iSixSigma

GRR acceptance criteria

Six Sigma – iSixSigma Forums General Forums Tools & Templates GRR acceptance criteria

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #53613

    Ng
    Participant

    Hi,

    I am thinking whether is there an alternative approach to define the GR&R acceptance criteria, besides the guidelines established by AIAG (<10% ~ acceptance, 10%-30% ~ marginally acceptance, >30%- unacceptable)?

    Kindly advise, and thanks.

    0
    #190865

    Quesnel
    Participant

    AIAG – MSA 3rd edition clearly defines criteria for the GR&R acceptance criteria as a standard.

    Why would you want to redefine the guidelines? If you clearly know that the design or process is able to handle more tolerance, then you should redefine the USL & LSL to a more reasonable spread to allow for process to fit within limits.

    This is just my suggestion – I am sure there will be others.

    Thanks.
    Eric

    0
    #190871

    Ng
    Participant

    Hi Eric,

    Appreciate your thoughts.

    Redefine the USL and LSL is indeed a good idea to accommodate the measurement errors, and the idea is precisely works well for %tolerance (P/T ratio).

    However, the USL and LSL idea might not work for %StudyVar, as %StudyVar only taking Total variation into consideration instead of tolerance. It is likely to fail the GR&R (%StudyVar) if the part-to-part variation is relatively ‘low’ compare to measurement variation.

    Any advice will be great.

    Thanks,
    regards

    0
    #190872

    Mikel
    Member

    Do you understand the criteria? Maybe that would be a starting point vs. changing the criteria.

    0
    #190876

    Quesnel
    Participant

    I agree – that is why I made the statement:”If you clearly know that the design or process is able to handle more tolerance, then you should redefine the USL & LSL to a more reasonable spread to allow for process to fit within limits.”

    The key operative word being “IF” – you cannot just indiscriminately change a process you know nothing about.

    0
    #191417

    Brooke
    Participant

    I just took a new position in a company where the customer STA has driven the use of ANOVA method of GR&R and everyone is used to other methods. The results on the GRR report from our software dont specify %GRR as a specific output, it just gives %of variation and standard dev etc. What is the criteria to interpret ANOVA vs other methods which show EV/AV and TV as a percentage of tolerance?

    0
    #191421

    khatri
    Participant

    i think you should check the below three standards for Annova GR&R

    1. No of distinct categories should be >= 4
    2. Gage R&R as a %age of tolerance
    less than 10% – accept
    between 10 – 30% – can take it with caution, again depends upon the nature of business &
    >30% reject
    3. Gage R&R as a %age of contribution should be less than part to part contribution.

    hope this help

    0
    #191431

    Brooke
    Participant

    Thanks, nowhere on the Minitab, AIAG or PLEX software options does it report R&R as % of tolerance, which is what I’m used to seeing.

    0
    #191432

    Jephcott
    Participant

    Something often missed (that may account for lack of detected difference between parts) is to make certain your sample (presumably 10 parts – yes?) includes parts from the limits of what your process produces … ie the smallest and the largest.

    Often, people take consecutive samples which are in fact more likely to be similar (short term variation) than random samples taken over a period of time.

    Note this does not mean you have to make an adjustment (we’re not talking about the specification limits … it all about what your process delivers under “normal” conditions).

    0
    #191433

    Severino
    Participant

    You could look at Wheeler’s “Honest Gage R&R” study if you want an alternative.
    Presumably you are asking about alternatives because you are failing the current requirements. The suggestion about ensuring that your parts represent the normal variation in the process is a good one. I’ve seen more than one of my QE’s have a GRR fail on them because they had pieces that were autocorrelated.

    Your problem may also be that you don’t know how to break down the problem to improve your measurement system. You need to understand which R is the larger contributor to the problem. What type of measurement are you performing and what exactly is the issue?

    0
Viewing 10 posts - 1 through 10 (of 10 total)

You must be logged in to reply to this topic.