A KNOWLEDGE BASED FRAMEWORK FOR PROCEDURE SYNTHESIS AND ITS APPLICATION TO THE EMERGENCY RESPONSE IN A NUCLEAR POWER PLANT (ARTIFICIAL INTELLIGENCE, EXPERT SYSTEM).

摘要

In this dissertation a nuclear power plant operator is viewed as a knowledge based problem solver. It is shown that in responding to an abnormal situation an operator typically solves several problems, for example, plant status monitoring, diagnosis, sensor data validation, consequence prediction and procedure synthesis. It is proposed that in order to respond to unexpected situations and handle procedure failures the capability to synthesize and modify procedures dynamically or in runtime is required. A knowledge based framework for dynamically synthesizing procedures (DPS), a knowledge representation language to apply DPS framework for real problems (DPSRL), and a framework for emergency procedure synthesis (FEPS) for nuclear power plants based on DPS are developed.; The DPS framework views the task of synthesis as a process of selecting predefined procedures to match the needs of the dynamically changing plant conditions. The DPSRL language provides knowledge organization and representation primitives required to support the DPS framework. Specifically, it provides mechanisms to build various plan libraries and procedures to access information from them.; The DPS based framework for emergency procedure synthesis for nuclear power plants provides following capabilities: (1) instead of a few safety functions it utilizes a hierarchy of safety goals thus providing the capability to detect and control safety threatening situations at various levels in the plant; (2) the safety goals hierarchy also implements a type of defense-in-depth philosophy for reactor operations; (3) it also provides a smoother transition between the function based procedures and abnormal event procedures leading to an effective intergration between the two approaches; and it provides various problem solving capabilities required for procedure selection, synthesis, execution, and failure recovery.; Finally, the capabilities and the use of DPS, DPSRL, and FEPS are demonstrated by developing an experimental expert system for a typical boiling water reactor and analyzing its performance for various selected abnormal incidents. It is concluded that the knowledge based approach provides an adequate methodology to develop approaches to aid the operator in knowledge based decision making. (Abstract shortened with permission of author.)