Automated Model Generation Using the KIF Declarative Language

Abstract

Abstract Logic-based artificial intelligence researchers propound the declarative programming paradigm as a solution to problems arising from conventional procedural programming methods. A formal language possessing a declarative semantics, called Knowledge Interchange Format (or KIF), has been used for interchanging information between disparate programs, each containing specialized internal representations to support specific requirements. A simple system has been developed to evaluate the utility of KIF and declarative programming. The domain chosen for this evaluation was lumped-parameter dynamic systems analysis, due to its well-established vocabulary and concepts. In particular, bond graph theory formed the basis of the knowledge representation which was written in KIF. Models were generated to analyze the physical dynamics of a servomechanism used in a compact disc player. Fully automated model construction and solution was achieved, beginning with a set of library elements written in KIF and ending with solution of a set of first-order differential equations which characterize dynamic behavior. Work has begun to include assembly as well as bond graph information in order to evaluate the system’s utility for managing constraints in multiple domains. Based on demonstrated success in the dynamics and assembly domains, the next step will be to apply declarative programming to more-open domains, such as functional tolerancing, for which a comprehensive vocabulary and conceptual framework is still lacking. The hope is that the declarative paradigm will contribute to the formalization of several domains which can then be more easily integrated within a concurrent engineering environment, thus fostering conceptual product designs which are more robust with respect to manufacturability constraints.