iSixSigma

GRR for GDT features

Six Sigma – iSixSigma Forums Old Forums General GRR for GDT features

This topic contains 8 replies, has 3 voices, and was last updated by  Scott 9 years, 10 months ago.

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #53051

    Scott
    Member

    Hello everyone!
    I’m looking for some guidance regarding calculating %tolerance measurement errors for a GR&R on a profile or true position type requirement.
    Basic study design:  10 parts, 3 operators, 3 trials.  We are able to get the GDT value and the X & Y coordinates for analysis.
    Problem:  If I input just the true position value for each inspection, I can get a 33% %Tol value on a 0.010 tolerance, but the number of categories is only 2.  This indicates that Minitab can’t tell the difference between some of the parts.  In looking at the data closer, the true position value doesn’t allow you to know which quadrent the part lies (+X +Y/ +X -Y/ -X -Y/ -X +Y).  So parts that are truely at opposite ends of the distribution (+X +Y vs -X -Y) are grouped together because the true position number is essentially the same value.
    I can look specifically at analyzing the measurement error in getting consistent X and Y coordinates, but I don’t have a tolerance I can compare the variation to.  So essentially I can get a partial answer.
    Performing an attribute study is always an option, but not the desired option.
    Can anybody offer some insight on how to approach this??
    Thanks in advance for any assistance I receive!!

    0
    #187646

    Taylor
    Participant

    Can you tell me specifically how this is being measured? I understand your delima, but need to know the measurement device before I can give you dome ideas

    0
    #187649

    Scott
    Member

    We are using a customized operator assisted camera system.  Basically something similar to an optical comparator.  Not sure how the method comes into play here, but I am interested to hear your ideas!!

    0
    #187650

    Taylor
    Participant

    Rick
    For each part you need to have an X and Y data point. Basically perform two R&R’s. One for X and One for Y. This will solve your qauderant issue.

    0
    #187651

    Scott
    Member

    That is a great idea and I have tossed that approach around. 
    Getting data to analyze isn’t the issue.  The question with that approach is what tolerance is used compare the X & Y coordinate to??  If there is a true position callout of 0.010, what does that translate to in terms of a X & Y coordinate tolerance?

    0
    #187652

    Taylor
    Participant

    Rick
    Your goal here is to establish whether or not you can measure the feature in both X & Y coordinates and that they are reproducible and repeatable. I personally do not like to use % tolerance for Gage R&R for the reason that if your Tolerance is inflated then your R&R will be better than the true capability of your measurement system. So my 2 cents worth is don’t use % Tolerance

    0
    #187653

    Taylor
    Participant

    One more thing. Make sure your samples cover the full range of your tolerance.

    0
    #187654

    Bower Chiel
    Participant

    Hi Rick
    Have you considered exploring the use of polar coordinates?
    A quick web search gave the following reference: –
    Gauge R&R analysis for two-dimensional data with circular tolerances

    Journal of Quality Technology | April 1, 2003|

    document.write(“Voelkel, Joseph G”);

    Voelkel, Joseph G | Copyright American Society for Quality Apr 2003.
     
    Good Luck!
    Bower Chiel

    0
    #187679

    Scott
    Member

    I agree that you have to be cautious looking at the coordinates compared to an inflated tolerance can give a false sense of security on the measurement system.  But you cannot only look at the system to itself (% Study in Minitab terms) because that does not give you an understanding if the measurement system is adequate to meet your needs.  It is possible to have a system that shows a low %Study result, but when compared to the tolerance it isn’t acceptable.  %Tolerance will allow you to understand if the system can meet the requirements and %Process will allow to see if your system is capable enough to help control your process.
    I think looking at the coordinates is the right way to go, but the question is what is used to construct a “window” to be able to determine if the amount of variation is acceptable or not.
    I hadn’t thought of polar coordinates, but will look into that idea more, too.
    I do appreciate the feedback and ideas!!

    0
Viewing 9 posts - 1 through 9 (of 9 total)

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