iSixSigma

Max tolerance

Six Sigma – iSixSigma Forums Old Forums General Max tolerance

Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #48609

    Darrell Nemeth
    Participant

    I have a question that maybe someone can shed some light on. I have a produst that calls for a .004 MAX on a balance characteristic, I have product that measures.0041, I am told that it is in specification and I say it is not, who is correct?

    0
    #164490

    Taylor
    Participant

    What is the % accuracy of the balance? And does .0041 fall within that accuracy scale? Where I’m going with this is you may have enough Bias in the balance to pass product at .0041
    Just a Thought

    0
    #164492

    Darrell Nemeth
    Participant

    I have not completed any type of cabability on this as of yet, but what I am saying is that my boss states that because the print calls for a 3 decimal point callout ( .004 MAX) we do not have to carry it out to 4 places and that anything over .004 would not have to be reported, I ‘m not buying this frame of thought as .0041, .0042, .0043 are all out.

    0
    #164493

    Taylor
    Participant

    Darrel in that case you are correct.  Ask your boss if he Considers .0035 to be the same as .004, technically that is what he is implying.

    0
    #164495

    Darrell Nemeth
    Participant

    Thanks………………..

    0
    #164496

    Brandon
    Participant

    Darrell, this may be an issue of symantics. The key question is does ANYTHING over .004 impact the performance of your product?
    I would recheck the spec limits to determine what my true USL is. And however many decimals are needed to assure you are in spec is what you must measure to.
    Don’t forget, the true objective is to produce acceptable products.
    Otherwise you’re just playing games with numbers.

    0
    #164505

    Ward
    Participant

    Do you know if there is any effort to managing to the target rather than to the spec? As Deming said, “Specifications are the starting point, not the ending point”.

    0
    #164507

    Sloan
    Participant

    Also, according to Taguchi, it is not realistic to believe that a part barely inside the spec limit has little or no risk of failure or that a part barely outside the spec limit has a maximum chance of failure. However, to minimize the risk of failure and maximize cost efficiency you should work to get your product as close to the target as possible. Taguchi’s premise is that the further the product’s characteristic varies from the target spec the greater the loss, even if the products are still within the outer limits.

    0
    #164514

    annon
    Participant

    As always, start with the customer.  Ask for the operational definition, unit of measure, and target or tolerance (if applicable).  It should resolve itself here. 
    If not, calculate the process capability using first .0040 and then .0049 as your USL.  Present your findings in financial terms.  What is the cost of prevention associated with maintaining a spec of .0040 verses the potential external failure cost associated with shipping units up to .0049?  Now you can frame it in terms of loss avoidance
    Ask your manager if he wants to assume responsability for this loss if his initial position proves incorrect.
    Good luck.

    0
    #164846

    AlexanderSO
    Participant

    Hi Darrel, I suggets that before starting an argument with your boss check the end user, for this you can use historical data that tells you if the parts over the 0.004 spec value performed well or not and if they did maybe tou need to validate the upper limit of the spec.  Also if the extra digit in the measure really helps you to control the quality of the product or is irrelevant to the end use performance.
    On the other hand if they didn´t perform as expected you can explain with data on hand your point.

    0
    #164849

    Darrell Nemeth
    Participant

    Thanks.

    0
    #165300

    Stan Alekman
    Member

    Test results are usually reported to the same number of significant figures as in the specification. In some industries, for example, pharmaceuticals, test results are reported to the same number of decimal places as in the specification.
    For very good reason, tests are designed to find results with a greater number of significant figures or decimal places than in the specification. The test result is then rounded to the specification, as in the case for pharmaceuticals.
    In the case you describe, rounding an observed and out of specification test result to the requirements of the specification is known as rounding into conformance. It is acceptable to some companies and unacceptable to others.
    Since precision error accompanies all measurements, replicate measurements are often made and the average rounded to the requirements of the specification. In this case, depending upon the size of the precision error (standard deviation), when the object measured is at or very close to the specification limit, were it to be measured many times, the object would meet specification some times and fail specifications some times. This is known as type 1 error. It is controllable by the number of replicates and the precision of the test method.
    To avoid this problem, some companies employ internal release limits to minimize the frequency of type 1 error. This is known as guardbanding in metrology.
    Check the ASTM web site for the published standard that addresses comparing observed data to specification.
    This reply to your question is based on arithmetic. The other replies speak to other valid approaches.

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

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