Identification of causes of human errors in support of the development of intelligent computer-assisted instruction systems for plant operator training

This paper proposes a methodology to identify causes of human error in the operation of plant systems to support the development of CAI system for operator training. The target task of this methodology is goal-driven and knowledge-based planning behaviour, the cognitive process of which is assumed to be modeled as means-end analysis. The methodology uses four criteria to classify errors in an operation into eight groups, and then asks the trainee several questions to prune the causes. To confirm the usefulness of this methodology, a prototype CAI system is developed for the operation of filling up sodium into the primary coolant system of a liquid-metal-cooled fast reactor. The experimental result indicates that the system has the capability of identifying causes of the traineeʹs error, and consequently of figuring out the characteristics of his/her defect. As a result of this study, several issues are identified for future research.