Casey Volino, Engineering Associate and Senior Statistician, Corning
This talk covers informative ways to describe misshape in round or cylindrical objects that go beyond traditional metrics, such as “out-of-round,” and even modern GD&T (geometric dimensional tolerancing) metrics, such as “circularity.” We show that traditional metrics of misshape act only as indicators of the degree of misshape, but are non-informative in terms of type and orientation. Three related but easy approaches are presented, which are at least partially based on PCA or Fourier analyses that provide practitioners with detailed descriptions of patterned deviations from a perfect circle. One approach is preferred in the case when measurement orientation is consistent. Two other easy approaches can be used when measurement orientation is thought to be random or unreliable. Only a basic familiarity with ordinary least squares regression is assumed, although some knowledge for Fourier analysis is helpful. JMP will be used as both the analysis and visualization software for this talk, but the technical details are not specific to any particular JMP feature or platform. A paper will be made available after the talk as well as all data sets used, which are not proprietary.
