Representing and Recognizing Features in Mechanical Designs
Abstract When experts view an object, they perceive it in terms of their own expertise. For example, manufacturers see features that affect the processes used to fabricate a part, while structural engineers see sources of stresses and other features that tend to reduce the life of a part. Features can be geometric, such as slots or chamfers; they can be quantitative, such as distances between holes; they can be functional, such as alignment; or they can be qualitative, such as a rough surface finish. Research in feature-based design systems for mechanical designers has been motivated by the realization that geometric models represent the design in greater detail than can be utilized by designers, process planners, assembly planners, or by systems that emulate these activities. Features provide abstractions to facilitate the creation, representation, and analysis of designs. Our goal is to enable designers to compose mechanical designs from high-level features that embody functional and geometric properties. In addition, we want to provide designers with feedback on the manufacturability, assemblability, functionality, cost, etc. of the design as it evolves. To support this process in an intelligent CAD environment requires the integration of geometric models, analysis tools, and synthesis tools so that all aspects of the design can be considered while it is in progress. We are developing a design environment based on a shared representation of the design in which we can extract and reason about features of the design from different perspectives. Our approach is to represent both the design and the features using graph grammars. By representing the features using the same grammar as the design, we can recognize features by parsing a feature against the graph that represents the design. We are exploring grammars for behavior as well as geometry in order to provide a link between behavioral and geometric representations. In this paper, we focus on the representation and recognition of features.