1. Introduction
During the operation of a nuclear power plant, the situation awareness of operators has a great influence on the safe, reliable, and efficient operation of the plant. Failure to correctly complete subsequent complex behaviors due to loss of situation awareness (LSA) may lead to disastrous consequences. For example, in the accident at Three Mile Island Nuclear Power Plant [
1], the operator failed to maintain a correct understanding of the status of the primary circuit, leading to disastrous consequences. Endsely reports that 88% of human errors are caused by situation awareness [
2]. In a nuclear power plant, if the operator’s situation awareness is improved, the human errors that cause accidents can be reduced.
So far, there are many methods to measure the level of situation awareness. Salmon and Stanton et al. [
3] have compared and analyzed a variety of measurement methods. The results show that different methods have advantages and disadvantages and are suitable for different application fields. The methods with the highest comprehensive scores are the situation awareness global assessment technique (SAGAT) and the situation awareness rating technique (SART).
In order to measure the situation awareness of the operators in the main control room of multi-module high-temperature gas-cooled reactor nuclear power plants, a comprehensive measurement method of situation awareness is designed that combines SART, NASA-TLX (NASA Task Load Index), and eye movement tracking, which makes up for the shortcomings of these three methods and improves the accuracy of data analysis.
SART is a subjective measurement method that requires operators to fill out the questionnaire after each experimental scenario, so there is no need to pause the simulator during the experimental process. It has the advantages of no interference with the main task, easy operation, convenience, high accuracy, and easy acceptance by operators.
Bustamante [
4] used SART to measure pilots’ situation awareness (SA) during critical weather events. The experimental results show that when the weather system fails to provide an indication of the weather event at the specified waypoint, the SA of pilots will decrease, and first officers (FO) will have higher situation awareness than Captains. Pilots’ workloads will significantly increase when they approach the weather event.
Stark [
5] used SART to analyze the influence of the size of the synthetic visual display on the situation awareness of pilots. The results show that there is no significant difference in the performance of monitors of different sizes, and smaller display sizes will improve the situation awareness of pilots.
Burke [
6] used SART to evaluate the situation awareness of 12 pilots using TAP software in a Piaggio Avanti flight test aircraft.
Although the SART method has many advantages, there are also some shortcomings, such as the fact that this method calculates the situation awareness of the operator in the experimental scenario based on the scale filled out by the operator. Therefore, it cannot achieve real-time monitoring of the situation awareness of the operator in the experimental scenario. The eye tracking method, as an objective measurement method, can record the operator’s performance in real time during the experimental process, making up for the shortcomings of the SART method and improving the accuracy of measurement.
Kilingaru [
7] used dwell time entropy to distinguish the behavior of expert pilots and novice pilots. The results show that the dwell time entropy of expert pilots is significantly higher than that of novice pilots.
Koen [
8] investigated pilots’ situation awareness using eye tracking. The results showed that pilots had low situation awareness and high gaze entropy during the experiment.
Ha et al. [
9] proposed a novel method to infer the operator’s thoughts from his eye movement data and evaluated it with a nuclear power plant simulator. It was concluded that the operator’s thoughts can be inferred from the operator’s eye movement data. Analyzing operator situation awareness through eye movement data has great application potential in nuclear power plants.
NASA-TLX, like SART, belongs to subjective measurement methods and also has the advantages of no interference with the main task, easy operation, high accuracy, and easy acceptance by operators. According to the scale filled out by the operator, the workload of the operator in this experimental scenario can be calculated.
Selcon and Taylor [
10,
11] compared the accuracy and reliability of SART and NASA TLX within a computer-graphics flight simulation environment. Twelve RAF pilots took part in the experiment and were asked to rate a videotape of an air combat flight simulation sequence using TLX and SART technology. The experimental results show that the calculation results of SART and NASA TLX can reflect the difficulty of the task, and the results of the two scales have a certain correlation.
Dember et al. [
12] used NASA-TLX to measure the workload of visual alert tasks. They reported that the workload would increase with the increase in working time.
The HTR-PM600 nuclear power plant is composed of six identical NSSS modules, as shown in
Figure 1. The steam generated by the six NSSS modules is gathered in the main pipe to promote a steam turbine to generate electricity, which not only ensures the inherent safety of the high temperature reactor but also improves the economy of the power plant. It is the development trend of the high-temperature gas-cooled reactor nuclear power plant. The main control room of the HTR-PM600 nuclear power plant is different from that of the HTR-PM nuclear power plant [
13,
14,
15]. The main control room operator team consists of five operators: (1) Reactor operator (1#RO), (2) reactor operator (2#RO), (3) reactor operator (3#RO), conventional island operator (CO) and shift supervisor (SO). Each RO needs to monitor two NSSS modules; the CO monitors conventional island systems and equipment; and the SO is the duty supervisor in the main control room. Each NSSS module has the same monitoring parameters and operating procedures, so the problem of situation awareness among operators in the multi-modular high-temperature gas-cooled reactor (HTGR) nuclear power plant is more prominent. When in power operation, due to the advantages of the digital control system, the system status and parameters change little, and the workload of the operator is relatively small. The main work of the operator during the operation of the nuclear power plant is “monitoring, decision, control”. When a reactor accident occurs, the system parameters change dramatically. Due to the “keyhole effect” [
16], a large amount of information cannot be displayed on the same interface at the same time. If the operator cannot observe the change in parameters in time or cannot detect important alarm information in time, the possibility of accidents will increase.
In this study, SART, NASA-TLX, and gaze entropy were used to measure the SA of operators, and a series of accident handling experiments were carried out on a full-scale simulator. Through the experimental data, the possibility of the operator monitoring two NSSS modules in the case of emergency failure of a nuclear power plant was analyzed, the law of eye transfer when the operator monitors two NSSS modules at the same time was explored, and the operation experience of a multi-modular nuclear power plant was summarized.
4. Results and Analysis
We use the S2 experimental scenario as an example to introduce the actions that three ROs need to perform during the experiment. The S2 experimental scenario is that the main helium fan of the 5#NSSS module mistakenly accelerates, causing an increase in primary circuit flow, reactor nuclear power, and cold helium temperature due to the unexpected acceleration of the main helium fan. If 3#RO found that the relevant parameters changed, he would timely adjust the rotating speed of the main helium fan to put the nuclear power plant in a safe operating state. If the adjustment failed, the action of the nuclear reactor protection device would be triggered, and the reactor emergency shutdown would be triggered, 3#RO would monitor the reactor shutdown action to ensure that all control rods dropped safely. Because the mHTGR nuclear power plant generates steam from six NSSS modules to jointly promote the turbine generator to generate electricity, the coupling induction will cause changes in the steam parameters of the other five NSSS modules when the 5#NSSS module experiences an accident. The three ROs need to adjust the steam parameters of the five NSSS modules in a timely manner to ensure the safe operation of the reactor.
4.1. Situation Awareness and Workload
Through the calculation of the SART and NASA-TLX scales filled in by operators. The results are shown in
Figure 3. The results obtained through calculation are all dimensionless values, and the magnitude of the values can reflect the level of situation awareness and workload of the operator in the experimental scenario. Taking
Figure 3 as an example, the Abscissa (1–10) of all figures in this paper only represent experimental scenarios under different classes, not a continuously increasing variable. The connecting lines between coordinate points are only for the convenience of distinguishing them and have no continuous relationship. Through the results of the SART and NASA-TLX scales, it can be found that the RO’s situation awareness decreases with the increase in difficulty of different classes of experimental scenarios, but the difference is not significant. It shows that the number of NSSS modules with accidents has little impact on operators. Due to the advanced degree of automation in the nuclear power plant, when accidents occur in the plant, the ROs do not need too much operation; they just need to constantly monitor the process parameter changes and confirm the status of the power plant. When operation accidents occur in nuclear power plants, ROs can monitor two NSSS modules and handle the corresponding tasks at the same time.
Figure 4 is a graph showing the average situation awareness and workload of ROs in different classes of accident scenarios. It can be seen that in the class M accident scenario, the operator’s workload is the highest and the situation awareness is the lowest, while in category S, the operator’s workload is the lowest and the situation awareness is the highest. Through the calculation of correlation:
and
are the two groups of values involved in the calculation, and
and
are the average values of the two groups of values involved in the calculation. It can be calculated that there is a negative correlation between the operator’s situation awareness and workload (as shown in
Table 2).
4.2. Eye Tracker Hot Spot Map
We selected the experimental data from three experimental scenarios to show the difference in hot spot maps (as shown in
Figure 5) of operators under different operating conditions. It can be seen from the figure that the operator can monitor the two NSSS under normal operating conditions.
In case of an accident in 2#NSSS, the operator’s attention is focused on the malfunctioning 2#NSSS module, but the 1#NSSS module without an accident can also be monitored at the same time. When both NSSS modules have accidents, the RO will strengthen the monitoring of the two NSSS modules, and the points of view are also scattered.
Compared with the hot spot maps of the operators in the power operation phase and the accident operation phase, we know that the operators monitor more parameters in the accident operation phase, not just some key parameters.
Through the comparison of the three pictures, it can be found that the RO can monitor the two NSSS modules at the same time, regardless of whether there are accidents in the two NSSS modules.
4.3. Markov Entropy and Dwell Time Entropy
During the experiment, it was found that when the operator is monitoring two NSSS modules, the large display screens are usually used to monitor the process parameters, while the computer display screens are more used to operate the equipment of the nuclear power plant, and the interface will be switched frequently. In order to calculate the gaze entropy of the operator during the monitoring process, the two large display screens of each NSSS module are divided into 18 areas of interest (AOIs) according to the parameter content, so each operator needs to monitor 36 AOIs in total. The division of AOIs is shown in
Figure 6.
Calculate the ROs’ Markov entropy and dwell time entropy in all experimental scenarios, and the results are shown in
Figure 7. By comparison, it can be found that there is no significant difference in the ROs’ Markov entropy and dwell time entropy in different classes of experimental scenarios. This analysis result once again confirms the previous conclusion: “Even if two NSSS modules have accidents, ROs can monitor the two NSSS modules at the same time”.
Calculate the average value of the operators’ Markov entropy and situation awareness in each class of experimental scenario, and the results are shown in
Figure 8. By comparison, it can be found that there is a negative correlation between the operator’s situation awareness and Markov entropy. When the operator’s situation awareness is high, the operator’s eyes do not need to pay attention to a lot of AOI; they only need to switch their eyes back and forth between some key AOI, so their Markov entropy will be relatively low.
Each experimental scenario includes two phases: The power operation phase and the accident operation phase. The Markov entropy and dwell time entropy of the power operation phase and the accident operation phase in all experimental scenarios are calculated separately, and the results are shown in
Figure 9.
Through the comparison of the data, it can be seen that in all the accident scenarios, the Markov entropy in the accident operation phase is greater than that in the power operation phase, indicating that the operators’ situation awareness in the accident operation phase is less than that in the power operation phase. ROs need to pay attention to more parameters to understand the status of the power plant. The dwell time entropy in the accident operation phase is less than that in the power operation phase, which indicates that the operator is more focused on a few AOIs and more time is spent changing key parameters in the accident operation phase. Combined with the hot spot maps in
Figure 4, it can be seen that although the number of parameters concerned by operators during the accident operation phase has increased, the attention time is more concentrated on a few key parameters. Through comparison, it can be concluded that the parameters to which the operator has paid the most attention on the large screens are the primary system and pressure parameters, main feedwater parameters, hot helium parameters, primary circuit pressure, and steam parameters.
5. Conclusions
Combined with the characteristics of multi-module control rooms and multi-module running tasks, this paper designs a comprehensive situation awareness measurement method that combines SART, NASA-TLX (NASA Task Load Index), and eye movement tracking methods. A series of accident handling experiments are performed on a full-scale simulator to gain enough data for analysis. Through this method, the possibility of one operator monitoring two NSSS modules in the case of operation accidents in nuclear power plants is increased, which contributes to the further development of multi-modular high-temperature gas-cooled reactors.
The SART and NASA-TLX scales are used to measure SA and the workload of operators, while gaze entropy is used to indicate gaze movement. Under different accident scenarios, the ROs’ situation awareness and workload do not change significantly, indicating that even in the class M scenario with the heaviest tasks, the ROs still have the ability to monitor two NSSS modules at the same time. This conclusion can also be obtained from the analysis of hot spot map data and gaze entropy data. By analyzing the ROs‘ hot spot map data and gaze entropy data, we can know that: ① even if two NSSS modules have accidents at the same time, the RO can also monitor the two NSSS modules; ② when accidents occur, the number of parameters concerned by the RO will increase; and ③ when accidents occur, the RO will focus on some key parameters.
Therefore, from all the experimental data, we can observe that when accidents occur in the nuclear power plant, the RO can monitor the two NSSS modules and complete the relevant accident handling.