A CONCEPTUAL BASIS FOR INTEGRATED, MODEL-BASED SYSTEM DESIGN (SIMULATION, EXPERT SYSTEM

The thesis sets up a conceptual framework for constructing model-based, computer-aided environments for system design. The framework is based on the formal structures underlying the multifacetted modelling methodology, namely that of the system entity structure and experimental frame. The entity structure formalism recognizes three primary relationships in the design problem domain: the decomposition hierarchy, the taxonomic structure determining the design alternatives, and the coupling constraints restricting the combinations in which components can be synthesized into a target system. An experimental frame is a specification of circumstances under which a design model is to be observed and experimented with. The concept of experimental frame is generalized to define the so-called generic frame types i.e., structures that represent behavioral aspects of the design objectives in the form of standard variable types. The variable types express measures associated with performance indexes for evaluation of design models.;Both, the system entity structure and generic frame types are employed in a design framework characterized by the following fundamental activities: (a) the system entity structure is a basic means of organizing a family of possible configurations of the system being designed. (b) the objectives and requirements of the design project induce appropriate generic experimental frames. (c) the design entity structure is pruned with respect to the generic frames. This results in a family of model structures that conform to the design objectives. (d) the pruned substructures serve as skeletons for generating production rules for synthesis of design models. (e) resulting models are evaluated in respective experimental frames and the best design models are chosen on the basis of such evaluations.;Finally, architectures for experimentation control in design simulation are discussed. Both, centralized and distributed modes of experimentation are discussed. The specification of the coordination instruments and information sharing facilities are given.;The results of the thesis contribute to both simulation modelling methodology and system design disciplines. To the first area, because the procedures for entity structure and frame-based model construction set up a framework for development of expert simulation systems; to the second, because the presented design methodology offers representation schemes which are well structured and have formalized operations that exploit such structures. This greatly facilitates a uniform treatment of design at each level and lays down a groundwork for the construction of computer-aided, expert design environments.