Computational problems in feature research

The process of computing higher level information in order to more thoroughly understand the shape or other characteristics of a solid geometric model is referred to as geometric reasoning. Two forms of geometric reasoning are feature recognition and feature refinement. Feature recognition is the process of grouping the elements of an input data set into subsets such that the elements of each subset, taken together, provide additional information about the input data set. These subsets are denoted features. Feature refinement is the process of regrouping or transforming the elements of previously obtained features into new features. The research presented in this thesis addresses feature description methods, feature recognition methods, and feature refinement methods involving primarily solid geometric models. A preliminary feature recognition method is presented first. This research has resulted in a fast feature recognition methodology that is quite general and is aimed at a class of machining domain features. This is followed by the presentation of a declarative programming language for feature description. A search engine is described that has the ability to search the input data space for the described features. Additionally, a feature description development and testing facility is discussed. An example manufacturing domain application of the language and search engine is also presented. The thesis is concluded with the presentation of a feature refinement method that merges information available from a feature-based model with boundary-based feature recognition. This method provides an initial application for the language and search system. This thesis is appended by a description of the software implementation details needed by the engineer who wishes to integrate this system into a downstream application. Additionally, an appendix containing a complete set of syntax diagrams for the language is included