Real-Time Expert Systems and Their Applications

Abstract

Real-time expert systems differ from simple consultation type expert systems in their relation with the external world. While the latter prompt the user for input data and control the rate at which input will be absorbed and processed, the former are driven by external events at a rate which cannot be influenced by the expert system. Because sequences of data relating to different lines of reasoning interleave each other, multitasking operation is a key feature of such systemsRTEX is an expert system building tool for creating embedded real-time applications in process control, network management, manufacturing control, industrial robots and simulation. It has been written in Ada and integrates the advanced concepts used in real-time distributed software systems with many features of the well established A.I. toolsAs a solid test application a procedure tracking system far monitoring nuclear power plants has been built. Monitoring the behaviour of a nuclear power plant after an incident has occurred involves a high degree of parallel reasoning. While diagnosing the incident and forwarding corrective actions, constant monitoring of critical functions is of primary importance. In the RTEX application the behaviour of the nuclear power plant is modelized by a flow of messages carrying measurement values of important plant parameters. A hierarchy of micro expert systems cooperate at run-time to assist the operator