Architecture of a coupled expert system for optimum design of plate girder bridges

Abstract

A prototype knowledge-based expert system has been developed for optimum design of steel plate girders used in highway bridges. This expert system, called PG-BRIDGE1, is a coupled system in which AI-based symbolic processing is combined with the traditional numerical processing. Plate girders can be unstiffened or stiffened with single- or doublesided transverse stiffeners. They can be homogeneous or hybrid, made of high-strength flange plates and low-strength web plate. A mathematical optimization algorithm has been developed for minimum weight design of plate girders using the generalized geometric programming technique. The basis of design is the American Association of State Highway and Transportation Officials (AASHTO) specifications. The plate girders are subjected to the live (moving) loads of the AASHTO specifications. The knowledge base and symbolic processing has been developed using the Expert System Development Environment (ESDE). Numerical processing for structural analysis, optimization algorithm and graphics interface have been developed in FORTRAN 77.