iSixSigma

When GRR is within 10% to 30% – Acceptance?

Six Sigma – iSixSigma Forums Old Forums General When GRR is within 10% to 30% – Acceptance?

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #30627

    Pipkin
    Participant

    Hi all.1) In our MSA program, we find a question on the GR&R series. While we use a Caliper (Tolerance0.01mm), to measure a 1mm+/-0.1. We find that the GR&R value is 26.58%.Here is some info: 10 samples(Same Cavity,take sample per each 2 hrs),3 times measure and 3 guy measure.Bar X of A guy=1.007,Bar R=0.007Bar X of B guy=1.004,Bar R=0.007Bar X of C guy=1.005,Bar R=0.006BARBAR X=1.005 , Rp=0.047 , BARBAR R=0.0067 , BAR X diff=0.003—EV=0.02 / AV=0.007 / R&R=0.021 / PV=0.076 / TV=0.079 /  %EV=25.35 / %AV8.86 / %GR&R=26.58 / %PV=96.2—Also we use same caliper for Cpk study, the result is more than 1.33. But our Bar Range is so small, if is small, than the GR&R% will bigger. How can be slove it?In this analysis, we make sure no enviroment,human error,wrong measurement influence.2) In GR&R is 10%~30%, they say Based upon importance of application, cost of gage, cost of repair. Which kind of thing i can do to prove this case is acceptance? Can u explain.Unfortunably, auditor,consult also can not answer above problem. Hope i can know the answer this time.Thanks you so muchJack  

    0
    #80017

    Chris Seider
    Participant

    It is hard to fully understand your output, but I suspect that the gage study did not get samples that take up 80% of your typical process range or specifications.  This typical mistake will inflate your gage r&r results causing one to improperly question a good gage.  Of course, I’ve seen many a beginning BB or a non-trained individual pick 3 wildly different products and get a great gage relative to process variation–which would be misleading unless one was ONLY testing the ability to select one product from another.
    Hope this helps.

    0
    #80022

    Jim Athon
    Participant

    Jack,
    A good way to check if your samples were chosen properly is to perform the GRR in Minitab (assuming you are) and run the Percent Tolerance option.  This can be done by filling in the “Process Tolerance” field in the GRR menu pop-up under the options tab.
    Stat>Quality Tools>Gage RR>Options>”Process Tolerance” field
    Put in your USL-LSL for the total tolerance.  (aka specification range)
    You will get additional output in the session window under the heading % Tolerance.  This tells you (roughly) what amount of the spec rance is consumed by the gage, and gives you and indication of whether the gage can tell good parts from bad. 
    Use the same30% danger rules. 
    It is quite possible to have a gage that is works fine but shows high % study or high % contribution, esp when the process is very capable.
    It then depends on what you wnat to do with the gage.
    Hope this helps.
    Jim Athon
    The Athon Group, Inc.

    0
    #80031

    Mike Carnell
    Participant

    Jack,
    I’m the same as C Seider and didn’t really understand your data.
    Just a couple quick comments. You said you took the human error and environmental error out of the test. When you are using Calipers I am not sure how you tke people out of the equation. The variation in your data looks like it is out 3 decimal places which you can pick up depending on your environment (that is why things going into a metrology lab sit). If you parts came straight off a machining operation it is worse because of the heat from machining. Just a thought.
    Jim’s advice was good advice. Actually so was C’s. If you use the % tolerance you can actually quantify the effect of the measurement error and if your Cpk is 1 or less it is the only thing that makes sense. Just remember that when you have a % tolerance you lose it twice – once off each end. If your R&R is 10% you lose 20% of your workable tolerance.
    Good luck.

    0
    #80036

    Pipkin
    Participant

    Hi allAbout the GR&R problem, i still confusedWell i posted there, plz see whether have problem on it.http://us.f1.yahoofs.com/users/b2d0288b/bc/GR%26R/GR%26R.xls?bckdEv9Ac7zCj7xmAlso whats is the GR&R diff on PArts Tolerance,Combine R&R and PV, and Process 5.15sigma. I also mentioned on the fileHope you can fully understand what’s problem onThanksJack

    0
    #80037

    Jim Athon
    Participant

    Jack,
    The link you provided I can’t get to work (broken?)
    Your gage is probably fine, you selected your parts incorrectly for the study.  The GRR number you got was high because there was so little part-to-part variation.  When I estimate the % tolerance, that is, how much of the spec range is being consumed by the gage, I get ~10.7%, using the rbar/d2 method to get an estimate of sigma, and then using the % tolerance formula (USL-LSL)/(5.15*sigma).
    When you have a high “% study” and a low “% tolerance”, it means 1 of 2 things:
    1. Your process is very capable, eg 6 sigma or higher, (little part to part variation) and the gage has problems differnetiating part to part varition.  However, the gage can tell good parts to bad parts.
    2. You chose the samples wrong (sounds like this is your problem, hence the small part-to-part variation).  Since you chose parts off of the same cavity, you neccessarily eliminated some variation from the other cavities/dies/machines, and over a short span of time could have meant a single lot as well.
    As I mentioned, use the % tolerance as a check to insure the above doesn’t occur.
    By the way, what is the historic capability of the process in question?
    Good Luck!
     

    0
    #80039

    Pipkin
    Participant

    Thanks Jim,Mike,C seider
    I finished the GR&R result, but maybe something error posting process.
    Well, i calculated the Process 5.15 Sigma: TV=5.15xSigma=0.03047
    Also i made three kind of percentage Variation, there are
    Parts Tolerance: %RR=10.62%
    Combine RR and PV: %RR=27.06%
    Process 5.15Sigma: %RR=69.73%
    But which one should i use for GRR analysis, i confused on it.
    Also i will go to china plant(Production Line) for GR&R study again. And will follow yours valued idea to access. Thanks you so much.
    But would you mind send me a email, let me to send a GRR result to you. To see what is my stupid problem clearly. Thanks Alot.
    My email address is : [email protected]
    Thanks so much

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

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