iSixSigma

Is a Capability Study Appropriate?

Six Sigma – iSixSigma Forums General Forums Methodology Is a Capability Study Appropriate?

Viewing 21 posts - 1 through 21 (of 21 total)
  • Author
    Posts
  • #53999

    GL
    Participant

    I have a normal data set of measurements taken on the placement of a part using a ‘window type’ fixture that bounds the location so that it can never below or above the specification limits. The assembly goal is for the part to fall within the window not necessarily for it to fall on the target dimension. Is it appropriate to perform a capability study on this type of data set?

    0
    #192674

    MBBinWI
    Participant

    @glewis2012 – Greg: Rather than answering your question directly, I’m going to ask you to look at the equation for Cpk and how the variables are derived. When you look at those data items, and the equation for Cpk, you should be able to answer the question yourself.

    0
    #192675

    GL
    Participant

    Thanks. I just re-reviewed the Cpk equation however it’s still not clear to me if it’s appropriate in this case.

    0
    #192676

    MBBinWI
    Participant

    @glewis2012 – how much of the range (window) does the actual data cover?

    0
    #192677

    GL
    Participant

    The data is accross 47% of the range. (data max – min = 1.493, spec is +/- 1.6)

    0
    #192679

    MBBinWI
    Participant

    @glewis2012 – so, the actual data is not being constrained by the “window fixture.” Thus, it’s almost like it’s not even there.

    0
    #192680

    Mike Carnell
    Participant

    @glewis2012 Is there a business reason (financial) for the target? How did they decide on the width of the window?

    If you are staying inside the window and there is a financial advantage to staying on target then switch to Cpt.

    0
    #192684

    GL
    Participant

    MBBinWI: Good point.

    Mike: There is no business advantage for the part placement to be on target in this case

    0
    #192685

    Mike Carnell
    Participant

    @glewis2012 In that case I would leave it alone and let it run. This that goal post stuff. You get the points as long as it is between the posts.

    You might want to run the data just incase someone asks it you looked.

    Just my opinion.

    0
    #192690

    MBBinWI
    Participant

    @Mike-Carnell – Cpt? Educate me please. The only Cpt’s that I am familiar with I had to salute (’til I became one!).

    0
    #192694

    Thomas Whitney
    Member

    All capability analysis comes down to What Question are you trying to answer. If I need a target more than anything use Cpm, if I just need to stay within spec use Cpk. But, what question are any capability values trying to answer?

    Do I need to incur the cost of inspection seems a good one. How many parts pass this go-no-go test sequentially. If it has been 10,000 stop the test!

    What is the customer tolerance for defect escapes. No it is not zero. But if you believe your process can be under the tolerance; Stop the Test!

    In my early days as a quality engineer we stopped 90% of the testing in our factory because we could demonstrate with data our capability was sufficient enough to drop the cost of testing.

    Just my ranting…

    0
    #192695

    GL
    Participant

    @twhitney99: Thanks for the comments. In this case we use capability as one of the measures to qualify and release an assembly process (automotive PPAP). And we do systematically “step down and out” of inspection processes like this. My original question was just whether or not capability really has any significance if the target is not important and the process is poke-yoke’d so that it almost eliminates defects from being created.

    0
    #192696

    Thomas Whitney
    Member

    If the question you are trying to answer is not “How capable is my process?” , then capability measures are not needed.
    If your inspection methods answer the question “Is my customer protected from defects?” then they are sufficient.

    0
    #192698

    Mike Carnell
    Participant

    We are going to try this response one more time and see if my ISP can figure out how to keep me connected long enough to post it.

    @MBBinWI My apologies on the Cpt. Cpt is an old term that became Cpm.

    @Twhitney99 It is about time you showed up here and shared some of your experience. Just remember that most factories thes days have indoor plumbing and electricity.

    @glewis2012 Be careful when Whitney talks about when he was a quality engineer. The process could easily have been whittling spokes for wagon wheels.

    0
    #192699

    Gary Cone
    Participant

    @Twhitney99 Hey thanks for showing up here, this place actually needs someone who is passionate about fixing the system.

    @Mike-Carnell I didn’t skip a beat with Cpt, it looked right to me.

    Just a small history lesson here, there was a group of us who derived what has come to be called Cpm at least two years before the article in the Journal of Quality Technology (I think July 1989) that named it. We called it Cpt. The group included me, Mike and a guy named John Lupienski (famous in Western NY for good and bad reasons). We did it after reading Quality Engineering by Taguchi and challenging why his computations were not right. We found his standard deviation used target, not x-bar. We called his standard deviation s sub t (sorry hard to do subscript on here) meaning standard deviation with respect to target, and Cpt meaning capability with respect to target. Taguchi thought that a value of 1.33 or higher was all that was needed and in the subsequent 2+ decades I’ve not seen a case where he was wrong.

    0
    #192700

    Mike Carnell
    Participant

    Twhitney99 Does this mean we are going to have to listen to your “Red X” stuff? Again?

    @garyacone I am not sure if it is age or resistance to change but some of the old terms just don’t go away that easy.

    0
    #192702

    Thomas Whitney
    Member

    Oh yea, and pink X’s too. In fact, instead of lean I like JIT and all kinds of other whittling spoke wheel terms!

    0
    #192703

    MBBinWI
    Participant

    @twhitney99 – Tom: I’m going to jump on the bandwagon and welcome you as well.

    @Mike-Carnell – I can honestly say that during more than 14 hours of driving time, Tom repeated the stories only twice (but, I’m sorta dense, so probably thought it was needed to actually penetrate).

    @garyacone – please don’t take the fun out of my pulling Mike’s chain (although the story that I got was that the t stood for Taguchi).

    Finally, @glewis2012 – You didn’t start this off asking if a capability study was necessary if a target wasn’t important and the system was PY’d. You asked if it was appropriate to perform a capability study. My answers to you were to get you to evaluate whether the data was truly normally distributed or being truncated. If truncated by the “guards” then it may not have been appropriate (although even in that situation a capability study is highly tolerant of truncated data). When you responded that the distribution was only 47% of the overall allowable distribution, I suggested to you that it was almost as if the guards weren’t there (the hint being that they were not needed, and also that there was no problem with doing a capability study). Assuming that your values are reasonably centered (and with that low of a range of variation if they are not, then someone ought to be smacked upside the head), then you have a process with a Cpk approaching 2. Why would you be messing around with this process? Don’t you have more important things to deal with?

    0
    #192706

    Mike Carnell
    Participant

    @twhitney99 Somehow that pink X just isn’t a surprize. Still glad you showed up.

    @MBBinWI it is easy to miss the third time around. By then your eyes are rolling back in your head. Next time jus look at him and say “This is the third time you have told me that. Do you think I am stupid?” Use a heavy German accent. He will have flashbacks and will stop doing it.

    You were right on the 47%. I read it as 74%. I guess it is time to head for the optomologist.

    0
    #193147

    Hi All.. I’m new to Six Sigma

    Pls help me understand –
    a) While deriving process capability, for discrete data, we go with DPMO method and for continuous we adopt CP,CPK method. If i find the cpk just about 1 what steps to take or how to increase my CPK.
    b) If the data is not normal, what do i do ?

    0
    #193148

    Chris Seider
    Participant

    It depends…..
    If you have a practically significant defective based on actual data, don’t fret about finding a distribution. Just reduce the % defective produced out of the process since that’s the real issue.
    Others will tell you to find a distribution or use Weibull as the default and determine the Ppk but that’s not as important as finding the % defective from a process capability study and driving that number down with your process improvements.

    0
Viewing 21 posts - 1 through 21 (of 21 total)

You must be logged in to reply to this topic.