iSixSigma

Adjusting Specification Due To Gage ERROR

Six Sigma – iSixSigma Forums Old Forums General Adjusting Specification Due To Gage ERROR

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • #30184

    JoeB.
    Participant

    Someone has said to me the other day: ‘Since you have a  Gage R&R  error of 12%, therefore you should reduce your specification by 12 % accordingly.’  Don’t be silly, I said! What say you learned folk? Your wisdom on the subject please.
     

    0
    #78390

    Mike Carnell
    Participant

    JoeB,
    It doesn’t make any sense to me either.
    They could be refering to the concept of guardbanding which would take +/- 12% on either specification limit and make it an area that the results are questionable. This would have the net effect of reducing your tolerance by 24%.
    This may be an odd suggestion but someone could fix the gage or take mutiple measurements if they need the error reduced. Actually reducing the spec isn’t going to resolve anything.
    Good luck.

    0
    #78518

    brennan
    Participant

    Thanks
    What about  the rest  of  SS smurfs and smurfets in the six sigma kingdom, any opinions?

    0
    #78519

    James A
    Participant

    Just goes to show that Co-Co the clown’s offspring have pervaded all areas of industry.  I would say that there is more than a cultural issue at stake with this individual.  Intensive aversion therapy using electric cattle prod may be required.  Extended education in common sense and what gauging is for may also prove helpful.
    Barmy.
    James A

    0
    #78521

    Ed Van Haute
    Participant

    Fiddling with a spec will NEVER have any effect on a process output. Gages themselves do not have any effect on a process output, but are used to verify the output. I think the questions you need to ask and respond to are (not necessarily in order):
    1. Is the gage tolerance sufficient?
    2. Are the operators using and reading the gage properly?
    3. Is the gage calibrated? 

    0
    #78525

    Mikel
    Member

    Mike,
    If there is an area of uncertainty where you are not sure the product is good, even though your gauge gives you a reading inside the spec; you wouldn’t guardband it?
    Of course the guardband the spec by 12% is wrong, but if my gauge can say good and my customer’s can say bad and it is predicted by the GR&R, you have to compensate for it.

    0
    #78528

    O’Connell
    Participant

    Joe,
    When your in the Improve phase you need to adjust your spec limits using the 12% tolerance window you discovered in Measure.
    Example:
    Measure USL = 200, LSL=100, 12% tolerance.
    Improve – What are the new spec limits you will use for your transfer function? 200-100= 100, 12%(100) = 12/2 = +/-6, USL=200-6=194, LSL=100+6=106.
    Good luck.

    0
    #78529

    Ron
    Member

    This issues is common in industry. You say good the customer says bad.
    Before doing anything you need to study the entire measurement system including the operators at both your facility and at your customers.  I did a BB project on aircraft altimeters and found that the customer was correct.  I needed the data to prove it to our supplier who thought they did everything correctly.
    You did not give enough detail for  a more in depth answer, however, trying to change the results bt some “Fudge factor” is just wrong.
    Have you performed adetailed Gauge R&R on your process?  If the gauge is not capable get a new gauge if it is.. go to your supplier or have them send you a guage that they accept.
     

    0
    #78537

    denton
    Participant

    Best practice is a “can’t tell zone” that is two Effective Resolutions wide, centered on both the USL and LSL.  If something falls in that zone, you acknowledge that you can’t tell whether it is good or bad.  That gives you three output states:  good, bad, and can’t tell.  As you improve your measurement system, the “can’t tell zone” gets smaller.  If you go the other way, eventually the two “can’t tell zones” merge, and you can never tell if something is good.

    0
    #78542

    Scott
    Member

    I don’t have a lot to add, but I wanted to go back to Mike’s response.  First I think that Mike was correct.  The main point for me would be to determine if the error is mostly operator or gage and try to improve it from there.  If that is not possible, take Mike’s advice and try multiple readings.  Personally, I don’t like ‘gray’ areas. 
    Hope this helps a little or turns the direction towards R&R improvement.
    Rick

    0
    #78543

    brennan
    Participant

    I think we are getting warm. But the variability in the answers is too great. Six responses with 6 different opinions. What is the statistical or mathematical rational? I am meeting that someone next week, so I want to load up my six shooter, but need your help.

    0
    #78544

    brennan
    Participant

    I think we are getting warm. But the variability in the answers is too great. Six responses with 6 different opinions. What is the statistical or mathematical rational? I am meeting that someone next week, so I want to load up my six shooter, but need your help.

    0
    #78548

    Scott
    Member

    I know that you don’t have much time, but one of the best references on gage / machine qualification is “Statistical Procedures for Machine and Process Qualification” by Dietrich and Schulze.  It’s available through ASQ press, but again, I realize that you don’t have time for that now.  As I stated earlier in the thread, I basically agree with Mike.  The reference I mentioned follows the same advice, but does give an option.  First, their criteria for an exisiting measurement system or gage is 20% or less and 10% or less for a new system or gage.  If the gage is greater than 30% it should not be used.  If it is between 10% (20% for existing systems) you should work to improve it or possible take multiple readings , standardize measurement location, etc.  As far as reducing the specification, they prefer the use of an error chart to track error of each operator.  Again, this is assuming that most of your error is from the operators.  The error chart would code their reading to zero based on their capability.  If you want more info from the book I could send you a fax.  My email is [email protected].  One other item.  If you do decide to adjust your specification, remember that the R&R% is based on total tolerance.  If your level is 12% then you would only remove 12% from your total specification (6% from upper and 6% from lower for a bilateral spec).  I don’t believe that you would want to double dip and take away 12% from both the upper and lower specification.
    Hope this helps a little.  It sounds like you are under the gun next week.
     
    Rick
     

    0
    #78551

    Mike Carnell
    Participant

    Stan,
    Sorry about the delay. We were at a conference in Toronto. The city is great – it was difficult to stay inside at night.
    I am not sure I understand your comment/question. The % tolerance regardless of how large or small creates a grey area around the spec. You really do not know what the actual reading is (regardless of how good the gage is – it is a single point measurement).
    If I make a measurement that is within 6% of the spec limit and the % tolerance is 12% I am in that area and I really do not know where the actual reading is. If it is actually off by 12% – it could be outside the spec – now I am shipping bad product. If it is 6% outside the spec I can be rejecting good product and it is purely a function of the gage. You can take multiple measurements and reduce the error by increasing the value of n in the standard error of the mean calculation. Now I can ship or reject with more confidence.
    I wouldn’t reduce the spec – but I would work on decreasing the the % tolerance. As far a just purchasing a better gage – not all gage error is a function of the gage.
    It would help if the gage manufacturers would join the rest of us (since we are their customers) and start doing some of these studies themselves so you can get it when you purchase the gage. One way to get around their lack of attention to customers is make payment for your gages subject to acctabilty of a MSA at your facility.
    Good luck.

    0
    #78552

    Mike Carnell
    Participant

    Brian,
    Maybe I am not understanding your response.
    If I have 12% of tolerance and I reduce the spec width and do not do anything to improve the gage the calculation becomes the same amount of gage variation in the numerator divided by a smaller number (spec – 12%) in the denominator which means the % tolerance will be even larger for the new and tighter spec. This hasn’t improved anything.
    Ultimately I have to fix the gage or the process for using the gage (see Ed’s answer). The first thing I have to do is protect the customer and second protect my company which is what guardbanding does. Moving the specs may in some way protect the customer – you now believe you have hard failures outside the specs but where are you protecting the company against false failures. You are always going to have the alpha and beta type errors. You have to deal with both.
    Good luck.

    0
Viewing 15 posts - 1 through 15 (of 15 total)

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