within sample variation usually an estimate of the s.d. based on the r bar from the control chart and are often referred to as short term variation. The Pp and Ppk are calculated using the s.d. from the whole data set and is often referred to as long term variation.
0July 27, 2008 at 6:38 pm #174205
Paul F. JacksonParticipant@PaulF.Jackson Include @PaulF.Jackson in your post and this person will
be notified via email.Darth,
I think that we said the same thing albiet differently?
“Control is a precondition”… agreed
Cp and Cpk are evaluated from sequential subgroups…agreed
Pp and Ppk from “The whole data set” from point A to point B… the beginning to the end …finite… (considering one has data from the actual “approved” beginning and has encountered the end) or “The whole data set” from point A to point B… the beginning to the end… finite (whether or not one knows if the sample variation reflects the beginning and end nevertheless “the sample”) hence the confusion over “long term” vs. “short term” or “preapproval labled… preliminary”.
Paul0July 27, 2008 at 9:48 pm #174209
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.Paul, I was supporting and reinforcing your response. Yes, what defines long term and short term is always a point of contention.
0July 27, 2008 at 11:58 pm #174212
Joe PeritoParticipant@JoePerito Include @JoePerito in your post and this person will
be notified via email.Bear with me Suresh: A process capability study determines whether a process can meet the customer’s specifications on an ongoing basis. User’s of this methodology usually want the process to have a CpK of 1.33 or higher to be called (statistically) “CAPABLE”. A process with normal variation will produce a bell shaped curve with the process average at the center of the curve. Nonnormal processes will still produce a normal curve if sample sizes and their plotted averages are large enough. The area under the bell curve is usually quantified by a Z table, which is a table of the areas under the normal curve. This table is usually made up of one half of the bell curve. Therefore, a Z factor of 3.00 represents roughly 49.73% (just say 50%) of the bell curve. Now, if the whole bell curve fits inside of the customer’s spec limits, that’s good. However, with process variation, the bell curve is going to move around a little. Will the whole curve still stay within the spec limits? To answer that question, a capability study will calculate the difference between the center point of the bell curve (the “mean”) and the closest spec limit. This difference is then divided by 3 standard deviations to see how many times 3 standard deviations can be divided into the difference. If there are 4 standard deviations (divided by three std.dev.), then you get a CpK of 1.33. YOU HAVE A “CAPABLE” PROCESS. Why divide by 3 Std. Dev.? There are 3 standard deviations in one half of the normal curve. What you have now determined is that the bell curve (the process)can move around and that the mean (and all the rest of the curve) will not exceed the spec limit. The distance of the mean to the closest spec limit(subtract the two) may be more or less…but you will always divide by 3 standard deviations to get the CpK. Where do you get the standard deviation? That’s the calculated (Standard Deviation) from your process data. Note, by example, that the mean of a six sigma process would have to be six standard deviations from the closest spec limit and would have a CpK of 2.00.(6/3 = 2.00)
One example: I have a spec of 60 +/ 5. My process average is 58.5 and it has a std.dev. of 0.60. The upper and lower spec limit are 65 and 55. The mean of 58.5 is closer to 55 than 65. Subtract 55 from 58.5 to get 3.5 (the difference). Divide 3.5 by 3 times the standard deviation, 3.5/(3×0.60)=1.94 CpK. One last note. Many times when you subtract you get a negative number. Carry on the calculations as if they were positive. If the process mean exceeds a spec limit, “THEN”, the calculated CpK “WILL” be negative…which is always bad.0July 28, 2008 at 10:12 am #174228
Chris WhittakerParticipant@ChrisWhittaker Include @ChrisWhittaker in your post and this person will
be notified via email.Keeping it simple.
You have a garage and you want to park your car. You have a one car garage and the car you want to park is a landrover discovery.
Will the car ( the process) fit in the garage (the specs) , yes /no ? If it does not fit my garage (specs) it is not CAPABLE of housing my car and I will need to adjust my garage ( the specs).
If I have a small car and a big garage then my car will fit in there nicely?If the answer is yes it is in specification. That is CP whether your “car ” (the process) is in or out of the limits that you have been given to park within ( the specification).
If it does fit in the garage will I need to park it closer to the left wall or the right wall?
I need to do this so i can open the door and get out, or is it small enough to park in the middle so everyone can get out ? That is the Cpk?
Cp is the Fit of your process into the customers specs., Cpk is the way it fits into those specs.0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.