iSixSigma

Design of experiment question

Six Sigma – iSixSigma Forums Old Forums General Design of experiment question

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #52577

    Keith Rupert
    Participant

    I am having difficulty identifying the steps in solving a Design of Experiment problem.  I have software that internally designs the problem.  I need to identify the factors that affect the response.  The range of the factors are from A through F.  Each factor has a number range of 0 through 50.  When data is entered into each factor, the software generates a response value.  What are the steps in solving a problem like this?  I have Minitab available, but I do not have much experience with its use.
    1. How many responses are needed for this type of problem?
    2. How do I determine the values to input for the generated responses?
    3.  What steps in Minitab do I use in analyzing this data?
    4.  How do I interpret the data to indentify the factors that affect the response?

    0
    #185045

    Craig
    Participant

    1) depends on how many critical outputs you have
    2) Depends what the process is. are you baking pies or machining parts, etc?
    3) Read up on model fitting
    4) Read up on hypothesis testing, anova, etc

    0
    #185048

    Severino
    Participant

    How can you have software that “designs the problem?”  Are you taking a course?  Or do you mean you have software that designs the study?

    0
    #185062

    mep
    Participant

    1. How many responses are needed for this type of problem?
    If you have just 1 response variable, then you only need 1 to identify the important factors that impact that response.
    2. How do I determine the values to input for the generated responses?
    Not sure what this is asking. Once you create the design, run each experiment at the given settings and record your responses in a column in Minitab.
    3.  What steps in Minitab do I use in analyzing this data?
    For a 2-level experiment, use Stat > DOE > Factorial > Analyze Factorial Design (after using Create Factorial Design to setup your worksheet).
    4.  How do I interpret the data to indentify the factors that affect the response?
    After you run the analysis, hit Shift+F1 to use StatGuide, or you may want to take a look at Quality Trainer.

    0
    #185081

    anupam
    Participant

    For each Factor A through F fix High & Low values.
    Then enter the data in Minitab.
    It should create the Runs.
    For each random Run, set the values and note the response.
     
     

    0
    #185085

    Forrest W. Breyfogle III
    Member

    Because of how you defined the problem, I am confident that this is the same simulation software that we use in one of our classes.
     
    The solution to this problem involves the use of a two factor DOE to eliminate four of the six factors as being not significant.  You then need to use response surface method design in conjunction with steepest accent/decent trials for the remaining two factors to locate the optimum response settings for the two inputs.

    0
    #185095

    Velasquez
    Participant

    You have 6 factors which can be set at a range of levels and as mentioned previously, you can have as many responses as you want.
    I was taught to use a screening design like a placket burman to identify which of the 6 factors are critcal (this method will not show interactions between the factors due to the aliassing within the DOE) with the factors set at low and high levels.  Once the critical factors were identified, a 2 level factorial would be used with the factors set at low and high levels and this would identify the interactions between the critical factors.  This would be followed by a surface response design like a D-optimal or central composite design and the design configured with low and high levels but intermediate levels will be selected as part of the design.
    If you have one response, then you will be optimising the levels of your input factors to this response.  If you have more than one response then you will optimise to obtain the best outcome from each response.
    As one of the previous posters stated, use the statguide within minitab and follow the process.  I have not used minitab for DOE in a while but use software specifically designed for DOE and still use the above process when needed.
    Hope this helps.

    0
    #185096

    Mikel
    Member

    Forrest, A two factor experiment to eliminate Four factors?You are losing it. To much pushing control chart management
    reports?

    0
    #185103

    k.bhadrayya
    Participant

    As explained by Mr.martin, it is better to use Placket burman designs for screening significant factors taht affect the process . On screening signifcoicnat factors you can use Response surface methodology for any number of responses. Minitab will be able to solve for optimising all responses simultaniously.
    k.bhadrayya

    0
    #185107

    Forrest W. Breyfogle III
    Member

    TYPO
    POSTED: The solution to this problem involves the use of a two factor DOE to eliminate four of the six factors as being not significant. 
    SHOULD HAVE POSTED: The solution to this problem involves the use of a two-LEVEL DOE to eliminate four of the six factors as being not significant. 
    Sorry

    0
Viewing 10 posts - 1 through 10 (of 10 total)

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