Gage rr for Distractive test

Six Sigma – iSixSigma Forums Old Forums General Gage rr for Distractive test

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #52653


    I have a gage which is used to pull welded plastic and record the pull pressure it took to break the plastic piece.  I was asked to conduct gage r&r. this is a distractive test and the housing for the plastic pieces are coming from three different welders.  If I cut Nine pieces out of one housing and repeat for all 10 housings I will get 90 pieces.  I can assume the nine pieces as homogeneous samples.  Do you think this is the right way to conduct nested gage r&r?  Thanks


    Ken Feldman

    Sure that will work but only for a Destructive Test. Not sure it is good for a Distractive Test since it will be hard to focus.


    Cesar Arcila

    When u r conducting a R&r study for destructive test, u have get 15 pieces (5 for each person) but to assuming that each piece is the same to the other in order to get good results.
    I hope this help you a little,



    Possibly random thoughts :
    1) Testing for homogeneity.  Chicken and egg do you test the variation to prove homogeneity or satisfy homogeneity to prove the parts are sufficiently similar for a reasonable destructive test.
    2) What is the reason for the test? homogeneity between machines?
    3) How many test machines?  I am assuming 1.  Therefore are you performing a type 1 test: repeatability but no bias.
    4) If your assuming a difference between bonding machines. 
    5) Seek local proffessional support for this.  Have you heard of an IsoPlot.  Don Wheeler describes a similar analysis. Shainin and Wheeler may not like being associated or approve of the suggestion here.  Basically if you bonded another 10, cut in half (this reduces of product variation for a repeat test), test one half then the other. Each test pull weight forms the magnitude of a Scatter-Plot X-Y axis.  Randomise test: don’t repeat pull end 1 & 2 for a sample after one-another as you endangering hiding patterns in your analysis.  Analysis: Best line fit to a 45 Deg (linearity and bias); the scatter of dots has a length (part var) and width (gauge var) (sausage shape), length to width should be greater then 5 figure.  If less (approaching a circle) forget it: improve the gauge.  The width gives an quantitative figure.
    6) Assuming that the measurement process includes the process of bonding the plastic, mount into a fixture then pulling with 1 pull testing machine.  Take another 5 or more (economic considerations) and again cut in half.  Again the two halves provide the nearest manuf conditions (end to end variation) and reduce variation.  There is greater uncertainy in the test as you are not testing the exact conditions as for a cross-tab test, you need to take  samples associated with a particular machine.  This needs >= 5 samples for each machine.
    7) It sounds as though you are trying to translate a cross-tab test into a heirachical / nested test. I think you need to slightly change the approach.
    8) Randomise, Randomise, Randomise.  Keep track of parts and position of test piece, but randomise.
    9) I don’t know what stats package you are using, I use Minitab and it hints at the data type and the analysis and presentation gives a reasonably good presentation.
    Hope these thoughts help.



    Go back to the person who made this request. Ask him or her how to put the samples back together again perfectly (kinda like humpty dumpty), and then assess the repeatability of the measurement system with some precision. 
    I must confess that I did an experiment in 1989 where the response was a destructive test and I did not do a GRR! However I did randomize the order in which I measured the experimental samples.The signal far outweighed the noise in the DOE. Plotting residuals versus measurement order didn’t highlight any non random behavior.
    You are in fact stuck with a nested approach with no way of repeating a measurement.



    For conducting the Gage R&R two important aspects to be considered

    selection of parts
    Above aspects are pretty easy where the measurements are repeatable in same part, for measurement which can not be repeated in same parts (destructive type measurement) additional consideration required.
    Let us see how to select part for destructive type measurements

    There are two requirement
    1)     10 different parts which covers entire measurement range, even we can select parts go above or below specification.
    2)     For operators to repeat during each measurement trial, consider a scenario of 3 operators, 3 trials and with 10 parts, for each measurement trial 9 similar parts to be selected.
    Parts selection
    1)     10 different parts
    a.      Parts from different supplier or different raw material batch or different days of production
    2)     Similar parts to repeat for each measurement trial
    a.      Parts manufactured continuously such that only common cause variation only present, parts produced in one batch of raw material, incase of welding process parts manufactured with one raw material bar and produced by one operator continuously.
    Analogy to understand this situation
    ·        In control chart the sub group sample collection can be analogy to the sample selection for destructive R&R
    a.      10 different part – Between Subgroup sample are taken such that process allow to vary
    b.      Similar parts to repeat measurement trial – With in subgroup sample are selected continuously such that only common cause variation present.
    ·        Using the Gage R&R Nested in Minitab.

Viewing 6 posts - 1 through 6 (of 6 total)

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