Coordinate Measuring Machines (CMMs) collect a sampling of points on measured features for use in dimensional metrology. Conformance to specified geometric tolerances is done by analyzing the point cloud to fit the corresponding feature to the point cloud to determine if the simulated feature lies within the specified tolerance limits. Different types of feature fitting algorithms are needed: nominal, minimal/maximal, circumscribing/inscribing, and zone. Studies have shown that the same point cloud data sent to different vendors CMM software, produces different results. It is suspected that some of these algorithms may be inconsistent with the tolerance class definitions in tolerance standards and, in some cases, with shop floor conventional practices. We have previously reported on the development of normative algorithms and a feature fitting library that could be used by all CMMs. This paper gives a summary of those algorithms and then reports on methods used for verification. Three different types of verification methods were used to validate the algorithms developed. The scope of the current work is limited to linear, planar, circular, and cylindrical features. This set of algorithms described conforms to the international Standards for GD&T. In order to reduce the number of points to be analyzed, and to identify the possible candidate points for linear, circular and planar features, 2D and 3D convex hulls are used. For minimum, maximum, and Chebyshev cylinders, geometric search algorithms are used. Algorithms are divided into three major categories: least square, unconstrained, and constrained fits. Primary datums require one sided unconstrained fits for their verification. Secondary datums require one sided constrained fits for their verification. For size and other tolerance verifications, we require both unconstrained and constrained fits. Use of three different methods has validated the robustness, efficiency and accuracy of the algorithms.

This content is only available via PDF.
You do not currently have access to this content.