A graph-based framework for feature recognition

This paper discusses a feature recognition system for recognizing User Defined Features (UDF). The feature recognizer uses a graph-based approach to represent and recognize features. An attributed face adjacency graph consisting of topological and geometric attributes is used to represent UDF's. The feature recognition step involves finding similar subgraphs in the part graph. The novelty of the framework lies in the usage of a rich set of attributes to recognize a wide range of features efficiently. A unique representation using graph grammars has also been developed to define family of features such as pockets with variable number of side faces. The feature recognizer also addresses many kinds of feature interactions by progressive suppression of the identified features. New techniques have been implemented for suppressing degenerate or virtual features. The feature recognizer also consists of a parameterization module to extract user-defined parameters from the recognized features.