Issues involved in a knowledge-based approach to procedure synthesis

Many knowledge-based systems (KBSs) have been built to assist human operators in managing nuclear power plant operating functions, such as monitoring, fault diagnosis, alarm filtering, and procedure management. For procedure management, KBSs have been built to display and track existing written procedures or to dynamically follow procedure execution by monitoring plant data and action execution and suggesting recovery steps. More recent works build KBSs able to synthesize procedures. This paper addresses and examines the main issues related to the implementation of on-line procedure synthesis using KBSs. A KBS for procedure synthesis can provide a more robust and effective procedural plan during accidents. Currently existing procedures for abnormal plant conditions, written as precompiled step sets based on the event and symptom approaches, are inherently not robust because anticipation of all potential plant states and associated plant responses is not possible. Thus, their failure recovery capability is limited to the precompiled set. Procedure synthesis has the potential to overcome these two problems because it does not require such precompilation of large sets of plant states and associated recovery procedures. Other benefits obtained from a complete procedure synthesis system are providing (a) a methodology for off-line procedure verification and (b) a methodology for the eventual automation of plant operations