iSixSigma

Calculating Cpk with USL or LSL?

Six Sigma – iSixSigma Forums Old Forums General Calculating Cpk with USL or LSL?

Viewing 18 posts - 1 through 18 (of 18 total)
  • Author
    Posts
  • #51540

    dic09
    Participant

    Hi,
    I have a product that have a spec of 0 to 30.
    Meaning, if more than 30, will be a failure, but if the reading is 0 or nearer to 0, it means good. I guess this is a unilateral tolerance thingy.
    So if I want to calculate the Cpk , should I use as option (a) or (b )stated below ?
    a) (USL – Avg)/3 * Std
    b) (Avg – LSL)/3 * Std
    Thanks.

    0
    #178894

    tcm
    Member

    Hi,
    U mention that specs of 0 to 30, is it possible to have a reading that is below 0?
    If it’s not possible, then you will have no “LSL” since a 0 is a natural bound and not a Spec given by customer.
    To calculate cpk, 1st make sure that you process is in control. Take special care of which formular you use for ‘Std’… (Look at the formula for CPK n PPK)
     
    Tcm
    19 Dec 2008

    0
    #178895

    dic09
    Participant

    Hi Tcm,
    There won’t be reading below 0. So I shall take the
    (USL – Avg)/3 * Std.
    By the way, what kind of things that I should take special care of the formula calculating the standard deviation (Std).?
    Thanks for replying & help.

    0
    #178896

    tcm
    Member

    Hi,
    For CPK n PPK the Std calculation is diff… And if your process is not normal, u might want to do more research on reporting non-normal capability indexes….
    Rule 1: Make sure your gage, measurement methods etc…. is good enuf to measure (Gage R&R) the specs.
    Rule 2: Make sure your process is in control… (No special trends in the outputs, within limits etc…….)
     
    *Disclaimer: I have >1.5 yrs exp… therefore you should wait for the pros to reply before doing anything*
     
    Tcm
    19 Dec 2008

    0
    #178897

    Kubal
    Participant

    cpk is calculated where the process is under statistical control, and within subgroup variation  substantial,
     
    where ppk is calculated for the process which is not under the statistical control and total variation  of sample is taken for the calculation purpose
     
     
     

    0
    #178901

    Ron
    Member

    The Cpk is the distance to the nearest specification. therefore in your case your equation (a) is the correct response.
     

    0
    #178907

    Ken Feldman
    Participant

    Are you sure that Ppk is for a process NOT under statistical control? You are correct in stating that Cpk uses within and Ppk uses overall but please confirm whether Cpk is under control and Ppk is not.

    0
    #178913

    Mikel
    Member

    Wrong

    0
    #178928

    SS
    Member

    (Which is less!)

    0
    #179028

    Kubal
    Participant

    Hi Stan.
    would you like to clarify why it is wrong. it will help me better understanding of cpk and ppk concept.
     
    looking for your guidance
    regds
    rajiv
     

    0
    #179029

    Kubal
    Participant

    Hi Darth,
    Conceptually the Cpk used to predict the long term process behaviour as whether my process is capable for producing the product on long term basis. so you can’t predict on unstable process.
     
    ppk is attempting to answer does my current production sample meet specification, ppk is used where statistical control can not be evaluated.
     
    I hope this will help in clarifying twhat i meant.
    regards
    rajiv

    0
    #179033

    Remi
    Participant

    Hai Rajiv,
    Minitab (I’m not 100% sure about the other software packages but from the forum they seem to use the same ideas) uses Pp(k) for longterm Capability: In the formula the Sigma of all data is used. The assumption is made that in the data set all the long term variation is caught because the data should be (is) representative for the products you send to your customer.
    For the calculation of Cp(k) Minitab uses the short-term sigma: Sigma-within. If you have collected your data in subgroups this is the average variation of these subgroups. If you have individual measurements (subgroup size =1) it combines two consecutive datapoints into a subgroup and calculates from that the Sigma within.
    So Ppk would then be a quality verdict about what you deliver to the customer with all variation sources included (= longterm) and Cp would be a quality verdict about the best possible production level that can be achieved when only special causes are attacked in the improvement project (the assumption used is that special causes show up in long term but not in short term variation). If Cp is not good enough for your wished-situation then redesign is necessary.
    Hope this clarifies
    Remi

    0
    #179034

    Anonymous
    Guest

    Remi,Some of us don’t agree American authors should be able to change the definition of a Japanese concept. In Japan, Cp and Cpk have a different interpretation.To change it to try and account for an erroneous theory seems quiet arrogant. If these author’s want a new metric – call it something else. (I never ever heard of any Japanese colleague referring to a shift or drift …As I’m sure you appreciate, if SPC is set up correctly and run properly, the process average does not shift by much more than 0.6 sigma – even less when using Cusum charts. Furthermore, the idea that processes set up in a robust condition can shift – even without SPC is incongruous.If one takes your definition, than a subgroup as small as 2 will have a very low entropy, and you’ll see substantial differences between the subgroup average and the population average, simply on the basis that the latter only uses one estimate of the mean.To be honest, the fact that this howler has remained in the Six Sigma domain for this length of time is beyond comprehension – no wonder we’re in the midst of Quality’s most spectacular failure … the financial industry.Just my opinion …

    0
    #179035

    Remi
    Participant

    Hi Andy,
    Thanks for your post.
    You confirm an idea I have already for a long time: “99.9 % of all problems are due to poor communication”.
    Apparantly it is not known well enough that Americans use a different definition than the Japanese. This could explain some of the confusion that occurs when people from India/Pakistan/Indonesia/… ask questions and Americans answer.
    I didn’t state my personal view, I only explained how Minitab (and probably most of the othe US S/W packages) uses Cp and Pp since that is the software package that I use.
    Whether it is correct or not was also not the issue. Rajiv asked for an explanation of the ‘wrong’ of Stan, and I expected he would not get one from Stan (I never got one for my ‘wrong’ posts). So I gave him my version of why Stan would find it wrong (yes yes, I know: pretentious)
    And if you give me a Euro for every manager who believes that when you send people to a BB course they will bring the total reject level of the production chain down to 3.4 ppm I would be a millionaire.
    I generally use the tools as a help to find out what is wrong with the process and what may be done about the problem. I don’t believe they can tell me the ‘true’ status of the process; only a certain version of it. And I (try to) use the tools consistently in the language my customer is used to (if he used Cp so do I; if he uses ppm so do I and if he uses Sigma-level so do I). And if he uses it wrongly (imnsho) I try to correct him (very carefully).
    Often SPC is set up ok but then ignored for a long time. And since people often keep changing things then the results will erode over time. It is very hard to consistently use it well.
    On your remark about the subgroup size 2: I don’t know what the exact calculation of Minitab is (I could look it up). I generally don’t care because when I have single data points I don’t use the Cp and Cpk because they have no meaning for me (I even remove them from any output I show my customer).
    Have a nice Christmas and a Nice New Year: “may the old problems be solved and new ones found”
    Remi

    0
    #179036

    Craig
    Participant

    Andy,
    Who changed the definition of cpk? Not sure I have read that one. I’d like to see the Japanese definition and the new one you are referring to.
    And speaking of new metrics beyond cpk, what do you call ppk and “sigma” level? I am not saying that I agree with these metrics, but you are asking the arrogant authors to “call it something else”, and that is what they have done.
    HACL

    0
    #179039

    Anonymous
    Guest

    HaCl,The original use of Cp and Cpk were as summary metrics on X-bar and R charts. These charts had a minimum of 30 subgroups or size n = 2 to 5. I first visited Japan on behalf of Motorola in 1986. Later, we had several Japanese engineers join MOS 8 in Austin, including one of Hitachi’s ex-directors – so I believe I’m familiar with their practice, but I don’t have any copies of Japanese articles.I have to concede your second points – thank you for pointing out Ppk and Process Sigma (Sigma level) are new metrics.What I do find refreshing in your reply is someone of your caliber admitting you don’t necessarily agree with these metrics. Surely what we need now is a group of like minded people coming together to forge the new Quality.Just my opinion ….

    0
    #179041

    Mikel
    Member

    There is no requirement that Ppk is for out of control. In fact if it is
    out of control, the capability of the process will be understated.It is just as explained by another poster – within and overall.

    0
    #179042

    Mikel
    Member

    If Cp is not good enough, it doesn’t mean redesign is necessary.It means your process knowledge is not adequate.

    0
Viewing 18 posts - 1 through 18 (of 18 total)

The forum ‘General’ is closed to new topics and replies.