iSixSigma

Which factor for doing measure?

Six Sigma – iSixSigma Forums Old Forums General Which factor for doing measure?

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #32074

    Wara
    Member

    Hi all,
    How can I know, which factor should be selected for measurement? Which factor should be, or should not be selected to do DOE? 
    When I don’t have experience in designing or process system, I might measure and record all factors, then I could make an analysis for seeking the effect of each factor. Finally, I will know which factor should be mearsured next time, which factor has got the effect for all system.
    I would really like to know there is someone make guiding solution to select which appropriated factor that should be measured when I am designing new system.
    Best regards,
    Wara S.

    0
    #85290

    Michael Schlueter
    Participant

    Dear Wara,
    I do understand your question: I had the same questions in the beginning. The anwer is:

    state first what your new design/process is intended to do
    state its (one or few) intended functions.
    This is important to do because there is no universal answer to your question. For example, what is the intended function of a knife? Well, it depends (on the customer); it can be:

    the knife cuts a slice of bread (intention: to cut)
    the knife protects me (intention: to protect)
    the knife impresses me (intention: to impress)
    and so on.
    An excellent cutting knife can differ from an excellent protecting knife, which can differ considerably from an excellently impressing knife. By DOE you can approach this state of excellence, but you need to state beforehand, what is (and what is not) “excellence” in your case, viewed from your customers perspective.
    On the opposite, following a “measure it all” strategy, you may end up with a long list of missing-the-point measurements, like:

    knife density
    knife weight
    knife appearance
    knife color
    etc.
    Those are neither wrong nor right, but they do not really express the intended function(s) of this specific (customer wanted) knife. Does a customer care about a certain ?optimal? knife-density, when he/she wants to impress the neighbors? Perhaps there is something more important about this kind of knife.
    Perhaps we can have a closer look at your situation? Please post a reply to me here , if you are interested. I am very experienced in setting up this kind of investigations.
     
    Best regards,
    Michael Schlueter

    0
    #85291

    Wara
    Member

    Thank you very much, Michael. Your answer makes me clear one step. You point me to concentrate in the intended function(s) of system or process that I’m studing. But I have still got few questions,
    1. Are you mean to adapt the QFD method to interested system or process, right?
    2. How can I know the factor that I have investigated, it is the correct factor? Need much time to learn? For learning to know, it isn’t an effect factor. And then finally, I will found the factor that I should measure, it is “noise”. This case can not be reached by the other quicker ways?
    3. About the Murphy’s law that shown Type III error, it can be applied for this case?
    Thank you and best regards,
    Wara S.

    0
    #85292

    Michael Schlueter
    Participant

    Dear Wara,
    You raise good questions. Let’s sort it a bit.
    It is good to do QFD from a systematic point of view. But you may already be in a position where you do know enough about what your customer wants.
    Once you know your customers intention we have to identify 2 important items :

    Intent
    Perceived Result
    The Intention is my input to the process, reflecting what I, the customer, want to do with my product (say the knife again). The Perceived Result is the ouput, indicating how well my intention was met by applying the product (i.e. running its process, utilizing its function).
    Let’s try this on the knife example, which main function is “to cut bread”. My intention then is “to cut a slice”; my perceived result is “a well or mediocre cut slice of bread”. – Next we have to translate these descriptions from the customer domain into the engineering domain: making things measurable as you need.
    To do this try figuring out:

    how do different people cut a slice of bread?
    how do their intention vary? (e.g. thickness of slice, slicing speed etc.)
    how does their result vary?
    do they all cut the same bread or different sorts of?
    etc.
    Here we learn about 2 important factors:

    intention-related (-> signal factor)
    depending on usage conditions (-> noise factor)
    Let’s assume you found out that your main customer group simply wants to cut slices of different thickness. That’s what they expect. That’s their intention. So the “cutting system (bread, knife, user)” should be able to cut slices of different thickness …
    Noticed? We just formulated a measurement system:

    take a bread
    cut slices of different thickness (e.g. 20mm, 10mm, 5mm, 2mm, 1mm) => signal-factor M (process-input)
    measure thickness of slice at various points, e.g middle, rim => output response y (process output)
    So you can compare the quality, performance, effectiveness of different knife designs by just running this series of tests (there is also a way to qunatify the findings). The ideal knife will:

    show no variability in y (always a constant thickness)
    provide always the target thickness (20mm intended IS 20 mm cut)
    A real knife will deviate from this ideal one, of course. We are happy to find a quite-ideal one as a starting point ;-)  – We also just mentioned the most relevant noise-factors:

    variability in bread
    variability in user
    (those probably outperform all variability in slice thickness the knife manufacturing can ever introduce)
    So all you need to do is defining 2 extreme noise situations, which will reflect reasonable variability in usage conditions, e.g.:

    N+ : careful user, bread with good consistency
    N-:  lazy user, bread with very instable consistency
    Run the test under those 2 exterme conditions => that’s your evaluation system to quantify the cutting performance of this knife for this target group of users.
     
    Now, just one group of factors is left: the control- or design-factors. What can you, the engineer, provide so that the cutting result will be most stable and on-target? Regardless of N+ or N- situations at the user? This kind of DOE will result quickly in a knife design, which leads to excellent results in the hand of most users; a market success, so to say.
    Be creative: vary knife length, thickness, angles, blade shape, material etc. Use DOE to create variability in knife design. Evaluate each design by the measurement system described above. Select the best one (low variability, on-target cutting, least cost). Enjoy your ROI (return on investments).
    If you need more help, please let me know at  [email protected]
    Best regards,
    Michael Schlueter
     

    0
    #85294

    Wara
    Member

    Your vision is shown you have much experiences. I’m thankful for your explanation.
    Best regards,
    Wara S.

    0
    #85298

    Michael Schlueter
    Participant

    Thank you, Wara. I hope it helps you in your initial problem. Please let me know, when I can do anything else for you ([email protected]). Perhaps it is good to have a closer look at your problem.
    To summarize:

    noise-factors reflect variability in usage conditions
    output responses y should reflect what the user expects as a result
    signal-factor M reflects customer intention, which can change, of-course
    control-/design-factors reflect design changes
    Changes in design (do it this way or in a different way) are under your full control at design-time. Make a choice.
    Once your product is build and shipped to its user, you lost any control over your product. It has to withstand/survive usage conditions, which are reflected by noise-factors.
    Introducing noise-factors during the design-stage tries anticipating later trouble (stress situations for your product). Try selecting design-factors, which minimize the impact of (unforseable variability in) noise factors. This is a way to avoid later trouble, like scrap, returns etc.
    Best regards,
    Michael Schlueter

    0
    #85299

    Manee
    Participant

    Michael
    Great explanation.  Useful for everyone. 
    Manee

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

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