Ppk and True Position

  • Filter
  • Time
  • Show
Clear All
new posts

  • Ppk and True Position

    I read the never-ending thread “Cpk and True Position” and I am happy to see that y’all are as passionate about addressing this problem as I have been for some time now. I am recently retired so I have lots more time to browse.

    The attached files, explain the predictions as I see them, and they offer strategies to assess process potential and optimization of variable tolerance specifications.
    I named this thread “Ppk and True Position” because the analysis is intended to look at a snapshot of the process variation (finite sample) rather than from sequential sub-group averages (cumulative sample) that are intended to “normalize” the distribution occurring from the observations to predict Cpk.

    It is not perfect! It doesn’t address “datum shift” although it is possible to do so if all features in the “simultaneous requirement” are analyzed together for their “residual variable tolerance” from iterated “candidate” datum shifts. It assumes “normality” for the size and variable geometric tolerance distributions to simplify the analysis using the classic reliability model for “Strength vs. Stress,” although the analysis of interference of dissimilar or “non-normal” distributions could be more accurately predicted with more rigorous tools (Monte-Carlo, etc.). It doesn’t address whether the parameter observations exhibit predictable variation “whether they are within control limits.” I suggest as many others have that the size and X,Y,Z coordinate parameters be monitored individually so that the analyzer can see which parameters are the significant contributors to variation for potential process improvement. It cannot distinguish whether a variable tolerance modifier has been non-functionally applied to a specification! Furthermore I suggest that those coordinate parameters be measured with respect to the orientation of the process or machine’s adjustable parameters rather than the dimensional orientation depicted on the specification (if different), by doing so it will enable the process owner to adjust his process more effectively.

    I have studied the few competing analysis solutions out there (one of them my own) and let me just say that I think that this one is the best so far. It can be improved by addressing the prediction inaccuracies listed above. So why employ it? Because!
    • It enables…even highlights the parameters relevant to process optimization and improvement with variable tolerance specifications.
    • It advances the statistical analysis geometric dimensioning and tolerancing beyond the current practice of ignoring the variable portion of tolerance in continuous data analysis.
    • It significantly improves the correlation of the predicted “probability of a defect” from attribute gages and statistical specification conformance predictions from continuous data.
    • It significantly…no dramatically reduces sample size requirements required for demonstration of process conformance to variable tolerance specifications.
    • It promotes understanding (or the lack there of) Geometric Dimensioning and Tolerancing and the gaps that exist with quality assessments.

    (Note: I refuse to call attribute gauges “functional gauges” unless the modifiers were selected according to the functional liabilities of location or orientation variation.) You can use this criterion to test the application; if the function of a feature worsens as it is permitted to deviate from its ideal location or orientation measurement then its location or orientation tolerance should not functionally be variable. If MMC or LMC modifiers are “none-the-less” designated, they are selected for attribute gauging purposes rather than function, which is occasionally desirable.

    To those who say that process capability estimations cannot distinguish whether the position deviations are clustered in one quadrant of the perimeter of the tolerance zone or evenly disbursed about the perimeter itself, see the part of the presentation “ppkmmc.pdf” that explains how to determine process potential with constant or variable tolerance specifications.

    To analyze processes where the coordinates can be independently adjusted use ppkmmcXY.xls and when they cannot use ppkmmc.xls. They have macros so I am sending them zipped. If you are just figuring the capability of a position tolerance (RFS) and you do not have individual coordinate data to examine process potential then by all means use Minitab or some other analysis tool to transform the typically skewed distribution or compare it to the best-fitting distribution function.

    Happy Analyzing!

    Paul F. Jackson
    Former Layout Inspector, STA Engineer, Quality Engineer, and Product Development Engineer in CAD for GD&T Strategy (although not resolved to stop working!)
    Certified Senior GDTP and Six Sigma Black Belt.

    mailto:[email protected]

    I cannot upload the PDF presentation because it is 600kb as PDF and 400kb zipped but you can download it from a cove discussion if you want to see it. http://elsmar.com/Forums/showthread.php?t=16607&page=2
    Attached Files

  • #2
    Thanks Paul, and welcome aboard!!!
    Kevin A. Botta
    Last edited by kbotta; 05-08-2007, 07:36 AM.
    RFS Means Really Fussy Stuff

    When all you have is a hammer - everything looks like a nail....


    • #3
      Wow! That's some very good stuff you have just shared w/us Paul...

      That should help sum it up for some of the other threads. Welcome aboard and thank you very much for sharing a piece of your knowledge with us. Enjoy that retirement and please keep in touch with us on the forum...
      Xcel 15-20-10 - PFXcel 7-6-5 - Merlin 11-11-7 - Romer Absolute 7525SI
      PCDMIS 2012
      Windows Office XP


      • #4
        Thank you, Paul, nice work...


        • #5
          Good to hear from you Paul!

          Another excellent piece on true position capability!

          Xcel & MicroVal Pfx & Global 37mr4 thru 2012mr1sp3
          Contura Calypso 5.4

          Lord, keep Your arm around my shoulder and Your hand over my mouth. Amen.


          Related Topics