iSixSigma

Batch Consistency DOE

Six Sigma – iSixSigma Forums Old Forums General Batch Consistency DOE

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #30324

    Mike Beauregard
    Participant

    I’m working with a team to reduce batch-to-batch variability as measured by mechanical properties. The team has designed a 15-factor L-16 Taguchi screening experiment. The experiment and testing will cost $200K for a single replicate per treatment combination. We can only spend $100K for the experimentation. The team can’t agree on cutting the number of factors down to 7 so that we can run an L-8. Would it be legitimate to run the L-16 with a single run per treatment combination and then compare for each factor the range of the responses for TCs run at its high level and the range of the responses for TCs run at its low level? And then set the factors to the levels that gave the lowest range (after balancing them out to ensure the mechanical properties remained in spec) for the confirmation runs? Thanks for any input!

    0
    #78873

    Robert Butler
    Participant

    Β Replication does not necessairly mean that you have to replicate an entire design.Β  When you are interested in screening for variables that have a major impact on your process and/or when the experiments are too costly or time consuming to run, you can opt for running a saturated design with replication restricted to only one or two of the points in the design.
    Β  Based on my understanding of your problem it would appear that a way to address the issue of screening 15 variables would be for you to generate the usual 2 level full factorial design for 4 variables and then assign the other 11 variables to the second, third, and 4th level interactions.Β  If the variables are such that there is a natural center point for each one then I would recommend adding a center point to this design and then replicate just the center point. This would give a total of 18 runs.Β  If there isn’t a natural center point (a type variable instead of a continuous variable, for example) then toss the numbers 1-16 in a hat and draw the number of the experiment that you will replicate.Β  If you can afford to replicate more than one experimental condition then do so.Β 
    Β  It is understood that this approach will not give you the degree of precision with respect to an estimate of your error as would a complete replication of the entire design but it is statistically sound. Indeed, replication of just the center point for an estimate of error is standard practice with full or fractional composite designs
    Β  This approach willΒ  permit a check of the 15 variables of interest while keeping the cost of the experimental effort within your budget.
    Β  If you want more information on nonreplicated experiments I would recommend looking at Analysis of Messy Data-Vol 2 – Nonreplicated Experiments by Milliken and Johnson

    0
    #78874

    melvin
    Participant

    You may be able to improve experimental efficiency by using a single array rather than a crossed array of the Taguchi style.Β  Expt. design for this is well explained in
    1.Β  Wu and Hamada – Experiments – Planning, Analysis, and Parameter Design Optimization
    2.Β Β Myers and Montgomery, Response Surface Methodology – Process and Product Optimization Using Designed Experiments

    0
    #78889

    Mikel
    Member

    Mike,
    Two thoughts
    Running a non replicated design is done all of the time. Go for it.
    15 factors! Give me a break. Your team hasn’t done their homework. What do you think the analyze phase is about?

    0
    #78896

    girish
    Participant

    Dear Mike,
    Your objective is to reduce batch to batch variability w. r. t. mech properties.I suggest the following approach:
    Before going for DOE, select 2 batches which have high variation between them.Test both batches & measure the important parameters in both of them which can be the cause of variation in mech properties.Compare the parameters in both batches . At this stage you will get clue of likely factors which generally speaking will not be as high as 15.
    After this screening of factors is completed , please proceed with the DOE on the shortlisted factors
    Best Luck & Best Regards
    Girish
    my email ID is: [email protected]

    0
    #78917

    Robert Butler
    Participant

    Β  While 15 variables may be a result of a poor effort in the measure phase of a project it is also quite possible that 15 is the minimum number.Β 
    Β  If you have no knowledge of statistics and the power of designed experiments you will find it very difficult to deal with much more than 4-5 variables.Β  Under such circumstances it can become an article of faith that there can’t be more than 4-5 critical variables in a process.Β  When the no-more-than-five-variables barrier is removed all sorts of interesting things come out of the woodwork and people start to tell you what they REALLY think is impacting the process. Having done this sort of thing for 20+ years I have found that even after doing all of the things that are lumped under Measure in DMAIC it is not uncommon to have a final variable list in the 10-20 range.
    Β  The benefits of a design that investigates the top however many (even if some of them are questionable) far outweighs any other concerns.Β  A screen that checks them will not only aid in an identification of theΒ really important variables it will also lay to rest a lot of myth, misunderstanding and disagreement about the process itself.

    0
    #78942

    Jordan
    Member

    Run your L16 with your 15 variables. Be very careful and don’t delegate too much. Too many DOE are unconclusive because the engineers are biased and and don’t put enough factors thus missing opportunities. Many engineers like DOE to prove their best educated guess. I have seen some screening DOEs where the most important factors was least suspected.
    For the specific batch consistency, I would recommend to analyse the variance of residuals as it is possible to do now by declaring a “Logvariance effect” in the new version 5 of JMP

    0
    #79045

    James Thacker
    Participant

    The designs you discussed (L16 with 15 factors, L8 with 7 factors) are known as saturated designs, I believe. The next step would be to use a supersaturated design (Linn,1993.) A supersaturated design can examine dozens of factors using half as many runs. There are caveats; your design should be aproximately four times larger than the number of active factors (usually, there’s only two or three significant factors, but you can’t tell them from the trivial many.) The Pareto and Effect Sparcity principles usually play a significant role in making these designs fruitful, however.

    0
Viewing 8 posts - 1 through 8 (of 8 total)

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