Multimodal tailings dam breach digital twin emergency simulation and decision optimization method

Abstract

The present invention proposes a multimodal tailings dam break digital twin emergency simulation and decision optimization method. First, a high-precision three-dimensional scenario model is constructed. Based on the finite element model, the high-precision three-dimensional tailings dam model is simulated and calculated to obtain the dam body stress-strain field. Based on ANSYS+reinforcement learning, the dam body stress-strain field and the current environmental status are used to predict the dam break probability distribution map. Based on the dam break probability distribution map and resource distribution data, a rescue scheduling plan is generated. Then, a dynamic virtual simulation scene is built for exercises, and the rescue scheduling plan is verified and optimized based on the exercise record data to obtain the final rescue scheduling plan. The present invention uses the ANSYS+reinforcement learning hybrid engine to enhance the simulation accuracy, realize dam break path prediction, optimal resource scheduling and plan self-optimization closed loop, and solves the problems of traditional technology data fusion distortion, simulation rigidity, decision lag and other problems. It is particularly suitable for variable working conditions such as sudden changes in the tailings dam seepage field and extreme weather.

Description

Technical Field

[0001] The present invention relates to the field of emergency management technology, and in particular to a multi-modal tailings dam break digital twin emergency simulation and decision optimization method. Background Art

[0002] Traditional disaster drills primarily rely on field exercises and tabletop exercises. While field exercises (such as firefighting and earthquake evacuation) can provide realistic experiences, they are often difficult to organize, costly, and limited in scale. They also struggle to simulate complex or extreme disaster scenarios and pose certain safety risks. Tabletop exercises, on the other hand, rely primarily on paper plans, maps, and verbal discussions, lacking interactivity, immersion, and realism, making them ineffective in training participants' emergency response and decision-making skills. Technological advancements have introduced computer simulations, virtual reality (VR), and augmented reality (AR) into the drill field, providing more controllable, repeatable, and safe simulation environments. However, existing technical solutions still have significant shortcomings in terms of scenario fidelity, multi-role collaboration, dynamic evaluation and feedback during the drill, dynamic matching and optimization of plans, and low-cost, large-scale deployment. In particular, the lack of integrated solutions that efficiently integrate multiple technical approaches, intelligently generate dynamic disaster scenarios, and accurately quantify and evaluate drill effectiveness limits the depth and breadth of drill effectiveness. Therefore, there is an urgent need to develop more advanced, intelligent, efficient, and easily scalable disaster drill methods.

[0003] Currently, existing technologies mainly implement emergency simulations through the following two types of solutions:

[0004] Static parameter deduction system: relies on preset parameters (such as fixed rainfall and dam height) for finite element simulation (such as ANSYS), but cannot respond to environmental changes (such as sudden rainstorms) in real time.

[0005] Single-modal data-driven deduction: Generates risk models based on a single data source (such as GNSS displacement monitoring), but does not address the fusion accuracy issues of multi-source data (satellite imagery, LiDAR point clouds, sensors).

[0006] For example:

[0007] The document number is CN119248108A, and the patent name is a solution for an emergency drill system based on big data. It is used to use cameras, drone equipment and mobile terminals to collect on-site images and video data in real time, use image recognition algorithms to identify key information at the accident scene, process and analyze real-time data obtained from the data acquisition and recognition module, and predict the development of emergency events, and based on this, formulate decisions and simulation tasks at different levels.

[0008] The document number is CN118504399A, and the patent name is a solution for an emergency drill method and system based on digital twins and knowledge graphs. It uses a virtual environment constructed through digital twin technology, and utilizes external parameter measurement devices to collect various environmental data of the site in real time, and then transmits it to the system. Through data processing and analysis, combined with the reasoning ability of the knowledge graph, it can simulate various complex emergency scenarios and emergencies, thereby comprehensively testing and evaluating the feasibility and effectiveness of the emergency plan; it can also automatically optimize and improve the emergency plan based on the drill results, thereby improving the efficiency and accuracy of responding to emergencies.

[0009] From the above, it can be seen that the existing technology has the following defects:

[0010] 1) The emergency drill system based on big data has the following defects:

[0011] ① Through data-driven deduction, based on traditional big data analysis, the problem of multimodal data size differences is not solved;

[0012] ② In terms of interactivity, it relies on preset scenario deductions and has no user dynamic operation interface;

[0013] ③General emergency scenario analysis has poor industry adaptability to tailings ponds;

[0014] ④ Lack of dynamic modeling capabilities for scenarios (dynamic seepage field of tailings dam breach, fusion of infiltration line data and point cloud data);

[0015] 2) The emergency drill system based on digital twins and knowledge graphs has the following defects:

[0016] ①During the data fusion process, there is also the problem of data scale differences;

[0017] ② Knowledge graph reasoning generates emergency decision-making thematic maps through entity relationship analysis to optimize the emergency plan process, but there is a long response delay time;

[0018] ③ Real-time coupling between physical simulation and large-scale model prediction has not been achieved;

[0019] ④ There is a lack of multi-objective resource matching algorithms for emergency scenario problems. Summary of the Invention

[0020] In view of the above situation, the main purpose of the present invention is to propose a multi-modal tailings dam break digital twin emergency simulation and decision optimization method and system to solve the above technical problems.

[0021] The present invention proposes a multi-modal tailings dam breach digital twin emergency simulation and decision optimization method, which includes the following steps:

[0022] Step 1: Collect remote sensing data, meteorological data, and tailings pond monitoring data to build a high-precision three-dimensional scene model;

[0023] Step 2: Based on the finite element model, a high-precision three-dimensional tailings pond model is simulated to simulate the mechanical properties of the tailings pond structure and obtain the stress-strain field of the dam body;

[0024] Step 3: Based on the PPO reinforcement learning algorithm, the dam stress-strain field and the current environmental status are used to predict the dam break probability distribution map;

[0025] Step 4: Generate a rescue dispatch plan based on the dam failure probability distribution map and resource distribution data;

[0026] Step 5: Build a VR scene based on the high-precision three-dimensional tailings pond model, then import the dam break probability distribution map and resource distribution data into the VR scene to obtain a dynamic virtual simulation scene;

[0027] Step 6: In a dynamic virtual simulation scenario, use VR equipment to conduct exercises based on the rescue dispatch plan, and verify and optimize the rescue dispatch plan based on the exercise record data to obtain the final rescue dispatch plan.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. Advantages of dynamic scene generation: Multimodal data fusion achieves a breakthrough in accuracy.

[0030] Traditional data applications often struggle with multimodal data fusion due to differences in spatiotemporal datums and dimensionality, which can easily lead to distortion. By eliminating sensor dimensional differences and aligning coordinate systems using the UTM projection, cross-modal GNSS and InSAR data fusion is achieved with millimeter-level errors (<1mm), significantly improving dam displacement monitoring accuracy and further enhancing reliability in sudden change scenarios.

[0031] 2. Advantages of intelligent deduction: Real-time deduction closed loop driven by dynamic interaction

[0032] Traditional emergency simulations rely on pre-set scripts, making it impossible to dynamically adjust key parameters (such as the flood discharge channel width ΔW), resulting in simulation rigidity. Using a drag-and-drop user interface, dragging operations trigger real-time updates to the hybrid engine (ANSYS finite element calculations + ROM order reduction models + reinforcement learning heatmap warning generation), forming a closed loop of "operation-simulation-feedback." This technical solution significantly improves decision-making efficiency, reduces response time for dam failure path prediction, and supports dynamic intervention of influencing parameters, enhancing system flexibility.

[0033] 3. Plan risk-resource dynamic matching

[0034] Traditional risk management approaches generally rely on manual experience and do not quantify the trade-off between risk probability and cost. By dynamically linking risk heat maps with resource distribution, the NSGA-II multi-objective algorithm generates a Pareto optimal solution set. Based on the dam failure probability distribution (reinforcement learning output) and material location data, it fine-tunes the balance between losses and costs.

[0035] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of the multi-modal tailings dam breach digital twin emergency simulation and decision optimization method proposed in the present invention;

[0037] Figure 2 This is the overall framework diagram of the multi-modal tailings dam break digital twin emergency simulation and decision optimization method proposed in this invention. DETAILED DESCRIPTION

[0038] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0039] These and other aspects of the embodiments of the present invention will become clear with reference to the following description and accompanying drawings. In these descriptions and accompanying drawings, some specific implementations of the embodiments of the present invention are specifically disclosed to illustrate some ways of implementing the principles of the embodiments of the present invention, but it should be understood that the scope of the embodiments of the present invention is not limited thereto.

[0040] See also Figure 1 This embodiment provides a multi-modal tailings dam breach digital twin emergency simulation and decision optimization method, which includes the following steps:

[0041] Step 1: Collect remote sensing data, meteorological data, and tailings pond monitoring data to build a high-precision three-dimensional scene model;

[0042] As a preferred embodiment of the present invention, collecting remote sensing data, meteorological data and tailings pond monitoring data and constructing a high-precision three-dimensional scene model specifically includes the following steps:

[0043] Acquire GNSS displacement meter data, piezometer data, weather station data, tailings pond LiDAR point cloud, and satellite remote sensing images, align the data streams by timestamp, and remove outliers;

[0044] The tailings pond LiDAR point cloud is denoised, and the key point clouds of the tailings pond dam and surrounding surface are retained to obtain the denoised tailings pond LiDAR point cloud;

[0045] The satellite remote sensing image and the denoised tailings pond LiDAR point cloud are aligned in spatial scale through bilinear interpolation to obtain a satellite image in a unified coordinate system.

[0046] The WGS84 coordinates of the GNSS displacement meter data and the piezometer data are converted to the tailings pond LiDAR point cloud coordinate system through UTM projection to obtain the GNSS displacement meter data and the piezometer data in a unified coordinate system;

[0047] The WGS84 coordinates of the geometrically corrected satellite image are converted to the UTM projection, and then aligned to the tailings pond LiDAR point cloud coordinate system through affine transformation and cropped to obtain a satellite image in a unified coordinate system;

[0048] The meteorological station data is mapped to the tailings pond LiDAR point cloud coordinate system through inverse distance weighted interpolation to obtain the meteorological parameters of the unified coordinate system;

[0049] Each band of the satellite image in the unified coordinate system is independently normalized, and the contrast of the surface cover characteristics is retained to obtain the normalized satellite image;

[0050] Normalize the elevation part of the denoised tailings pond LiDAR point cloud to obtain the normalized LiDAR elevation field;

[0051] The GNSS displacement meter data in the unified coordinate system, the osmometer data in the unified coordinate system, and the meteorological parameters in the unified coordinate system are respectively calculated using a sliding window method within the window and linearly mapped to the [-1,1] interval to obtain the normalized GNSS displacement meter data, normalized osmometer distribution, and normalized meteorological parameters;

[0052] The normalized LiDAR elevation field, normalized satellite imagery, normalized GNSS displacement meter data, normalized piezometer data, and normalized satellite imagery are spatiotemporally aligned using an improved Kalman filter to obtain spatiotemporally aligned fused data. A high-precision three-dimensional tailings pond model is constructed based on the spatiotemporally aligned fused data.

[0053] As a preferred embodiment of the present invention, remote sensing data, meteorological data, and tailings pond detection data are collected to construct a high-precision three-dimensional scene model, specifically comprising the following steps: normalized LiDAR elevation field, normalized satellite imagery, normalized GNSS displacement meter data, normalized piezometer data, and normalized satellite imagery are spatially and temporally aligned using an improved Kalman filter to obtain spatially and temporally aligned fused data; and a high-precision three-dimensional tailings pond model is constructed based on the spatially and temporally aligned fused data, specifically comprising the following steps:

[0054] The state vector and physical model are constructed based on the normalized GNSS displacement meter data, normalized seepage pressure distribution, and normalized meteorological parameters, and the weights of the state vector and physical model are dynamically adjusted according to the sensor accuracy;

[0055] The state vector and physical model are used to predict the current dam displacement and pressure trends, and the prediction results are used for correction. After the prediction-correction cycle is completed, the time-space aligned fusion data is obtained. The time-space aligned fusion data includes millimeter-level displacement, continuous pressure distribution, and gridded meteorological parameters.

[0056] The normalized LiDAR elevation field is superimposed with the normalized GNSS displacement meter data, and a centimeter-level resolution 3D surface mesh is generated through Delaunay triangulation.

[0057] Normalized piezometer data are mapped to the internal nodes of the three-dimensional surface network to construct the pore water pressure field; normalized meteorological parameters are interpolated to the three-dimensional surface grid to generate rainfall / wind speed distribution; based on the time-space aligned fusion data, the grid node coordinates are adjusted in real time to simulate the dam deformation, and a high-precision three-dimensional tailings pond model that integrates displacement, pressure, meteorology, and terrain is obtained.

[0058] Step 2: Based on the finite element model, a high-precision three-dimensional tailings pond model is simulated to simulate the mechanical properties of the tailings pond structure and obtain the stress-strain field of the dam body;

[0059] As a preferred embodiment of the present invention, a high-precision three-dimensional tailings pond model is simulated based on a finite element model to simulate the mechanical properties of the tailings pond structure and obtain the stress-strain field of the dam body, which specifically includes the following steps:

[0060] Extract the dam body geometry from the high-precision 3D tailings pond model, perform material differentiation, load the material constitutive model parameters for the material, and obtain the finite element geometry model;

[0061] The potential plastic zone in the finite element geometric model is encrypted with a hexahedron mesh, the far-field foundation is encrypted with a tetrahedron mesh, and the contact element and target element pairing mesh is generated at the interface between the dam and the water body to obtain a non-uniformly encrypted finite element mesh model.

[0062] The displacement-pore pressure coupling equation is used to treat the rockfill in the non-uniformly densified finite element mesh model. The pore water pressure field is used as the body force boundary condition, and displacement constraints are applied to obtain a finite element solution model that includes complete physical equations and boundary conditions.

[0063] A high-precision three-dimensional tailings pond model is input as an external dynamic load into a finite element solution model containing complete physical equations and boundary conditions to obtain a full-order finite element model. The full-order finite element model is then iteratively solved to obtain the critical reduction factor, unstable slip surface coordinates, and strain field evolution data.

[0064] The full-order finite element model is mapped to a reduced-order model and encapsulated as a callable module. The stress-strain field distribution cloud map of the dam body is generated according to the critical reduction coefficient, the coordinates of the unstable sliding surface and the strain field evolution number.

[0065] Step 3: Based on the PPO reinforcement learning algorithm, the dam stress-strain field and the current environmental status are used to predict the dam break probability distribution map;

[0066] As a preferred embodiment of the present invention, based on the PPO reinforcement learning algorithm, using the dam stress-strain field and the current environmental state, predicting the dam break probability distribution map specifically includes the following steps:

[0067] Based on the current environmental state and current action, the reduced-order model is used to output the dam body stress-strain field distribution cloud map, equivalent plastic strain zone distribution, and strength reduction factor. The current environmental state includes water level elevation, rainfall intensity, and soil saturation. The current action includes flood discharge channel operation instructions and personnel evacuation operation instructions.

[0068] A reward function is constructed using the strength reduction coefficient as a stability reward, an exponential penalty triggered when the equivalent plastic strain zone distribution area exceeds a threshold as a risk penalty, and the matching result between the flood discharge channel opening status and the personnel evacuation heat zone as an artificial rule reward to calculate the reward for the output of the reduced-order model.

[0069] Based on the Actor-Critic network architecture, the actor network generates online actions based on the current environmental state, taking the flood discharge channel opening, sub-dam reinforcement strength, and drainage pump start-stop frequency as actions. The reduced-order model output is converted into target actions for learning. The critic network is used to predict expected rewards based on online actions and uses the rewards of the reduced-order model output as expected values ​​for learning, guiding the generation of online actions by the actor network. After learning is complete, the optimal action is output.

[0070] Random perturbations are superimposed on the optimal action to perform Monte Carlo simulation. The frequency of dam failure events during the Monte Carlo simulation is counted to generate a two-dimensional probability distribution in space and time.

[0071] The spatiotemporal two-dimensional probability distribution is interpolated to the grid nodes of the full-order finite element model, and the dam body stress-strain field distribution cloud map is superimposed for weighted rendering to obtain the dam break probability distribution thermal map.

[0072] Step 4: Generate a rescue dispatch plan based on the dam failure probability distribution map and resource distribution data;

[0073] As a preferred embodiment of the present invention, generating a rescue dispatch plan based on the dam failure probability distribution map and resource distribution data specifically includes the following steps:

[0074] The dam failure probability distribution heat map is divided into risk areas of different levels according to the threshold value;

[0075] According to the obtained resource distribution data, the corresponding resource points are obtained and the corresponding utility weights are given according to the resource categories;

[0076] Based on the road network topology and real-time traffic conditions, resource points and risk areas are associated with the nearest path, and the estimated travel time is marked to obtain a risk-resource association matrix that includes risk level, resource type, path time, and utility weight.

[0077] A multi-objective optimization model was constructed with the optimization objectives of minimizing the probability of dam failure, minimizing the affected population density, minimizing the economic loss coefficient, minimizing the average arrival time from resource points to risk areas, and maximizing the matching degree between material allocation and demand forecast, and with the total resource limit and time window limit as constraints.

[0078] Based on the risk-resource correlation matrix, the multi-objective optimization model is iteratively optimized to obtain the Pareto optimal solution set;

[0079] Calculate the comprehensive score of each solution in the Pareto optimal solution set based on economic losses, arrival time, and material allocation utilization;

[0080] The Pareto optimal solution set is sorted according to the comprehensive score and divided into priority execution plan and backup plan to obtain a hierarchical rescue dispatch plan.

[0081] Step 5: Build a VR scene based on the high-precision three-dimensional tailings pond model, then import the dam break probability distribution map and resource distribution data into the VR scene to obtain a dynamic virtual simulation scene;

[0082] As a preferred embodiment of the present invention, a VR scene is built based on a high-precision three-dimensional tailings pond model, and then a dam break probability distribution map and resource distribution data are imported into the VR scene to obtain a dynamic virtual deduction scene, which specifically includes the following steps:

[0083] Map the hierarchical rescue dispatch plan to a high-precision 3D tailings pond model to generate a planned route;

[0084] Mechanical properties are added to the virtual objects of the high-precision three-dimensional tailings pond model, and the scene timelines of all terminals are aligned based on the cloud synchronization server to obtain a dynamic virtual deduction scene.

[0085] Step 6: In a dynamic virtual simulation scenario, use VR equipment to conduct exercises based on the rescue dispatch plan, and verify and optimize the rescue dispatch plan based on the exercise record data to obtain the final rescue dispatch plan.

[0086] As a preferred embodiment of the present invention, in a dynamic virtual deduction scenario, a rescue dispatch plan is exercised using VR equipment, and the rescue dispatch plan is verified and optimized based on the exercise record data to obtain a final rescue dispatch plan, which specifically includes the following steps:

[0087] Assign VR equipment to each staff member and determine their position in the dynamic virtual simulation scene;

[0088] In the dynamic virtual simulation scene, the optimal path is generated according to the location of the staff and the planned route, and a highlighted arrow is generated;

[0089] Staff members move and perform drills according to the highlighted arrows and record the drill data;

[0090] The commander makes real-time interventions and adjustments to the exercise process based on the current exercise process and the overall dispatch situation, and records the commander's command data;

[0091] Based on the exercise data, the VR equipment operation records, actual resource consumption, actual resource scheduling time, and dam break suppression effects were obtained, and the commander's command data was manually labeled.

[0092] A multi-dimensional evaluation matrix is ​​constructed based on the deviation rate between the actual resource scheduling time and the planned time, the matching between the actual resource consumption and the demand forecast, and the reduction in the probability of dam failure after the simulation.

[0093] Based on a multi-dimensional evaluation matrix, a causal reasoning engine was used to evaluate and analyze the rescue dispatch plans, classifying them into high-quality and low-quality plans. The core operations that affected the results during the exercise were also identified, resulting in key decision points.

[0094] The best solutions are stored in the case knowledge base and used as the final rescue dispatch plan;

[0095] The key decision points of the inferior solutions are obtained, and the parameters of the multi-objective optimization model are optimized according to the key decision points of the inferior solutions, so as to realize parameter tuning by using the inferior solutions as negative examples.

[0096] The architecture of the present invention is as follows Figure 2As shown in the figure, it includes a data acquisition layer, an intermediate processing layer, and an output linkage layer. The data acquisition layer is used for data collection, primarily from field equipment such as GNSS displacement meters, piezometers, and weather stations. Remote sensing data includes InSAR satellites and UAV LiDAR. The output linkage layer primarily includes desktop and mobile terminals for display and operation, as well as the emergency equipment control command flow for sending commands.

[0097] The middle processing layer mainly includes three key functions:

[0098] 1) Dynamic scenarios: Equipment data and remote sensing data from the data acquisition layer are collected and key features are extracted. For example, GNSS displacement meters (accuracy ±1 mm), piezometers (range 0-2 MPa), and weather stations (wind speed / rainfall monitoring) are deployed in the reservoir area. Kalman filtering algorithms are used to align the spatiotemporal references of multimodal data to construct a high-precision 3D scenario model.

[0099] 2) Intelligent deduction engine: Physical simulation core: Calculates the stress-strain field of the dam body based on the finite element model;

[0100] AI deduction core: The PPO reinforcement learning algorithm is used to train the intelligent agent. The input parameters include real-time water level, rainfall intensity, and soil saturation, and the output is a dam break probability distribution map.

[0101] 3) Plan self-optimization: Digital twin editor: provides a dam break impact radius calculator, such as dam height, reservoir capacity, geological correction coefficient, etc.

[0102] Closed-loop feedback mechanism: Reversely optimize the reward function of the AI ​​deduction core through exercise results, such as adjusting the casualty weight coefficient.

[0103] In this embodiment, the present invention adopts a drag-and-drop data fusion interface.

[0104] During the tailings pond emergency drill, the user dragged the drone LiDAR point cloud into the scene, and the system automatically aligned the GNSS displacement data and meteorological data to generate a dynamic 3D model of the dam break path.

[0105] ① Visual drag tool: Users can drag satellite images, point clouds, and sensor data into the 3D scene editor to trigger the feature alignment algorithm (spatiotemporal difference technology)

[0106] ② Automatic feature recognition: The algorithm automatically extracts key features from the data (such as the edge of the tailings dam and crack displacement points) and updates the 3D model in real time.

[0107] The overall business operation logic is as follows:

[0108] 1) First, data is collected through sensors.

[0109] Through the improved Kalman filtering algorithm: the adaptive instance normalization (AdaIN) module is introduced, and the following formula is used to eliminate the scale difference of sensor data.

[0110] ;

[0111] in, x Represents GNSS displacement data; y Represents the LiDAR point cloud coordinate system;

[0112] 2) Cross-modal calibration process: UTM projection conversion is used to align the WGS84 coordinate system (GNSS) with the local grid coordinate system (LiDAR). GNSS displacement data (millimeter-level accuracy) and LiDAR point clouds are combined using oblique photography to generate a 3D grid. Real-time meteorological data is mapped to the grid nodes using an interpolation algorithm.

[0113] 3) Hybrid deduction engine

[0114] ① Hybrid deduction engine: Physical simulation (ANSYS finite element model) simulates the structural mechanical properties of the tailings pond. Hybrid interpolation (uP hybrid formula) is used for the dam body to deal with incompressible deformation. The CONTA174+TARGE170 equations are used to simulate the potential slip surface on the contact surface with water. The potential plastic zone (dam slope, core wall) is encrypted to 0.5~1 times the dam high resolution, and a gradually sparse grid is used for the far-field foundation to reduce the calculation pressure. The strength reduction method is used to simulate the dynamic process of geological changes such as water level changes and earthquakes. Finally, based on the calculation results, the maximum principal stress, equivalent plastic strain zone, and the reduction factor FS when the strength is reduced to non-convergence are predicted, and finally a visual cloud map and predicted data are output. In addition, the simulation model is mapped to a reduced-order model for fast calculation and encapsulated as a callable simulation module. On-site personnel can customize the input parameters and obtain instant simulation feedback.

[0115] ② User interaction driven: The reinforcement learning module predicts the risk probability distribution (such as the curve of dam failure probability changing over time).

[0116] 4) Risk dynamic modeling

[0117] ② User interaction driven: By dragging and dropping to adjust parameters (such as sub-dam reinforcement strength and flood discharge channel width), AI is triggered to dynamically update the risk heat map.

[0118] 5) Dynamic matching of risks and resources

[0119] ① Dynamic matching algorithm: Generates a Pareto optimal solution set based on risk probability distribution and resource distribution data (such as the location of emergency supplies and the number of rescue personnel).

[0120] ② Recommendation logic: Prioritize matching high-risk areas (such as sub-dams upstream of dam breaches) with the nearest resource points, and calculate the response time and success rate weights.

[0121] 6) Closed-loop optimization mechanism

[0122] ① Digital Twin Editor: Simulation results are fed back to the emergency plan library, allowing users to modify strategies using a drag-and-drop interface (using sliders to adjust evacuation route priorities and hotspot resource deployment areas). An emergency decision-making model is constructed based on reinforcement learning, optimizing disaster avoidance route planning and rescue resource scheduling through algorithms such as deep networks. The system evaluates the effectiveness of different response plans in real time, rehearses the execution of each plan in a virtual environment, and uses augmented reality to mark the optimal course of action. Commanders use a 3D panoramic view to coordinate the overall situation, while rescue personnel wear VR equipment for immersive operational training. Force feedback devices simulate the operating resistance of rescue machinery, and cloud synchronization technology ensures consistency of multi-terminal drill scenarios.

[0123] ② Self-optimization logic: The system records historical simulation data, optimizes parameter weights using a genetic algorithm, generates a new version of the emergency plan, and pushes it to the mobile command system. It establishes an evaluation matrix encompassing dimensions such as response timeliness and resource utilization, and uses a causal reasoning engine to analyze key decision points during the drill. The system automatically generates a case knowledge base, driving the continuous optimization and updating of the emergency plan, forming a closed loop of "drill-evaluation-improvement."

[0124] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0125] It should be understood that various components of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0126] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0127] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A multi-modal tailings dam breach digital twin emergency simulation and decision optimization method, characterized by: The method comprises the following steps: Step 1: Collect remote sensing data, meteorological data, and tailings pond monitoring data to build a high-precision three-dimensional scene model; Step 2: Based on the finite element model, a high-precision three-dimensional tailings pond model is simulated to simulate the mechanical properties of the tailings pond structure and obtain the stress-strain field of the dam body; Step 3: Based on the PPO reinforcement learning algorithm, the dam stress-strain field and the current environmental status are used to predict the dam break probability distribution map; Step 4: Generate a rescue dispatch plan based on the dam failure probability distribution map and resource distribution data; Step 5: Build a VR scene based on the high-precision three-dimensional tailings pond model, then import the dam break probability distribution map and resource distribution data into the VR scene to obtain a dynamic virtual simulation scene; Step 6: In a dynamic virtual simulation scenario, use VR equipment to conduct an exercise based on the rescue dispatch plan, and verify and optimize the rescue dispatch plan based on the exercise record data to obtain the final rescue dispatch plan; In step 1, collecting remote sensing data, meteorological data, and tailings pond detection data to construct a high-precision three-dimensional scene model specifically includes the following steps: Acquire GNSS displacement meter data, piezometer data, weather station data, tailings pond LiDAR point cloud, and satellite remote sensing images, align the data streams by timestamp, and remove outliers; The tailings pond LiDAR point cloud is denoised, and the key point clouds of the tailings pond dam and surrounding surface are retained to obtain the denoised tailings pond LiDAR point cloud; The WGS84 coordinates of the GNSS displacement meter data and the piezometer data are converted to the tailings pond LiDAR point cloud coordinate system through UTM projection to obtain the GNSS displacement meter data and the piezometer data in a unified coordinate system; The WGS84 coordinates of the geometrically corrected satellite image are converted to the UTM projection, and then aligned to the tailings pond LiDAR point cloud coordinate system through affine transformation and cropped to obtain a satellite image in a unified coordinate system; The meteorological station data is mapped to the tailings pond LiDAR point cloud coordinate system through inverse distance weighted interpolation to obtain the meteorological parameters of the unified coordinate system; Each band of the satellite image in the unified coordinate system is independently normalized, and the contrast of the surface cover characteristics is retained to obtain the normalized satellite image; Normalize the elevation part of the denoised tailings pond LiDAR point cloud to obtain the normalized LiDAR elevation field; The GNSS displacement meter data in the unified coordinate system, the osmometer data in the unified coordinate system, and the meteorological parameters in the unified coordinate system are respectively calculated using a sliding window method within the window and linearly mapped to the [-1,1] interval to obtain the normalized GNSS displacement meter data, normalized osmometer distribution, and normalized meteorological parameters; The normalized LiDAR elevation field, normalized satellite imagery, normalized GNSS displacement meter data, normalized piezometer data, and normalized satellite imagery are spatiotemporally aligned using an improved Kalman filter to obtain spatiotemporally aligned fused data. A high-precision three-dimensional tailings pond model is constructed based on the spatiotemporally aligned fused data.

2. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 1, characterized in that: The normalized LiDAR elevation field, normalized satellite imagery, normalized GNSS displacement meter data, normalized piezometer data, and normalized satellite imagery are temporally and spatially aligned using an improved Kalman filter to obtain temporally and spatially aligned fused data. A high-precision three-dimensional tailings pond model is constructed based on the temporally and spatially aligned fused data. The specific steps include: The state vector and physical model are constructed based on the normalized GNSS displacement meter data, normalized seepage pressure distribution, and normalized meteorological parameters, and the weights of the state vector and physical model are dynamically adjusted according to the sensor accuracy; The state vector and physical model are used to predict the current dam displacement and pressure trends, and the prediction results are used for correction. After the prediction-correction cycle is completed, the time-space aligned fusion data is obtained. The time-space aligned fusion data includes millimeter-level displacement, continuous pressure distribution, and gridded meteorological parameters. The normalized LiDAR elevation field is superimposed with the normalized GNSS displacement meter data, and a centimeter-level resolution 3D surface mesh is generated through Delaunay triangulation. The normalized piezometer data are mapped to the internal nodes of the three-dimensional surface network to construct the pore water pressure field; Normalized meteorological parameters are interpolated into the three-dimensional surface grid to generate rainfall / wind speed distribution. Based on the time-space aligned fusion data, the grid node coordinates are adjusted in real time to simulate the dam deformation, and a high-precision three-dimensional tailings pond model that integrates displacement, pressure, meteorology, and terrain is obtained.

3. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 2, characterized in that: In step 2, simulation calculation is performed on a high-precision three-dimensional tailings pond model based on a finite element model to simulate the mechanical properties of the tailings pond structure and obtain the stress-strain field of the dam body. Specifically, the steps include: Extract the dam body geometry from the high-precision 3D tailings pond model, perform material differentiation, load the material constitutive model parameters for the material, and obtain the finite element geometry model; The potential plastic zone in the finite element geometric model is encrypted with a hexahedron mesh, the far-field foundation is encrypted with a tetrahedron mesh, and the contact element and target element pairing mesh is generated at the interface between the dam and the water body to obtain a non-uniformly encrypted finite element mesh model. The displacement-pore pressure coupling equation is used to treat the rockfill in the non-uniformly densified finite element mesh model. The pore water pressure field is used as the body force boundary condition, and displacement constraints are applied to obtain a finite element solution model that includes complete physical equations and boundary conditions. A high-precision three-dimensional tailings pond model is input as an external dynamic load into a finite element solution model containing complete physical equations and boundary conditions to obtain a full-order finite element model. The full-order finite element model is then iteratively solved to obtain the critical reduction factor, unstable slip surface coordinates, and strain field evolution data. The full-order finite element model is mapped to a reduced-order model and encapsulated as a callable module. The stress-strain field distribution cloud map of the dam body is generated according to the critical reduction coefficient, the coordinates of the unstable sliding surface and the strain field evolution number.

4. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 3, characterized in that: In step 3, based on the PPO reinforcement learning algorithm, using the dam stress-strain field and the current environmental state, the prediction of the dam break probability distribution map specifically includes the following steps: Based on the current environmental state and current action, the reduced-order model is used to output the dam body stress-strain field distribution cloud map, equivalent plastic strain zone distribution, and strength reduction factor. The current environmental state includes water level elevation, rainfall intensity, and soil saturation. The current action includes flood discharge channel operation instructions and personnel evacuation operation instructions. A reward function is constructed using the strength reduction coefficient as a stability reward, an exponential penalty triggered when the equivalent plastic strain zone distribution area exceeds a threshold as a risk penalty, and the matching result between the flood discharge channel opening status and the personnel evacuation heat zone as an artificial rule reward to calculate the reward for the output of the reduced-order model. Based on the Actor-Critic network architecture, the actor network generates online actions based on the current environmental state, taking the flood discharge channel opening, sub-dam reinforcement strength, and drainage pump start-stop frequency as actions. The reduced-order model output is converted into target actions for learning. The critic network is used to predict expected rewards based on online actions and uses the rewards of the reduced-order model output as expected values ​​for learning, guiding the generation of online actions by the actor network. After learning is complete, the optimal action is output. Random perturbations are superimposed on the optimal action to perform Monte Carlo simulation. The frequency of dam failure events during the Monte Carlo simulation is counted to generate a two-dimensional probability distribution in space and time. The spatiotemporal two-dimensional probability distribution is interpolated to the grid nodes of the full-order finite element model, and the dam body stress-strain field distribution cloud map is superimposed for weighted rendering to obtain the dam break probability distribution thermal map.

5. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 4, characterized in that: In step 4, generating a rescue dispatch plan based on the dam failure probability distribution map and resource distribution data specifically includes the following steps: The dam failure probability distribution heat map is divided into risk areas of different levels according to the threshold value; According to the obtained resource distribution data, the corresponding resource points are obtained and the corresponding utility weights are given according to the resource categories; Based on the road network topology and real-time traffic conditions, resource points and risk areas are associated with the nearest path, and the estimated travel time is marked to obtain a risk-resource association matrix that includes risk level, resource type, path time, and utility weight. A multi-objective optimization model was constructed with the optimization objectives of minimizing the probability of dam failure, minimizing the affected population density, minimizing the economic loss coefficient, minimizing the average arrival time from resource points to risk areas, and maximizing the matching degree between material allocation and demand forecast, and with the total resource limit and time window limit as constraints. Based on the risk-resource correlation matrix, the multi-objective optimization model is iteratively optimized to obtain the Pareto optimal solution set; Calculate the comprehensive score of each solution in the Pareto optimal solution set based on economic losses, arrival time, and material allocation utilization; The Pareto optimal solution set is sorted according to the comprehensive score and divided into priority execution plan and backup plan to obtain a hierarchical rescue dispatch plan.

6. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 5, characterized in that: In step 5, a VR scene is built based on a high-precision three-dimensional tailings pond model, and then the dam break probability distribution map and resource distribution data are imported into the VR scene to obtain a dynamic virtual deduction scene, which specifically includes the following steps: Map the hierarchical rescue dispatch plan to a high-precision 3D tailings pond model to generate a planned route; Mechanical properties are added to the virtual objects of the high-precision three-dimensional tailings pond model, and the scene timelines of all terminals are aligned based on the cloud synchronization server to obtain a dynamic virtual deduction scene.

7. A multi-modal tailings dam break digital twin emergency simulation and decision optimization method according to claim 6, characterized in that: In step 6, in a dynamic virtual simulation scenario, a rescue dispatch plan is exercised using VR equipment, and the rescue dispatch plan is verified and optimized based on the exercise record data to obtain a final rescue dispatch plan, which specifically includes the following steps: Assign VR equipment to each staff member and determine their position in the dynamic virtual simulation scene; In the dynamic virtual simulation scene, the optimal path is generated according to the location of the staff and the planned route, and a highlighted arrow is generated; Staff members move and perform drills according to the highlighted arrows and record the drill data; The commander makes real-time interventions and adjustments to the exercise process based on the current exercise process and the overall dispatch situation, and records the commander's command data; Based on the exercise data, the VR equipment operation records, actual resource consumption, actual resource scheduling time, and dam break suppression effects were obtained, and the commander's command data was manually labeled. A multi-dimensional evaluation matrix is ​​constructed based on the deviation rate between the actual resource scheduling time and the planned time, the matching between the actual resource consumption and the demand forecast, and the reduction in the probability of dam failure after the simulation. Based on a multi-dimensional evaluation matrix, a causal reasoning engine was used to evaluate and analyze the rescue dispatch plans, classifying them into high-quality and low-quality plans. The core operations that affected the results during the exercise were also identified, resulting in key decision points. The best solutions are stored in the case knowledge base and used as the final rescue dispatch plan; The key decision points of the inferior solutions are obtained, and the parameters of the multi-objective optimization model are optimized according to the key decision points of the inferior solutions, so as to realize parameter tuning by using the inferior solutions as negative examples.

Images