Investigation on Applicability of Information Theory to Prediction of Operator Performance in Diagnosis Tasks at Nuclear Power Plants

This paper investigates the applicability of information theory to predicting operator performance in diagnosis tasks at nuclear power plants (NPPs). Some general descriptions about diagnosis tasks in a main control room (MCR) are provided and information theory conventionally used in the studies of human information processing is briefly reviewed. Then, as an alternative to classical information theory, this paper proposes a method to quantify the cognitive information flow of diagnosis tasks, integrating a stage model (a qualitative approach) with information theory (a quantitative approach). An information flow model is developed based on operating procedures and quantified using Conant’s model. Finally, three experiments were conducted to evaluate the effectiveness of the proposed approach to predicting human performance, especially average diagnosis time. The results of two experiments show that subjects’ diagnosis time has a proportional relationship to the quantified information flow. The other investigated operator performance variation in terms of accuracy under system-paced information presentation. As a result of the third experiment, subjects showed better performances in a specific range of information rate, compared with in other ranges.