Pp, PPk vs. Cp and Cpk
Six Sigma – iSixSigma › Forums › Old Forums › General › Pp, PPk vs. Cp and Cpk
 This topic has 6 replies, 5 voices, and was last updated 13 years, 5 months ago by Obiwan.

AuthorPosts

February 28, 2009 at 12:10 am #51930
Felipe DiasParticipant@FelipeDias Include @FelipeDias in your post and this person will
be notified via email.Hello all, I would like to understand the correct concept to use the Pp, Ppk and Cp and Cpk. I have one situation to qualify a mould that generates plastic bottles. The mould contains 10 cavities and when I ask for 5 shots, it will give me 50 samples (10 * 5). I will measure a critical dimension of this bottle with a high precision equipment (2 decimal places).Under this situation, what is the process to be applied? Pp and Ppk or Cp and Cpk? I do not see any benefit to apply Cp and Cpk because this method uses a standard deviation that came from a range (maximum – minimum) of subgroups. So, by using Cp / Cpk to calculate the standard deviation, I am currently using just 10 samples as I am dividing all samples in 5 subgroups ( subgroup 1 = shot 1, subgroup 2= shot 2, subgroup 3 = shot 3, subgroup 4 = shot 4, subgroup 5= shot 5) and using the maximum and minimum value of each subgroup. Afterwards, it is necessary to apply a constant d2 in the calculation to determine the standard deviation. So, for me, I am “killing” all precision that came from my measurement equipment because I do not have idea where this “constant d2” came from. Hope someone can give direction and clarification on this subject. Thanks in advance, Felipe Dias, PMP
0February 28, 2009 at 12:49 am #181803First…you don’t have 50 samples…you have 5 samples from 10 separate processes. Once the melt stream diverts into its individual cavities, it becomes an individual process…and treating them as one process does not allow for homogeneity.
Second…why are you suggesting using an estimate of Cp/Cpk via the Rbar/d2 method? Why not simply calculate the short term standard deviation?
Third…have you established control of your process? With 5 shots?
Fourth…your statement that says “I do not have idea where this ‘constant d2’ came from” shows that you need to back up and really learn a little bit about control and capability statistics.
Finally…with Cp/Pp/Cpk/Ppk being estimates of process capability, why are you worried about killing the precision from your equipment?
Statistics are not exact…they are estimates!
Obiwan0February 28, 2009 at 3:32 am #181809
Rajeev sethParticipant@Rajeevseth Include @Rajeevseth in your post and this person will
be notified via email.Hi Felipe Dias, PMP
Concept of Cpk & Ppk is very simple
Cpk : Process capability, will give you an indication what your process is capable to deliver. It uses estimated sigma for the calculation & d2 as a constant.
Ppk : Process performance, will tell you actual process performance. It uses actual calculated sigma for the calculation.
In short if you want to design/approve a process you have to use Cpk & if you want to measure running process performance use Ppk.
Please calculate Cpk of each cavity separately.
Rajeev, HongKong
0March 5, 2009 at 10:43 pm #182065
Felipe DiasParticipant@FelipeDias Include @FelipeDias in your post and this person will
be notified via email.1) No Obwian, I have 50 samples as it is the same packaging materials. 2) Yes, there is a previous process before the 5 shots. 3) No, I am not suggesting, I am trying to understand the difference between both method. 4) I am looking for help… so, if you do not want to help, just ignore. 5) Because if I apply a constant in the methodology, I will have great differences in CP and Cpk values. So, if I use an approval criteria > 1.33, this can lead to incorrect conclusions.
0March 5, 2009 at 11:06 pm #182066
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.You do realize that the difference between Pp/Ppk and Cp/Cpk is that the Pp/Cp refer to the potential capability if the process is centered. It only considers the width of the process and the width of the specs and not to where the process is relative to the specs. The Ppk/Cpk consider the process relative to the mean, that is whether the process is centered or not. The Cp and Cpk will only be the same if the process is centered in the middle of the specs. If they are different it only tells you that the process is not centered with the correct calculations being against the worse case scenario which means the closest spec.
0March 6, 2009 at 1:09 am #182069Hi Felipe Dias,
Care to explain how the ‘d2’ constant kills your precision? I really can’t figure that out… :(
From your description, it would seems that you will have not enough subgrps for statistical significance…
There is no mention of process stability (*Note that “process stability” is not the same as “process capability”).
Process stability = “Is the process capable of producing parts with controlled variation; Can i know the next part is going to be produced within what readings?” <== Regardless of specs
Process capability=”Can the process produce part constantly within Specs?”
If you are worrying about the range being not able to represent the spread of your process, you might want to consider using SD but you will need to have a larger amount of data.
Tcm0March 6, 2009 at 3:04 am #182070Felipe
Help is what I was attempting to provide…but you obviously did not want to listen.
On #1 in your response. In injection molding, or any multiple stream process, you do not have homogeneity of the sample if you take random (or in this case, not so random) samples across 10 cavities. When you say that you are taking 10 samples from 5 shots, you no longer have homogeneity…you MUST, if you are to do it correctly…take 50 samples from EACH cavity…while many individuals lead companies astray when it comes to this process, I have done a lot of research on this situation, as I spent over 20 years of my past working in injection molding.
(2) I don’t understand this bullet point.
(3) What I was trying to suggest via questions was that, instead of estimating Cp/Cpk, that you actually capture the data and calculate the short term standard deviation. Unless you have already captured a lot of data on control charts.
(4) As I said, I am trying to help if you will listen…my fourth bullet point was suggesting that you do some basic research in the literature about where the d2 constant comes from…most basic SPC books have tons of information on that.
(5) Again, I have no idea what you are getting to with this bullet point.
Obiwan0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.