A hybrid expert system combining AI techniques with a neural-net

Abstract

This paper proposes an architecture for hybrid expert systems that combines AI techniques with a neural-net. These interactive hybrid expert systems are more powerful than existing systems because they can have a premise the veracity of which depends on the recognition of an object/curve/pattern.The operation of a hybrid expert system using a neural-net is described. This expert system is intended to be a diagnostic tool in a nuclear energy generation application. The architecture of a special neural-net designed for curve/object recognition is discussed.