A method for evaluating risk of a node in a diffusion network outside a nuclear accident site

Abstract

The application discloses a kind of nuclear accident off-site diffusion road network node risk assessment method, belong to information processing technical field.For solving the problem of accurate risk assessment of nuclear accident off-site road network node.The application establishes intersection node adjacency matrix, surveys the actual road network passing distance of each connected intersection node pair, obtains the standardized passing distance of all connected intersection node pairs;High-dimensional feature vector extraction and minute-level aggregation are carried out on video image data to obtain minute-level average high-dimensional feature vector and corresponding nuclear accident off-site radiation dose value are reduced in dimension, and the low-dimensional effective feature vector obtained is used to construct the basic pedestrian state assessment score, and then the basic pedestrian state assessment score is dynamically corrected using the relative change rate of nuclear accident off-site radiation dose value to obtain the corrected pedestrian state assessment score;The dynamic influence coefficient of each intersection node is calculated, and then the cross-node coupling correction value of the intersection node is calculated;Calculate bimodal fusion nuclear accident off-site radiation risk value.

Description

Technical Field

[0001] This invention belongs to the field of information processing technology, specifically relating to a method for risk assessment of road network nodes in the off-site spread of nuclear accidents. Background Technology

[0002] Following a nuclear accident, radioactive materials spread off-site via the atmosphere and pedestrian traffic. As the core carriers for personnel evacuation and emergency rescue, accurate radiation risk assessment at each intersection (node) of the urban road network is a crucial prerequisite for emergency decision-making. In the off-site nuclear accident monitoring system, radiation dose monitoring data and video surveillance data at key intersections are two types of fundamental data that are universally accessible, highly real-time, and of stable quality: radiation dose data is collected by fixed-point sensors in the off-site radiation monitoring network, covering major urban road network nodes, and consists of discrete values ​​at the minute level, directly reflecting the degree of radiation hazard at each node; video image data is collected by existing high-definition cameras in the urban road network, providing continuous frame visual information, and can capture real-time pedestrian dynamics, congestion levels, and other conditions at nodes. Furthermore, the deployment of both types of equipment at road network nodes requires no additional hardware investment, offering advantages such as low cost and ease of implementation.

[0003] Current methods for assessing radiation risks at nuclear accident sites still have significant technical shortcomings: First, the assessment data sources are limited, relying heavily on discrete numerical data from fixed-point radiation dose sensors. This lack of real-time awareness of pedestrian dynamics and crowd distribution at nodes leads to risk assessments that only consider radiation dose itself, ignoring the cumulative risks of prolonged radiation exposure due to factors such as crowding and congestion. Second, multimodal data fusion is challenging. Existing technologies require the integration of multiple types of data, including radiation, meteorological, traffic, and personnel location data, to achieve comprehensive risk assessment. Some data collection is difficult, lacks real-time performance, and suffers from modal heterogeneity and spatiotemporal asynchrony, making practical application difficult. Third, the risk transmission between road network nodes is not adequately considered. The impact of radiation diffusion and pedestrian shifts from adjacent nodes on target nodes is not reflected in the road network connectivity, resulting in assessment results that are disconnected from actual accident scenarios and cannot provide a basis for precise evacuation scheduling and risk management. Summary of the Invention

[0004] The problem this invention aims to solve is to achieve accurate risk assessment of road network nodes outside nuclear accident sites, and proposes a risk assessment method for road network nodes in the context of nuclear accident spread.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for risk assessment of off-site nuclear accident propagation intersection nodes includes the following steps:

[0007] S1. Based on the synchronous deployment of nuclear accident site radiation dose monitoring sensors and high-definition cameras at each intersection node, establish an intersection node adjacency matrix, survey the actual road network travel distance of each connected intersection node pair, and perform standardization processing to obtain the standardized travel distance of all connected intersection node pairs.

[0008] S2. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node, extract high-dimensional feature vectors and aggregate them at the minute level to obtain the minute-level average high-dimensional feature vector and the corresponding radiation dose value outside the nuclear accident site;

[0009] S3. The minute-level average high-dimensional feature vector obtained in step S2 is subjected to dimensionality reduction processing. The obtained low-dimensional effective feature vector is used to construct a basic pedestrian status assessment score. Then, the basic pedestrian status assessment score is dynamically corrected using the relative change rate of radiation dose values ​​outside the nuclear accident site to obtain the corrected pedestrian status assessment score.

[0010] S4. Based on the intersection node adjacency matrix obtained in step S1, the nuclear accident field radiation dose value obtained in step S2, and the corrected pedestrian state assessment score obtained in step S3, calculate the dynamic influence coefficient of each intersection node, and then calculate the cross-node coupling correction value of the intersection node.

[0011] S5. Normalize the off-site radiation dose value of the nuclear accident obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3, and then calculate the dual-modal fusion off-site radiation risk value of the nuclear accident by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4.

[0012] Furthermore, the specific implementation method of step S1 includes the following steps:

[0013] S1.1. Establish the adjacency matrix of intersection nodes, and obtain

[0014]

[0015] in, for An adjacency matrix of intersection nodes of order 1. These are elements in the adjacency matrix of intersection nodes. , These represent the intersection node numbers and the connectivity between intersection nodes. If the value is 1, then the intersection nodes are not connected. Set to 0;

[0016] S1.2. Survey all Actual travel distance of intersection node pairs with a value of 1 Then, the maximum actual travel distance between all connected intersection nodes in the off-site monitoring road network of the nuclear accident site is calculated. Standardization processing is performed to obtain the standardized travel distance for intersection node pairs. .

[0017] Furthermore, the specific implementation method of step S2 includes the following steps:

[0018] S2.1. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node to obtain the first... The first intersection node Minute-level off-site radiation dose scalar data of nuclear accidents , No. Video image data of the kth frame of the intersection node ;

[0019] S2.2. Perform image preprocessing, image segmentation, positional encoding, sequence construction, multi-head self-attention calculation, and feature aggregation on the video image data obtained in step S2.1 to obtain a high-dimensional feature vector. The expression for feature aggregation is:

[0020]

[0021] in, For the first Node number High-dimensional feature vectors of frame videos, Feature extraction operator for visual Transformer;

[0022] S2.3. Perform dimension-wise mean aggregation on the high-dimensional feature vectors obtained in step S2.2 to obtain minute-level averaged high-dimensional feature vectors. The expression is:

[0023] .

[0024] Furthermore, the specific implementation method of step S3 includes the following steps:

[0025] S3.1. Calculate the minute-level average high-dimensional feature vector obtained in step S2. The covariance matrix is ​​obtained by performing eigenvalue decomposition on it, and selecting the eigenvectors corresponding to the first 16 largest eigenvalues ​​as principal components, thus obtaining 16-dimensional low-dimensional effective eigenvectors. ;

[0026] S3.2. Collect the first... Low-dimensional feature vector of normal pedestrian state at each intersection node when there is no nuclear accident Then, a basic pedestrian state assessment score is constructed, expressed as:

[0027]

[0028] in, For the first The first intersection node A basic pedestrian status assessment score within minutes. It is a function with maximum value. It is a 2-norm;

[0029] S3.3. Calculate the first... Node number Relative change rate of off-site radiation dose per minute during a nuclear accident The formula is as follows:

[0030]

[0031] in, The relative rate of change of radiation dose. For the first The first intersection node Minute-level scalar data of radiation dose outside the nuclear accident site;

[0032] use The absolute value of the value is used to dynamically adjust the basic pedestrian status assessment score, resulting in the adjusted pedestrian status assessment score. , For the first The first intersection node The adjusted pedestrian status assessment score for the minute.

[0033] Furthermore, the specific implementation method of step S4 includes the following steps:

[0034] S4.1. Calculate the dynamic impact coefficient of each intersection node. The calculation formula is as follows:

[0035]

[0036] in, For the first The intersection node corresponds to the first The first intersection node The dynamic impact coefficient per minute;

[0037] S4.2. Based on the dynamic influence coefficient of each intersection node obtained in step S4.1, calculate the cross-node coupling correction value of the intersection node. The calculation formula is as follows:

[0038]

[0039] in, For the first Node number Minute-level cross-node coupling correction value, It is the hyperbolic tangent function.

[0040] Furthermore, the specific implementation method of step S5 includes the following steps:

[0041] S5.1. Normalize the off-site radiation dose value obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3 to obtain the... The first intersection node Normalized value of off-site radiation dose per minute in a nuclear accident , No. The first intersection node Normalized value of pedestrian status assessment score per minute ;

[0042] S5.2. Calculate the attention weights, including the attention weights for the radiation dose factor, the pedestrian state factor, and the cross-node coupling correction factor. The calculation formula is as follows:

[0043]

[0044]

[0045]

[0046] in, For the first The first intersection node Attention weight of the radiation dose factor per minute. It is the minimum value. For the first The first intersection node Attention weights for pedestrian state factors at the minute level. Attention weights for cross-node coupling correction factors;

[0047] S5.3. Calculate the dual-modal fusion nuclear accident off-site radiation risk value by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4. The calculation formula is as follows:

[0048]

[0049] in, For the first Node number Minutes of off-site radiation risk value for a dual-modal fusion nuclear accident.

[0050] The beneficial effects of this invention are:

[0051] The present invention discloses a risk assessment method for road network nodes outside nuclear accident sites, which breaks through the technical bottlenecks of traditional assessment methods. Through multimodal data feature extraction, spatiotemporal synchronization, and cross-node coupling fusion, it achieves accurate risk assessment of road network nodes outside nuclear accident sites. The entire process requires no other auxiliary data and completes the entire calculation from data collection to risk value output using only two types of commonly available basic data. Multimodal heterogeneous data directly supports risk decision-making for road network nodes outside nuclear accident sites.

[0052] This invention discloses a risk assessment method for road intersection nodes in the context of nuclear accident spillover. Primarily targeting emergency scenarios involving the spillover of radioactive materials from nuclear accident sites in urban road networks, it addresses the modal heterogeneity and spatiotemporal asynchrony issues present in the monitoring data and video image data of nuclear accident spillover at various road intersections (nodes) within the urban road network. It also considers the connectivity between road network nodes and the cross-node transmission impact of radioactive materials and pedestrian flow. Through a five-step progressive calculation process—road network topology calibration, dual-modal data feature extraction, node pedestrian status assessment, connected node coupling correction, and multi-factor risk fusion—it achieves a quantitative assessment of the nuclear accident spillover risk at any node in the urban road network. This method uses urban road intersections as basic nodes, with nuclear accident spillover radiation dose monitoring sensors and high-definition cameras simultaneously deployed at each intersection node. First, it completes the calibration of the road network node topology and basic parameters. Then, based on the two types of core multimodal data collected from the nodes, it completes the full-link calculation, ultimately outputting a quantitative risk value for nuclear accident spillover radiation at each node. This provides accurate and practical technical support for emergency evacuation of personnel, road network traffic control, and rescue route planning in the event of a nuclear accident.

[0053] The present invention discloses a risk assessment method for road intersection nodes in the off-site spread of nuclear accidents. It completes the risk assessment based solely on two core data types: radiation dose monitoring data and video image data from off-site nuclear accidents. It does not require the integration of other auxiliary data such as meteorological, traffic, and personnel positioning data. Both types of data are basic data in off-site nuclear accident monitoring and urban road network management. They are collected by existing sensors and cameras, and are characterized by widespread deployment, real-time collection, and stable data. No additional hardware equipment or data collection system is required, which greatly reduces the hardware investment and operation and maintenance costs for technology implementation. It can be quickly promoted and applied in off-site nuclear accident monitoring road networks, solving the pain points of traditional methods such as high data collection difficulty and poor real-time performance.

[0054] The present invention provides a risk assessment method for road intersection nodes in the context of nuclear accident spillover. This method overcomes the limitations of traditional single radiation dose data assessment by deeply integrating radiation dose hazard data with personnel status data reflected in video images. It also considers the connectivity between road network nodes outside the nuclear accident site and designs dynamic coupling computation with spatiotemporal attenuation and feature attention-weighted fusion computation. This method accurately captures the synergistic risk of radiation dose and personnel status at the target node, while fully reflecting the cross-node risk transmission caused by the spread of radioactive materials and the transfer of people between adjacent nodes. The assessment results not only reflect the radiation hazard itself but also consider the cumulative risk of prolonged radiation exposure time due to personnel congestion and overcrowding. This method is highly consistent with the actual scenario of nuclear accident spillover outside the site, significantly improving the accuracy and reliability of radiation risk assessment.

[0055] This invention discloses a risk assessment method for road intersection nodes at the site of a nuclear accident. It employs a five-step progressive technical chain, encompassing road network topology calibration, dual-modal data acquisition and feature extraction, single-node state quantification, cross-node coupling correction, and final risk value calculation. This technical chain is complete, reproducible, and operable. The final output of the quantitative risk values ​​for each node can directly form a road network risk statistics table, providing clear and accurate quantitative basis for emergency evacuation of personnel outside the nuclear accident site (e.g., prioritizing evacuation from high-risk nodes), road network traffic control (e.g., closing passage to high-risk nodes), and rescue route planning (e.g., avoiding high-risk nodes). This method achieves seamless integration from dual-modal data acquisition to emergency decision support applications, resolving the disconnect between traditional assessment methods and actual emergency decision-making. Attached Figure Description

[0056] Figure 1 This is a flowchart of a method for risk assessment of off-site diffusion intersection nodes in a nuclear accident, as described in this invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0058] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0059] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 Detailed explanation is as follows:

[0060] Example 1:

[0061] A method for risk assessment of off-site nuclear accident propagation intersection nodes includes the following steps:

[0062] S1. Based on the synchronous deployment of nuclear accident site radiation dose monitoring sensors and high-definition cameras at each intersection node, establish an intersection node adjacency matrix, survey the actual road network travel distance of each connected intersection node pair, and perform standardization processing to obtain the standardized travel distance of all connected intersection node pairs.

[0063] Furthermore, the specific implementation method of step S1 includes the following steps:

[0064] S1.1. Establish the adjacency matrix of intersection nodes, and obtain

[0065]

[0066] in, for An adjacency matrix of intersection nodes of order 1. These are elements in the adjacency matrix of intersection nodes. , These represent the intersection node numbers and the connectivity between intersection nodes. If the value is 1, then the intersection nodes are not connected. Set to 0;

[0067] S1.2. Survey all Actual travel distance of intersection node pairs with a value of 1 Then, the maximum actual travel distance between all connected intersection nodes in the off-site monitoring road network of the nuclear accident site is calculated. Standardization processing is performed to obtain the standardized travel distance for intersection node pairs. .

[0068] Furthermore, the maximum actual travel distance between all connected nodes in the off-site monitoring road network for nuclear accidents was statistically analyzed. (Unit: m), for each valid After standardization, the formula is as follows:

[0069]

[0070] in, For the first The intersection node and the first Standardized travel distance at each intersection node, dimensionless, range of values. ; For the first , Actual road network travel distance at a node (unit: m); The maximum actual travel distance between all connected nodes (in meters) is a fixed scalar.

[0071] S2. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node, extract high-dimensional feature vectors and aggregate them at the minute level to obtain the minute-level average high-dimensional feature vector and the corresponding radiation dose value outside the nuclear accident site;

[0072] Furthermore, this step is a fundamental step in single-node data processing, which focuses on solving the modal heterogeneity problem between scalar data of radiation dose at the nuclear accident site and non-numerical data of video images. Based on the premise that "the default data is normal and valid", there is no need for tedious calculations such as outlier removal and noise reduction. It only completes the synchronous acquisition of the two types of core raw data, the extraction of high-dimensional feature vectors of video images, and the minute-level aggregation of video frame-level features, so as to achieve time synchronization with radiation dose data.

[0073] Furthermore, the specific implementation method of step S2 includes the following steps:

[0074] S2.1. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node to obtain the first... The first intersection node Minute-level off-site radiation dose scalar data of nuclear accidents , No. Video image data of the kth frame of the intersection node ;

[0075] S2.2. Perform image preprocessing, image segmentation, positional encoding, sequence construction, multi-head self-attention calculation, and feature aggregation on the video image data obtained in step S2.1 to obtain a high-dimensional feature vector. The expression for feature aggregation is:

[0076]

[0077] in, For the first Node number High-dimensional feature vectors of frame videos, Feature extraction operator for visual Transformer;

[0078] Furthermore, regarding the first Each frame of video image within minutes ( (1800 represents the total number of frames in a 1-minute video: 30 frames / second × 60 seconds) Perform high-dimensional feature extraction using Visual Transformer (ViT). The specific practical steps are as follows:

[0079] S2.2.1. Image Preprocessing: Preprocessing single-frame video images Resize to 224×224 pixels and normalize the pixel values ​​(map the pixel value range to...). );

[0080] S2.2.2. Image Patching: The preprocessed image is divided into 196 image patches of size 16×16 pixels, and each image patch is flattened into a 768-dimensional vector;

[0081] S2.2.3. Add positional encoding: Add a learnable positional encoding vector to each image patch vector to preserve the spatial positional information of the image;

[0082] S2.2.4. Constructing the sequence: Add a classification label (CLS) vector before all image patch vectors to form a 197×768-dimensional input sequence;

[0083] S2.2.5. Multi-head self-attention computation: The input sequence is fed into the multi-head self-attention layer, and the global dependencies between image patches are captured through parallel computation by 8 attention heads, and a 197×768-dimensional feature sequence is output.

[0084] S2.2.6. Feature Aggregation: Extract the output vector corresponding to the classification label (CLS), and map it to a 512-dimensional vector through a fully connected layer, which is the high-dimensional feature vector of the image frame. , For feature dimensions.

[0085] S2.3. Perform dimension-wise mean aggregation on the high-dimensional feature vectors obtained in step S2.2 to obtain minute-level averaged high-dimensional feature vectors. The expression is:

[0086] .

[0087] S3. The minute-level average high-dimensional feature vector obtained in step S2 is subjected to dimensionality reduction processing. The obtained low-dimensional effective feature vector is used to construct a basic pedestrian status assessment score. Then, the basic pedestrian status assessment score is dynamically corrected using the relative change rate of radiation dose values ​​outside the nuclear accident site to obtain the corrected pedestrian status assessment score.

[0088] Furthermore, this step is a single-node state quantization step, which mainly solves the problem that the high-dimensional feature vector of the video has too high a dimension and cannot be fused with scalar radiation dose data. By reducing the dimension, the core effective information of the video features is extracted, and then the node pedestrian state assessment score is quantitatively calculated based on the dimensionality-reduced feature vector, thereby realizing the transformation of abstract image features into calculable quantitative scores.

[0089] Furthermore, the specific implementation method of step S3 includes the following steps:

[0090] S3.1. Calculate the minute-level average high-dimensional feature vector obtained in step S2. The covariance matrix is ​​obtained by performing eigenvalue decomposition on it, and selecting the eigenvectors corresponding to the first 16 largest eigenvalues ​​as principal components, thus obtaining 16-dimensional low-dimensional effective eigenvectors. The formula is as follows:

[0091]

[0092] in, For the first Node number A low-dimensional, dimensionless, effective feature vector per minute; To use the dimensionality reduction operator, follow the practical steps outlined above; This is the average high-dimensional feature vector output from step 2; The dimension after dimensionality reduction; The dimension of the original high-dimensional feature vector.

[0093] S3.2. Collect the first... Low-dimensional feature vector of normal pedestrian state at each intersection node when there is no nuclear accident Then, a basic pedestrian state assessment score is constructed, expressed as:

[0094]

[0095] in, For the first The first intersection node A basic pedestrian status assessment score within minutes. It is a function with maximum value. It is a 2-norm;

[0096] Furthermore, Given a 16-dimensional feature vector, calculate and L2 norm (Reflecting the degree of characteristic deviation), then divided by the historical maximum value of that deviation. (Fixed scalar) to obtain the basic pedestrian state assessment score Dimensionless, range of values The closer the value is to 1, the more abnormal the pedestrian state of the node; the L2 norm calculation operator, the calculation rule is as follows: ( for (dimensional vector); This is the low-dimensional effective feature vector calculated in this step; The low-dimensional feature vector of a normal pedestrian state; This represents the historical maximum value of the characteristic deviation.

[0097] S3.3. Calculate the first... Node number Relative change rate of off-site radiation dose per minute during a nuclear accident The formula is as follows:

[0098]

[0099] in, The relative rate of change of radiation dose. This indicates an increase in dosage. This indicates a decrease in dosage. For the first The first intersection node Minute-level scalar data of radiation dose outside the nuclear accident site;

[0100] use The absolute value of the value is used to dynamically adjust the basic pedestrian status assessment score, resulting in the adjusted pedestrian status assessment score. , For the first The first intersection node The adjusted pedestrian status assessment score for the minute.

[0101] Furthermore, using The absolute value of the basic pedestrian state assessment score Dynamic correction is performed by amplifying the impact of dose changes on pedestrian status using a natural exponential function, resulting in a corrected pedestrian status assessment score. ;

[0102] S4. Based on the intersection node adjacency matrix obtained in step S1, the nuclear accident field radiation dose value obtained in step S2, and the corrected pedestrian state assessment score obtained in step S3, calculate the dynamic influence coefficient of each intersection node, and then calculate the cross-node coupling correction value of the intersection node.

[0103] Furthermore, this step is a core component of cross-node data fusion, fundamentally addressing the issue of traditional assessment methods neglecting risk transmission between nodes in the off-site road network during nuclear accidents. Based on calibrated topology parameters, it fuses the dual-modal data of all connected nodes to calculate the... The cross-node coupling correction value of the node reflects the cross-node impact of radioactive material diffusion and human movement.

[0104] Furthermore, the specific implementation method of step S4 includes the following steps:

[0105] S4.1. Calculate the dynamic impact coefficient of each intersection node. The calculation formula is as follows:

[0106]

[0107] in, For the first The intersection node corresponds to the first The first intersection node The dynamic impact coefficient per minute;

[0108] Furthermore, Dimensionless, range of values ,satisfy ; These are elements of the adjacency matrix; for , Standardized travel distances between nodes; This is the spatial decay term; This is the sum of the spatial decay terms for all connected nodes; For the first Node number Radiation dose value per minute; For the first Node number The corrected pedestrian status assessment score after minutes; This is the sum of (radiation dose + pedestrian state score) for all connected nodes; , For node sequence number ( );

[0109] S4.2. Based on the dynamic influence coefficient of each intersection node obtained in step S4.1, calculate the cross-node coupling correction value of the intersection node. The calculation formula is as follows:

[0110]

[0111] in, For the first Node number Minute-level cross-node coupling correction value, It is the hyperbolic tangent function.

[0112] Furthermore, based on the dynamic influence coefficient Nonlinear coupling operations are performed on the bimodal data of all connected nodes. First, the square root of the product of radiation dose and pedestrian state score of each connected node is calculated (co-quantization). Then, the results are summed according to weights. Finally, the amplitude is limited by the hyperbolic tangent function to obtain the first... Cross-node coupling correction value of nodes ;

[0113] S5. Normalize the off-site radiation dose value of the nuclear accident obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3, and then calculate the dual-modal fusion off-site radiation risk value of the nuclear accident by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4.

[0114] Furthermore, this step is the final output stage of the method, which corely solves the problem of deep fusion of the target node's own dual-modal data and cross-node coupling correction values. Through weighted fusion operations with feature attention, the first output is obtained. Quantitative risk value of off-site radiation from nuclear accidents at nodes.

[0115] Furthermore, the specific implementation method of step S5 includes the following steps:

[0116] S5.1. Normalize the off-site radiation dose value obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3 to obtain the... The first intersection node Normalized value of off-site radiation dose per minute in a nuclear accident , No. The first intersection node Normalized value of pedestrian status assessment score per minute ;

[0117] Furthermore, the radiation dose value of the target node itself. Corrected pedestrian status assessment score To normalize and eliminate dimensional differences, the formula is as follows:

[0118]

[0119] in, This is a normalized value for radiation dose, dimensionless, with a range of values. ; The normalized value of the pedestrian status assessment score is dimensionless and has a range of values. ; The historical maximum radiation dose (unit: μSv / h) of all nodes in the off-site monitoring network for nuclear accidents is a fixed scalar quantity. The historical maximum value of the pedestrian state assessment score after correction for all nodes is a fixed scalar.

[0120] S5.2. Calculate the attention weights, including the attention weights for the radiation dose factor, the pedestrian state factor, and the cross-node coupling correction factor. The calculation formula is as follows:

[0121]

[0122]

[0123]

[0124] in, For the first The first intersection node Attention weight of the radiation dose factor per minute. It is the minimum value. For the first The first intersection node Attention weights for pedestrian state factors at the minute level. Attention weights for cross-node coupling correction factors;

[0125] Furthermore, Dimensionless, range of values ,satisfy ; This is the normalized value of the radiation dose; Normalized value of pedestrian state score; Minimum value (take) ), to prevent the denominator from being 0.

[0126] S5.3. Calculate the dual-modal fusion nuclear accident off-site radiation risk value by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4. The calculation formula is as follows:

[0127]

[0128] in, For the first Node number Minutes of off-site radiation risk value for a dual-modal fusion nuclear accident.

[0129] Furthermore, by integrating the self-factor and the cross-node coupling correction value, the weighted sum of the self-factor is first calculated, then the impact of the cross-node correction is amplified through an exponential function, and finally the weight contribution of the cross-node correction factor is superimposed to obtain the final risk value.

[0130] The following is an example illustrating its application:

[0131] Specifically, the road network surrounding a nuclear facility in a certain city is planned as an off-site emergency monitoring road network for nuclear accidents. Four core intersections are selected as radiation risk monitoring nodes (nodes 1-4). The node topology meets the requirements: node 1 is the initial impact point of the nuclear accident (closest to the nuclear facility); nodes 2 and 3 are directly connected to node 1 (with direct roads); node 4 is directly connected to nodes 2 and 3 (a key transit node for evacuation routes); there are no other cross-node connections. The basic node parameters are calibrated using GIS: the actual travel distance between nodes 1 and 2. Nodes 1-3 Nodes 2-4 Nodes 3 and 4 All nodes have four lanes, and the monitoring coverage area is 15 meters around each node. After a nuclear accident, emergency monitoring is activated: high-definition cameras (30 frames / second) are deployed at nodes 1-4, and an additional external radiation dose sensor (1 time / minute) is deployed at node 1, collecting data for 15 minutes. to Multimodal data (minutes) are used to calculate the radiation risk values ​​of four nodes using this method, providing a basis for emergency evacuation decisions. The final results are shown in Table 1.

[0132] Table 1

[0133]

[0134] This embodiment implements a real road network topology with four nodes. The method is fully validated through 15 minutes of dual-modal data acquisition and step-by-step calculation. In this case study, only existing road network radiation dose sensors and high-definition cameras are used (no additional hardware required) to solve the core problem of "frequency heterogeneity and modal heterogeneity between radiation dose data (1 time / minute) and video data (30 frames / second)." Through standardization, dimensionality reduction, and coupling correction, the dynamic risk value of each node is quickly output.

[0135] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0136] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for risk assessment of off-site nuclear accident propagation intersection nodes, characterized in that, Includes the following steps: S1. Based on the synchronous deployment of nuclear accident site radiation dose monitoring sensors and high-definition cameras at each intersection node, establish an intersection node adjacency matrix, survey the actual road network travel distance of each connected intersection node pair, and perform standardization processing to obtain the standardized travel distance of all connected intersection node pairs. S2. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node, extract high-dimensional feature vectors and aggregate them at the minute level to obtain the minute-level average high-dimensional feature vector and the corresponding radiation dose value outside the nuclear accident site; S3. The minute-level average high-dimensional feature vector obtained in step S2 is subjected to dimensionality reduction processing. The obtained low-dimensional effective feature vector is used to construct a basic pedestrian status assessment score. Then, the basic pedestrian status assessment score is dynamically corrected using the relative change rate of radiation dose values ​​outside the nuclear accident site to obtain the corrected pedestrian status assessment score. S4. Based on the intersection node adjacency matrix obtained in step S1, the nuclear accident field radiation dose value obtained in step S2, and the corrected pedestrian state assessment score obtained in step S3, calculate the dynamic influence coefficient of each intersection node, and then calculate the cross-node coupling correction value of the intersection node. S5. Normalize the off-site radiation dose value of the nuclear accident obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3, and then calculate the dual-modal fusion off-site radiation risk value of the nuclear accident by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4.

2. The method for risk assessment of off-site diffusion intersections in a nuclear accident, as described in claim 1, is characterized in that... The specific implementation method of step S1 includes the following steps: S1.

1. Establish the adjacency matrix of intersection nodes, and obtain ； in, for An adjacency matrix of intersection nodes of order 1. These are elements in the adjacency matrix of intersection nodes. , These represent the intersection node numbers and the connectivity between intersection nodes. If the value is 1, then the intersection nodes are not connected. Set to 0; S1.

2. Survey all Actual travel distance of intersection node pairs with a value of 1 Then, the maximum actual travel distance between all connected intersection nodes in the off-site monitoring road network of the nuclear accident site is calculated. Standardization processing is performed to obtain the standardized travel distance for intersection node pairs. .

3. The method for risk assessment of off-site diffusion intersections in a nuclear accident, as described in claim 2, is characterized in that... The specific implementation method of step S2 includes the following steps: S2.

1. Collect scalar data of radiation dose outside the nuclear accident site and video image data at each intersection node to obtain the first... The first intersection node Minute-level off-site radiation dose scalar data of nuclear accidents , No. Video image data of the kth frame of the intersection node ; S2.

2. Perform image preprocessing, image segmentation, positional encoding, sequence construction, multi-head self-attention calculation, and feature aggregation on the video image data obtained in step S2.1 to obtain a high-dimensional feature vector. The expression for feature aggregation is: ； in, For the first Node number High-dimensional feature vectors of frame videos, Feature extraction operator for visual Transformer; S2.

3. Perform dimension-wise mean aggregation on the high-dimensional feature vectors obtained in step S2.2 to obtain the minute-level average high-dimensional feature vectors. The expression is: 。 4. The method for risk assessment of off-site diffusion intersections in a nuclear accident, as described in claim 3, is characterized in that... The specific implementation method of step S3 includes the following steps: S3.

1. Calculate the minute-level average high-dimensional feature vector obtained in step S2. The covariance matrix is ​​obtained by performing eigenvalue decomposition on it, and selecting the eigenvectors corresponding to the first 16 largest eigenvalues ​​as principal components, thus obtaining 16-dimensional low-dimensional effective eigenvectors. ; S3.

2. Collect the first... Low-dimensional feature vector of normal pedestrian state at each intersection node when there is no nuclear accident Then, a basic pedestrian state assessment score is constructed, expressed as: ； in, For the first The first intersection node A basic pedestrian status assessment score within minutes. It is a function with maximum value. It is a norm 2; S3.

3. Calculate the first... Node number Relative change rate of off-site radiation dose per minute during a nuclear accident The formula is as follows: ； in, The relative rate of change of radiation dose. For the first The first intersection node Minute-level scalar data of radiation dose outside the nuclear accident site; use The absolute value of the value is used to dynamically adjust the basic pedestrian status assessment score, resulting in the adjusted pedestrian status assessment score. , For the first The first intersection node The adjusted pedestrian status assessment score for the minute.

5. The method for risk assessment of off-site nuclear accident propagation intersection nodes according to claim 4, characterized in that, The specific implementation method of step S4 includes the following steps: S4.

1. Calculate the dynamic impact coefficient of each intersection node. The calculation formula is as follows: ； in, For the first The intersection node corresponds to the first The first intersection node The dynamic impact coefficient per minute; S4.

2. Based on the dynamic influence coefficient of each intersection node obtained in step S4.1, calculate the cross-node coupling correction value of the intersection node. The calculation formula is as follows: ； in, For the first Node number Minute-level cross-node coupling correction value, It is the hyperbolic tangent function.

6. The method for risk assessment of off-site diffusion intersections in a nuclear accident, as described in claim 5, is characterized in that... The specific implementation method of step S5 includes the following steps: S5.

1. Normalize the off-site radiation dose value obtained in step S2 and the corrected pedestrian status assessment score obtained in step S3 to obtain the... The first intersection node Normalized value of off-site radiation dose per minute in a nuclear accident , No. The first intersection node Normalized value of pedestrian status assessment score per minute ; S5.

2. Calculate the attention weights, including the attention weights for the radiation dose factor, the pedestrian state factor, and the cross-node coupling correction factor. The calculation formula is as follows: ； ； ； in, For the first The first intersection node Attention weight of the radiation dose factor per minute. It is the minimum value. For the first The first intersection node Attention weights for pedestrian state factors at the minute level. Attention weights for cross-node coupling correction factors; S5.

3. Calculate the dual-modal fusion nuclear accident off-site radiation risk value by combining the attention weight and the cross-node coupling correction value of the intersection node obtained in step S4. The calculation formula is as follows: ； in, For the first Node number Minutes of off-site radiation risk value for a dual-modal fusion nuclear accident.

Images