Identification of Multiple Feature Representations by Volume Decomposition for 2.5-Dimensional Components

Decomposition of computer-aided design models into features that can be directly manufactured and easily translated from one representation to another has been recognized as a necessity for robust automated process planning systems. The algorithms presented here yield multiple representations of features that can be used to generate plans that are easily and efficiently manufacturable by milling processes. Features are recognized from the faces of a prismatic stock by first identifying missing regions on the external faces of the stock and then recursively descending into the part. Each missing region corresponds to a feature. As the projection of the feature is swept into the part, changes in the cross-section are identified. These denote the beginning of new features which become children of the original feature. This process yields a set of six trees, each of which resembles a depth-first search tree and is a partial ordering of features within a setup. Multiple representations are investigated for those features that are accessible from more than one direction. Corresponding features in the different trees are linked by arcs, thus yielding a feature graph. The feature graph can then be used to generate optimal features for machining depending on design or manufacturing requirements such as fixtures, tolerances, corner radii, and tool accessibility. [S1087-1357(00)01901-8]