Network Modeling of Nuclear Operator Procedures

Abstract

Research and evaluation on human factors issues can be very expensive owing to 1) the high cost of running experiments and 2) high inter-team variability which makes it necessary to run large numbers of subjects to get stable estimates of performance. Increasingly, the engineering disciplines are looking towards computer modeling as a means of predicting performance as a function of engineering design. Human factors engineering has that goal as well. This paper presents the results of a validation study that evaluated a human performance modeling technology termed task network modeling. Task network models were built of a crew executing two emergency procedures and one normal procedure. For each of these three procedures, one model was built reflecting the use of paper procedures and one reflecting the use of computerized procedures. Model predictions were then compared to data on actual crews performing under identical conditions. In general, the model predictions were representative of actual performance, although a number of issues arose that should be addressed prior to using these models as a technical basis for regulatory action.