iSixSigma

Analyzing DOE response for attribute data

Six Sigma – iSixSigma Forums Old Forums General Analyzing DOE response for attribute data

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #49268

    dcvetkov
    Participant

    Hello
     
    I am sure this was mentioned here before,, but each case may be unique..
     
    This the scenario
    -Defect is surface defect ( wrinkle) on moulded applique
     – Assume that 3 factors are considered ( injection velocity, part thickness and temperature) at two levels
    – Total of 8 runs required for full factorial ( what is the required / best sample size in each run, given the fact historical defect rate is about 15%)?
    – Length of defect can be measured, to convert to variable, but I dont see immediate benefit of this… there is no big difference between bad part of 2 mm wrinkles or 20 mm wrinkles, both are not acceptable, and using variable data like this in Minitab can give misleading results, wont it, trying to quitify the stregth of each factor?
    Any other consideration / feedback is welcome
    Thanks

    0
    #168171

    HF Chris
    Participant

    Why not measure humidity and material set up rate time? This would allow for proper cure time and possible allow the material to settle prior to wrinkling.

    0
    #168173

    Bower Chiel
    Participant

    Hi dcvetkov
    You might find it useful to look at Process Quality Control by Ellis Ott. In the third edition Chapter 11 is entitled “Troubleshooting with Attributes Data” and he analyses data from a 2^3 factorial in a case history. Approximately 45 components were manufactured under each of the 8 possible combinations. He uses analysis of means to analyse the results (the book has software on the CD that comes with it) with the response being the proportion defective. However you may gain useful insights by examining main effects and interaction plots prior to any formal analysis.Good Luck
    Bower Chiel

    0
    #168174

    PB
    Participant

    Hello,When you say “applique” do you mean component or something applied onto a molded component? If it is a molded compoent – how did you derive on the 3 factors? What about mold cooling, injection pressure, back pressure, etc?What I typically do is to do a Taguchi DOE with most of the relevant molding factors and with limited experimentation derive main effects of the experiments and then study the ones that really affect the part quality in a broader experiment which can also be a Taguchi DOE. I agree with one of the responder – is the material hygroscopic? Did you consider drying time? Is the defect from one or two cavities? Have you tried blocking them and study effect on other cavities? Hopefully, I may have shed some light.PB

    0
    #168179

    Chris Seider
    Participant

    Consider using power and sample size for 2 proportion tests and this will give you an idea of how many observations to make at each condition.  However, divide the answer using the 2-proportion sample size test by four and make that the number of replicates at each corner since in your case, each level has 4 runs at each condition.
    Then I would use % defective as the output to analyze with Minitab.

    0
    #168188

    Hylton
    Participant

    The ASQ published a paper on this called Quality Quandries. It describes the correct way to use variable data and appropriate transformatons in both Defects and Defects per unit. It was in Quality Engineering  1994-5 7(2) pp 429-443. I saved mega-bux using these methods launching new products, in short time frames.

    0
    #168190

    dcvetkov
    Participant

    Thank you for the responces. I tried measuring the length of the defect and treat is as variable, but not sure if the results will be valid.
     
    Applique is a thin plastick film placed in the mold cavity and then moulded,  it is used extensively in automotive interiors ( think of fake wood). I dont believe drying and cure time has an effect, so far my main suspect is variability of film thickness. Other process parameters are chosen from process expert, able to minimize the defect with process parameter tweaking.
    Any one has some quick 101 tutorial how to analyze DOE in Mintab with proportion defective?
    Quality Engineering , that edition where can I get it? Can it be purchased from ASQ?
    Thanks

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

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