Inference of feature grammars for feature-based modeling in CAD systems

Abstract

The introduction of features concept enables the association of shape and knowledge in understanding a CAD model. Grammar is considered potentially powerful formalism by its ability to represent and generate features. However, effective methods of computational must consider both feature modeling and feature learning in order to update the feature language. This paper presents an approach for the inference of Feature Grammars including three main phases. In the first phase, based on the hypothesis of the robustness, we search for terminals of features structures. In the second phase, from the terminals and their interrelationship, we represent the structures of features by corresponding canonical matrices. In the third phase, we infer the production rules of Features Grammars. Production rules are inferred based on the clustering of features structures as well as on their ordering relation. Examples and application illustrate the steps involved and the advantages of this approach.