iSixSigma

typical sample in r%r study

Six Sigma – iSixSigma Forums Old Forums General typical sample in r%r study

Viewing 11 posts - 1 through 11 (of 11 total)
  • Author
    Posts
  • #33535

    ROSS
    Member

    How can we get the typical sample in r%r study? As we all know, it will do great effect in r%r result. Thank you!

    0
    #90908

    ROSS
    Member

    Who can help me? How can I get typical samples fot r%r study.

    0
    #90909

    Reinhold Toerek
    Participant

    Samples used for Gauge R&R should represent the variation you expect to see in your process.  It is typical to chose 10 samples with variation covering your process width.  The purpose of the Gauge R&R is to determine  whether or not the gauge you are using is good enough.  The measurement error needs to be small enough that you can distinguish differences within your tolerance band and also within the process variation. 
    I

    0
    #90911

    ROSS
    Member

    Thanks for your feedback, we all know the sample should represent the variation for your process, but know can we do it?

    0
    #90913

    Hemanth
    Participant

    Thats a very good question..it is indeed difficult to characterise variation in your process even before you have done a gr&r. I use the tolerance band as my guiding light. I take parts which represent the whole tolerance band to start with.
    Another way (in cases where you have a very capable process) we handled this issue by randomly taking pieces from a days production. That way we atleast had a good sample.
     

    0
    #90921

    Doc
    Participant

    Though we all know we’re supposed to do the gage R&R before the capability anlaysis, just 10 samples will provide a very poor estimate of variability. Through local simulations I found you need 50 to 100 samples in order to get a good estimate of the standard deviation. On top of that, most users hand-select those samples, so they are far from random.
    This methodology is OK for assessing R&R, but it is terrible for assessing process variation – a critial part of the .
    If at all possible, try to obtain an estimate of your process sigma from a larger sample (n=30, 50, or preferably 100 or more).
     I know Minitab software allows users to “inject” a “historical” sigma into the anlaysis. Hopefully other software does that as well. If not, the final %-GR&R metric using the historical sigma is easy to hand calculate.

    0
    #90923

    Chris Butterworth
    Participant

    Tony,
    Take a large sample of parts and measure them all. Build a tally sheet of the dimension you are concerned with and then simply take the lowest, highest and a couple values in between. Those four or five are the parts used in your gauge R & R study. This is adequate.

    0
    #90926

    Reinhold Toerek
    Participant

    Let’s not confuse Gauge R&R with process capability.  I have found this to be a common point of confusion.  Process capability has nothing to do with Gauge R&R.  Once you have determined that a gauge is capable (repeatable and reproducable with an acceptable measurement error), then that gauge can be used to access process capability.  Obviously, the larger the sample size, the more accurate the assessement will be. 

    0
    #90928

    Doc
    Participant

    No confusion here.
    The denominator of the %GR&R metric IS an estimate of process variation – the very same variation that is used to assess process capability. In most GR&R studies it is a very poor estimate. A poor estimate of the process variation (due a small or poorly selected GR&R sample) could have a dramatic affect on the %GR&R metric.
    If the process variation is underestimated, the user might spend unnecessary effort to improve the measurement system’s resolution. If the process variation is over estimated, the user might think their measurement system is OK, when it is not. Which is worse??
    My point was that, if available, a large-sample estimate of the process variation should be when calculating the %GR&R metric.
    Yes, to do this a user would essentially have to do the process capability study BEFORE doing the GR&R.   Radical eh??
    The user would then use the the results of the GR&R to assess the validity of the capability study. If the measurement system proves inadequate, then it should be fixed/improved and then a new capability study must be completed.
    It may seem backwards from what is usually taught, but it is the only way to ensure a valid MSA, which is necessary to ensure a valid process capability assesment

    0
    #90929

    Reinhold Toerek
    Participant

    I agree with your statements.  However, to conduct a MSA with large sample sizes proves to be very costly.  The Gage R&R indices of % Tolerance and % Study would give information as to whether or the not gauge tested is adequate for each case.  If  % Tol is good and % Study is not good, one conlusion would be that the process is running much tighter than the allowable tolerance.  Maybe not a bad thing.  Maybe you don’t need to measure that close.  If however the % Study is ok, and the % Tol is not, then the process is running wider than the allowable tolerance and the gage may be able to be used to measure differences in the wide varying process.  Once the process variation is reduced, the gauge may no longer be useful. 
     
    In any event, I do agree with your comments.  My only point is that one can determine quite a bit of information with smaller sample sizes.  Gauge R&R’s are very time consuming to conduct and thus are very costly.  Process capability assessment can be easily done with data once the R&R is done. 

    0
    #90931

    ROSS
    Member

    Hello!
            I have got great information from your feedback. But, I still do not know how to get typical samples in r%r study clearly, how do your get your samples in your r%r study to make it typical and present process variation enough? Thank you very much!
        

    0
Viewing 11 posts - 1 through 11 (of 11 total)

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