Computational Design Synthesis (CDS) is used to enable the computer to generate valid and even creative solutions for an engineering task. Graph grammars are a CDS approach in which engineering knowledge is formalized using graphs to represent designs and rules that describe possible graph transformations, i.e. changes of designs. For most engineering tasks two different kinds of rules are required: rules that change the topology and rules that change parameters of a design. One of the main challenges in CDS using both topologic and parametric rules is to decide a priori which type of rule to apply in which stage of the synthesis process. The research presented in this paper describes different strategies for the combination of topologic and parametric rules during automated design synthesis. A graph grammar for the design of gearboxes is investigated in which topologic rules change the structure, i.e. the number and connections of gears and shafts, whereas parametric rules change the layout and sizing, i.e. the dimensions and positions of gears and shafts, in the gearbox. For the generation of new designs, two simple multi-objective stochastic search algorithms are used and compared. Four different strategies are presented that determine in different ways which type of rule (topologic or parametric) to apply in which stage of the synthesis process. The presented strategies are compared considering the quantity of the generated designs, i.e. the number of topologically different designs, and their quality, i.e. their objective function values. Results show a significant influence of the chosen strategy only in an early stage of the synthesis process. The discussion examines the adaptability of the proposed strategies to other engineering tasks.

View PDF No PDF? Try Arxiv!