A FRAMEWORK FOR COMPUTATIONAL DESIGN SYNTHESIS BASED ON GRAPH-GRAMMARS AND FUNCTION-BEHAVIOR-STRUCTURE

摘要

Computational design synthesis supports the knowledge-intensive process of developing new products. However, most approaches to date are often limited to a narrow domain and viewpoint of a synthesis task. The framework introduced in this paper aims to respond to the need for a method that integrates a richer product representation for computational synthesis within a framework that includes simulation, performance evaluation, and search. A computational and parameterized product model is presented that combines the Function-Behavior-Structure levels of abstraction. Graph-grammars are then used to create a formal definition of vocabulary and valid graph transformation rules. This approach offers the possibility to harness the large knowledge source of design catalogues in order to formulate vocabulary in a viewpoint-independent and thus, flexible way. In addition to manually entered and problem-specific rules, a class of generic rules is introduced that is instantiated computationally based on the defined vocabulary and has the advantage of being problem independent and re-usable. Finally, validation of the method is given through the synthesis of electric power-trains. This research goes beyond prior work in the field as it allows for synthesis and reasoning on different layers of abstraction, including function, behavior and structure, and introduces methods to alleviate the encapsulation of engineering knowledge for synthesis methods.