Task network modeling of human operators in nuclear power plant control rooms

Studying nuclear power plant operators is expensive and often impossible. There are a limited number of operators whose time is in great demand and a limited number of simulators or plants in which experimentation could be conducted. Yet, we often need to make predictions about operator behavior in the control room. Whenever the control room changes (e.g., panel or system modifications) or the plant procedures change, there is an essential need to predict how these changes will impact operator performance and, ultimately, plant performance. The question is, if we can't study real operators, what are the alternatives? The human engineering community within the nuclear industry needs to develop predictive methods. The approaches proposes in this paper, task network modeling and Micro SAINT, present promising opportunities. First, they represent a logical outgrowth of current technologies such as task analysis and probabilistic risk analysis. Second, there are data from which these models can be constructed. However, before any extended investment is made in this technology, it should be proven. A modest study could be run to directly address the issues defining the utility of this technology to determine whether we can collect the required data to build the models and the models can be used to accurately predict changes in operator performance. From the answers to these questions and the lessons learned in addressing them, future control room operator modeling research and development can be better directed