ABSTRACT
System entity structures (SESs) compose a formal ontology framework that captures system aspects and their properties (Zeigler, 1984; Zeigler & Zhang, 1989). In the early 1990s, researchers working at the intersection of artificial intelligence and modeling and simulation began to design and implement environments that automated the process of design space exploration (Rozenblit et al., 1990; Kim et al., 1990; Zeigler et al., 1991; Zeigler & Chi, 1993) and solved engineering problems (Rozenblit & Huang, 1991). SES ontologies were originally used to represent system configuration alternatives in these environments. These SESs were in no way specifying the internal behavior of the entities that were modeled through some other mechanism (e.g., Discrete Event Systems Specification [DEVS] modeling). The SES was primarily used to specify the relations between these entities (Rozenblit & Zeigler, 1993). In addition to capturing aspects, entities, taxonomic relationships, variable values, and structural/configuration alternatives, these SESs included information about how entities in the SES could be realized in the DEVS formalism and composed into an executable model (Zeigler et al. 2000). To systematically explore design spaces in these environments: (1) rule-based search processes were used to derive all valid pruned entity structures (PES) captured by the SES; (2) information based on entities and aspects in each PES was used to compose an executable model using DEVS components stored in a model repository; (3) each composed model was simulated; and (4) simulation results were analyzed in order to identify the most desirable design alternative. The scope of SES described a complete solution set that contained all the permutations and combinations available for determining the actual design in a formal manner. This actual design is a result of the intersection of the requirements set and the optimal solution from the solutions set. The solutions set is determined by the pruning process on the SES and the optimal solution was determined by the simulation of each of the designs in the solutions set.
