As the operational paradigm of nuclear power plants shifts toward multi-module systems, such as Small Modular Reactors (SMRs), a single operator faces significantly higher cognitive demands to monitor and control multiple modules simultaneously. Maintaining an adequate level of Situation Awareness (SA) is critical for operational safety in these complex environments. While eye movement data has been increasingly utilized to evaluate SA, existing quantitative validations remain predominantly restricted to single-module operations. Although some studies have explored multi-module environments, they primarily focus on qualitative observations or subjective assessments rather than establishing definitive mathematical relationships.
To address this research gap, this study proposes a robust quantitative SA measurement framework utilizing a fixation-based Markov entropy model. An experiment was conducted using a Compact Nuclear Simulator (CNS) to emulate various operational modes. A total of 15 participants with relevant thermohydraulic system knowledge were recruited to perform procedure-based tasks. Participants managed simulated transient scenarios, specifically Loss of Coolant Accidents (LOCA), Steam Generator Tube Ruptures (SGTR), and Excess Steam Demand Events (ESDE). These scenarios were structured into single- and multi-module configurations, requiring participants to perform the diagnosis of accidents and initiate appropriate mitigation strategies based on emergency operating procedures. Operator SA was evaluated using the Situation Awareness Control Room Inventory (SACRI) during simulator freeze-probes, which were triggered immediately after the diagnosis of accidents but prior to mitigatory actions. Concurrently, their visual scanning complexities were captured via eye-tracking to calculate Markov entropy. Finally, regression analysis was performed to evaluate the statistical relationship and construct a mathematical model for predicting the SACRI SA score (A') from eye movement data.
The result from linear regression analysis revealed that fixation-based Markov entropy is a highly significant predictor of SA. Specifically, the model identified a definitive negative relationship (B = -0.208, p = 0.024), indicating that higher entropy—which reflects a fragmented and unorganized gaze transition pattern—quantitatively degrades the overall SA level. By integrating these statistical findings, this study established a comprehensive mathematical equation that quantifies eye movement-based SA across both single- and multi-module operations. Ultimately, this data-driven formulation provides an objective tool for evaluating operator cognitive performance, offering foundational insights for the design of advanced human-machine interfaces and training programs in next-generation multi-module control rooms.