Flood risk assessment method for dam-break flood considering multi-dimensional parameters

Abstract

This invention discloses a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. First, it simultaneously collects basic data at micro, meso, and macro scales, along with resilience parameters of the affected structures, to construct a standardized parameter library. Then, it uses a BREACH physical model to generate multi-condition simulation data, builds a prediction model based on a Transformer architecture, embeds physical constraint rules, and outputs dynamic data on breach evolution. Multiple simulations quantify the uncertainty parameters of breach evolution. A multi-scale spatiotemporal coupling fusion algorithm is used to complete deep linkage calculations between parameters. Combined with comprehensive flood hazard parameters, a comprehensive risk index for various affected structures under different scenarios is obtained. Finally, the index is converted to the corresponding risk level using an up-rounding function to generate a comprehensive assessment result. This invention achieves a multi-dimensional and dynamic assessment of dam-break flood risk, improving the accuracy and response capability of risk assessment and providing a scientific basis for disaster prevention and emergency management.

Description

Technical Field

[0001] This invention belongs to the field of disaster risk assessment technology, specifically involving a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. Background Technology

[0002] Against the backdrop of global climate change and frequent extreme hydrological events, the safe operation of water conservancy projects and flood disaster prevention have become key issues in ensuring stable socio-economic development. As an important water conservancy infrastructure, the risk assessment of dam failure involves complex factors across multiple dimensions, requiring comprehensive consideration of parameters such as dam structural characteristics, hydrological dynamics, regional geological conditions, and the disaster-bearing capacity of the dam body. With the development of monitoring technology, the ability to acquire multi-source data, including micro-scale structural damage monitoring, meso-scale watershed confluence analysis, and macro-scale climate trend prediction, has significantly improved. However, how to effectively integrate multi-scale and multi-type data, and combine physical laws with data-driven models to achieve dynamic prediction of the breach evolution process, and thus accurately quantify flood risk, has become a core issue that urgently needs to be addressed in the field of water conservancy project disaster assessment. At the same time, the resilience of the disaster-bearing body during a disaster directly affects the risk consequences, and it needs to be included in the assessment system to enhance the practical application value of the results and provide a scientific basis for disaster prevention and emergency management.

[0003] Traditional dam failure risk assessment methods have significant limitations, failing to meet the need for accurate assessment in complex scenarios. Some methods rely on a single physical model for breach simulation, only able to output prediction results based on fixed operating conditions, and cannot fully integrate multi-scale monitoring data. This results in insufficient adaptability to dynamic parameter changes in the actual environment and difficulty in quantifying uncertainties in the prediction process, thus limiting the reliability and comprehensiveness of the assessment results. Other methods focus on data-driven model construction but lack the embedding of core rules such as physical conservation laws, which can easily lead to prediction results that contradict actual physical processes, reducing model credibility. In addition, traditional risk assessments often neglect the deep coupling between the resilience parameters of the disaster-bearing body and flood hazard parameters, determining the risk level only through a single risk indicator without dynamically adjusting the assessment standards according to different regional characteristics. This leads to a disconnect between the assessment results and actual disaster response needs, making it difficult to effectively guide the formulation of differentiated prevention and control measures. Summary of the Invention

[0004] The purpose of this invention is to provide a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters, thereby achieving a multi-dimensional and dynamic assessment of dam-break flood risks, improving the accuracy of risk assessment and response capabilities, and providing a scientific basis for disaster prevention and emergency management.

[0005] The technical solution adopted in this invention is a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters, which is implemented according to the following steps:

[0006] Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0007] The invention is further characterized in that, In step 1, basic data and disaster-bearing body resilience parameters at three scales—micro, meso, and macro—are collected simultaneously. At the micro scale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and terrain scanning equipment, respectively. At the meso scale, tributary confluence data and township-level population distribution data of the watershed are collected through hydrological monitoring stations and population information management systems, respectively. At the macro scale, regional climate trend data, watershed geological structural stability data, and relevant basic data of the dam body are collected through meteorological monitoring systems, geological monitoring equipment, and dam body archive systems, respectively. Disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records, and regional economic and disaster recovery archive compilation.

[0008] Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

[0009] In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: A Transformer prediction model containing an input layer, encoder, decoder, and output layer is built. The input layer performs feature encoding on the standardized dam body and hydrological parameters, transforming them into feature vectors recognizable by the model. The encoder adopts a multi-layer stacked structure, capturing the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is followed by a layer normalization and activation function to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receiving the feature information output by the encoder and generating a prediction sequence related to the breach evolution through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to mass conservation and energy conservation. Constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

[0010] In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core predicted value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core predicted value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thereby achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

[0011] In step 3, multi-scale spatiotemporal coupling, the mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is: , in, For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time Correction term for breach and leakage of space unit.

[0012] Step 4 is implemented in the following steps: In the resilience-risk coupling quantification, the mathematical expression of the resilience-risk adversarial coupling algorithm is: , in, For the first Time node, number Within the space unit The comprehensive risk index of disaster-bearing bodies For comprehensive parameters of flood risk, For the first Engineering toughness parameters of disaster-bearing structures For the first Social resilience parameters of disaster-bearing entities For the first Emergency response correction coefficient for disaster-bearing bodies This is the toughness balance coefficient.

[0013] Step 5 is implemented in the following steps: In the risk level output, the mathematical expression for the risk level-uncertainty coupling algorithm is: , in, For the first Time node, number Within the space unit Risk level of disaster-bearing bodies As a comprehensive risk index, For uncertainty coefficient, For the uncertainty parameters of the breach evolution, This is the floor function.

[0014] In step 5, the thresholds used to determine the risk level are set as follows: a comprehensive risk index ≥ 0.8 is considered extremely high risk, 0.6 ≤ comprehensive risk index < 0.8 is considered high risk, 0.4 ≤ comprehensive risk index < 0.6 is considered medium risk, and a comprehensive risk index < 0.4 is considered low risk. For areas with a population density exceeding 500 people / square kilometer or containing large transportation hubs, the threshold for extremely high risk is lowered to 0.75, and the threshold for high risk is lowered to 0.55. For sparsely populated mountainous or desert areas, the threshold for extremely high risk is raised to 0.85, and the threshold for high risk is raised to 0.65.

[0015] The beneficial effects of this invention are that the comprehensive risk assessment method for dam-break floods, which integrates multi-dimensional parameters, achieves full-dimensional coverage and accurate prediction of flood risk assessment through the deep integration of multi-scale parameter collection and integration with physical models and machine learning architectures. It relies on the standardized processing of multi-source data to build a comprehensive parameter foundation, and trains the prediction model with multi-condition data generated by the physical model. The embedded physical constraint rules ensure the scientificity and rationality of the prediction results. At the same time, by quantifying the uncertainty parameters of breach evolution, it makes up for the shortcomings of traditional assessment methods in not considering risk fluctuation factors. The multi-scale spatiotemporal coupling algorithm realizes the deep linkage of parameters of different dimensions, enabling flood hazard assessment to accurately match the actual scenarios of different time nodes and spatial ranges, providing high-precision data support for subsequent risk quantification, and significantly improving the comprehensiveness and dynamic adaptability of risk assessment.

[0016] This invention constructs a comprehensive assessment system that considers both the characteristics of the disaster-bearing body and risk fluctuations by quantifying resilience-risk coupling and incorporating uncertainty into risk level determination. It correlates and calculates multi-dimensional resilience parameters of the disaster-bearing body with flood hazard parameters, fully considering the impact of the disaster-bearing body's own disaster resistance and recovery capabilities on risk. This makes the risk assessment more aligned with practical application scenarios. Uncertainty factors are incorporated into the risk level calculation, and a scientific algorithm transforms the predicted dispersion into a basis for risk level adjustment. Simultaneously, risk thresholds are dynamically optimized for different regional characteristics to ensure the relevance and practicality of the assessment results. Through tiered risk response suggestions, differentiated prevention and control solutions are provided for regions with different risk levels, effectively improving the accuracy and efficiency of flood disaster prevention and control, and providing reliable decision support for disaster emergency management. Attached Figure Description

[0017] Figure 1 This is a flowchart of the comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters according to the present invention; Figure 2This diagram illustrates the data transmission between the steps of the comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters, as presented in this invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0019] This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. In step 1, at the microscale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and topographic scanning equipment, respectively; at the mesoscale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively; at the macroscale, regional climate trend data, basin geological structural stability data and dam-related basic data are collected through meteorological monitoring systems, geological monitoring equipment and dam body archive systems, respectively; and disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records and regional economic and disaster recovery archives.

[0020] Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

[0021] In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: A Transformer prediction model containing an input layer, encoder, decoder, and output layer is built. The input layer performs feature encoding on the standardized dam body and hydrological parameters, transforming them into feature vectors recognizable by the model. The encoder adopts a multi-layer stacked structure, capturing the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is followed by a layer normalization and activation function to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receiving the feature information output by the encoder and generating a prediction sequence related to the breach evolution through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to mass conservation and energy conservation. Constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

[0022] In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core predicted value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core predicted value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thereby achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

[0023] Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; In step 3, multi-scale spatiotemporal coupling, the mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is: , in, For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time Correction term for breach and leakage of space unit.

[0024] Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 4 is implemented in the following steps: In the resilience-risk coupling quantification, the mathematical expression of the resilience-risk adversarial coupling algorithm is: , in, For the first Time node, number Within the space unit The comprehensive risk index of disaster-bearing bodies For comprehensive parameters of flood risk, For the first Engineering toughness parameters of disaster-bearing structures For the first Social resilience parameters of disaster-bearing entities For the first Emergency response correction coefficient for disaster-bearing bodies This is the toughness balance coefficient.

[0025] Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0026] Step 5 is implemented in the following steps: In the risk level output, the mathematical expression for the risk level-uncertainty coupling algorithm is: , in, For the first Time node, number Within the space unit Risk level of disaster-bearing bodies As a comprehensive risk index, For uncertainty coefficient, For the uncertainty parameters of the breach evolution, This is the floor function.

[0027] In step 5, the thresholds used to determine the risk level are set as follows: a comprehensive risk index ≥ 0.8 is considered extremely high risk, 0.6 ≤ comprehensive risk index < 0.8 is considered high risk, 0.4 ≤ comprehensive risk index < 0.6 is considered medium risk, and a comprehensive risk index < 0.4 is considered low risk. For areas with a population density exceeding 500 people / square kilometer or containing large transportation hubs, the threshold for extremely high risk is lowered to 0.75, and the threshold for high risk is lowered to 0.55. For sparsely populated mountainous or desert areas, the threshold for extremely high risk is raised to 0.85, and the threshold for high risk is raised to 0.65.

[0028] Example 1 This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0029] Example 2 This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. In step 1, at the microscale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and topographic scanning equipment, respectively; at the mesoscale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively; at the macroscale, regional climate trend data, basin geological structural stability data and dam-related basic data are collected through meteorological monitoring systems, geological monitoring equipment and dam body archive systems, respectively; and disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records and regional economic and disaster recovery archives.

[0030] Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0031] Example 3 This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. In step 1, at the microscale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and topographic scanning equipment, respectively; at the mesoscale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively; at the macroscale, regional climate trend data, basin geological structural stability data and dam-related basic data are collected through meteorological monitoring systems, geological monitoring equipment and dam body archive systems, respectively; and disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records and regional economic and disaster recovery archives.

[0032] Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

[0033] In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: A Transformer prediction model containing an input layer, encoder, decoder, and output layer is built. The input layer performs feature encoding on the standardized dam body and hydrological parameters, transforming them into feature vectors recognizable by the model. The encoder adopts a multi-layer stacked structure, capturing the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is followed by a layer normalization and activation function to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receiving the feature information output by the encoder and generating a prediction sequence related to the breach evolution through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to mass conservation and energy conservation. Constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

[0034] In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core predicted value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core predicted value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thereby achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

[0035] Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0036] Example 4 This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. In step 1, at the microscale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and topographic scanning equipment, respectively; at the mesoscale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively; at the macroscale, regional climate trend data, basin geological structural stability data and dam-related basic data are collected through meteorological monitoring systems, geological monitoring equipment and dam body archive systems, respectively; and disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records and regional economic and disaster recovery archives.

[0037] Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

[0038] In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: A Transformer prediction model containing an input layer, encoder, decoder, and output layer is built. The input layer performs feature encoding on the standardized dam body and hydrological parameters, transforming them into feature vectors recognizable by the model. The encoder adopts a multi-layer stacked structure, capturing the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is followed by a layer normalization and activation function to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receiving the feature information output by the encoder and generating a prediction sequence related to the breach evolution through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to mass conservation and energy conservation. Constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

[0039] In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core predicted value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core predicted value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thereby achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

[0040] Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; In step 3, multi-scale spatiotemporal coupling, the mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is: , in, For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time Correction term for breach and leakage of space unit.

[0041] Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0042] Example 5 This invention provides a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters. The flowchart is as follows: Figure 1 As shown, please follow these steps: Step 1: Multi-source parameter acquisition and integration: Simultaneously collect basic data and disaster-bearing body resilience parameters at three scales: micro, meso, and macro. Perform dimensional unification, outlier removal, and standardization transformation on all collected data to construct a standardized parameter library. The disaster-bearing body resilience parameters are used for subsequent risk coupling quantification. In step 1, at the microscale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and topographic scanning equipment, respectively; at the mesoscale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively; at the macroscale, regional climate trend data, basin geological structural stability data and dam-related basic data are collected through meteorological monitoring systems, geological monitoring equipment and dam body archive systems, respectively; and disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records and regional economic and disaster recovery archives.

[0043] Step 2, Breach Evolution Prediction and Analysis: Using the dam body and hydrological parameters in the standardized parameter library as input, multi-condition simulation data is generated with the help of the BREACH physical model. A prediction model is built based on the Transformer architecture and physical constraint rules are embedded to output dynamic data of breach evolution. The uncertainty parameters of breach evolution are quantified through multiple simulation operations. Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

[0044] In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: A Transformer prediction model containing an input layer, encoder, decoder, and output layer is built. The input layer performs feature encoding on the standardized dam body and hydrological parameters, transforming them into feature vectors recognizable by the model. The encoder adopts a multi-layer stacked structure, capturing the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is followed by a layer normalization and activation function to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receiving the feature information output by the encoder and generating a prediction sequence related to the breach evolution through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to mass conservation and energy conservation. Constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

[0045] In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core predicted value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core predicted value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thereby achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

[0046] Step 3, Multi-scale spatiotemporal coupling: Extract multi-scale parameters from the standardized parameter library and the breach evolution prediction results, and use the multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage operation between parameters, and output the comprehensive flood hazard parameters at different time nodes and in different spatial ranges; In step 3, multi-scale spatiotemporal coupling, the mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is: , in, For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time Correction term for breach and leakage of space unit.

[0047] Step 4, Resilience-Risk Coupling Quantification: Extract the resilience parameters of disaster-bearing bodies from the standardized parameter library, combine them with the comprehensive flood hazard parameters, use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 4 is implemented in the following steps: In the resilience-risk coupling quantification, the mathematical expression of the resilience-risk adversarial coupling algorithm is: , in, For the first Time node, number Within the space unit The comprehensive risk index of disaster-bearing bodies For comprehensive parameters of flood risk, For the first Engineering toughness parameters of disaster-bearing structures For the first Social resilience parameters of disaster-bearing entities For the first Emergency response correction coefficient for disaster-bearing bodies This is the toughness balance coefficient.

[0048] Step 5, Risk Level Output: Integrate the comprehensive risk index and the uncertainty parameter of the breach evolution, use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation, and transform it into the corresponding risk level through the rounding function to generate a comprehensive assessment result.

[0049] Step 5 is implemented in the following steps: In the risk level output, the mathematical expression for the risk level-uncertainty coupling algorithm is: , in, For the first Time node, number Within the space unit Risk level of disaster-bearing bodies As a comprehensive risk index, For uncertainty coefficient, For the uncertainty parameters of the breach evolution, This is the floor function.

[0050] Example 6 Example of a comprehensive risk assessment method for dam-break floods that integrates multi-dimensional parameters.

[0051] Taking a mountainous concrete gravity dam as the assessment object, the dam is 45m high with a concrete compressive strength of 30MPa. The basin includes three tributaries and is surrounded by two townships with population densities of 320 and 610 people / km² respectively. The basin also includes a county-level highway transportation hub. Within a 5km radius downstream of the dam, there are villages, small factories, and farmland, which are also potential sources of flooding. The overall geological structure of the basin is stable, but small fissures exist in some areas. The recent regional climate shows a trend towards higher rainfall. Therefore, a comprehensive risk assessment of the potential dam break flood is needed to provide a basis for flood control and disaster reduction decisions in the basin.

[0052] Implementation steps: such as Figure 1 As shown Step 1, Multi-source parameter acquisition and integration: Basic data were collected simultaneously at three scales: micro, meso, and macro, along with parameters on the resilience of the disaster-bearing body. At the micro scale, structural damage data, such as the length and depth of local cracks in the dam foundation and face, were collected using dam structure monitoring equipment. Topographic roughness data within a 1km radius of the dam was obtained using terrain scanning equipment. This data accurately reflects the local structural condition of the dam and the details of the surrounding terrain, providing micro-level support for subsequent analysis of dam stability and flood discharge paths. At the meso scale, runoff data, such as flow rate and velocity, from the confluence of tributaries was collected from three hydrological monitoring stations within the basin. Township-level population distribution data for two townships were extracted from the local population information management system to clarify the resident population in each area. The data collection and distribution range provide a clear understanding of the basin's hydrological dynamics and population density, laying the foundation for assessing the impact of floods on the population. At the macro scale, regional meteorological monitoring systems collect data on precipitation, frequency, and temperature trends over the past five years. Geological monitoring equipment provides stability data on geological structures within the basin, including rock displacement and fissure development. A dam archive system retrieves basic data such as construction time, design standards, and material properties, allowing for a holistic understanding of the basin's climate background, geological conditions, and basic dam attributes, providing a macro-level basis for overall risk assessment. Regarding the resilience parameters of disaster-bearing bodies, data on engineering resilience are obtained by reviewing archives of village residences, small factory buildings, and county-level highway bridges. Daily maintenance records of infrastructure such as water supply, power supply, and communications are collected, and historical economic data and recovery records from past minor floods are compiled to form a complete set of disaster-bearing body resilience parameters, providing a comprehensive understanding of the disaster resistance and reconstruction capabilities of various disaster-bearing bodies. All collected data undergoes unitization processing to eliminate computational interference caused by differences in the units of different parameters, removes outliers caused by equipment failure or human error in recording, ensuring data accuracy, and then performs standardization transformation to bring the data to the same order of magnitude for easy subsequent calculations. Finally, a standardized parameter library containing all valid parameters is constructed to provide comprehensive and reliable data support for subsequent evaluation stages.

[0053] Step 2, Breach Evolution Prediction Analysis: The dam body parameters and hydrological parameters from the standardized parameter library were used as input data to ensure that the input data closely matched the actual situation of the assessment object. Using the BREACH physical model, and combining 60 sets of basic working conditions generated by combining five gradient values ​​for concrete compressive strength, three gradient values ​​for dam height, and four gradient values ​​for water level rise rate, the entire process of the breach from its initial width of 0m to its stable width was simulated for the actual parameters of the assessment object. The simulation time step was set to 5 minutes, and the output sequence data of breach width and discharge flow changing over time were used as training data to provide samples consistent with actual working conditions for the training of the prediction model. A prediction model based on the Transformer architecture is constructed, comprising an input layer, encoder, decoder, and output layer. The input layer encodes the standardized dam body and hydrological parameters, transforming them into model-recognizable feature vectors to match the data format with the model's requirements. The encoder employs a multi-layer stacked structure, using a multi-head attention mechanism to capture the spatiotemporal correlation features between parameters, fully exploring the intrinsic relationships between different parameters in time and space. Each encoder layer is followed by a normalization and activation function to optimize training performance and avoid gradient vanishing or overfitting issues during model training. The decoder also uses a multi-layer stacked structure, receiving feature information from the encoder output and generating a prediction sequence related to breach evolution through masking and cross-attention mechanisms, ensuring the rationality and continuity of the prediction results. Physical constraint rules containing calculation logic related to mass and energy conservation are embedded into the model training loss function to construct the total loss function, the mathematical expression of which is: ;in Represents the total loss function. This represents the prediction error term. The deviation between the predicted and actual breach parameters is calculated using the mean squared error formula. This represents the physical constraint term, constructed by calculating and summing the deviations of the predicted results from the laws of conservation of mass and energy. This represents the weight of the prediction error term. Represents the weight of the physical constraint term, and satisfies During training, the minimum value is minimized through backpropagation. By optimizing model parameters through backpropagation, the model's prediction results are made to conform to both data patterns and physical laws, thus improving prediction accuracy. Simultaneously, a Dropout layer is added to the prediction model with a fixed dropout probability. For the same set of input dam body and hydrological parameters, while keeping the model structure and core parameters unchanged, different network nodes are randomly dropped through the Dropout layer, and the prediction calculation is repeated multiple times to simulate prediction results under different network conditions. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on this data, calculating the mean, variance, and standard deviation of each indicator. The mean serves as the core predicted value for breach evolution, while the variance and standard deviation reflect the degree of prediction dispersion, determining prediction intervals at different confidence levels. This integrates to form a complete set of uncertainty parameters for breach evolution, outputting dynamic data on breach evolution and quantified uncertainty parameters, providing information on the breach development process and risk fluctuation range for subsequent risk assessment.

[0054] Step 3, Multi-scale spatiotemporal coupling: Relevant parameters at three scales—micro, meso, and macro—are extracted from a standardized parameter library. Micro-scale parameters reflect local details, meso-scale parameters reflect regional dynamics, and macro-scale parameters present overall background conditions. Combined with breach evolution dynamic data and uncertainty parameters obtained from breach evolution prediction analysis, breach data can clearly identify the source and intensity changes of floods, while uncertainty parameters can indicate the range of data fluctuations. A multi-scale spatiotemporal coupling fusion algorithm is used for deep linkage calculations. The mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is as follows: ,in For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time The algorithm for breach discharge correction terms in spatial units breaks down barriers between parameters of different scales and spatiotemporal dimensions, achieving organic integration and synergistic effects among parameters. Through algorithmic calculation, it fully considers the parameter characteristics at different time points and spatial ranges, integrating the comprehensive values ​​of micro-scale parameters, meso-scale parameters, macro-scale time and spatial correction coefficients, as well as scale coupling coefficients, spatiotemporal matching coefficients, breach impact coefficients, and breach discharge correction terms. This eliminates spatiotemporal differences and scale contradictions among parameters, ultimately outputting comprehensive flood hazard parameters for the downstream of the mountainous concrete gravity dam at different time points and spatial ranges. These parameters intuitively reflect the degree of flood hazard under different spatiotemporal conditions, providing crucial input for subsequent resilience-risk coupling quantification.

[0055] Step 4, Quantification of Resilience-Risk Coupling: Resilience parameters of disaster-bearing bodies are extracted from the standardized parameter library, including engineering resilience parameters, social resilience parameters, and emergency response correction coefficients for various types of disaster-bearing bodies. These parameters comprehensively reflect the ability of disaster-bearing bodies to cope with flood disasters. Combined with the comprehensive flood hazard parameter output by multi-scale spatiotemporal coupling, this parameter clarifies the degree of threat posed by floods to disaster-bearing bodies. A resilience-risk adversarial coupling algorithm is used to construct and calculate the correlation model. The mathematical expression of the resilience-risk adversarial coupling algorithm is: ,in For the first Time node, number Within the space unit The comprehensive risk index of disaster-bearing bodies For comprehensive parameters of flood risk, For the first Engineering toughness parameters of disaster-bearing structures For the first Social resilience parameters of disaster-bearing entities For the first Emergency response correction coefficient for disaster-bearing bodies As a resilience balance coefficient, this algorithm effectively correlates flood hazard with the resilience of disaster-bearing bodies, quantifying the interaction between the two. During the calculation, the resilience balance coefficient is used to balance and adjust social resilience parameters and emergency response correction coefficients, ensuring that different types of resilience parameters play a reasonable role in risk assessment. It fully considers the interaction between comprehensive flood hazard parameters and various disaster-bearing body resilience parameters. When flood hazard is high and disaster-bearing body resilience is strong, the risk level will decrease accordingly, and vice versa. Ultimately, it yields comprehensive risk indices for various disaster-bearing bodies such as villages, small factories, farmland, and county roads in the region at different time points and in different spatial units under different scenarios. These indices can accurately measure the magnitude of risk faced by various disaster-bearing bodies under different conditions, providing a core basis for risk level determination.

[0056] Step 5, Risk Level Output: The comprehensive risk index, obtained by integrating resilience-risk coupling quantification, and the uncertainty parameter of breach evolution are used. The comprehensive risk index forms the basis for risk level determination, while the uncertainty parameter of breach evolution reflects the reliability and fluctuation range of the risk assessment results. Using a risk level-uncertainty coupling algorithm, the uncertainty parameter is incorporated into the risk level calculation process through uncertainty coefficients. This allows the risk level to not only reflect the core risk level but also the volatility of the risk, improving the comprehensiveness and practicality of risk assessment. The mathematical expression of the risk level-uncertainty coupling algorithm is as follows: ,in For the first Time node, number Within the space unit Risk level of disaster-bearing bodies As a comprehensive risk index, For uncertainty coefficient, For the uncertainty parameters of the breach evolution, The calculation result is then converted into a corresponding risk level using a rounding function to ensure that the risk level is an integer, facilitating intuitive understanding and decision-making. The risk level determination thresholds are implemented according to the following standards: a comprehensive risk index ≥ 0.8 indicates extremely high risk, 0.6 ≤ comprehensive risk index < 0.8 indicates high risk, 0.4 ≤ comprehensive risk index < 0.6 indicates medium risk, and a comprehensive risk index < 0.4 indicates low risk. This standard clearly delineates different risk levels. Because one township in this area has a population density of 610 people / square kilometer and includes a large transportation hub, these areas are densely populated and have crucial transportation links, making them more susceptible to severe consequences from floods. Therefore, the extremely high risk threshold for this area is lowered to 0.75, and the high risk threshold is lowered to 0.55, increasing vigilance for highly sensitive areas. Other areas with lower population density and no large transportation hubs follow the original standards to ensure that risk assessments are consistent with the actual conditions of different areas. The final comprehensive assessment results of the potential dam-break flood of the concrete gravity dam in the mountainous area were generated, clarifying the risk levels of different regions and different disaster-bearing bodies under different scenarios, and providing clear and specific decision-making references for the formulation of flood control and disaster reduction plans, emergency resource allocation, and personnel evacuation arrangements within the basin.

[0057] In summary, this embodiment focuses on a concrete gravity dam in a specific mountainous area, conducting an assessment in five steps: multi-source parameter acquisition and integration, breach evolution prediction and analysis, multi-scale spatiotemporal coupling, resilience-risk coupling quantification, and risk level output. Each step is closely linked. First, a standardized parameter library is constructed through multi-scale data acquisition and processing. Then, the BREACH model and Transformer architecture are used to predict breach evolution and quantify uncertainty. Subsequently, flood hazard parameters are obtained through a multi-scale spatiotemporal coupling fusion algorithm. Combined with resilience parameters, a comprehensive risk index is calculated using a corresponding algorithm. Finally, uncertainty is incorporated, and the risk level is determined based on differentiated thresholds. Figure 2 As shown, this provides comprehensive support for the dam's flood control and disaster reduction decision-making.

Claims

1. A comprehensive risk assessment method for dam-break floods integrating multi-dimensional parameters, characterized in that, The specific steps are as follows: Step 1: Construct a standardized parameter library; Step 2: Generate multi-condition simulation data using the BREACH physical model, build a prediction model based on the Transformer architecture and embed physical constraint rules, and output dynamic data on the evolution of the breach. Step 3: Use a multi-scale spatiotemporal coupling fusion algorithm to complete the deep linkage calculation between parameters; Step 4: Use the resilience-risk adversarial coupling algorithm to construct and calculate the correlation model, and obtain the comprehensive risk index of various disaster-bearing bodies under different scenarios; Step 5: Use the risk level-uncertainty coupling algorithm to incorporate uncertainty into the risk level calculation and generate a comprehensive assessment result.

2. The comprehensive risk assessment method for dam-break floods integrating multi-dimensional parameters according to claim 1, characterized in that, In step 1, basic data and disaster-bearing body resilience parameters at three scales—micro, meso, and macro—are collected simultaneously. At the micro scale, local structural damage data and small-scale topographic roughness data of the dam body are collected through dam structure monitoring equipment and terrain scanning equipment, respectively. At the meso scale, tributary confluence data and township-level population distribution data of the basin are collected through hydrological monitoring stations and population information management systems, respectively. At the macro scale, regional climate trend data, basin geological structural stability data, and relevant basic data of the dam body are collected through meteorological monitoring systems, geological monitoring equipment, and dam body archive systems, respectively. Disaster-bearing body resilience parameters are obtained through building archive verification, infrastructure operation and maintenance records, and regional economic and disaster recovery archive compilation.

3. The comprehensive risk assessment method for dam-break floods integrating multi-dimensional parameters according to claim 2, characterized in that, Step 2 is implemented in the following steps: In the predictive analysis of breach evolution, the specific process of generating training data for the BREACH physical model is as follows: Five gradient values ​​are set for the concrete compressive strength: 20MPa, 25MPa, 30MPa, 35MPa, and 40MPa; three gradient values ​​are set for the dam height: 30m, 40m, and 50m; and four gradient values ​​are set for the water level rise rate: 0.5m / h, 1m / h, 1.5m / h, and 2m / h. Sixty sets of basic working conditions are generated. For each set of working conditions, the entire process of the breach from its initial width of 0m to its stable width is simulated. The simulation time step is set to 0 to 5 minutes, with the interval represented as (0,5]. The sequence data of the breach width and discharge flow rate changing over time are output, forming 60 complete training datasets.

4. The comprehensive risk assessment method for dam-break floods integrating multi-dimensional parameters according to claim 3, characterized in that, In step 2, the specific process of building a prediction model based on the Transformer architecture and embedding physical constraint rules is as follows: a Transformer prediction model containing an input layer, encoder, decoder and output layer is built. The input layer performs feature encoding processing on the standardized dam body and hydrological parameters, and transforms them into feature vectors that the model can recognize. The encoder adopts a multi-layer stacked structure and captures the spatiotemporal correlation features between parameters through a multi-head attention mechanism. Each encoder layer is configured with layer normalization and activation functions to optimize the training effect. The decoder also adopts a multi-layer stacked structure, receives the feature information output by the encoder, and generates a prediction sequence related to the evolution of the breach through a masking mechanism and a cross-attention mechanism. Physical constraint rules are embedded in the loss function of the model training. These rules cover the calculation logic related to the conservation of mass and energy. The constraint terms are constructed by calculating the deviation between the prediction results and the physical conservation laws. The constraint terms and the prediction error terms are weighted and fused as the total loss function, and the model parameters are optimized through backpropagation.

5. The comprehensive risk assessment method for dam-break floods integrating multi-dimensional parameters according to claim 4, characterized in that, In step 2, the specific process of quantifying the uncertainty parameters of the breach evolution through multiple simulations is as follows: A Dropout layer is added to the prediction model based on the Transformer architecture, with a fixed dropout probability. For the same set of input dam body and hydrological parameters, the model structure and core parameters remain unchanged, and different network nodes are randomly dropped through the Dropout layer, repeating the prediction calculation multiple times. Data on the breach width over time, the final breach size, and the maximum discharge flow are collected from all prediction calculations. Statistical analysis is performed on the collected prediction data, calculating the mean, variance, and standard deviation of each indicator. The mean is used as the core prediction value for breach evolution, and the standard deviation reflects the degree of prediction dispersion. Based on the statistical results, prediction intervals at different confidence levels are determined. The core prediction value, the degree of dispersion of each indicator, and the prediction intervals at different confidence levels are integrated to form a complete set of uncertainty parameters for breach evolution, thus achieving a quantitative characterization of the uncertainty of the breach evolution prediction results.

6. The method for comprehensive risk assessment of dam-break floods integrating multi-dimensional parameters according to claim 5, characterized in that, In step 3, multi-scale spatiotemporal coupling, the mathematical expression of the multi-scale spatiotemporal coupling fusion algorithm is: ， in, For the first Time node, number Comprehensive parameters of flood hazard of spatial units for The combined value of microscale parameters at any given time. for Comprehensive value of mesoscale parameters of spatial unit. for Time correction factor at the macroscopic scale. for Spatial correction coefficient at the macroscopic scale of a spatial unit. The scale coupling coefficient is... For spatiotemporal matching coefficients, The factor representing the impact of the breach is denoted as . for time Correction term for breach leakage in space units.

7. The method for comprehensive risk assessment of dam-break floods integrating multi-dimensional parameters according to claim 6, characterized in that, Step 4 is implemented in the following steps: In the resilience-risk coupling quantification, the mathematical expression of the resilience-risk adversarial coupling algorithm is: ， in, For the first Time node, number Within the space unit The comprehensive risk index of disaster-bearing bodies For comprehensive parameters of flood risk, For the first Engineering toughness parameters of disaster-bearing structures For the first Social resilience parameters of disaster-bearing entities For the first Emergency response correction coefficient for disaster-bearing bodies This is the toughness balance coefficient.

8. The method for comprehensive risk assessment of dam-break floods integrating multi-dimensional parameters according to claim 7, characterized in that, Step 5 is implemented in the following steps: In the risk level output, the mathematical expression for the risk level-uncertainty coupling algorithm is: ， in, For the first Time node, number Within the space unit Risk level of disaster-bearing bodies As a comprehensive risk index, For uncertainty coefficient, For the uncertainty parameters of the breach evolution, This is the floor function.

9. The method for comprehensive risk assessment of dam-break floods integrating multi-dimensional parameters according to claim 8, characterized in that, In step 5, the thresholds used to determine the risk level are set as follows: a comprehensive risk index ≥ 0.8 is considered extremely high risk, 0.6 ≤ comprehensive risk index < 0.8 is considered high risk, 0.4 ≤ comprehensive risk index < 0.6 is considered medium risk, and a comprehensive risk index < 0.4 is considered low risk. For areas with a population density exceeding 500 people per square kilometer or containing large transportation hubs, the extremely high risk threshold will be lowered to 0.75 and the high risk threshold will be lowered to 0.55; for sparsely populated mountainous or desert areas, the extremely high risk threshold will be raised to 0.85 and the high risk threshold will be raised to 0.65.

Images