3.3 FEATURE RECOGNITION ISSUES
To counter the difficulties of combinatorial explosion and model variations, it becomes necessary to filter the information available so that attention can quickly be focussed upon only the elements that are of importance. The DDPF does this by explicitly obtaining edges which are of interest to the feature recognition process. The process of skeletonization achieves an equivalent goal, although through a different technique.
Both approaches substantially reduce (although do not necessarily eliminate) the combinatoric explosion problem as a direct consequence of the filtering process. Also, the problems with topological and geometric variation are reduced when features are defined in terms of abstracted entities, instead of local face-edge-vertex adjacency relationships.
The feature graph grammar approach is not without merit. Within design synthesis, the generation of features from functional requirements is well accomplished [Brown et al 92][Wolter & Chandrasekaran 91][Suh et al 91]. A design-with-features approach allows features to have been pre-defined, to be selected for inclusion into a part by the designer with only a modification of parameters[Cunningham & Dixon 88][Libardi et al 86][Luby et al 86].
The feature generation and design-with-features approach work under the assumption of known features. But secondary features may be created by a combination of primary features, and these must be recognized and evaluated. Additionally, the features generated by functional or other requirements may not be the specifically same features that must be recognized for manufacturing, recycling, or other concerns. Some hybrid systems, containing a design with feature and a feature recognition approach have been developed [Falcidieno et al 91].