Visual simulation and prediction system for mine ground subsidence risk

Abstract

This invention relates to the field of geological disaster monitoring and early warning technology, and discloses a visualization simulation and prediction system for mine ground subsidence risk. The system includes: a data acquisition and management module, which mainly acquires monitoring data and basic data related to mine ground subsidence risk in real time or periodically through multiple heterogeneous data sources, and the monitoring data and basic data are uniformly stored and managed; and a data processing and fusion module, which is connected to the data acquisition and management module. This invention integrates multi-source heterogeneous data from satellite remote sensing, aerial remote sensing, ground monitoring, and geological mining through the data acquisition and management module, and combines the preprocessing, spatiotemporal reference registration, and feature fusion technology of the data processing and fusion module to solve the shortcomings of traditional single technologies: it overcomes the problems of limited coverage and high cost of ground measurement, and makes up for the defects of aerial photography being constrained by weather and lacking flexibility.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, specifically a visualization simulation and prediction system for mine ground subsidence risk. Background Technology

[0002] Effective assessment and prevention of mine surface subsidence risk are crucial in mining operations. Currently, the relevant technical means for addressing mine surface subsidence risk have certain limitations.

[0003] In terms of traditional monitoring technologies, ground surveying relies on tools such as levels and total stations to directly measure changes in the earth's surface position. While offering high accuracy over small areas, it suffers from time-consuming and costly applications in large mining areas. Furthermore, its application is limited by terrain and surface conditions, making it difficult to comprehensively cover the entire mining area. Aerial photography, using aircraft and helicopters equipped with photographic equipment to acquire images of the mining area to generate topographic maps and analyze surface changes, can cover large areas and provide high-resolution images. However, it is costly, lacks flexibility, and its feasibility is severely limited in areas with heavy cloud cover or adverse weather conditions. Satellite remote sensing technology, while capable of continuous monitoring of vast mining areas from a global perspective, typically has lower data resolution than ground surveying and aerial photography. This makes it insufficient for detailed analysis of small-scale surface changes, and the frequency of data acquisition is constrained by satellite orbits and mission scheduling, easily leading to data acquisition delays.

[0004] In the field of predictive models, existing probabilistic integral methods are based on the laws governing rock strata movement to establish mathematical models. These are suitable for layered geological conditions and can estimate parameters such as surface subsidence and tilt, but their adaptability to complex geological situations is limited. Numerical simulation software such as FLAC3D and UDEC can consider factors such as rock joints, faults, and groundwater to simulate stress-strain responses during mining and predict surface displacement and potential fracture zones. However, the calculation process is complex and requires significant computational resources. Artificial neural network / machine learning models are trained using historical subsidence data and combined with geological parameters to achieve dynamic predictions. However, model training requires a large amount of high-quality data, and the accuracy and stability of the model are significantly affected by factors such as data quality and model parameter settings. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a visualization simulation and prediction system for mine ground subsidence risk, solving the problem of "low work efficiency" in the aforementioned background technologies.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a visualization simulation and prediction system for mine surface subsidence risk, the system comprising:

[0009] The data acquisition and management module mainly acquires monitoring data and basic data related to the risk of mine ground subsidence in real time or periodically through multiple heterogeneous data sources. The monitoring data and basic data are stored and managed in a unified manner.

[0010] The data processing and fusion module, connected to the data acquisition and management module, mainly performs preprocessing, spatiotemporal reference registration, and feature fusion on the monitoring data and basic data to generate a standardized multidimensional spatiotemporal dataset. The preprocessing includes format conversion, noise filtering, outlier removal, and data interpolation of the raw data. The spatiotemporal reference registration includes unifying all data from all sources to the same geographic coordinate system and time reference frame.

[0011] The hybrid risk prediction module, connected to the data processing and fusion module, mainly receives the standardized multidimensional spatiotemporal dataset and generates a ground subsidence prediction field and collapse risk level distribution map for a specific future time period in the mining area through a collaborative prediction mechanism composed of a machine learning prediction unit and a physical numerical simulation unit.

[0012] The 3D visualization and early warning interaction module is connected to the hybrid risk prediction module. It mainly overlays the ground subsidence prediction field and the collapse risk level distribution map onto a 3D comprehensive model that includes the surface topography, underground geological structure and mining activity process of the mining area. This enables dynamic visualization simulation of the spatiotemporal evolution of risk, and automatically generates and issues graded early warning information when the predicted risk indicators exceed the preset threshold.

[0013] Preferably, the data acquisition and management module includes:

[0014] The satellite remote sensing data acquisition unit mainly acquires interferometric data, differential interferometric data, and multispectral image data covering the entire mining area from synthetic aperture radar satellite systems and high-resolution optical satellite systems, for large-scale, long-term monitoring of surface deformation fields.

[0015] The aerial remote sensing data acquisition unit mainly uses lidar sensors and high-definition digital cameras carried by drones or aircraft to acquire high-precision digital elevation models, digital surface models and orthophoto maps of key areas of interest in the mining area, for refined analysis of surface morphology changes.

[0016] The ground monitoring data acquisition unit mainly receives and integrates high-frequency, high-precision point displacement and deformation data collected by continuously operating reference stations of the Global Navigation Satellite System, automated total stations, high-precision levels, crack gauges, and inclinometers deployed at key locations in the mining area.

[0017] The geological and mining data input unit mainly imports and structures geological exploration report data, rock mechanics parameters, hydrogeological conditions data, historical mining records, current mining face locations, and future mining plan data for the mining area, providing physical constraints and driving parameters for the prediction model.

[0018] Preferably, in the data processing and fusion module, the satellite remote sensing data, aerial remote sensing data, and ground monitoring data are synchronized in the time dimension. A deformation time series with a unified time interval is generated through interpolation or extrapolation algorithms. In the spatial dimension, discrete point displacement data and areal deformation data of different resolutions are fused using general kriging interpolation or radial basis function interpolation methods to generate a high-resolution three-dimensional deformation field that is spatiotemporally continuous and covers the entire mining area. In addition, key static and dynamic characteristic parameters affecting surface collapse are extracted from the structured geological and mining data. The static characteristic parameters include stratigraphic lithology, geological structural complexity index, fault activity, and rock mass quality classification. The dynamic characteristic parameters include time-varying functions of mining depth, mining thickness, advance speed, and goaf volume.

[0019] Preferably, the collaborative prediction mechanism in the hybrid risk prediction module is implemented using the following steps:

[0020] Step 1: The machine learning prediction unit first receives the standardized multidimensional spatiotemporal dataset and uses a pre-trained deep learning network model to perform a rapid and global preliminary prediction of the ground subsidence rate and a preliminary classification of the collapse risk probability for the entire mining area, thereby quickly identifying potential high-risk areas.

[0021] Step 2: The physical numerical simulation unit receives the boundary conditions and refined feature parameters of the high-risk area identified by the machine learning prediction unit, and performs high-precision stress-strain coupling analysis and failure process simulation on the high-risk area through the rock strata movement calculation engine, outputting detailed displacement field, stress field and plastic zone distribution;

[0022] Step 3: The hybrid risk prediction module further feeds back the accurate simulation results of the physical numerical simulation unit in Step 2 as high-quality samples to the machine learning prediction unit for online model updates and continuous optimization. At the same time, the accurate prediction results of the physical numerical simulation unit and the global prediction results of the machine learning prediction unit are weighted and fused to generate the final ground subsidence prediction field and collapse risk level distribution map that takes into account both global coverage and local accuracy.

[0023] Preferably, the machine learning prediction unit employs a spatiotemporal convolutional long short-term memory network model, mainly including the following steps:

[0024] Step 1: Through its convolutional layer structure, automatically extract the spatial correlation features of the fused three-dimensional deformation field and geological characteristic parameters to capture the spatial continuity and clustering patterns of surface deformation;

[0025] Step 2: By using the long short-term memory network layer structure, the long-term time dependence of the surface deformation time series and mining activity dynamic parameters is effectively learned, and the nonlinear dynamic trend of the subsidence evolution process is captured.

[0026] Step 3: The training process of the model uses historical multi-source monitoring data and corresponding mining activity records as input. Through the backpropagation algorithm and adaptive moment estimation optimizer, the root mean square error loss function between the predicted deformation and the actual observed deformation is minimized, thereby obtaining the optimal model weight parameters.

[0027] Preferably, the physical numerical simulation unit includes:

[0028] The three-dimensional geomechanical model builder mainly uses the data provided by the geological and mining data input unit to automatically or semi-automatically build a three-dimensional solid geological model that can accurately reflect the stratigraphic structure, rock joints, fault distribution and material properties of the mining area.

[0029] The dynamic simulator of the mining process mainly simulates the gradual advancement of the mining face, the formation and expansion of the goaf, and the application of the filling material on the three-dimensional physical geological model, and applies them as time-varying boundary conditions and loads to the calculation model.

[0030] The stress-seepage-damage coupled solver mainly uses constitutive models that can characterize the elastoplasticity, strain softening and damage evolution of rock mass, including but not limited to the Mohr-Coulomb criterion or the Hoek-Brown criterion, and couples the influence of groundwater seepage field to solve the entire process of stress redistribution, deformation, failure and resulting surface displacement of rock mass under mining disturbance.

[0031] Preferably, the 3D visualization and early warning interaction module includes:

[0032] The integrated scene rendering unit mainly integrates the three-dimensional geological model, the high-precision surface terrain model, and the mine tunnel and mining space model to create a high-fidelity digital twin scene of the mining area.

[0033] The risk spatiotemporal evolution playback unit mainly overlays a series of ground subsidence prediction fields and collapse risk level distribution maps at different time points output by the hybrid risk prediction module onto the digital twin scene in the form of cloud maps, contour lines, vector arrows or color blocks. It supports users to drag the timeline, rewind, and fast forward, and intuitively displays the evolution process and expansion trend of the risk area.

[0034] The multi-dimensional information query unit is mainly designed to allow users to select any spatial point, line, or surface on the three-dimensional visualization interface and instantly query the historical deformation curve, future deformation prediction curve, current risk level, main disaster-causing factors, and related monitoring data source information for that location.

[0035] Preferably, the collapse risk level distribution map is generated through a multi-factor comprehensive evaluation model, specifically including the following steps:

[0036] Step 1: Establish a risk assessment index system, which should include at least the maximum settlement value, maximum settlement rate, maximum tilt, maximum curvature, and horizontal deformation;

[0037] Step 2: Calculate the values ​​of each indicator in the indicator system from the ground subsidence prediction field output by the hybrid risk prediction module;

[0038] Step 3: Based on relevant national standards or specific safety requirements of the mining area, set corresponding multi-level risk thresholds for each indicator, such as four levels: "Safe", "Caution", "Danger", and "Extremely Dangerous".

[0039] Step 4: Using the analytic hierarchy process (AHP) or fuzzy comprehensive evaluation method, weighted calculations are performed on each indicator to obtain the comprehensive risk index of each spatial grid unit. Based on the range of the comprehensive risk index, it is mapped to the final collapse risk level, thereby generating the collapse risk level distribution map.

[0040] Preferably, the tiered early warning information includes:

[0041] A blue alert is triggered when the predicted comprehensive risk index falls within the "attention" level range, and the system sends a routine attention notification to the administrator's terminal.

[0042] A yellow alert is triggered when the predicted comprehensive risk index enters the "dangerous" level range or when the predicted value of a key indicator exceeds its danger threshold for the first time. The system automatically generates a detailed report containing the risk location, scope of impact, and predicted evolution trend, and pushes it to the mobile terminal of the on-site management and technical personnel.

[0043] An orange alert is triggered when the predicted comprehensive risk index is in the "dangerous" level range and continues to rise, or when the predicted rate of change of key indicators exceeds the level 2 rate threshold. The system will activate the audible and visual alarm device and issue evacuation warnings to people in and around the risk area through the emergency broadcast system.

[0044] A red alert is triggered when the predicted comprehensive risk index enters the "extremely dangerous" level range, or when the system determines that a collapse or instability is about to occur. The system automatically sends the highest level of alarm information to the mine's top decision-making level and the local emergency management department, and activates the preset emergency response plan.

[0045] (III) Beneficial Effects

[0046] This invention provides a visual simulation and prediction system for mine surface subsidence risk. It has the following beneficial effects:

[0047] (1) When using the visualization simulation and prediction system for mine ground subsidence risk, the system integrates heterogeneous data from multiple sources, including satellite remote sensing, aerial remote sensing, ground monitoring, and geological mining, through the data acquisition and management module. Combined with the preprocessing, spatiotemporal reference registration, and feature fusion technologies of the data processing and fusion module, it solves the shortcomings of traditional single technologies: it overcomes the problems of limited coverage and high cost of ground measurement, and makes up for the defects of aerial photography that are constrained by weather and lack flexibility. At the same time, it makes up for the disadvantages of low resolution and insufficient small-scale analysis of satellite remote sensing data by generating high-resolution three-dimensional deformation fields. Finally, a standardized multi-dimensional spatiotemporal dataset covering the entire mining area is formed, realizing full-scale monitoring of surface deformation and risk characteristics from macroscopic large-scale to microscopic fine-scale, and significantly improving the continuity and comprehensiveness of data.

[0048] (2) When using the visualization simulation and prediction system for mine ground subsidence risk, the system employs a hybrid risk prediction module combining machine learning and physical numerical simulation. The machine learning prediction unit, based on a spatiotemporal convolutional long short-term memory network, quickly achieves a preliminary global prediction of ground subsidence and risk across the entire mining area, solving the problems of poor adaptability of the probability integral method to complex geology and the reliance of artificial neural networks on massive amounts of data. The physical numerical simulation unit, targeting high-risk areas, accurately simulates the stress-strain and failure processes under complex geological conditions through three-dimensional geomechanical modeling and stress-seepage-damage coupling solution, thus overcoming the shortcomings of complex and inefficient pure numerical simulation calculations. The synergistic optimization and weighted fusion of the two methods ensures both the high efficiency of global prediction and improves the simulation accuracy of local high-risk areas, significantly enhancing the accuracy and reliability of ground subsidence risk prediction in complex mining areas.

[0049] (3) When using the visualization simulation and prediction system for mine ground subsidence risk, the system constructs a digital twin scenario of the mining area through a three-dimensional visualization and early warning interaction module, presenting the prediction results in a dynamic evolution form, thus solving the problem of abstract and difficult-to-interpret results of traditional models. At the same time, a four-level graded early warning mechanism is established, which automatically triggers early warnings based on the comprehensive risk index and key indicator thresholds, covering the entire scenario response from routine attention to emergency response. Compared with the traditional manual analysis mode that is lagging and the early warning is vague, this module supports the dynamic playback of the risk spatiotemporal evolution process and the real-time query of multi-dimensional information. Moreover, the early warning information is accurately associated with the risk location, evolution trend and response measures, improving the timeliness and pertinence of mine safety management. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0051] Figure 2 This is a detailed block diagram of the data acquisition and management module of the present invention;

[0052] Figure 3 This is a detailed block diagram of the hybrid risk prediction module of the present invention;

[0053] Figure 4 This is a detailed block diagram of the 3D visualization and early warning interaction module of the present invention.

[0054] In the diagram: 1. Data acquisition and management module; 101. Satellite remote sensing data acquisition unit; 102. Aerial remote sensing data acquisition unit; 103. Ground monitoring data acquisition unit; 104. Geological and mining data input unit; 2. Data processing and fusion module; 3. Hybrid risk prediction module; 4. 3D visualization and early warning interaction module; 401. Comprehensive scene rendering unit; 402. Risk spatiotemporal evolution playback unit; 403. Multidimensional information query unit. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 - Figure 4 This invention provides a visualization simulation and prediction system for mine ground subsidence risk, including a data acquisition and management module 1, a data processing and fusion module 2, a hybrid risk prediction module 3, and a three-dimensional visualization and early warning interaction module 4.

[0057] The main function of the data acquisition and management module 1 is to acquire multi-source heterogeneous data related to the risk of mine surface subsidence and to store and manage it uniformly. This includes a satellite remote sensing data acquisition unit 101, an airborne remote sensing data acquisition unit 102, a ground monitoring data acquisition unit 103, and a geological and mining data input unit 104. The satellite remote sensing data acquisition unit 101 acquires interferometric data, differential interferometric data, and multispectral image data through synthetic aperture radar satellite systems and high-resolution optical satellite systems. This data is used to monitor changes in the large-scale surface deformation field in the mining area. The airborne remote sensing data acquisition unit 102 uses lidar sensors and high-definition digital cameras mounted on UAVs or aircraft to generate high-precision digital elevation models, digital surface models, and orthophoto maps to support refined analysis of surface morphology changes. The ground monitoring data acquisition unit 103 uses continuously operating reference stations of the global navigation satellite system, automated total stations, high-precision levels, crack gauges, and inclinometers deployed at key locations in the mining area to acquire high-frequency, high-precision point displacement and deformation data in real time. The geological and mining data input unit 104 is responsible for importing geological exploration report data, rock mechanics parameters, hydrogeological conditions data, historical mining records, current mining face locations, and future mining plan data for the mining area, and for structuring this data for subsequent use. All of the above units are connected to the core server of the data acquisition and management module 1 via a network. The server adopts a distributed storage architecture to ensure efficient storage and rapid retrieval of massive amounts of data.

[0058] The data processing and fusion module 2 is directly connected to the data acquisition and management module 1. It primarily performs preprocessing, spatiotemporal reference registration, and feature fusion on the acquired multi-source data, ultimately generating a standardized multidimensional spatiotemporal dataset. Further, it first performs preprocessing operations such as format conversion, noise filtering, outlier removal, and data interpolation on satellite remote sensing data, aerial remote sensing data, and ground monitoring data to improve data quality. In the time dimension, it generates deformation time series with uniform time intervals using interpolation or extrapolation algorithms. In the spatial dimension, it uses general kriging interpolation or radial basis function interpolation methods to fuse discrete point displacement data and areal deformation data of different resolutions to generate a high-resolution three-dimensional deformation field covering the entire mining area. In addition, it extracts key static and dynamic characteristic parameters affecting surface collapse from structured geological and mining data. Static characteristic parameters include stratigraphic lithology, geological structural complexity index, fault activity, and rock mass quality grading; dynamic characteristic parameters include time-varying functions of mining depth, mining thickness, advance speed, and goaf volume. These feature parameters, together with the three-dimensional deformation field, constitute a standardized multidimensional spatiotemporal dataset, providing a foundation for subsequent predictions.

[0059] The hybrid risk prediction module 3 receives a standardized multidimensional spatiotemporal dataset from the data processing and fusion module 2, and generates a predicted ground subsidence field and a collapse risk level distribution map for a specific future time period in the mining area through the collaborative work of the machine learning prediction unit and the physical numerical simulation unit. Specifically, the machine learning prediction unit adopts a spatiotemporal convolutional long short-term memory network model. This model automatically extracts spatial correlation features from the three-dimensional deformation field and geological characteristic parameters through the convolutional layer structure, and effectively learns the long-term time dependence of surface deformation time series and mining activity dynamic parameters through the long short-term memory network layer structure. During model training, historical multi-source monitoring data and corresponding mining activity records are used as input. The backpropagation algorithm and adaptive moment estimation optimizer minimize the root mean square error loss function between the predicted deformation and the actual observed deformation, thereby obtaining the optimal model weight parameters. Based on the trained model, the machine learning prediction unit performs a preliminary global prediction of ground subsidence rate and a preliminary classification of collapse risk probability for the entire mining area, quickly identifying potential high-risk areas. Subsequently, the physical numerical simulation unit receives the boundary conditions and refined feature parameters of the high-risk area output by the machine learning prediction unit. It then uses a rock strata movement calculation engine to perform stress-strain coupling analysis and failure process simulation on this area, outputting detailed displacement fields, stress fields, and plastic zone distributions. The hybrid risk prediction module 3 feeds back the accurate simulation results from the physical numerical simulation unit as high-quality samples to the machine learning prediction unit for online model updates and continuous optimization. Simultaneously, it weights and fuses the accurate prediction results from the physical numerical simulation unit with the global prediction results from the machine learning prediction unit to generate the final ground subsidence prediction field and collapse risk level distribution map.

[0060] Furthermore, the physical numerical simulation unit includes: a three-dimensional geomechanical model builder, a dynamic simulator for the mining process, and a stress-seepage-damage coupled solver. The three-dimensional geomechanical model builder mainly uses the data provided by the geological and mining data input unit (104) to automatically or semi-automatically establish a three-dimensional solid geological model that can accurately reflect the stratigraphic structure, rock joints, fault distribution, and material properties of the mining area. The dynamic simulator for the mining process mainly simulates the gradual advancement of the mining face, the formation and expansion of the goaf, and the application process of the filling body on the three-dimensional solid geological model, and applies it to the calculation model as a boundary condition and load that changes with time. The stress-seepage-damage coupled solver mainly adopts a constitutive model that can characterize the elasticity, strain softening, and damage evolution of the rock mass, including but not limited to the Mohr-Coulomb criterion or the Hoek-Brown criterion, and couples the influence of the groundwater seepage field to solve the stress redistribution, deformation, failure, and resulting surface displacement of the rock mass under mining disturbance.

[0061] The 3D visualization and early warning interaction module 4 is directly connected to the hybrid risk prediction module 3. It primarily overlays the ground subsidence prediction field and collapse risk level distribution map onto a 3D integrated model that includes the surface topography, underground geological structure, and mining activity process of the mining area. This enables dynamic visualization and simulation of the spatiotemporal evolution of risks. When the predicted risk indicators exceed preset thresholds, it automatically generates and issues graded early warning information. The module includes a comprehensive scene rendering unit 401, a risk spatiotemporal evolution playback unit 402, and a multi-dimensional information query unit 403. Specifically:

[0062] The integrated scene rendering unit 401 integrates the rendering of the 3D geological model, the high-precision surface terrain model, and the mine tunnel and mining space model to construct a high-fidelity digital twin scene of the mining area. The risk spatiotemporal evolution playback unit 402 dynamically overlays a series of ground subsidence prediction fields and collapse risk level distribution maps at different time points, output by the hybrid risk prediction module 3, onto the digital twin scene using cloud maps, contour lines, vector arrows, or color blocks. Users can drag, rewind, and fast-forward the timeline, intuitively displaying the evolution and expansion trend of the risk area. The multi-dimensional information query unit 403 allows users to arbitrarily select spatial points, lines, or surfaces on the 3D visualization interface and instantly query the historical deformation curves, future deformation prediction curves, current risk level, main disaster-causing factors, and related monitoring data source information for that location. Furthermore, the method for generating the collapse risk level distribution map includes setting a risk assessment index system, which at least includes maximum settlement value, maximum settlement rate, maximum tilt, maximum curvature, and horizontal deformation. The values ​​of each index in the system are calculated from the ground settlement prediction field output by the hybrid risk prediction module 3, and multi-level risk thresholds are set for each index according to relevant national standards or specific safety requirements of the mining area. The analytic hierarchy process (AHP) or fuzzy comprehensive evaluation method is used to weight and calculate the various indicators, obtaining a comprehensive risk index for each spatial grid unit. The comprehensive risk index is then mapped to the final collapse risk level based on its range, thereby generating the collapse risk level distribution map.

[0063] The tiered early warning system includes four levels: blue, yellow, orange, and red. A blue warning is triggered when the predicted comprehensive risk index falls within the attention level range, and the system sends a routine notification to management personnel. A yellow warning is triggered when the predicted comprehensive risk index enters the danger level range or the predicted value of a key indicator exceeds its danger threshold for the first time. The system automatically generates a detailed report including the risk location, impact range, and predicted evolution trend, and pushes it to the mobile terminals of on-site management and technical personnel. An orange warning is triggered when the predicted comprehensive risk index is within the danger level range and continues to rise, or the predicted rate of change of a key indicator exceeds the secondary rate threshold. The system activates audible and visual alarms and issues evacuation warnings to personnel in and around the risk area via the emergency broadcast system. A red warning is triggered when the predicted comprehensive risk index enters the extremely dangerous level range or the system determines that collapse and instability are imminent. The system automatically sends the highest-level alarm information to the mine's top decision-making level and local emergency management departments, and activates the pre-set emergency response plan.

[0064] In summary, the data processing and fusion module 2 generates a high-resolution 3D deformation field through spatiotemporal reference registration and interpolation algorithms, ensuring the spatiotemporal consistency and continuity of the data and providing a reliable foundation for subsequent predictions. The hybrid risk prediction module 3, through the synergistic effect of the machine learning prediction unit and the physical numerical simulation unit, ensures both rapid global prediction capabilities and improves the simulation accuracy of local high-risk areas, enhancing the accuracy and reliability of the prediction results. The 3D visualization and early warning interaction module 4 constructs a high-fidelity mining area scenario using digital twin technology, presenting the prediction results in a dynamic and intuitive way, helping users fully grasp the risk evolution process. The tiered early warning mechanism automatically triggers different levels of response measures based on the risk level, improving the timeliness and effectiveness of mine safety management. The modules achieve efficient collaboration through clear data transmission and logical connections, breaking the isolation of traditional prediction systems, enhancing the overall performance and adaptability of the system, and providing strong technical support for mine surface subsidence risk prevention and control.

[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A visualization simulation and prediction system for mine surface subsidence risk, characterized in that, The system includes: The data acquisition and management module (1) mainly acquires monitoring data and basic data related to the risk of mine ground subsidence in real time or periodically through multiple heterogeneous data sources. The monitoring data and basic data are stored and managed in a unified manner. The data processing and fusion module (2) is connected to the data acquisition and management module (1). It mainly performs preprocessing, spatiotemporal reference registration and feature fusion on the monitoring data and basic data to generate a standardized multidimensional spatiotemporal dataset. The preprocessing includes format conversion, noise filtering, outlier removal and data interpolation of the original data. The spatiotemporal reference registration includes unifying all data from all sources to the same geographic coordinate system and time reference frame. The hybrid risk prediction module (3) is connected to the data processing and fusion module (2). It mainly receives the standardized multidimensional spatiotemporal dataset and generates a ground subsidence prediction field and collapse risk level distribution map of the mining area in a specific future time period through a collaborative prediction mechanism composed of machine learning prediction unit and physical numerical simulation unit. The three-dimensional visualization and early warning interaction module (4) is connected to the hybrid risk prediction module (3). It mainly overlays the ground subsidence prediction field and the collapse risk level distribution map onto a three-dimensional comprehensive model that includes the surface topography, underground geological structure and mining activity process of the mining area, so as to realize the dynamic visualization simulation of the risk spatiotemporal evolution process. When the predicted risk indicators exceed the preset threshold, it automatically generates and issues graded early warning information.

2. The visualization simulation and prediction system for mine surface subsidence risk according to claim 1, characterized in that: The data acquisition and management module (1) includes: The satellite remote sensing data acquisition unit (101) mainly acquires interferometric measurement data, differential interferometric measurement data and multispectral image data covering the entire mining area from synthetic aperture radar satellite system and high-resolution optical satellite system, for large-scale, long-term monitoring of surface deformation field; The aerial remote sensing data acquisition unit (102) mainly uses lidar sensors and high-definition digital cameras carried by UAVs or aircraft to acquire high-precision digital elevation models, digital surface models and orthophoto maps of key areas of concern in the mining area for refined analysis of surface morphology changes. The ground monitoring data acquisition unit (103) mainly receives and integrates high-frequency, high-precision point displacement and deformation data collected by the continuously operating reference station of the global navigation satellite system, automated total station, high-precision level, crack gauge and inclinometer deployed in key locations in the mining area. The geological and mining data input unit (104) mainly imports and structures geological exploration report data, rock mechanics parameters, hydrogeological conditions data, historical mining records, current mining face location and future mining plan data of the mining area, providing physical constraints and driving parameters for the prediction model.

3. The visualization simulation and prediction system for mine surface subsidence risk according to claim 2, characterized in that: In the data processing and fusion module (2), the satellite remote sensing data, aerial remote sensing data and ground monitoring data are synchronized in the time dimension. The deformation time series with a unified time interval is generated by interpolation or extrapolation algorithm. In the spatial dimension, the discrete point displacement data and the planar deformation data with different resolutions are fused to generate a high-resolution three-dimensional deformation field that covers the entire mining area, is continuous in time and space. In addition, the key static and dynamic characteristic parameters affecting surface collapse are extracted from the structured geological and mining data. The static characteristic parameters include stratum lithology, geological structure complexity index, fault activity and rock mass quality classification. The dynamic characteristic parameters include time-varying functions of mining depth, mining thickness, advance speed and goaf volume.

4. A visualization simulation and prediction system for mine surface subsidence risk according to claim 1 or 3, characterized in that: The specific implementation steps of the collaborative prediction mechanism in the hybrid risk prediction module (3) are as follows: Step 1: The machine learning prediction unit first receives the standardized multidimensional spatiotemporal dataset and uses a pre-trained deep learning network model to perform a rapid and global preliminary prediction of the ground subsidence rate and a preliminary classification of the collapse risk probability for the entire mining area, thereby quickly identifying potential high-risk areas. Step 2: The physical numerical simulation unit receives the boundary conditions and refined feature parameters of the high-risk area identified by the machine learning prediction unit, and performs high-precision stress-strain coupling analysis and failure process simulation on the high-risk area through the rock strata movement calculation engine, outputting detailed displacement field, stress field and plastic zone distribution; Step 3: The hybrid risk prediction module (3) further feeds back the accurate simulation results of the physical numerical simulation unit in Step 2 as high-quality samples to the machine learning prediction unit for online updating and continuous optimization of the model. At the same time, it weights and fuses the accurate prediction results of the physical numerical simulation unit with the global prediction results of the machine learning prediction unit to generate the final ground subsidence prediction field and collapse risk level distribution map that takes into account both global coverage and local accuracy.

5. The visualization simulation and prediction system for mine surface subsidence risk according to claim 4, characterized in that: The machine learning prediction unit employs a spatiotemporal convolutional long short-term memory network model, mainly including the following steps: Step 1: Through its convolutional layer structure, automatically extract the spatial correlation features of the fused three-dimensional deformation field and geological characteristic parameters to capture the spatial continuity and clustering patterns of surface deformation; Step 2: By using the long short-term memory network layer structure, the long-term time dependence of the surface deformation time series and mining activity dynamic parameters is effectively learned, and the nonlinear dynamic trend of the subsidence evolution process is captured. Step 3: The training process of the model uses historical multi-source monitoring data and corresponding mining activity records as input. Through the backpropagation algorithm and adaptive moment estimation optimizer, the root mean square error loss function between the predicted deformation and the actual observed deformation is minimized, thereby obtaining the optimal model weight parameters.

6. The visualization simulation and prediction system for mine surface subsidence risk according to claim 4, characterized in that: The physical numerical simulation unit includes: The three-dimensional geomechanical model builder mainly uses the data provided by the geological and mining data input unit (104) to automatically or semi-automatically build a three-dimensional solid geological model that can accurately reflect the stratigraphic structure, rock joints, fault distribution and material properties of the mining area; The dynamic simulator of the mining process mainly simulates the gradual advancement of the mining face, the formation and expansion of the goaf, and the application of the filling material on the three-dimensional physical geological model, and applies them as time-varying boundary conditions and loads to the calculation model. The stress-seepage-damage coupled solver mainly uses constitutive models that can characterize the elastoplasticity, strain softening and damage evolution of rock mass, including but not limited to the Mohr-Coulomb criterion or the Hoek-Brown criterion, and couples the influence of groundwater seepage field to solve the entire process of stress redistribution, deformation, failure and resulting surface displacement of rock mass under mining disturbance.

7. The visualization simulation and prediction system for mine surface subsidence risk according to claim 1, characterized in that: The three-dimensional visualization and early warning interaction module (4) includes: The integrated scene rendering unit (401) mainly integrates the three-dimensional geological model, the high-precision surface terrain model, and the mine tunnel and mining space model for integrated rendering to construct a high-fidelity digital twin scene of the mining area; The risk spatiotemporal evolution playback unit (402) mainly overlays a series of ground subsidence prediction fields and collapse risk level distribution maps at different time nodes output by the hybrid risk prediction module (3) onto the digital twin scene in the form of cloud maps, contour lines, vector arrows or color blocks. It supports users to drag, rewind, and fast forward the time axis, and intuitively displays the evolution process and expansion trend of the risk area. The multidimensional information query unit (403) is mainly designed to allow users to select any spatial point, line or surface on the three-dimensional visualization interface and instantly query the historical deformation curve, future deformation prediction curve, current risk level, main disaster-causing factors and related monitoring data source information of that location.

8. A visualization simulation and prediction system for mine surface subsidence risk according to claim 1 or 7, characterized in that: The collapse risk level distribution map is generated through a multi-factor comprehensive evaluation model, specifically including the following steps: Step 1: Establish a risk assessment index system, which should include at least the maximum settlement value, maximum settlement rate, maximum tilt, maximum curvature, and horizontal deformation; Step 2: Calculate the values ​​of each indicator in the indicator system from the ground subsidence prediction field output by the hybrid risk prediction module (3); Step 3: Based on relevant national regulations or specific safety requirements of the mining area, set corresponding multi-level risk thresholds for each indicator, such as four levels: "Safe", "Caution", "Danger", and "Extremely Dangerous". Step 4: Using the analytic hierarchy process (AHP) or fuzzy comprehensive evaluation method, weighted calculations are performed on each indicator to obtain the comprehensive risk index of each spatial grid unit. Based on the range of the comprehensive risk index, it is mapped to the final collapse risk level, thereby generating the collapse risk level distribution map.

9. The visualization simulation and prediction system for mine surface subsidence risk according to claim 8, characterized in that: The tiered early warning information includes: A blue alert is triggered when the predicted comprehensive risk index is in the "attention" level range, and the system sends a routine attention notification to the administrator's terminal. A yellow alert is triggered when the predicted comprehensive risk index enters the "dangerous" level range or when the predicted value of a key indicator exceeds its danger threshold for the first time. The system automatically generates a detailed report containing the risk location, scope of impact, and predicted evolution trend, and pushes it to the mobile terminal of the on-site management and technical personnel. An orange alert is triggered when the predicted comprehensive risk index is in the "dangerous" level range and continues to rise, or when the predicted rate of change of key indicators exceeds the level 2 rate threshold. The system will activate the audible and visual alarm device and issue evacuation warnings to people in and around the risk area through the emergency broadcast system. A red alert is triggered when the predicted comprehensive risk index enters the "extremely dangerous" level range, or when the system determines that a collapse or instability is about to occur. The system automatically sends the highest level of alarm information to the mine's top decision-making level and the local emergency management department, and activates the preset emergency response plan.

Images