Using Artificial Neural Networks for Predicting Mental Workload in Nuclear Power Plants Based on Eye Tracking

The development of a model for mental workload (MWL) prediction of an operator in nuclear power plants (NPPs) is necessary but challenging. In this study, the validity, sensitivity, and relationship between the four indices of eye tracking (i.e., pupil dilation, blink rate, fixation rate, and saccadic rate) and subjective rating method (i.e., the National Aeronautics and Space Administration-Task Load Index) of both experts and nonexperts when they are operating the state-oriented procedure system in NPPs are analyzed. An artificial neural network (ANN) is used to develop the MWL prediction model using the data of nonexperts. The correlation analysis results indicate that four eye tracking indices are sensitive to the subjective MWL, but there is no significant difference in the pupil diameter and saccadic rate between the experts and nonexperts. The validity of the proposed ANN-based prediction model is proven by the high correlation coefficient (higher than 0.95) between the original and predicted data. However, when the proposed ANN model was applied to the experts’ data, there was a significant difference between the original and predicted data. Therefore, the proposed prediction model can be applied to the experts’ data but with a certain adjustment to obtain the most possibly reasonable results.