Feature based computer aided design systems are now becoming commercially available for mechanical design. Features are recurring stereotypical patterns with which knowledge related to various aspects of the product is associated. Feature modelers make it easy to modify the geometry as associations between entities of the part are encoded in the feature definition. Since features are viewpoint dependent, it is necessary to map/transform them from one viewpoint to another for concurrent engineering. However, feature mapping presents some difficult problems, particularly with modelers that do not restrict one to a limited set of predefined features, and when there are interactions between features that change their generic meaning. This research develops techniques which will enable manufacturing application programs to understand design features from a manufacturing viewpoint and automonously generate feasible machining process sequences for undocumented features. In effect, the system will be a self learning system that will automatically update its own knowledge about machining options for a feature it has never seen before. Techniques developed will include: constructive solid geometry tree reconstruction to localize interactions, intelligent representation of manufacturing features, and a meta knowledge base for encoding capabilities of machining processes from which feasible machining alternatives can be generated.
