A 3D object classifier for discriminating manufacturing processes

Automated classification of artifacts produced by mechanical computer-aided design (CAD) is a unique research frontier for 3D matching and mesh processing. Unlike general graphical models, mechanical CAD artifacts have a physical realization via a variety of manufacturing processes as well as functional and behavioral attributes. The general problem of how to best correlate low-level shape data with the higher-order manufacturing and mechanical properties remains an open area of research with many practical applications (cost estimation, design archival, variational design and process selection).This paper addresses the problem of manufacturing process discrimination, i.e., determination of the best (or most likely) manufacturing process from shape feature information. Specifically, we introduce a new curvature-based shape descriptor and show its applicability to manufacturing process discrimination using a publicly available set of artifacts from the National Design Repository. Statistics on surface curvatures are used to construct the curvature-based shape descriptor; and a supervised machine learning classifier, based on support vector machines, is applied to learn a separator for models that are “prismatic machined” and “cast-then-machined”. The authors believe that this work can be the basis for practical new techniques for manufacturing cost estimation, engineering analysis and design archival.