iSixSigma

Cpk calc when outside the LSL to USL Range

Six Sigma – iSixSigma Forums Old Forums General Cpk calc when outside the LSL to USL Range

Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #37391

    Fontanilla
    Participant

    Please forgive what is probably a very amateur question but:
    I have a process which in an ideal world would have a mean within the a LSL of 60 and a USL of 90
    However the mean at the moment is a measly 21 with a standard dev of 11.
    Hence, I am getting a negative result for Cpk which is not very helpful.
    Is this the best way of analysing variance in this case or should I not even be reporting this performance becuase it is so awful??
     
    All help appreciated!!
    Dan

    0
    #110037

    RubberDude
    Member

    Your process mean is way off target and the variation is excessive as well (Cp = 0.45).  Does any of your output fall within tolerance?  From your figures, it appears it does not.
    Is this a characteristic that is critical to customer quality?  Are you reporting this to prove it needs to be “fixed” or just to be reporting?

    0
    #110038

    Fontanilla
    Participant

    I guess we are just going to use this as a reporting figure at this stage – we all know that our process is way off base (none of the results are within tolerance), but I was just trying to plug in the figures to Cpk in order increase the effect of this result as a ‘burning platform’. I guess in this case the burning platform is that results are so bad we can’t even get a good cpk measurement!!!
    To be honest at the moment our measurement is not great, so I think as the process matures, we will be able to create a revised measure (based on this) that will be more aligned to the CTQs.

    0
    #110039

    RubberDude
    Member

    Years ago I plotted Cpk and Cp values on a monthly basis during improvement projects as a “moral booster” to say “hey… we’re getting better!”  It did give the improvement team something to cheer about when we were seeing a trend towards improvement.  And at the same time gave them incentive to keep improving.
    If you are on this same type journey, I would suggest you keep the figures to show improvement over time.
    Good luck in your endeavors!
    RubberDude – Certified Grand Master Holiday Inn Express Black Belt

    0
    #110040

    Dog Sxxt
    Participant

    Two posibilities here,
    1. The designers set the spec arbitrarily without engineeing and process basic.
    2. Your process is damned bad.
    In either way, this kind of Cpk will never get through the qual build stages in Japanese companies. The program either be dropped or the design team has to go back to basic and improve Cpk by hook or by crook before it can be released to production. 
     

    0
    #110044

    mjones
    Participant

    Dan-
    Dog & RDude are right on with their cmts.
    And you are doing the right thing too. Cp & Cpk are great metrics to track how you are doing. Presuming (as Dog suggested) that this is a useful measure with valid requirements, you clearly have a need and opportunity to improve. Don’t dump it just because it is so bad right now. Facts are facts.
    Looks like you are blessed with opportunities to celebrate…as you improve. This could be a very useful and memorable lesson for you and your team as you work together to overcome this daunting problem. Once you do that, this will be your story, like RDude’s story, which is burned into peoples’ memories to be told again and again, with learning further multiplied. Take advantage of this opportunity!
    Only thing (other than encouragement and cheers) I can add: it may be useful to put a histogram on a chart showing were you are vs. where you need to be, and adding new histograms (with dates) as you improve. A “picture” of what the Cp & Cpk numbers mean, and of your progress over time can help understanding.

    0
    #110238

    Dan Benig
    Participant

    Hi Dan, Assuming your spec limits are correct, perhaps your process is operating under the influence of an assignable cause. You need to figure this out somehow.The 6 sigma spread is currently wider than the spec width which means that you not only need to work on driving the process average towards the spec center but also on reducing the standard deviation.Also, take a look at the shape of your distribution. If it departs significantly from a normal distribution, a Cpk value is actually meaningless. Cpk should not even be computed until a process is in a state of statistical control. To get there, you need to be convinced that special (assignable) causes are gone and only random causes are present.Regards.Dan Benig

    0
    #110264

    DHogue
    Participant

    The only thing a capabaility analysis will tell you is if the process is or is not capable. It tells you that only if the system is stable/in statistical control. You need to check this stability issue. Also check your data collection R&R if you haven’t already.
    I would report that the process is “not capable” at present.  It may be that whatever machinery/SW/whatever generating your results may be completely inappropriate for the process.

    0
    #110297

    melvin
    Participant

    If objective of calculating Cpk is to establish sense of urgency, it sounds like you have one.
    However, there are a critical issue you raise here – the adequacy of the measurement to characterisation and measurement system analysis.  Spend the time making sure you have these nailed down before moving too much further in process aanalysis.
    Also, consider use of a control chart (e.g., I-MR) to develop an understanding of process behaviour.
    At the stage of looking at behaviour of Y – there can be some verey good clues in the simple graphics.

    0
    #110304

    rman84
    Member

    Hi Dan,Your Cp value, which is about 0.45, is an indication how well your process will perform with respect to your specification limits assuming your process is centered. It is obvious in your example that your current process is just too wide for your limits. This could be due to several reasons – maybe the specs was defined arbitrarily or your process variation is really out of control. Here’s one way I would approach your problem:Assuming the spec limits were defined correctly, given the Cp of 0.45. I will focus my efforts first in reducing my process variability. This could be done thru characterization / optimization activities. If your process potential (Cp) is already acceptable (Cp>1,5), then work on centering your process which will then improve your Cpk. 1,5), then work on centering your process which will then improve your Cpk. 1,5), then work on centering your process which will then improve your Cpk. I would say that reporting these numbers is still a good way to show how your process behaves with respect to your limits and what course of action should be taken (variation reduction / process centering).Regards.
    Ronnie

    0
    #110321

    JMR
    Participant

    Just a suggestion.. go to http://www.minitab.com and download the demo version. its fully functional.
    then do this. enter your data accordingly with the customer given specs. graph. then observe your existing data. do a descriptive statistic analysis and acquire your mean. observe your SDev and apply to your existing data to create your own USL and LSL. graph. minitab has an excellent function that will physically display your current data in relation to the existing customer spec limits. like they say, “a picture is worth a thousand words”. currently i am in a similar situation on one particular product. remember…the customer is not always correct. therefore i have submitted recommendations that our customer and my company hammer out updated specs that are realistically achievable (in the neighborhood of Cp = 1.33 to 1.67). of course part functionality also plays a role and in this instance it is on my side. many times the customer does not monitor their specs expecting the supplier to provide what is best. hope this helps.

    0
    #110322

    JMR
    Participant

    good answer! it is never a good idea to hide your findings. if the process is out of control then chances are the customer or end-user has already had problems and rejects are on their way back to you. as “dhogue” stated, tell it like it is and focus on what the problem may be after first properly defining the problem. good luck! 

    0
Viewing 12 posts - 1 through 12 (of 12 total)

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