Feature Extraction for Casting Core Development

Abstract

Feature extraction techniques are presented for the generation of casting core patterns from a boundary representation (B-Rep) solid model. Techniques are presented which would allow for automatic extraction of three classes of core features (internal voids, single and multi-surface holes, and boundary perturbations). The task of extracting casting cores from solid models involves recognizing a collection of entities (i.e., slots, bosses, undercut surfaces, local and global concavities, etc.) from the set of lower level entities (i.e., the B-Rep structure). To this end, a combination of solid modeling B-Rep and graph structures and their associated methods will be used for casting core development. Appropriate local features are identified and extracted from the original object, and are grouped into one or more new object(s) (termed a core-object). If the core-object is multiply connected (i.e., composed of multiple objects), it is graph separated into global feature objects. Each of these global feature objects represents a core in the final pattern. Lastly, the geometry of the original part is augmented to add core prints where core geometries were extracted. The core print, as currently developed, combines the extracted core geometry and its convex hull.