Expert systems are attractive for a variety of plant fault diagnosis applications because of their use of symbolic reasoning to narrow possibilities quickly. Programming expert systems, however, requires a manual, post hoc analysis of fault conditions. This paper presents a conceptual design for translating CAD databases in a format called Initial Graphics Exchange Specification (IGES) into diagnostic rule bases. The Process Intelligent CONtroller (PICON) expert system rule base is examined as a typical diagnostic rule base. Part of the rule base translates directly from IGES, the remainder is "instanced" by it from archived knowledge. The result is a direct interface between CAD and expert systems.