iSixSigma

Sample Size requirement for Cpk

Six Sigma – iSixSigma Forums Old Forums General Sample Size requirement for Cpk

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #30696

    rams
    Participant

    Anyone can provide me insights on this situation:
    Our current sampling frequency can not meet the required sample size of 25 subgroups per month to compute monthly Cpk indices. Assuming we can not increase sampling frequency (due to cost/manpower isssues), what can be done to estimate the Cpk?
    Here are some relevant info:
    Plotting of points is per machine.
    Each machine contains around 5 different charts for 5 products with different capabilities.
    Thanks

    0
    #80336

    Mike Carnell
    Participant

    Rams,
    Why do feel you are required to do this much monthly? Most people are looking for 25 subgroups but if you don’t have them you have one of three choices:
    1. Calculate it when you get 25 subgroups.
    2. Use what you have and don’t worry about it.
    3. Decide that the information that you need to control your process isn’t as costly as shipping defects or producing defects and find a way to collect the data.
    Basically you have built yourself a pretty small box to live in (one month and 25 subgroups) and yiou are causing yourself a lot of stress over nothing of critical importance. I would be more concerned that someone belives they can rationalize the cost/manpower issue.
    Good luck.

    0
    #80344

    RSloop
    Member

    Rams,
     
    I was just wondering if you had considered any Short Run SPC techniques.  Based on the data that you posted you may be able to use normalized control charts or possibly group charts.  The normalized would allow multiple products to be plotted by deviation from target.  In many machining processes this works very well allows calculation of accurate Cpk numbers when tolerances are the same but the target size is different.  There are several ways to code short run data to solve your problem.  If you are interested we can go into more detail.  Let me know if the same characteristics are measured on different parts and if the tolerances (as deviation from nominal) are the same and I’ll post more info.
     
    I guess my main concern here is that if you are only plotting 25 subgroups per month (I’m assuming per product), that you have enough frequency to actually use the data to improve the process (again assuming that you look at one product at a time).  By using short run methods you should be able to generate charts that will give you the proper sampling frequency.
     
    Rick

    0
    #80419

    rams
    Participant

    Mike/Rick,
    Thanks a lot.
    I am interested in the Short Run SPC techniques. I would like to know how would you use the techniques on one-sided parameters (which means I can’t plot “deviation from target” as the characteristic is higher the better).
    Please advise.
    By the way, the characteristic is ballshear strength on semicon devices.
    thanks
    Rams

    0
    #80440

    Loehr
    Member

    Rams,
    A good book on Short-Run SPC is Innovative Control Charting, by Steve Wise and Doug Fair.  You can order a copy (around $40) from ASQ, 1-800-248-1946.  Ask for item H0974.  They cover your situation of unilateral specifications in their book.
    Good luck.

    0
    #80443

    RSloop
    Member

    Rams,
    Off hand all I can think of is coding the data to a historical average.  I’ll try to think about it a little over the weekend. 
    Ross,
    I agree with your comments on “Innovative Control Charting”.  Steve is a friend of mine and he always makes light of it when I tell him what a useful book it is.  I’m sure he will enjoy your comment though.
    Best Regard,
    Rick

    0
    #80455

    rams
    Participant

    Ross,
    thank you so much. let me know once you have the details.
    Rick,
    thanks for the info. will purchase the book ASAP.
    Rams

    0
Viewing 7 posts - 1 through 7 (of 7 total)

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