A Plan Induction System for Monitoring and Interpreting Operator Interventions in Process Control Environments

This paper describes the architecture and behavior of a prototype intelligent decision support system for monitoring operations in complex process control environments. Development of the underlying model required an examination of the various influences on process outcomes, including not only the causal nature of physical processes themselves, but also the role of human interventions and the associated impact of operating procedures on human behavior. The empirical study of nuclear power plant operations used in this research indicates that procedures are an important, but not necessarily deterministic, influence on the intervening behavior of an operator. Operators will deviate from procedures when the requirements of a situation render a procedure inadequate or counterproductive. Goal- and plan-based knowledge structures were derived from physical processes, operating procedures, and human operators. These structures were incorporated into the model's knowledge base, which serves as the basis for interpretation and prediction of operator interventions in a series of emergency scenarios in simulated real-time. The eventual goal of this research is to enhance management oversight and control of complex, dynamic task environments by providing both management and operators with advice that is informed by an understanding of the constituent influences on process outcomes.