Method and system for evaluating and optimizing visual attention load of interface in main control room of nuclear power plant

Abstract

The application discloses a nuclear power plant main control room interface visual attention load evaluation and optimization method and system, belongs to the technical field of human-computer interface of nuclear power plant, and the method comprises the following steps: collecting eye movement data of an operator; calculating a heat map entropy, a gaze duration weighted coefficient and a pupil change rate coefficient based on the eye movement data, and generating a comprehensive attention load index; establishing a correlation model of attention entropy and response time, identifying a high-load area and an inefficient information area, and marking a safety-critical visual bottleneck; identifying an operation condition type, and generating an interface optimization suggestion adaptive to the operation condition; reevaluating after applying the optimization suggestion, and forming a closed-loop optimization; and the application accurately quantifies visual load of the operator through a multi-dimensional weighted fusion algorithm, automatically identifies interface design defects based on an entropy-time correlation model, generates adaptive optimization schemes for different operation conditions, and continuously improves through closed-loop feedback, so that the operator's key parameter identification time is shortened, and the operation error rate is reduced.

Description

Technical Field

[0001] This invention belongs to the field of human-machine interface technology for nuclear power plants, specifically relating to a method and system for assessing and optimizing visual attention load on the interface of the main control room of a nuclear power plant. Background Technology

[0002] The main control room of a nuclear power plant is the core location for operators to monitor and control the plant's operation. The quality of the human-machine interface (HMI) design within the control room directly affects the operator's situational awareness, decision-making efficiency, and operational accuracy, thus impacting the safe operation of the nuclear power plant. According to statistics from the International Atomic Energy Agency (IAEA), human factors account for a significant proportion of nuclear power plant accidents, a large portion of which is related to problems such as excessive cognitive load on operators and difficulties in accessing information caused by inadequate control room interface design.

[0003] With the development of nuclear power technology, the main control room has gradually evolved from the traditional analog panel control method to a digital, computerized human-machine interface system, including a DCS distributed control system, an SPDS safety parameter display system, and an alarm management interface. While digital interfaces provide richer information display capabilities, they also bring new problems such as information overload, unreasonable layout of interface elements, and difficulty in identifying key information. These problems can lead to excessive visual attention load on operators, prolong information search time, and increase the risk of operational errors. Especially in abnormal operating conditions and emergency situations, an unreasonable interface design can seriously affect the operator's emergency response capabilities.

[0004] Existing technologies include systems for equipment maintenance and experience feedback in nuclear power plants. For example, Chinese patent document CN207440824U discloses an equipment maintenance and experience feedback system for nuclear power plant safety, comprising a display screen connected to a central server, a central server operating platform, and a portable handheld device wirelessly connected to the central server. This system allows information exchange between the central server and the portable handheld device, sending maintenance procedure information to the device and receiving feedback data. Maintenance personnel complete the equipment maintenance process at the nuclear power plant site based on the information received by the portable handheld device and send feedback data to the central server. The feedback data includes maintenance result data, environmental condition data of the maintenance environment, and vital sign information of the maintenance personnel. The system uses a real-time analysis module to automatically check the feedback data in real time, including a process tracking and inspection module, a medical diagnosis and environmental safety monitoring module, to confirm that the maintenance work meets the specifications, and to confirm the environmental safety of the maintenance environment and the normal health status of the maintenance personnel.

[0005] However, the aforementioned existing technologies primarily focus on the management of equipment maintenance processes and the monitoring of the physical characteristics of maintenance personnel, without addressing the issue of assessing the visual attention load of nuclear power plant main control room operators during human-machine interface (HMI) interactions. Specifically, the existing technologies have the following shortcomings: First, they lack in-depth quantitative analysis methods for operator visual behavior, making it impossible to accurately assess the cognitive burden on operators caused by different areas of the main control room interface; second, they have not established a correlation model between visual attention characteristics and operational performance, failing to effectively identify visual bottleneck areas in the interface design; third, they do not consider the differences in operator visual behavior under different operating conditions of nuclear power plants, lacking targeted interface optimization strategies; and finally, they lack a closed-loop optimization mechanism based on actual evaluation data, making it impossible to verify and iteratively improve the interface optimization effect.

[0006] Therefore, accurately quantifying and assessing the impact of the nuclear power plant's main control room interface on the operator's visual attention load, identifying safety-critical visual bottlenecks in the interface design, and generating adaptive interface optimization schemes for different operating conditions are important technical problems that urgently need to be solved in the field of nuclear power plant human factors engineering. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and system for assessing and optimizing visual attention load on the main control room interface of a nuclear power plant. By integrating eye-tracking technology and machine learning models, it aims to achieve accurate quantitative assessment of the visual attention load on the main control room interface, automatically identify safety-critical visual bottlenecks in the interface design, generate adaptive interface optimization suggestions for different operating conditions, and continuously improve the quality of the interface design through a closed-loop feedback mechanism. This, in turn, enhances the operator's reaction speed and decision-making accuracy, reduces the risk of human error, and provides technical support for the safe operation of nuclear power plants.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] This invention provides a method for assessing and optimizing visual attention load on the interface of a nuclear power plant's main control room. The method includes an eye-tracking data acquisition step, a visual attention load quantification assessment step, an interface visual bottleneck identification step, and an operational condition adaptability optimization step. The eye-tracking data acquisition step collects eye-tracking data from operators interacting with the nuclear power plant's main control room control system, including gaze trajectory, gaze duration, pupil diameter changes, and saccade paths. The visual attention load quantification assessment step generates a visual attention heatmap based on the eye-tracking data, calculates heatmap entropy, gaze duration weighting coefficient, and pupil change rate coefficient, and generates a comprehensive attention load index through a multi-dimensional weighted fusion algorithm. The interface visual bottleneck identification step establishes a correlation model between attention entropy and response time based on the comprehensive attention load index and response time data of the interface area, identifies high-load and inefficient information areas in the interface, and automatically marks safety-critical visual bottlenecks using nuclear power industry human factors engineering standards. The operational condition adaptive optimization steps identify the current operational condition type of the nuclear power plant, generate interface layout optimization suggestions based on the visual bottleneck characteristics under different operational condition types, apply the interface layout optimization suggestions to the main control room interface, re-collect eye-tracking data for iterative evaluation, and form a closed-loop optimization.

[0010] This invention also provides a system for assessing and optimizing visual attention load in the main control room of a nuclear power plant, including an eye-tracking data acquisition module, a visual attention load quantification assessment module, an interface visual bottleneck identification module, and an operating condition adaptability optimization module. The modules are deeply coupled through data flow, with the output of one module serving as the key input to the next, ultimately forming a closed-loop feedback optimization mechanism.

[0011] Compared with the prior art, the present invention has the following beneficial effects:

[0012] 1. This invention integrates multiple physiological indicators such as heatmap entropy, fixation duration weight, and pupil change rate through an innovative multidimensional weighted fusion algorithm, which can comprehensively and accurately quantify and assess the operator's visual attention load. Compared with traditional methods that rely on only a single indicator, the comprehensive assessment accuracy of this invention is improved by more than 35%.

[0013] 2. Based on the correlation model between attention entropy and response time, this invention can automatically identify high-load areas and low-efficiency information areas in the main control room interface. Combined with the nuclear power industry human factors engineering standards (NUREG-0700, IEC 61772), it can accurately locate safety-critical visual bottlenecks, providing clear data support and scientific basis for interface optimization, and avoiding the subjectivity and blindness of traditional experience-based design.

[0014] 3. This invention establishes operator visual behavior characteristic models for three typical operating conditions of nuclear power plants: normal operation, abnormal operation, and emergency shutdown. It can generate interface layout optimization suggestions that are adaptive to the operating conditions, ensuring that the interface design can meet the operator's cognitive needs under various operating conditions, thereby improving the applicability and robustness of the interface design.

[0015] 4. This invention constructs a complete closed-loop feedback optimization mechanism, which re-evaluates the optimized interface using eye-tracking and dynamically adjusts the optimization strategy based on the evaluation results, achieving continuous improvement in interface design. Practical application shows that after 3-5 rounds of closed-loop optimization, the average operator key parameter recognition time is reduced by 42%, and the operation error rate is reduced by 63%, significantly improving the safety operation management level and operational efficiency of nuclear power plants.

[0016] 5. This invention deeply integrates eye-tracking technology with professional knowledge in the nuclear power field, filling the technical gap in quantitative evaluation and intelligent optimization of the interface design of the main control room of nuclear power plants. It provides a new methodology and technical tool for human factors engineering research in the nuclear power industry, and has important theoretical value and broad application prospects. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the method for assessing and optimizing visual attention load on the interface of a nuclear power plant main control room according to the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of the nuclear power plant main control room interface visual attention load assessment and optimization system of the present invention. Detailed Implementation

[0019] Please refer to the attached document. Figures 1-2 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] Reference Figure 1 and Figure 2 This invention provides a method and system for assessing and optimizing visual attention load on the interface of a nuclear power plant's main control room. The system includes an eye-tracking data acquisition module 1, a visual attention load quantification assessment module 2, an interface visual bottleneck identification module 3, and an operating condition adaptability optimization module 4. These modules are deeply coupled at both the parameter and state levels, forming a complete assessment-identification-optimization-feedback closed loop.

[0021] Reference Figure 2 The eye-tracking data acquisition module 1 is used to collect real-time eye-tracking data of operators when interacting with the control system in the nuclear power plant's main control room. This module integrates a high-precision eye-tracking device, capable of capturing the operator's eye-tracking information at a sampling frequency of over 60Hz.

[0022] The eye-tracking data acquisition module 1 includes a gaze trajectory acquisition unit, a gaze duration statistics unit, a pupil diameter monitoring unit, and a saccade path recording unit. The gaze trajectory acquisition unit uses an infrared camera and corneal reflection technology to record the operator's gaze point's two-dimensional coordinate position on the main control room interface in real time, with a coordinate accuracy within 0.5° of the field of view. The gaze duration statistics unit calculates the dwell time at each gaze point; a gaze point is considered valid when the eye remains relatively still at a certain position for more than 100ms. The pupil diameter monitoring unit calculates the operator's pupil diameter in real time using image processing algorithms, with a monitoring accuracy of 0.1mm and a sampling frequency of 60Hz. Changes in pupil diameter reflect the operator's cognitive load level, and research shows a positive correlation between pupil diameter and task difficulty. The saccade path recording unit records the operator's rapid movement trajectory of their gaze between different interface areas; the length and complexity of the saccade path reflect the rationality of the interface information layout.

[0023] During eye-tracking data acquisition, the type of control system the operator is currently interacting with (DCS distributed control system, SPDS safety parameter display system, or alarm management interface), the interface switching time, and the type of operation task are recorded simultaneously. This contextual information provides necessary data support for subsequent operational adaptability analysis.

[0024] The eye-tracking data acquisition module 1 also includes a data preprocessing unit, which filters the raw eye-tracking data to remove abnormal data such as blinks and lost gaze points. Preprocessing employs a combination of three-point median filtering and Kalman filtering to ensure data quality. Simultaneously, the coordinate system of the eye-tracking device is precisely mapped to the coordinate system of the main control room interface screen, achieving accurate correspondence between gaze points and interface elements.

[0025] The collected eye-tracking data is stored in time-series format, with data structures including fields such as timestamp, fixation point coordinates, fixation duration, pupil diameter, and saccade start and end points. This data serves as input to the visual attention load quantification assessment module 2, achieving parameter-level coupling between modules.

[0026] The visual attention load quantification assessment module 2 receives eye-tracking data from the eye-tracking data acquisition module 1 and generates a comprehensive attention load index through an innovative multi-dimensional weighted fusion algorithm, thereby achieving a precise quantitative assessment of the operator's visual cognitive load.

[0027] The visual attention load quantification assessment module 2 includes a visual heatmap generation unit, a heatmap entropy calculation unit, a fixation duration weight calculation unit, a pupil change rate calculation unit, and a multi-dimensional weighted fusion unit.

[0028] The visual heatmap generation unit generates a visual attention heatmap of the operator based on the gaze trajectory and gaze duration data provided by the eye-tracking data acquisition module 1. The heatmap generation employs a Gaussian kernel density estimation method, dividing the main control room interface into... A grid cell, typically taken as , This corresponds to a standard display resolution of 1920×1080 pixels. For each fovea... Its contribution to the surrounding grid cells follows a two-dimensional Gaussian distribution. gaze density per grid cell The density values ​​are obtained by summing up the contributions of all fixations, then normalizing them and mapping them to a color heatmap, typically using a gradient spectrum from blue (low density) to red (high density).

[0029] The heatmap entropy calculation unit calculates heatmap entropy based on visual heatmaps. This is used to quantify the degree of visual distraction of the operator. The calculation of heatmap entropy is based on the Shannon entropy concept in information theory. It assumes the gaze point is at a pixel location... The location follows a two-dimensional Gaussian distribution, with the fixation point as the reference point. Centered on, the standard deviation is The Gaussian probability density function is:

[0030] ,

[0031] in, and These are the pixel coordinates on the interface. and The center coordinates of the gaze point. The standard deviation is the range of recognizable pixels under the viewing angle, and is usually set as follows: Pixels correspond to a viewing angle of approximately 1°. Pi is the mathematical constant of a circle.

[0032] For including The gaze sequence of foci points has a joint probability density function that is the superposition of Gaussian distributions of all foci points:

[0033] ,

[0034] in, The total number of fixations. and For the first The coordinates of the gaze point.

[0035] Based on this joint probability density function, heatmap entropy By performing integral calculations across the entire interface area:

[0036] ,

[0037] The integration range covers the entire pixel area of ​​the interface. In the actual calculation, a discretization method is used, dividing the interface into grid cells, and the probability density of each grid cell is summed:

[0038] ,

[0039] in, and These represent the number of grid rows and columns, respectively. For the first Normalized probability density values ​​of individual grid cells. Heatmap entropy. The higher the entropy value, the more scattered the operator's visual attention and the higher the cognitive load; the lower the entropy value, the more concentrated the attention and the relatively lower the cognitive load. According to a large amount of experimental data from nuclear power plant operators, the heat map entropy is usually in the range of 6.5-8.2 under normal operating conditions, the entropy value rises to 8.5-10.3 under abnormal operating conditions, and can reach 10.5-12.1 under emergency shutdown conditions.

[0040] The gaze duration weighting calculation unit calculates gaze duration weighting coefficients based on the operator's gaze duration distribution in different interface areas. The main control room interface is divided according to function into: Calculate the values ​​for each of the following areas (e.g., reactor parameter display area, safety system status area, alarm information area, etc.). gaze duration percentage :

[0041] ,

[0042] in, For belonging to the first The set of fixation points in each region For the first The duration of each gaze point This represents the total number of fixation points.

[0043] Calculate the fixation duration weighting coefficient based on the fixation duration percentage. For critical safety areas (such as reactor coolant pressure, core outlet temperature, and emergency cooling system status), a low percentage of gaze time indicates that operators are not paying sufficient attention to this important information, and the visual load allocation is unreasonable. The gaze time weighting coefficient is calculated using an exponential decay function.

[0044] ,

[0045] in, A collection of critical security areas. For the first The importance weight of each region (set according to human factors engineering standards, usually with a value of 0.15-0.25). This represents the optimal fixation time percentage for this area (usually set to 0.2-0.3%). This is the decay time constant (with a value of 0.1). The further the actual fixation duration deviates from the optimal value, the more... The higher the value, the greater the visual load caused by unreasonable attention allocation.

[0046] The pupil change rate calculation unit calculates the pupil change rate coefficient based on time-series data of pupil diameter. Changes in pupil diameter are an important physiological indicator of operator cognitive load. During task execution, increased cognitive load leads to pupil dilation, while decreased cognitive load causes pupil constriction. The calculation of the pupillary diameter change rate coefficient considers two factors: the rate of change and the magnitude of change in pupil diameter.

[0047] ,

[0048] in, and Weighting coefficients (usually taken as...) , ), This represents the total number of pupil diameter sampling points. For the first Pupil diameter at each sampling time, This is the sampling time interval (usually 16.7ms, corresponding to a 60Hz sampling frequency). and These represent the maximum and minimum pupil diameters during the observation period, respectively. The baseline pupil diameter (measured in a relaxed state, typically 4-5 mm). The first term reflects the average rate of change of pupil diameter, and the second term reflects the magnitude of change of pupil diameter. The higher the value, the more drastic the fluctuation in cognitive load, and the greater the cognitive challenge faced by the operator.

[0049] The multi-dimensional weighted fusion unit will convert the heat map entropy Fixation duration weighting coefficient and pupillary change rate coefficient Weighted fusion is performed to generate a comprehensive attention load index. :

[0050] ,

[0051] in, , and To integrate the weighting coefficients, satisfy the following conditions: The optimal weights are determined through machine learning training on a large amount of operator experimental data, and preferably selected as follows: , , , and These are the minimum and maximum values ​​of the heatmap entropy, used for normalization. (Comprehensive attention load index) The value ranges from 0 to 1. The larger the value, the higher the operator's visual attention load, the greater the cognitive pressure, and the more likely operational errors will occur.

[0052] Comprehensive Attention Load Index As the core input parameter of the interface visual bottleneck recognition module 3, it realizes deep coupling between modules.

[0053] The interface visual bottleneck identification module 3 receives the comprehensive attention load index from the visual attention load quantification assessment module 2, and combines it with the operator's response time data to establish a correlation model between attention entropy and response time. It automatically identifies high-load areas and inefficient information areas in the main control room interface, and marks safety-critical visual bottlenecks in accordance with nuclear power industry human factors engineering standards.

[0054] The interface visual bottleneck identification module 3 includes a response time data acquisition unit, an entropy-time correlation modeling unit, a high-load area identification unit, an inefficient information area identification unit, and a security bottleneck labeling unit.

[0055] The response time data acquisition unit collects operator response time data through standardized information search tasks. These tasks are categorized into three difficulty levels: trend judgment (requiring operators to determine the trend of a parameter, such as whether reactor power is increasing or decreasing), search comparison (requiring operators to find and compare specific outliers among multiple parameters), and inference prediction (requiring operators to infer potential future trends based on the current system state). For each task, the time interval from the task prompt to the operator's correct response is recorded; this is the response time. Simultaneously, the operator's eye-tracking data is recorded during task execution, and the local heatmap entropy of the corresponding interface area is calculated. .

[0056] The entropy-time correlation modeling unit establishes a linear correlation model based on the collected heatmap entropy and response time data. Extensive experimental studies have shown a significant positive correlation between heatmap entropy and response time, with a correlation coefficient... Typically, it is between 0.55 and 0.65. The linear correlation model is expressed as:

[0057] ,

[0058] in, For response time, The local heatmap entropy of the task-related interface area. and These are the model parameters, determined by fitting experimental data using the least squares method. Typical parameter values ​​are... Second, The model shows that for every unit increase in local heatmap entropy, the response time increases by an average of 2.3 seconds. The model's slope varies for tasks of different difficulty levels. There are some differences: the slope of the trend judgment task is about 1.8 seconds, the search comparison task is about 2.5 seconds, and the reasoning prediction task is about 3.2 seconds, reflecting the impact of the cognitive complexity of the task.

[0059] The high-load area identification unit identifies high-load areas in the interface based on an entropy-time correlation model. This is achieved when the local heatmap entropy of a certain interface area... Exceeding the first threshold (Preferably set to 8.5, corresponding to the mean entropy of the heat map under normal operating conditions plus 1 standard deviation), and the corresponding response time Exceeding the second threshold (Preferably set to 25 seconds, corresponding to the upper limit of acceptable response time), then this area is marked as a high-load area. The existence of a high-load area indicates that the information layout or display method of this interface area causes excessive visual distraction for the operator, making information retrieval difficult, and requires optimization design.

[0060] The inefficient information region identification unit identifies inefficient information regions in the interface. This is determined by the local heatmap entropy of a certain interface region. Below the third threshold (Preferably set to 6.0, corresponding to a state of high concentration), but the corresponding response time Still exceeds the fourth threshold (Preferably set to 20 seconds), this area is then marked as an inefficient information area. The existence of an inefficient information area indicates that although the operator focuses their attention on this area, it is still difficult to quickly obtain the required information due to unclear information expression, ambiguous parameter units, insufficient numerical precision, etc., resulting in a prolonged response time. Such areas require improvements in the way information is presented and the precision of its expression.

[0061] The safety bottleneck labeling unit matches identified high-load and inefficient information areas with the critical safety areas of the nuclear power plant, labeling them as safety-critical visual bottlenecks. The critical safety areas on the main control room interface of a nuclear power plant include: safety system status display areas (such as the operating status of the emergency core cooling system, containment spray system, and residual heat removal system), reactor core parameter display areas (such as reactor power, core outlet temperature, coolant flow rate, and coolant pressure), and emergency alarm information display areas (such as reactor emergency shutdown signals, safety system activation signals, and radiation exceedance alarms). If a high-load or inefficient information area is located within one of these critical safety areas, that area is labeled as a safety-critical visual bottleneck.

[0062] The risk level assessment of safety bottlenecks is based on NUREG-0700 (Guidelines for Design Review of Nuclear Power Plant Control Rooms) and IEC 61772 (Design Principles for Nuclear Power Plant Main Control Rooms). The assessment considers three factors: the security importance of the information (classified as A-level – directly related to reactor safe shutdown, B-level – related to accident mitigation, C-level – related to normal operation), the operator's response time margin (the ratio of actual response time to the required response time), and the severity of human error consequences (classified as severe – potentially causing core damage, moderate – potentially causing loss of safety system functionality, minor – affecting operational efficiency but not endangering safety). Based on these three factors, a risk matrix method is used to classify safety bottlenecks into three levels: high-risk, medium-risk, and low-risk. High-risk safety bottlenecks require immediate interface optimization, medium-risk bottlenecks need improvement during the next major overhaul, and low-risk bottlenecks can be gradually improved during routine maintenance.

[0063] The list of safety-critical visual bottlenecks and their risk levels output by the interface visual bottleneck identification module 3 serve as input to the operating condition adaptability optimization module 4, guiding subsequent interface optimization work.

[0064] The operating condition adaptive optimization module 4 receives safety-critical visual bottleneck information from the interface visual bottleneck identification module 3, identifies the current operating condition type of the nuclear power plant, generates adaptive interface layout optimization suggestions based on the visual bottleneck characteristics under different operating conditions, and verifies the optimization effect through a closed-loop feedback mechanism.

[0065] The operating condition adaptive optimization module 4 includes an operating condition identification unit, a historical data analysis unit, an optimization strategy generation unit, an optimization effect prediction unit, and a closed-loop feedback control unit.

[0066] The operating condition identification unit determines the current operating condition type by acquiring real-time operating parameters of the nuclear power plant. Operating parameters include reactor power, primary loop pressure, coolant flow rate, core outlet temperature, and safety system status. The operating condition judgment rules are as follows: If the reactor power is within 95%-105% of the rated power, the primary loop pressure, coolant flow rate, and core outlet temperature are all within the normal range, and there is no safety system activation signal, it is determined to be a normal operating condition; if any critical parameter deviates from the normal range by more than 10% but no reactor protection signal is triggered, or a non-safety level alarm signal appears, it is determined to be an abnormal operating condition; if the reactor protection system activates, an emergency shutdown signal is triggered, or the safety system automatically activates, it is determined to be an emergency shutdown condition.

[0067] The historical data analysis unit extracts historical eye-tracking data for corresponding operating conditions, analyzing the typical visual behavior characteristics and attention allocation patterns of operators under these conditions. Under normal operating conditions, operators' visual attention is primarily focused on routine monitoring parameters, resulting in relatively low heatmap entropy and a low fixation shift frequency, reflecting a relatively relaxed monitoring state. Under abnormal operating conditions, operators' attention quickly shifts to abnormal parameter indicators and related system status display areas, leading to increased heatmap entropy and a higher fixation shift frequency, reflecting an increased cognitive load. Under emergency shutdown conditions, operators' attention is highly focused on safe shutdown procedures and key safety parameters, further increasing heatmap entropy and making the visual search path more complex, reflecting high cognitive stress under emergency conditions.

[0068] The optimization strategy generation unit generates interface layout optimization suggestions based on different operating conditions and identified safety-critical visual bottlenecks. These suggestions include adjustments to the location of key information, color coding optimization, and information display priority adjustment.

[0069] For normal operating conditions, the optimization strategy focuses on reducing the visual complexity of the interface and minimizing interference from non-critical information. Specific measures include: grouping routine monitoring parameters according to system functions and using clear visual separators or background colors to distinguish different functional areas; displaying important parameters that do not require continuous monitoring using trend charts to help operators quickly grasp the changing trends; and reducing the visual weight of non-critical alarm information by using smaller font sizes and lighter colors to avoid distracting the operator.

[0070] For abnormal operating conditions, the optimization strategy focuses on highlighting abnormal parameters and related system statuses to shorten the information search path. Specific measures include: automatically placing parameters that deviate from the normal range in prominent positions and annotating them with high-contrast color coding (such as yellow background with black text); automatically displaying relevant system flowcharts and operating procedure prompts next to abnormal parameters to reduce the number of times operators switch between different interfaces; and enhancing alarm prompts for abnormal states by using dynamic flashing or highlighted borders to attract operator attention, but avoiding excessive use of dynamic effects that could cause visual fatigue.

[0071] For emergency shutdown scenarios, the optimization strategy focuses on restructuring the interface information hierarchy, placing safe shutdown operation procedures and key safety parameters in the center of the main view. Specific measures include: automatically switching to a dedicated emergency operation interface with a simple layout displaying only the most critical information related to safe shutdown; presenting operation steps as a list on the left side of the interface, automatically highlighting the next step upon completion to guide operators in following the procedure; displaying key safety parameters (such as reactor coolant pressure, core water level, and safety injection flow rate) in large font in the center of the interface, using a red-yellow-green three-color coding system to indicate the safety status of the parameters; and removing all non-safety-related information to minimize visual interference and reduce cognitive load.

[0072] The optimization effect prediction unit predicts the effect of interface optimization based on historical data and machine learning models. Predictive indicators include: the expected reduction in the overall attention load index, the expected reduction in key parameter recognition time, and the expected improvement in operation response time. The prediction model uses the Support Vector Regression (SVR) algorithm. Input features include the overall attention load index before optimization, the heatmap entropy distribution of the interface area, and the type and intensity of optimization measures. The output is the predicted value of each indicator after optimization. By comparing the predicted values ​​with the optimization target, the feasibility of the optimization scheme is determined. If the predicted reduction in the overall attention load index is less than 20%, or the reduction in key parameter recognition time is less than 30%, the optimization scheme is considered unsatisfactory, and the optimization strategy needs to be adjusted.

[0073] The closed-loop feedback control unit verifies and iteratively improves the optimization effect. After applying the interface layout optimization suggestions to the main control room interface, the eye-tracking data acquisition module 1 is triggered to re-acquire the operator's eye-tracking data, and the optimized comprehensive attention load index is calculated by the visual attention load quantification and evaluation module 2. and compared with the values ​​before optimization A comparison was made. Simultaneously, the optimized key parameter identification time was measured through standardized task testing. and operation error rate The value was compared with the baseline value before optimization.

[0074] like compared to The decline If the overall attention load index decreases by less than a preset threshold (preferably set at 0.15, meaning a decrease of at least 15%), the optimization effect is considered unsatisfactory, and the optimization strategy needs to be adjusted. Adjustment methods include: increasing the intensity of optimization measures (such as further enlarging the font size of key information or improving color contrast), expanding the optimization scope (adding optimization to more interface areas), or changing the optimization method (such as changing from position adjustment to color coding optimization). The adjusted optimization scheme is then reapplied and evaluated, forming an iterative optimization cycle.

[0075] like Furthermore, the time for identifying key parameters has been significantly reduced. (That is, the recognition time is reduced by at least 35%), and the error rate is reduced by a significant amount. If the error rate is reduced by at least 55%, the optimization effect is considered to have achieved the expected goal. The current interface layout optimization scheme is then solidified and applied as a formal interface design standard in actual operation.

[0076] The closed-loop feedback control unit also includes a long-term effect tracking function. Within six months of the interface optimization solution's implementation, operator eye-tracking data and operational performance data are collected regularly (monthly) to monitor the stability and sustainability of the optimization effect. If a rebound trend is observed in the overall attention load index, or a decline in operational performance indicators, a re-evaluation and adjustment procedure for the optimization solution is initiated.

[0077] The entire system forms a complete closed-loop feedback optimization mechanism: eye-tracking data acquisition module 1 collects raw data → visual attention load quantification assessment module 2 performs quantification assessment → interface visual bottleneck identification module 3 identifies problem areas → operational condition adaptability optimization module 4 generates optimization solutions → optimization solutions are applied → eye-tracking data is re-acquired → effect verification → iterative optimization or solution solidification. This closed-loop mechanism ensures the scientific nature and effectiveness of interface optimization work, and achieves continuous improvement based on empirical data.

[0078] This embodiment provides a specific application scenario to illustrate the actual application effect of the method and system of the present invention.

[0079] A nuclear power plant used the system of this invention to evaluate and optimize the interface of the SPDS safety parameter display system in the main control room. In the initial evaluation phase, 10 certified operators were invited to participate in the experiment. Each operator completed 30 standardized information search tasks on the SPDS interface, including 10 trend judgment tasks, 10 search comparison tasks, and 10 inference prediction tasks. During the experiment, eye-tracking data acquisition module 1 collected the operators' eye-tracking data at a frequency of 60Hz, recording the gaze point trajectory, gaze duration, and pupil diameter changes.

[0080] The visual attention load quantification and assessment module 2 analyzes the collected eye-tracking data to generate a visual attention heatmap. The heatmap shows that the operator's attention is primarily focused on the upper-middle area of ​​the interface, which displays core reactor parameters. The calculated average heatmap entropy... The value was 9.8, higher than the typical value of 8.2 under normal operating conditions, indicating relatively scattered visual attention. Gaze duration analysis showed that the operator's gaze time in critical safety parameter areas (such as emergency core cooling system flow rate and containment pressure) accounted for only 12%, far below the optimal value of 25%. The calculated gaze duration weighting coefficient... A value of 0.45 indicates inappropriate attention allocation. Pupil diameter data shows that the average rate of change in pupil diameter during task execution was 0.08 mm / s, with a change amplitude of 1.2 mm. The calculated pupil change rate coefficient... A score of 0.35 reflects a moderate level of cognitive load fluctuation. Combining these three indicators, the comprehensive attention load index is calculated. The value is 0.68, which is at a medium-to-high load level.

[0081] The interface visual bottleneck identification module 3 analyzes task response time. Results show that the average response time for trend judgment tasks is 18 seconds, for search and comparison tasks it's 32 seconds, and for inference and prediction tasks it's 45 seconds. Based on the entropy-time correlation model, three high-load areas were identified: the auxiliary water supply system flow display area (local heatmap entropy 9.5, response time 28 seconds), the containment isolation valve status display area (local heatmap entropy 10.2, response time 35 seconds), and the emergency power switching status area (local heatmap entropy 9.8, response time 30 seconds). Two inefficient information areas were also identified: the main steam flow display area (local heatmap entropy 5.8, response time 22 seconds) and the condensate pump operating status area (local heatmap entropy 6.2, response time 24 seconds). Matching these areas with critical safety areas revealed that the auxiliary water supply system flow display area and the containment isolation valve status display area are safety-critical visual bottlenecks, assessed as high-risk according to the NUREG-0700 standard, requiring priority optimization.

[0082] The Operational Condition Adaptability Optimization Module 4 generates interface optimization suggestions for identified visual bottlenecks. For the auxiliary feedwater system flow display area, optimization measures include: moving the parameter from the lower right corner of the interface to a prominent position in the upper middle, displaying it alongside core reactor parameters; increasing the font size of the numerical display from 14 to 20, and changing the color from gray to black to improve readability; adding trend arrows next to the numerical values ​​to visually show the direction of flow changes. For the containment isolation valve status display area, optimization measures include: changing the status of multiple isolation valves from a list format to a graphical display, using green to indicate valve open and red to indicate valve closed, making the status clear at a glance; adding an abnormal valve status alarm function, automatically displaying a prominent prompt when any isolation valve fails to operate as required by regulations; displaying a simplified containment system flowchart next to the graphics to help operators understand the function and location of the isolation valves. For inefficient information areas, optimization measures include improving the accuracy of numerical display and unit annotations, and adding data update timestamps to ensure the accuracy and timeliness of information.

[0083] The optimization effect prediction unit, based on a machine learning model, predicted that after implementing the above optimization measures, the overall attention load index would decrease to 0.52 (a 24% reduction), the key parameter identification time would be shortened to 11 seconds (a 39% reduction), and the operation error rate would be reduced to 1.2% (a 67% reduction). The prediction results met the optimization objectives, and the optimization plan was approved for implementation.

[0084] The optimized solution was tested for two weeks on a simulator in the main control room of the nuclear power plant. During the test, the original 10 operators were invited to complete the same standardized tasks, and optimized eye-tracking and performance data were collected. The results showed that the optimized overall attention load index decreased to 0.49 (an actual reduction of 28%, better than the predicted value), the key parameter recognition time was shortened to 10 seconds (an actual reduction of 44%), and the operation error rate decreased to 0.8% (an actual reduction of 73%). All indicators met or exceeded the expected targets. The operators' subjective evaluation questionnaire showed that 95% of the operators believed that the optimized interface was clearer and easier to read, and the information acquisition efficiency was significantly improved.

[0085] Based on the successful experience gained from simulator testing, the nuclear power plant applied the optimized scheme in the actual main control room. A six-month follow-up evaluation showed that the overall attention load index remained around 0.51, operational performance indicators were stable, and no rebound occurred. The plant's human error incident rate decreased by 58%, and the safety and efficiency of main control room operations were significantly improved.

[0086] This embodiment fully verifies the effectiveness and practicality of the method and system of the present invention, and provides strong technical support for the scientific design and continuous optimization of the main control room interface of nuclear power plants.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing and optimizing visual attention load on the interface of a nuclear power plant's main control room, characterized in that... Includes the following steps: The eye-tracking data acquisition step involves collecting eye-tracking data from operators interacting with the control system in the main control room of the nuclear power plant. The eye-tracking data includes fixation point trajectory, fixation duration, pupil diameter changes, and saccade path. The visual attention load quantification assessment step involves generating a visual attention heatmap based on the eye movement data, calculating the heatmap entropy, fixation duration weighting coefficient, and pupil change rate coefficient, and generating a comprehensive attention load index through a multi-dimensional weighted fusion algorithm. The interface visual bottleneck identification step establishes a correlation model between attention entropy and response time based on the comprehensive attention load index and response time data of the interface area, identifies high-load areas and inefficient information areas in the interface, and automatically marks safety-critical visual bottlenecks in conjunction with the nuclear power industry human factors engineering standards. The operational condition adaptive optimization step involves identifying the current operational condition type of the nuclear power plant, which includes normal operation, abnormal operation, and emergency shutdown. Based on the visual bottleneck characteristics under different operational condition types, interface layout optimization suggestions are generated. These suggestions include key information location adjustment schemes, color coding optimization schemes, and information display priority adjustment schemes. After applying the interface layout optimization suggestions to the main control room interface, eye-tracking data is re-collected for iterative evaluation, forming a closed-loop optimization.

2. The method according to claim 1, characterized in that, In the visual attention load quantification assessment step, the multidimensional weighted fusion algorithm includes: Based on the visual attention heatmap, the interface is divided into multiple grid units, and the probability density function of the gaze point distribution is calculated for each grid unit to generate the heatmap entropy. Based on the gaze duration, calculate the gaze duration percentage of each interface region and generate the gaze duration weighting coefficient. Based on the change in pupil diameter, the rate of change and the magnitude of change in pupil diameter are calculated, and the pupil change rate coefficient is generated. The heatmap entropy, the gaze duration weighting coefficient, and the pupil change rate coefficient are weighted and fused to generate the comprehensive attention load index.

3. The method according to claim 2, characterized in that, The heatmap entropy is calculated as follows: Assuming that the gaze point follows a two-dimensional Gaussian distribution at the pixel position, a joint probability density function is constructed based on the coordinates of all gaze points. The information entropy value is calculated by integrating the probability density function. The information entropy value is the heatmap entropy. The larger the heatmap entropy value, the higher the degree of visual attention distraction.

4. The method according to claim 1, characterized in that, In the interface visual bottleneck identification step, the correlation model between attention entropy and response time is established as follows: Collect response time data of operators when performing information search tasks, including trend judgment tasks, search comparison tasks, and inference and prediction tasks; Calculate the attention entropy value and response time of the interface area corresponding to each task, and establish a linear correlation model between attention entropy value and response time; When the attention entropy value of an interface region exceeds the first threshold and the corresponding response time exceeds the second threshold, the interface region is marked as a high-load region. When the attention entropy value of an interface area is lower than the third threshold but the response time exceeds the fourth threshold, the interface area is marked as an inefficient information area.

5. The method according to claim 4, characterized in that, The labeling methods for the safety-critical visual bottlenecks include: The high-load area and the inefficient information area are matched with the nuclear power plant safety system status display area, reactor parameter display area and emergency alarm information display area; If the high-load area or the inefficient information area is located in the safety system status display area, the reactor parameter display area, or the emergency alarm information display area, then the area is marked as a safety-critical visual bottleneck. Based on the NUREG-0700 and IEC 61772 standards, the impact of the safety-critical visual bottlenecks on operator reaction speed and decision-making accuracy is assessed, and a safety risk level is generated.

6. The method according to claim 1, characterized in that, In the operational condition adaptive optimization step, the method for identifying the operational condition type is as follows: Acquire real-time operating parameters of the nuclear power plant, including reactor power, primary loop pressure, coolant flow rate, and safety system status; Based on the real-time operating parameters, determine whether the nuclear power plant is in normal operating condition, abnormal operating condition, or emergency shutdown condition. For different operating conditions, historical eye-tracking data for the corresponding conditions are extracted to analyze the typical visual behavior characteristics and attention allocation patterns of operators under those conditions.

7. The method according to claim 6, characterized in that, The methods for generating the interface layout optimization suggestions include: For normal operating conditions, optimize the display layout of routine monitoring parameters and reduce the visual weight of non-critical information; For abnormal operating conditions, highlight abnormal parameters and indicators, enhance the color coding contrast of abnormal states, and shorten the search path for key information. For emergency shutdown scenarios, the interface information hierarchy is restructured, placing safe shutdown operation steps and key safety parameters in the center of the main view area to reduce visual search time; The optimization suggestions for each working condition are quantitatively evaluated to predict the reduction in the overall attention load index and the shortening of the operation response time after optimization.

8. The method according to claim 1, characterized in that, The closed-loop optimization is implemented as follows: After applying the interface layout optimization suggestions, the operator's eye movement data was re-collected, and the optimized comprehensive attention load index was calculated. Compare the overall attention load index, key parameter recognition time, and operation error rate before and after optimization; If the decrease in the optimized overall attention load index is less than the preset threshold, the parameter weights of the optimization strategy will be adjusted, and interface layout optimization suggestions will be regenerated. If the reduction in key parameter recognition time and the decrease in operation error rate both reach the preset targets after optimization, then the current interface layout optimization scheme will be solidified.

9. The method according to claim 1, characterized in that, The nuclear power plant main control room control system includes a DCS distributed control system, an SPDS safety parameter display system, and an alarm management interface. In the eye-tracking data acquisition step, the type of control system currently being interacted with by the operator, the interface switching time, and the type of operation task are recorded synchronously.

10. A system for assessing and optimizing visual attention load on the interface of a nuclear power plant main control room, used to implement the method described in any one of claims 1-9, characterized in that, include: An eye-tracking data acquisition module is used to collect eye-tracking data of operators when interacting with the control system in the main control room of a nuclear power plant. The eye-tracking data includes fixation point trajectory, fixation duration, pupil diameter changes, and saccade path. The visual attention load quantification and assessment module is connected to the eye movement data acquisition module. It is used to generate a visual attention heatmap based on the eye movement data, calculate the heatmap entropy, fixation duration weighting coefficient and pupil change rate coefficient, and generate a comprehensive attention load index through a multi-dimensional weighted fusion algorithm. The interface visual bottleneck identification module is connected to the visual attention load quantification and evaluation module. It is used to establish a correlation model between attention entropy and response time based on the comprehensive attention load index and the response time data of the interface area, identify high-load areas and inefficient information areas in the interface, and automatically label safety-critical visual bottlenecks in combination with the nuclear power industry human factors engineering standards. The operating condition adaptive optimization module, connected to the interface visual bottleneck recognition module, is used to identify the current operating condition type of the nuclear power plant. The operating condition types include normal operating condition, abnormal operating condition, and emergency shutdown operating condition. For the visual bottleneck characteristics under different operating condition types, interface layout optimization suggestions are generated. The interface layout optimization suggestions include key information position adjustment scheme, color coding optimization scheme, and information display priority adjustment scheme. After the interface layout optimization suggestions are applied to the main control room interface, the eye-tracking data acquisition module is triggered to re-acquire eye-tracking data for iterative evaluation, forming a closed-loop optimization.

Images