Analysis of two-dimensional measurement data for automated inspection

Abstract

Abstract Automatic sensing machines often provide a mass of geometric boundary information that has to be analyzed in order to estimate desired geometric parameters of part features. This problem has been approached primarily by curve fitting and often ad-hoc and arbitrary definitions of geometric dimensioning and tolerancing terms such as “size” and “position”. As indicated in ANSI and ISO standards and as evident to designers, parts do not make contact on best fit surfaces but on their extreme functional boundaries. In this paper, definitions based on computational geometry are given to formalize the meaning of geometric imperfections according to the implications of the standards. Given two-dimensional data point patterns, the problem of determining the size, form, orientation, and position of circular, planar, and parallel-plane features are formally expressed. Those definitions which lead to known solution techniques in computational geometry are identified. Solutions to other problems not found in related studies are presented, such as the roundness problem and the size of parallel-plane features. A formal technique for dealing with datum related features is also presented. Finally, an example is given that illustrates how a typical position tolerance can be verified with a simple procedure. The application of these techniques is demonstrated in automated dimensional inspection and in modeling of manufactured part geometries.