Hybrid architecture of field-tested diagnostic expert system

Three approaches to diagnostic expert systems have emerged: rule-based, dependency models, and deep-knowledge models. A fourth approach is a hybrid architecture that combines two or more of the three pure forms so that the strengths of one support the weaknesses of the other and vice versa. The advantages and disadvantages of each of these approaches are examined. AI-Ferret, a system that was field-tested by US Army technicians troubleshooting operational equipment, and which implements a rule-dependency hybrid architecture, is described. The conclusions are that: the hybrid architecture will become the dominant approach to practical diagnostic expert systems; and that cost and performance benefits of hybrids can be determined by the level of interaction between rule and model parts.<
>