Digital-twin-based mine dynamic risk early warning and virtual drilling method and system

Abstract

The application provides a dynamic risk early warning and virtual exercise method and system based on digital twin mine, and relates to the technical field of mine safety management. The method comprises the following steps: establishing a multi-disaster space-time coupling risk propagation graph; obtaining a dynamic risk collapse skeleton network according to an underground risk propagation induction matrix; performing multi-agent competition deduction according to a plurality of counterfactual virtual exercise branches to obtain a virtual competition exercise result space; obtaining a corrected underground twin according to the virtual competition exercise result space; and obtaining a dynamic risk early warning result according to the corrected underground twin. The technical problem of poor mine underground safety management quality caused by the fact that the existing mine underground risk early warning relies on single-point threshold judgment, it is difficult to represent the multi-disaster coupling propagation relationship, the virtual exercise is disconnected with the real-time underground state, and the like is solved. The technical effects of improving the underground risk propagation prediction accuracy, improving the virtual exercise scene authenticity and adaptability, and improving the mine underground safety management quality are achieved.

Description

Technical Field

[0001] This application relates to the field of mine safety management technology, and in particular to a method and system for dynamic risk early warning and virtual drills based on digital twin mines. Background Technology

[0002] The underground environment in mines is complex, with prominent risks from the coupling of multiple hazards such as gas, roof collapse, water damage, and fire. Traditional early warning systems rely on single-point threshold judgments, which are difficult to characterize the propagation relationships of multiple hazards, resulting in low accuracy and strong lag in early warning.

[0003] The existing mine fire safety early warning monitoring and management system mainly collects environmental information such as temperature, humidity, gas, and dust through zoned monitoring, sets early warning weights in combination with structural damage coefficients, and realizes dynamic adjustment of monitoring points, which has made some progress in zoned early warning.

[0004] However, this scheme still has shortcomings: First, it does not consider the spatiotemporal coupling and propagation evolution among multiple hazards, and cannot characterize the diffusion and superposition process of risks in the tunnel network. Second, the adjustment of monitoring points relies on static coefficient comparison, lacking the ability to dynamically reconstruct the risk propagation chain. Third, it cannot verify the effectiveness of different emergency intervention measures, leading to a disconnect between the emergency plan and the real-time underground situation. Therefore, how to achieve dynamic early warning of multi-hazard coupled risks and improve the adaptability of emergency drills is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] This application provides a method and system for dynamic risk early warning and virtual drills in mines based on digital twins. It addresses the technical problems in existing technologies where underground mine risk early warning relies on single-point threshold judgments, making it difficult to characterize the coupled propagation relationships of multiple hazards, and resulting in a disconnect between virtual drills and real-time underground conditions, leading to poor quality of underground mine safety management. The application achieves the technical effects of improving the accuracy of underground risk propagation prediction, realizing dynamic linkage early warning for multiple hazards, enhancing the realism and adaptability of virtual drill scenarios, and improving the quality of underground mine safety management.

[0006] This application provides a method for dynamic risk early warning and virtual drills in mines based on digital twins. The method includes: constructing an underground digital twin based on a multi-source monitoring dataset from the mine; analyzing the multi-dimensional disaster coupling propagation relationship based on the underground digital twin to establish a multi-hazard spatiotemporal coupled risk propagation map; constructing an underground risk propagation induction matrix; performing dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map based on the underground risk propagation induction matrix to obtain a dynamic risk collapse skeleton network; constructing multiple counterfactual virtual drill branches based on the dynamic risk collapse skeleton network; performing multi-subject competitive deduction based on the multiple counterfactual virtual drill branches to obtain a virtual competitive drill result space; performing credibility attenuation calculation and virtual-real linkage reverse correction on the underground digital twin based on the virtual competitive drill result space to obtain a corrected underground twin; and performing feedback reinforcement and updating on the multi-hazard spatiotemporal coupled risk propagation map based on the corrected underground twin to obtain dynamic risk early warning results.

[0007] Furthermore, based on the underground digital twin of the mine, multi-dimensional disaster coupling propagation relationship analysis is performed to establish a multi-hazard spatiotemporal coupled risk propagation map, including: performing spatiotemporal mapping of multi-dimensional disaster factors based on the underground digital twin of the mine to obtain a spatiotemporal distribution map of multi-dimensional disaster factors; performing semantic extraction of underground risk nodes based on the spatiotemporal distribution map of multi-dimensional disaster factors to obtain a set of underground risk nodes; and identifying cross-hazard coupling propagation relationships based on the set of underground risk nodes to obtain the multi-hazard spatiotemporal coupled risk propagation map.

[0008] Furthermore, constructing the underground risk propagation induction matrix includes: performing propagation inertia analysis based on the multi-hazard spatiotemporal coupled risk propagation map to construct an underground risk propagation inertial field; predicting the migration of risk hotspots based on the underground risk propagation inertial field to obtain the underground dynamic risk migration trajectory; configuring criticality based on the underground dynamic risk migration trajectory to obtain the risk migration critical attention trajectory; and executing dynamic parameter configuration for induction intervention based on the risk migration critical attention trajectory to obtain the underground risk propagation induction matrix.

[0009] Furthermore, the dynamic risk propagation chain dynamic collapse reconstruction of the multi-hazard spatiotemporal coupled risk propagation map is performed based on the downhole risk propagation induction matrix to obtain a dynamic risk collapse skeleton network. This includes: performing propagation blocking analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix to obtain risk propagation blocking analysis results; performing propagation intensity attenuation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix to obtain risk propagation intensity attenuation analysis results; performing propagation direction reorientation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix to obtain risk propagation direction reorientation analysis results; and performing dynamic collapse reconstruction of the multi-hazard spatiotemporal coupled risk propagation map based on the risk propagation blocking analysis results, the risk propagation intensity attenuation analysis results, and the risk propagation direction reorientation analysis results to generate the dynamic risk collapse skeleton network.

[0010] Furthermore, multiple counterfactual virtual exercise branches are constructed based on the dynamic risk collapse skeleton network, including: identifying high-risk propagation main pathways based on the dynamic risk collapse skeleton network to obtain multiple baseline accident evolution chains; performing multi-node counterfactual perturbations on each baseline accident evolution chain to generate multiple counterfactual perturbation condition sets; evaluating and filtering the exercise value of each counterfactual perturbation condition set to obtain multiple preferred counterfactual perturbation conditions; and mapping and modeling the multiple baseline accident evolution chains with the multiple preferred counterfactual perturbation conditions to obtain the multiple counterfactual virtual exercise branches.

[0011] Furthermore, the virtual competitive exercise result space is obtained by performing multi-subject competitive simulations based on the multiple counterfactual virtual exercise branches, including: performing personnel evacuation simulations based on each counterfactual virtual exercise branch and obtaining the simulation results for each personnel evacuation; performing disaster spread simulations based on each counterfactual virtual exercise branch and obtaining the simulation results for each disaster spread; performing ventilation adjustment simulations based on each counterfactual virtual exercise branch and obtaining the simulation results for each ventilation adjustment; performing equipment interlocking simulations based on each counterfactual virtual exercise branch and obtaining the simulation results for each equipment interlocking; and competitively associating the simulation results for each personnel evacuation, each disaster spread, each ventilation adjustment, and each equipment interlocking to obtain the virtual competitive exercise result space.

[0012] Furthermore, the process of calculating credibility attenuation and performing virtual-real linkage reverse correction on the underground digital twin based on the virtual competition exercise result space to obtain a corrected underground twin includes: comparing the virtual competition exercise result space with the underground digital twin to obtain a virtual-real deviation feature set; calculating multiple twin credibility attenuation factors based on the virtual-real deviation feature set; performing attenuation correction on the underground digital twin based on the multiple twin credibility attenuation factors to obtain a first optimized underground twin; performing reverse correction on the underground digital twin based on the virtual-real deviation feature set to obtain a second optimized underground twin; and fusing the first optimized underground twin and the second optimized underground twin to generate the corrected underground twin.

[0013] Furthermore, the multi-source monitoring dataset includes environmental monitoring data, equipment operation data, personnel monitoring data, ventilation operation data, and geological monitoring data.

[0014] Furthermore, obtaining dynamic risk warning results includes: configuring an underground risk suppression mechanism based on the dynamic risk warning results, and executing the underground risk pre-suppression in the mine based on the underground risk suppression mechanism.

[0015] Furthermore, calculating multiple twin credibility decay factors includes: generating a twin credibility decay warning signal based on the multiple twin credibility decay factors.

[0016] This application provides a dynamic risk early warning and virtual drill system for mines based on digital twins. The system includes: a disaster coupling propagation module, used to construct an underground digital twin based on multi-source monitoring datasets from the mine, and to analyze multi-dimensional disaster coupling propagation relationships based on the underground digital twin to establish a multi-hazard spatiotemporal coupled risk propagation map; a dynamic collapse reconstruction module, used to construct an underground risk propagation induction matrix, and to perform dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map based on the underground risk propagation induction matrix to obtain a dynamic risk collapse skeleton network; and a multi-agent... The competitive simulation module is used to construct multiple counterfactual virtual exercise branches based on the dynamic risk collapse skeleton network, and to perform multi-agent competitive simulation based on the multiple counterfactual virtual exercise branches to obtain the virtual competitive exercise result space; the virtual-real linkage reverse correction module is used to perform credibility decay calculation and virtual-real linkage reverse correction on the underground digital twin of the mine based on the virtual competitive exercise result space to obtain the corrected underground twin; the feedback reinforcement update module is used to perform feedback reinforcement update on the multi-hazard spatiotemporal coupled risk propagation map based on the corrected underground twin to obtain dynamic risk early warning results.

[0017] This application proposes a method for dynamic risk early warning and virtual drills in mines based on digital twins. It analyzes the multi-dimensional disaster coupling and propagation relationships using an underground digital twin, establishing a multi-hazard spatiotemporal coupled risk propagation map. The method then uses an underground risk propagation induction matrix to dynamically collapse and reconstruct the risk propagation chain of the multi-hazard spatiotemporal coupled risk propagation map, obtaining a dynamic risk collapse skeleton network. Based on this dynamic risk collapse skeleton network, multiple counterfactual virtual drill branches are constructed, and multi-agent competitive simulations are performed using these branches to obtain a virtual competitive drill result space. Based on this virtual competitive drill result space, the credibility attenuation of the underground digital twin is calculated, and reverse correction based on virtual-real linkage is performed to obtain a corrected underground twin. Finally, the multi-hazard spatiotemporal coupled risk propagation map is updated and reinforced using the corrected underground twin, resulting in dynamic risk early warning results. This method solves the technical problems of existing technologies where underground mine risk early warning relies on single-point threshold judgments, making it difficult to characterize multi-hazard coupling and propagation relationships, and causing a disconnect between virtual drills and real-time underground conditions, leading to poor quality underground mine safety management.

[0018] By constructing a multi-hazard spatiotemporal coupled risk propagation map and using the downhole risk propagation induction matrix for dynamic collapse reconstruction, the induction, transmission and superposition relationships among various hazards such as gas, roof collapse, water damage, and fire can be accurately depicted, thereby significantly improving the prediction accuracy of downhole risk propagation paths and evolution trends.

[0019] Based on the feedback enhancement and update of the spatiotemporal coupled risk propagation map of multiple hazards using the corrected downhole twin, the risk early warning has been transformed from static threshold judgment to dynamic coupled analysis. It can output dynamic early warning results in real time, including risk location, risk level and evolution trend, effectively reducing false alarm and missed alarm rates.

[0020] Multiple counterfactual virtual exercise branches are constructed based on the dynamic risk collapse skeleton network, and multi-agent competitive simulations (personnel evacuation, disaster spread, ventilation adjustment, equipment interlocking) are executed, enabling the exercise scenarios to cover extreme disasters and multi-hazard coupling situations, significantly enhancing the matching degree between the exercise and the real-time underground status.

[0021] Multiple twin reliability decay factors are calculated by using a set of virtual and real deviation features. Then, decay correction and root cause reverse correction are performed separately. Finally, a high-precision and high-stability corrected downhole twin is generated by fusion, which solves problems such as model distortion, time lag, and behavioral bias.

[0022] Based on the dynamic risk warning results, an underground risk suppression mechanism is configured to automatically implement intervention measures such as ventilation adjustment, equipment power outage, and personnel evacuation at the nascent stage of a disaster, transforming safety management from "passive response" to "proactive prevention and control," effectively reducing the accident rate and losses. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the apparatus according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0024] Figure 1 A flowchart illustrating the method for dynamic risk early warning and virtual drills of digital twin mines provided in this application.

[0025] Figure 2 A schematic diagram illustrating the process of establishing a multi-hazard spatiotemporal coupled risk propagation map in the method of dynamic risk early warning and virtual drills for digital twin mines provided in this application.

[0026] Figure 3 A schematic diagram illustrating the generation process of the dynamic risk collapse skeleton network in the digital twin-based mine dynamic risk early warning and virtual drill method provided in this application.

[0027] Figure 4 A schematic diagram illustrating the process of obtaining the virtual competitive exercise result space in the digital twin-based dynamic risk early warning and virtual exercise method for mines provided in this application.

[0028] Figure 5 This is a flowchart illustrating the credibility decay calculation and virtual-real linkage reverse correction process in the digital twin-based dynamic risk early warning and virtual drill method for mines provided in this application.

[0029] Figure 6 A schematic diagram of the structure of the digital twin-based dynamic risk early warning and virtual drill system for mines provided in this application.

[0030] Figure labeling: 1. Disaster coupling propagation module; 2. Dynamic collapse and reconstruction module; 3. Multi-subject competitive simulation module; 4. Virtual-real linkage reverse correction module; 5. Feedback reinforcement and update module. Detailed Implementation

[0031] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or apparatuses. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for the purpose of describing this application only.

[0034] Example 1: This application provides a method for dynamic risk early warning and virtual drills based on digital twin mines, such as... Figure 1 As shown, the method includes: Step A100: Based on the multi-source monitoring dataset from underground mines, construct a digital twin of the underground mine, and analyze the multi-dimensional disaster coupling and propagation relationships based on the digital twin, establishing a multi-hazard spatiotemporal coupled risk propagation map. The multi-source monitoring dataset includes environmental monitoring data, equipment operation data, personnel monitoring data, ventilation operation data, and geological monitoring data. Environmental monitoring data includes environmental parameters such as methane concentration, carbon monoxide parameters, dust parameters, temperature and humidity parameters, and oxygen parameters in the mine. Environmental monitoring data is collected in real time using environmental sensors distributed throughout the mine. Equipment operation data includes the current, temperature, speed, start / stop, and fault status data of all equipment such as underground coal mining machines, conveyors, and ventilation fans. Equipment operation data is collected through the equipment's built-in sensing terminals. Personnel monitoring data includes real-time location, walking trajectory, on-duty status, and off-duty alarm information of underground workers. Personnel monitoring data is collected using positioning cards and intelligent monitoring systems. Ventilation operation data includes core ventilation parameters such as wind speed, air volume, air pressure, damper opening / closing, and ventilation power of the underground ventilation system. Geological monitoring data includes geological parameters such as underground roadway deformation, roof pressure, rock strata structure, in-situ stress, and hydrology. Specifically, a digital twin platform is used to dynamically model multi-source monitoring datasets, obtaining a digital twin of the mine's underground environment. This enables a realistic replication of all elements of the underground working conditions, including the environment, equipment, personnel, ventilation, and geology. The digital twin platform is an intelligent platform with 3D modeling, virtual-real driving, and simulation capabilities, among existing technologies. The underground digital twin is a virtual mirror model corresponding to the multi-source monitoring dataset.

[0035] like Figure 2As shown, step A100, "analyzing the multi-dimensional disaster coupling propagation relationship based on the underground digital twin of the mine, and establishing a multi-hazard spatiotemporal coupling risk propagation map," further includes steps A110 to A130. Step A110: Performing spatiotemporal mapping of multi-dimensional disaster factors based on the underground digital twin of the mine to obtain a spatiotemporal distribution map of multi-dimensional disaster factors. Multi-dimensional disaster factors refer to core disaster-causing indicators that can induce underground mine disasters. Specific multi-dimensional disaster factors include excessive gas concentration, carbon monoxide accumulation concentration, excessive dust concentration, surrounding rock deformation, roof pressure value, ground stress value, abnormal high temperature value of equipment, and roadway water accumulation hydrology, etc. Spatiotemporal mapping refers to matching, projecting, and filling discrete, fragmented, and asynchronous multi-dimensional disaster factor monitoring data onto the spatial location and time node corresponding to the underground digital twin of the mine, using the three-dimensional spatial coordinate system corresponding to the underground digital twin as the spatial reference and the acquisition time sequence of the multi-source monitoring dataset as the time reference. Multidimensional disaster factor monitoring data includes multiple monitoring parameters corresponding to the multidimensional disaster factors. Specifically, multidimensional disaster factor monitoring data is first extracted from the multi-source monitoring dataset. After time-series sorting, spatial coordinate matching, and removal of invalid and redundant data, the monitoring data of multidimensional disaster factors with different acquisition frequencies and spatial locations are made continuous, gridded, and spatiotemporally aligned. This ensures that each downhole three-dimensional spatial grid unit corresponds to the real-time value and time information of the matching disaster factor. Then, through color grading and spatial layer overlay visualization mapping, the abstract multidimensional disaster factor monitoring data is transformed into an intuitive and fully covered spatial distribution graphic. Finally, a multidimensional disaster factor spatiotemporal distribution map that can completely represent the intensity distribution characteristics of each disaster factor at different spatial locations and times downhole is obtained.

[0036] Step A120: Based on the spatiotemporal distribution map of the multidimensional disaster factors, perform semantic extraction of underground risk nodes to obtain a set of underground risk nodes. Underground risk nodes refer to various key locations within the three-dimensional underground space where disaster values ​​exceed safety standards, meet the conditions for disaster occurrence, and pose real safety hazards. Semantic extraction refers to using a mine disaster semantic rule base to perform feature identification, risk classification, location positioning, and attribute labeling on the spatiotemporal distribution map of the multidimensional disaster factors. Specifically, the mine disaster semantic rule base includes safety thresholds, disaster feature text labels, and risk level classification criteria for various disasters (such as gas disasters, roof fall disasters, water hazards, dust disasters, ventilation disasters, etc.). Subsequently, each spatial grid unit within the spatiotemporal distribution map of the multidimensional disaster factors is traversed and analyzed one by one. The analysis is compared with the mine disaster semantic rule base to verify the magnitude of disaster values, the rate of change of values, the degree of danger aggregation, and the abnormal fluctuation trend of data for each grid. High-risk grid areas with values ​​exceeding standards, sudden abnormal changes, and high concentrations of danger are selected. Next, each selected high-risk grid area is assigned a unique semantic tag, clearly indicating its corresponding disaster type, risk severity level, cause of risk, specific three-dimensional spatial location, and future risk evolution trend. Simultaneously, all invalid blank grids with normal values ​​and no safety hazards are automatically removed. Scattered small risk areas that are spatially adjacent, have the same disaster type, and consistent risk level are merged and standardized to obtain a set of downhole risk nodes. This set of downhole risk nodes includes multiple downhole risk nodes. Each downhole risk node corresponds to a unique semantic tag.

[0037] Step A130: Identify cross-hazard coupling propagation relationships based on the underground risk node set to obtain the multi-hazard spatiotemporal coupling risk propagation map. Cross-hazard coupling propagation relationship identification refers to breaking away from the traditional limitation of analyzing only one type of hazard individually. It identifies the interactions between different types of hazards, such as underground gas, roof collapse, water damage, dust, and abnormal ventilation, where "an accident in one hazard can trigger another, and multiple accidents are more dangerous than a single accident." Specifically, grey relational analysis is performed on the underground risk node set to obtain the node correlation strength results. The node correlation strength results include the correlation strength between each pair of different hazards. For example, the correlation strength between "insufficient ventilation" and "excessive gas concentration" is 0.9 (extremely strong correlation), and the correlation strength between "excessive roof pressure" and "water accumulation in the tunnel" is 0.3 (weak correlation). Based on the node association strength results, the set of underground risk nodes is divided into three categories: single independent risk nodes with only one type of hazard, dual-hazard coupled risk nodes where two hazards influence each other, and multi-hazard amplified risk nodes where three or more hazards overlap. Furthermore, combining the actual underground roadway connectivity, ventilation airflow direction, underground rock pressure transmission patterns, and the natural diffusion characteristics of different hazards such as gas, water, and dust, a digital twin platform is used to simulate and deduce the spread of single independent risk nodes, dual-hazard coupled risk nodes, and multi-hazard amplified risk nodes from their emergence over time, along which roadways they will spread, at what speed, which areas they will affect, and when they will encounter and overlap with other risks, thus obtaining cross-hazard coupling relationships. Cross-hazard coupling relationships include the inducing relationship, transmission relationship, and superposition relationship between multiple underground risk nodes. Finally, multiple underground risk nodes are used as core nodes of the graph, and cross-hazard coupling relationships are used as connecting node edges to obtain a multi-hazard spatiotemporal coupled risk propagation graph. By analyzing the multi-dimensional hazard coupling propagation relationships of the underground mine digital twin, the accuracy of underground risk propagation prediction is improved.

[0038] Step A200: Construct an underground risk propagation induction matrix, and perform dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map based on the underground risk propagation induction matrix to obtain a dynamic risk collapse skeleton network. Step A200 further includes: Step A210: Perform propagation inertia analysis based on the multi-hazard spatiotemporal coupled risk propagation map to construct an underground risk propagation inertia field. Propagation inertia analysis refers to deducing the subsequent propagation direction, priority coverage area, development rate changes, and duration of continuous development of the multi-hazard spatiotemporal coupled risk propagation map under conditions without external intervention. The underground risk propagation inertia field is a three-dimensional spatial distribution model characterizing the continuity trend of underground risk propagation, which can intuitively present the possibility, direction, and development trend of risk propagation in different areas. Specifically, firstly, from the multi-hazard spatiotemporal coupled risk propagation map, the propagation behavior characteristics of different hazards under the influence of different roadway types, ventilation conditions, geological environment, and obstacles (air doors, waterproof gates, support structures) are extracted to form a risk propagation behavior characteristic database. Secondly, based on the risk propagation behavior feature database of the digital twin platform, the continuity trend of risk propagation is obtained. The continuity trend of risk propagation includes the subsequent propagation direction parameters, priority coverage area parameters, development rate change parameters, and duration parameters of each underground risk node. Furthermore, based on the continuity trend of risk propagation, the underground digital twin of the mine is divided into three spatial units: inevitable propagation zone, potential propagation zone, and non-propagation zone. Each spatial unit is labeled with the corresponding propagation direction, propagation rate level (fast / medium / slow), and duration interval (short-term / medium-term / long-term), thus obtaining the inertial field of risk propagation in the mine.

[0039] Step A220: Based on the underground risk propagation inertial field, predict the migration of risk hotspots and obtain the dynamic risk migration trajectory underground. Risk hotspot migration prediction refers to using the underground risk propagation inertial field to deduce the spatial position change of risk hotspots over time. The dynamic risk migration trajectory underground refers to a continuous path formed by connecting the center positions of the same risk hotspot at different future time points, which can intuitively show the direction, speed, and range of influence of the risk hotspot. Specifically, firstly, in the underground risk propagation inertial field, all continuous spatial units belonging to the inevitable propagation zone and with a fast propagation rate are selected, and adjacent homogeneous units are merged to determine the boundaries and center coordinates of all current risk hotspots. Secondly, with a fixed time step of 10 minutes, the digital twin platform is called, and the propagation direction and rate level marked in each area of ​​the underground risk propagation inertial field are used to simulate and deduce the spatial movement process of each risk hotspot in the future. Finally, the center coordinates, boundary range, and risk intensity level of each risk hotspot are recorded at each time step. Finally, the center coordinates of the same risk hot zone at all time steps are connected by continuous lines, and the risk intensity and impact boundary at the corresponding time points are marked on the path. This generates a dynamic downhole risk migration trajectory that can dynamically display the spatiotemporal evolution of the risk hot zone.

[0040] Step A230: Configure criticality based on the underground dynamic risk migration trajectory to obtain the risk migration critical attention trajectory. Criticality configuration refers to scoring the importance of each underground roadway segment covered by the underground dynamic risk migration trajectory, quantitatively assessing the comprehensive loss caused by the risk attack on that roadway segment. The higher the score, the higher the prevention and control priority of that roadway segment. The risk migration critical attention trajectory refers to the set of core road segments with criticality scores higher than the criticality threshold selected from the underground dynamic risk migration trajectory, which are the priority intervention targets for risk prevention and control. Specifically, firstly, the underground dynamic risk migration trajectory is divided into several independent trajectory segments according to the natural segmentation of the actual underground roadway, with each trajectory segment corresponding to a physically independent underground roadway. Secondly, a four-dimensional criticality evaluation system is established. The four-dimensional criticality evaluation system consists of two dimensions: First, the risk intensity dimension, which assigns a value according to the risk level of the area covered by the trajectory segment, with higher levels resulting in higher scores. Second, the personnel exposure dimension, which assigns a value according to the average number of workers within the trajectory segment and a 50-meter radius, with more workers resulting in higher scores. Third, the equipment value dimension is used, assigning values ​​based on the importance of the electromechanical equipment deployed around the trajectory segment, with core production equipment scoring higher than auxiliary equipment. Fourth, the topological criticality dimension is used, assigning values ​​based on the node status of the trajectory segment in the underground roadway network, with multi-path intersections and unique passageways scoring the highest. Then, based on the weights of each dimension (risk intensity weight 0.4, personnel exposure weight 0.3, equipment value weight 0.2, topological criticality weight 0.1), a criticality score is calculated for each trajectory segment. Finally, trajectory segments with criticality scores higher than the criticality threshold are selected as critical attention trajectories for risk migration. The formula for calculating the criticality score is: ; in, Indicates criticality score, As a risk intensity weight, To score the risk intensity, For personnel exposure weight, To score personnel exposure, As a weighting factor for equipment value, Score the value of the equipment. For topological keyness weights, Topological criticality score. The risk intensity score characterizes the severity of disaster risk within the area covered by the trajectory segment. The risk intensity score is obtained by extracting the ratio of real-time monitored values ​​of multi-dimensional disaster factors (including gas concentration, roof pressure, dust concentration, etc.) within the trajectory segment to preset safety thresholds, and then selecting the maximum ratio. If 1 ≤ maximum ratio < 1.5, the risk intensity score is 1 point. If 1.5 ≤ maximum ratio < 2.5, the risk intensity score is 2 points. If 2.5 ≤ maximum ratio < 3.5, the risk intensity score is 3 points. If the maximum ratio ≥ 3.5, the risk intensity score is 4 points. The personnel exposure score characterizes the population density around the trajectory segment, reflecting the potential risk of casualties during a disaster. The personnel exposure score is obtained by counting the total number of real-time underground workers within a 50-meter radius outward from the geometric center of the trajectory segment. If 0 ≤ n < total number of real-time underground workers < 2, the personnel exposure score is 1 point. If 2 ≤ total number of real-time underground workers < 5, the personnel exposure score is 2 points. If 5 ≤ n < 10, the personnel exposure score is 3 points. If the total number of personnel working underground in real time is ≥ 10, the personnel exposure score is 4 points. The equipment value score characterizes the importance and asset value of electromechanical equipment surrounding the trajectory segment, reflecting equipment losses and production interruptions caused by the disaster. The equipment value score is obtained by extracting a list of electromechanical equipment within a 50-meter radius outward from the geometric center of the trajectory segment and assigning values ​​according to equipment function. That is, auxiliary equipment (including lighting and communication equipment) receives a value score of 1 point, general production equipment (including conveyors and water pumps) receives a value score of 2 points, core production equipment (including coal mining machines and main ventilation fans) receives a value score of 3 points, and key safety interlocking equipment (including gas cut-off devices and safety monitoring substations) receives a value score of 4 points. If multiple devices exist within this range, the highest score is taken as the equipment value score for the trajectory segment. The topological criticality score characterizes the node status of the trajectory segment in the underground roadway network, reflecting the degree of impact on the overall roadway connectivity after the roadway segment is attacked by risks. The topological criticality score is determined as follows: If the trajectory segment is a regular branch roadway, the topological criticality score is 1 point. If the trajectory segment is a critical intersection roadway, the topological criticality score is 2 points. If the trajectory segment is a unique passageway or a global hub node, the topological criticality score is 3 points.

[0041] Step A240: Configure dynamic parameters for induced intervention based on the risk migration critical attention trajectory to obtain the downhole risk propagation induction matrix. The dynamic parameter configuration for induced intervention refers to matching corresponding intervention parameters to the risk migration critical attention trajectory. The downhole risk propagation induction matrix includes multiple intervention parameters corresponding to the risk migration critical attention trajectory. Specifically, for different disaster types such as gas, roof, water, dust, and ventilation, corresponding standardized intervention measures (e.g., ventilation volume adjustment, gas extraction, and damper opening / closing for gas disasters; anchor bolt reinforcement and personnel evacuation for roof disasters) are pre-entered to obtain a downhole disaster intervention measure library. The risk migration critical attention trajectory is input into the downhole disaster intervention measure library, and corresponding intervention measures are matched according to the risk type of the risk migration critical attention trajectory. For example, for a gas diffusion trajectory, measures such as ventilation volume adjustment, damper opening / closing, and gas extraction are matched. For a roof risk propagation trajectory, measures such as anchor bolt reinforcement, roof monitoring, and personnel evacuation are matched. For a water hazard propagation trajectory, measures such as turning on the drainage pump and closing the waterproof gate are matched. Then, based on the risk intensity and expected arrival time of the critical attention trajectory of risk migration, the parameters of each intervention measure are dynamically configured. For example, the higher the risk intensity and the closer the expected arrival time, the larger the adjustment range of ventilation volume, the higher the power of gas extraction, and the earlier the personnel evacuation time. The critical attention trajectory of risk migration is matched with the parameters of multiple corresponding intervention measures to obtain the downhole risk propagation induction matrix.

[0042] After obtaining the downhole risk propagation induction matrix, step A200 further includes: Step A250: Performing propagation blocking analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtaining the risk propagation blocking analysis results. Propagation blocking analysis refers to the process of analyzing the range of risk propagation paths being cut off and risk nodes being isolated in the multi-hazard spatiotemporal coupled risk propagation map after simulating the execution of all propagation blocking intervention measures in the downhole risk propagation induction matrix. The risk propagation blocking analysis results include all successfully blocked propagation paths, isolated risk nodes, and unblocked residual paths. Specifically, firstly, all intervention measures belonging to the propagation blocking category are screened from the downhole risk propagation induction matrix, including closing air doors, lowering waterproof gates, temporary roadway sealing, and equipment power-off isolation. At the same time, the implementation location coordinates, execution time, and radius of influence of each measure are extracted. Secondly, in the multi-hazard spatiotemporal coupled risk propagation map, the topological nodes and connecting edges corresponding to each intervention measure are located, and the execution time of the intervention measures is synchronized with the time axis built into the map. The propagation edges corresponding to the locations where intervention measures are applied are marked as "blocked" and removed from the multi-hazard spatiotemporal coupled risk propagation map. Simultaneously, the connection edge states of all risk nodes are traversed; if all input and output propagation edges of a risk node are removed, the node is marked as "isolated." Finally, the number and location of all blocked propagation paths, the number and type of isolated risk nodes, and the number and direction of unblocked residual propagation paths are statistically analyzed to obtain the risk propagation blocking analysis results.

[0043] Step A260: Perform a propagation intensity attenuation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtain the results. The propagation intensity attenuation analysis simulates the execution of intensity control operations such as increasing ventilation volume, starting the gas extraction pump, and initiating dust suppression spraying according to the downhole risk propagation induction matrix, and analyzes the risk intensity reduction parameters of each risk node in the multi-hazard spatiotemporal coupled risk propagation map. The results of the risk propagation intensity attenuation analysis include the intensity attenuation ratio of all risk nodes, the distribution of remaining risk intensity, and the overall risk attenuation trend. Specifically, firstly, all intervention measures belonging to the propagation intensity control category are screened from the downhole risk propagation induction matrix, including ventilation volume adjustment, gas extraction, dust suppression spraying, roof pressure relief, and drainage pump acceleration. Simultaneously, the parameter settings, activation time, and impact coverage of each measure are extracted. Secondly, in the multi-hazard spatiotemporal coupled risk propagation map, the original risk intensity value is bound to each risk node. Based on the simulation of the risk intensity decay process at each risk node after the implementation of intervention measures using a digital twin platform, risk change data for each risk node is obtained. The intensity decay ratio includes the difference between the final risk intensity value within the risk change data of each risk node and the original risk intensity value of each risk node. The remaining risk intensity distribution includes the final risk intensity value within the risk change data of each risk node. The overall risk decay trend includes the risk intensity change rate and the direction of risk intensity change corresponding to the risk change data of each risk node.

[0044] Step A270: Perform a propagation direction reorientation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtain the risk propagation direction reorientation analysis results. The propagation direction reorientation analysis simulates the changes in the risk propagation direction of the multi-hazard spatiotemporal coupled risk propagation map after performing directional control operations such as adjusting damper opening, changing the direction of local ventilation fans, and arranging stress relief boreholes according to the downhole risk propagation induction matrix. The risk propagation direction reorientation analysis results include the reoriented propagation direction, newly added propagation paths, abandoned propagation paths, and affected risk nodes. Specifically, firstly, all intervention measures belonging to the propagation direction control category are screened from the downhole risk propagation induction matrix, including damper opening adjustment, local ventilation fan direction change, ventilation network reconstruction, and stress relief borehole arrangement, etc., and the parameter settings, implementation time points, and affected area coordinates of each measure are extracted. Secondly, the environmental parameters of the corresponding affected areas are updated in the multi-hazard spatiotemporal coupled risk propagation map. For example, adjusting the damper opening updates the ventilation airflow direction and velocity in the area, and arranging stress relief boreholes updates the surrounding rock stress distribution and transmission direction. Based on the digital twin platform's simulation of the propagation potential energy distribution of gaseous hazards such as methane and dust, and the digital twin platform's simulation of the stress potential energy distribution of geological hazards such as roof stress and ground pressure, the propagation direction of each risk node is redefined based on the potential energy gradient direction. Old paths that have lost propagation potential energy due to changes in environmental parameters are marked as "abandoned," and new paths formed due to changes in potential energy gradient are marked as "new propagation paths." The results of the risk propagation direction re-determination analysis are obtained by statistically analyzing all redefined propagation directions, the number and location of new propagation paths, the number and location of abandoned propagation paths, and the risk nodes affected by the direction change.

[0045] Step A280: Based on the risk propagation blocking analysis results, the risk propagation intensity attenuation analysis results, and the risk propagation direction reorientation analysis results, the multi-hazard spatiotemporal coupled risk propagation map is dynamically collapsed and reconstructed to generate the dynamic risk collapse skeleton network. The generation process of the dynamic risk collapse skeleton network is as follows: Figure 3 As shown, dynamic collapse and reconstruction refers to the trimming and updating of the spatiotemporal coupled risk propagation map of multiple hazards based on the results of risk propagation blocking analysis, risk propagation intensity attenuation analysis, and risk propagation direction reorientation analysis.

[0046] Specifically, based on the risk propagation blocking analysis results, all propagation edges marked "blocked" and risk nodes marked "isolated" are removed from the multi-hazard spatiotemporal coupled risk propagation map, while unblocked residual propagation paths and unisolated risk nodes are retained. The remaining risk intensity values ​​of all retained nodes are updated based on the risk propagation intensity attenuation analysis results, and low-risk nodes with remaining risk intensity values ​​below the safety threshold are deleted. Simultaneously, the propagation rate parameters of each retained propagation edge are updated according to the intensity attenuation ratio. Next, based on the risk propagation direction reorientation analysis results, all propagation edges marked "abandoned" are removed from the multi-hazard spatiotemporal coupled risk propagation map, and all edges marked "new propagation paths" are added with their direction attributes set. The direction attributes of all retained propagation edges are adjusted to match the reoriented propagation directions. Finally, spatially adjacent nodes with identical risk types and intensity levels are merged, redundant parallel propagation edges are deleted, and core high-risk nodes and critical propagation paths are retained. Simultaneously, corresponding intervention impact information and remaining risk assessment information are bound to each node and edge, resulting in a dynamic risk collapse skeleton network. By using the downhole risk propagation induction matrix to perform dynamic collapse reconstruction of the risk propagation chain of the spatiotemporal coupled risk propagation map of multiple hazards, a comprehensive and reliable characterization of the coupled propagation relationship of multiple hazards is achieved, further improving the accuracy of downhole risk propagation prediction.

[0047] Step A300: Construct multiple counterfactual virtual exercise branches based on the dynamic risk collapse skeleton network, and perform multi-agent competitive simulation based on the multiple counterfactual virtual exercise branches to obtain the virtual competitive exercise result space. Specifically, when constructing multiple counterfactual virtual exercise branches, step A300 includes: Step A310: Identify high-risk propagation main paths based on the dynamic risk collapse skeleton network to obtain multiple baseline accident evolution chains. High-risk propagation main path identification involves selecting the propagation path with the highest risk intensity, longest propagation link, widest impact range, and the potential to trigger a major accident if out of control from the dynamic risk collapse skeleton network. Each baseline accident evolution chain includes a complete causal evolution chain starting from the initial risk trigger node, sequentially passing through intermediate risk nodes along the high-risk propagation main path, and finally reaching the accident consequence node. Each baseline accident evolution chain corresponds to a typical downhole accident mode. Specifically, extract the remaining risk intensity value and the coupling strength value of the propagation edges from the dynamic risk collapse skeleton network. The product of the remaining risk intensity value and the coupling strength value is used as the path weight. Within the dynamic risk collapse skeleton network, multiple high-risk propagation main paths with path weights exceeding the path weight threshold are extracted, while short paths with a length of less than 3 nodes and duplicate paths are excluded. Finally, the causal relationships of each high-risk propagation main path are analyzed to clarify the triggering conditions, propagation logic, and accident consequences of each node. These are then structured and encoded in the order of "initial risk node → intermediate risk node → accident consequence node" to obtain multiple baseline accident evolution chains.

[0048] Step A320: Perform multi-node counterfactual perturbation on each baseline accident evolution chain to generate multiple counterfactual perturbation condition sets. Multi-node counterfactual perturbation refers to changing the intervention parameters of one or more nodes in each baseline accident evolution chain. Each counterfactual perturbation condition set includes multiple counterfactual perturbation conditions corresponding to each baseline accident evolution chain. Each counterfactual perturbation condition includes multiple intervention adjustment parameters corresponding to each baseline accident evolution chain. Specifically, each baseline accident evolution chain is classified into four categories: initial trigger nodes, intermediate propagation nodes, intervention nodes, and accident consequence nodes. For each intervention node and intermediate propagation node, the perturbable parameter types and value ranges are defined, including the implementation time, implementation intensity, equipment operating parameters, and environmental parameters of the intervention. For example, for the ventilation intervention node, the perturbable parameters are the ventilation volume adjustment range (0%-50%) and the implementation time (0-60 minutes in advance). Random values ​​are selected and combined according to the perturbable parameter types and value ranges to obtain multiple counterfactual perturbation condition sets.

[0049] Step A330: Perform a drill value evaluation and selection for each counterfactual disturbance condition set to obtain multiple preferred counterfactual disturbance conditions. First, establish a drill value evaluation index system: One is the severity of the accident consequences, assigned values ​​based on the potential casualties, property damage, and production interruption duration after the disturbance. The more casualties, the greater the property damage, and the longer the production interruption, the higher the accident consequence severity score. The second is the difficulty of intervention, assigned values ​​based on the complexity of emergency response, the amount of resources required, and the urgency of time after the disturbance. The higher the complexity of emergency response, the greater the amount of resources required, and the stronger the urgency of time, the higher the intervention difficulty score. Based on the drill value evaluation index system and the drill value evaluation index weights (accident consequence severity weight 60%, intervention difficulty weight 40%), calculate the comprehensive drill value score for each counterfactual disturbance condition. The formula for calculating the comprehensive drill value score is: Comprehensive Drill Value Score = Accident Consequence Severity Score × Accident Consequence Severity Weight + Intervention Difficulty Score × Intervention Difficulty Weight. Within each set of counterfactual perturbation conditions, the counterfactual perturbation condition corresponding to the maximum comprehensive exercise value score is selected as the preferred counterfactual perturbation condition.

[0050] Step A340: Map the multiple baseline accident evolution chains to the multiple preferred counterfactual perturbation conditions to obtain the multiple counterfactual virtual exercise branches. Match the multiple baseline accident evolution chains to the multiple preferred counterfactual perturbation conditions to obtain the mapping relationship of multiple perturbation evolution chains. Based on the mapping relationship of multiple perturbation evolution chains, use a digital twin platform to model the multiple baseline accident evolution chains and multiple preferred counterfactual perturbation conditions to obtain multiple counterfactual virtual exercise branches. The mapping relationship of multiple perturbation evolution chains includes the correspondence between the multiple baseline accident evolution chains and the multiple preferred counterfactual perturbation conditions. During the modeling process, first import the multiple baseline accident evolution chains into the digital twin platform, and then inject each preferred counterfactual perturbation condition into the corresponding baseline accident evolution chain one by one according to the mapping relationship of multiple perturbation evolution chains to obtain multiple counterfactual virtual exercise branches. Each counterfactual virtual exercise branch is a digital twin model corresponding to the baseline accident evolution chain and the preferred counterfactual perturbation condition. By constructing multiple counterfactual virtual exercise branches through a dynamic risk collapse skeleton network, the realism and adaptability of virtual exercise scenarios in underground mines are improved, and the accuracy of dynamic risk early warning in underground mines is enhanced.

[0051] like Figure 4As shown, after constructing multiple counterfactual virtual exercise branches, step A300 includes: Step A350: Performing personnel evacuation simulations based on each counterfactual virtual exercise branch, and obtaining the evacuation simulation results for each person. The personnel evacuation simulation refers to treating each underground worker as an independent "virtual intelligent agent," simulating their entire process from receiving a warning signal to evacuating to a safe area in a virtual accident scenario. The evacuation simulation results for each person include data information such as the evacuation path, evacuation time, trapped location, and injury / damage status of each virtual person within each counterfactual virtual exercise branch. Specifically, each counterfactual virtual exercise branch is imported into a digital twin platform. The system extracts real information on all personnel underground in the mine from the mine personnel positioning system, assigning each virtual person a unique identifier and real attribute parameters, including age, position, length of service (longer service indicates greater familiarity with the tunnels and priority in choosing the optimal evacuation route), movement speed (5 km / h for young adults, 3.5 km / h for middle-aged and elderly), and emergency knowledge level (whether they can use self-rescue devices and whether they have participated in emergency drills). All virtual personnel are placed at their actual initial positions at the time of the accident. Behavioral rules are set for virtual personnel to conform to real-world conditions: evacuation must be initiated within 3 seconds of receiving an audible and visual alarm signal; priority must be given to choosing the nearest safe exit; automatic detours must be taken when encountering smoke, water, or collapsed obstacles; stamina decreases linearly with movement time, and movement speed decreases by 50% when stamina is exhausted; and they will move with the group when encountering other personnel. Finally, with a fixed time step of 1 second, the location coordinates and status of each virtual person are updated in real time. The time they take to pass through each alleyway node, the time they take to reach the safety exit, and the time they are caught up by the disaster are recorded. The overall evacuation completion rate, the number of people who have not been evacuated, the location of trapped people and the casualties are statistically analyzed to obtain the evacuation simulation results for each person.

[0052] Step A360: Perform disaster diffusion subject simulation according to each counterfactual virtual exercise branch and obtain the simulation results for each disaster diffusion. Disaster diffusion subject simulation refers to treating the disaster in the accident as an independent evolutionary subject under each counterfactual virtual exercise branch, simulating its entire process of diffusion, intensity change, and coupling and superposition with other disasters in the mine over time. The simulation results for each disaster diffusion include the corresponding disaster type, diffusion boundary, propagation speed, time to reach each key area, and multi-hazard coupling effect data under each counterfactual virtual exercise branch. Specifically, firstly, the initial disaster parameters (such as initial gas concentration of 0.8%, initial roof displacement of 10mm), underground roadway topology, surrounding rock geological parameters, and initial ventilation conditions corresponding to the counterfactual virtual exercise branch are imported into the digital twin platform. Based on the digital twin platform, the movement and diffusion law of gas with ventilation airflow is simulated, the stress transmission, deformation and collapse process of surrounding rock is simulated, the flow of water from the permeable point to a lower place and the water level rise process are simulated, and the natural evolution process of the disaster under conditions without human intervention is simulated. Simultaneously, the simulation of multi-hazard coupling effects is performed: for example, the shock wave generated by a gas explosion can cause roof collapse, which in turn can block roadways and disrupt ventilation, leading to a further increase in gas concentration. Finally, the time it takes for the disaster to reach key areas such as the mining face, safety exits, substations, and pump rooms, as well as the maximum impact range and duration of the disaster, are recorded to obtain the diffusion projection results of each disaster.

[0053] Step A370: Perform ventilation adjustment simulations for each counterfactual virtual exercise branch and obtain the simulation results. Ventilation adjustment simulation refers to treating the underground ventilation system as an independent control entity under each counterfactual virtual exercise branch, simulating operations such as adjusting fan speed, damper opening, and fan start / stop, and recording changes in airflow, direction, and pressure of the underground ventilation system. The simulation results for each ventilation adjustment include the corresponding ventilation adjustment parameters and changes in airflow and direction in each roadway under each counterfactual virtual exercise branch. Specifically, first, import the initial state of the ventilation system (e.g., fan speed 1200 r / min, damper opening 100%), ventilation equipment parameters, and ventilation adjustment commands for each counterfactual virtual exercise branch into the digital twin platform. Based on the digital twin platform, simulate the execution process of different ventilation adjustment measures within each counterfactual virtual exercise branch: such as increasing the main fan speed to 1500 r / min, closing damper #2 leading to the disaster area, opening damper #4 leading to the safe area, and starting the backup local ventilation fan. The system synchronously collects parameters such as real-time air volume, wind direction and pressure, and changes in airflow distribution in each roadway after adjustment, and obtains the simulation results of each ventilation adjustment.

[0054] Step A380: Perform equipment interlocking entity simulations according to each counterfactual virtual exercise branch, and obtain the simulation results for each equipment interlocking entity. Equipment interlocking entity simulation refers to treating all underground electromechanical equipment as independent interlocking entities under each counterfactual virtual exercise branch, simulating the protection logic triggering and state change processes between equipment after an accident. The simulation results for each equipment interlocking entity include the equipment interlocking trigger time and equipment state change parameters corresponding to each counterfactual virtual exercise branch. Specifically, the initial operating state of the equipment (coal mining machine, conveyor, and ventilation fan are all operating normally), equipment interlocking logic (e.g., automatic power-off of the coal mining machine when gas concentration ≥ 1%, automatic shutdown when equipment temperature ≥ 85℃), and fault parameters within each counterfactual virtual exercise branch are imported into the digital twin platform. Simultaneously, the electrical connections, control signal transmission paths, and interlocking trigger conditions between various devices are clearly defined. For example, when a gas sensor detects an anomaly, it interlocks the coal mining machine control system, causing the coal mining machine to lose power. Once a local ventilation fan stops, the gas concentration at the working face will increase, triggering a power outage to the main power supply, causing all production equipment to stop. A digital twin platform is used to simulate the automatic triggering process of equipment interlocking: when the gas concentration rises to 1%, the gas sensor sends a power-off signal within 0.5 seconds, and the coal mining machine stops within 2 seconds. For another example, a local ventilation fan malfunction causing a gas concentration increase at the working face will trigger a power outage to the main power supply, stopping all production equipment. The interlocking trigger time and operating status change parameters of each device are recorded to obtain the interlocking simulation results for each device.

[0055] Step A390: Based on the simulation results of personnel evacuation, disaster spread, ventilation adjustment, and equipment interlocking, competitive correlation is established to obtain the virtual competitive exercise result space. Competitive correlation analyzes the mutual influence, mutual constraints, and dynamic game relationships among the four main entities—personnel evacuation, disaster spread, ventilation adjustment, and equipment interlocking—in the same time and space. The virtual competitive exercise result space includes the simulation results of multiple entities (i.e., the simulation results of personnel evacuation, disaster spread, ventilation adjustment, and equipment interlocking), as well as the mutual influence, mutual constraints, and dynamic game relationships among these simulation results.

[0056] Specifically, the first step is to align the spatiotemporal results of personnel evacuation simulations, disaster spread simulations, ventilation adjustment simulations, and equipment interlocking simulations. All data is unified to a timeline with the "accident occurrence time" as the zero point and within the same underground three-dimensional spatial coordinate system, ensuring that all events occurring at the same time and location correspond. For example, "Five minutes after the accident, the disaster spread to roadway 2, and three workers were evacuating from roadway 2 at the same time." The interaction logic between the various entities is defined: the spread of a disaster to a certain area will reduce the movement speed of personnel in that area and increase the probability of injury. Ventilation adjustment increases the airflow in a certain area, which will reduce the disaster spread speed in that area. Power outages in equipment interlocking will eliminate the risk of electrical sparks in that area, but will prevent personnel from using the transportation equipment in that area, etc. Based on the interaction logic, the relationships between the personnel evacuation simulation results, disaster spread simulation results, ventilation adjustment simulation results, and equipment interlocking simulation results are identified to obtain the competitive relationships among multiple entities. The multi-entity competitive relationships include the mutual influence, mutual constraints, and dynamic game relationships among the simulation results of personnel evacuation, disaster spread, ventilation adjustment, and equipment interlocking. Finally, based on the multi-entity competitive relationships, the simulation results of personnel evacuation, disaster spread, ventilation adjustment, and equipment interlocking are marked to obtain the virtual competitive exercise result space.

[0057] By conducting multi-subject competitive simulations through multiple counterfactual virtual exercise branches, the real evolution process of personnel evacuation, disaster spread, ventilation control, and equipment interlocking in underground accident scenarios can be fully reproduced, and the dynamic game between mutual constraints can be fully verified. This will comprehensively verify the implementation effect and adaptability of different emergency actions, thereby improving the quality of underground safety management in mines.

[0058] Step A400: Based on the virtual competition exercise result space, perform credibility attenuation calculation and virtual-real linkage reverse correction on the underground digital twin of the mine to obtain the corrected underground twin. Wherein, as... Figure 5As shown, step A400 includes: Step A410: Compare the virtual competition exercise result space with the underground mine digital twin to obtain a set of virtual-real deviation features. The set of virtual-real deviation features includes data source deviation, spatial area deviation, time lag deviation, and subject response deviation between the virtual competition exercise result space and the underground mine digital twin. First, unify the spatial coordinates, time step, data sampling frequency, and parameter dimensions of the virtual competition exercise result space and the underground mine digital twin to achieve data benchmark unification. Second, conduct data source deviation identification: compare each initial disaster parameter, environmental parameter, and equipment rated parameter to obtain the data source deviation. Conduct spatial area deviation identification: compare the disaster spread coverage area, risk node spatial location, personnel distribution area, and equipment deployment location of the virtual competition exercise result space and the underground mine digital twin to obtain the spatial area deviation. Time lag bias identification: Compare the disaster spread arrival time, personnel evacuation time, ventilation adjustment response time, and equipment interlocking action time between the virtual competition exercise result space and the underground mine digital twin to obtain time lag bias. Subject response bias identification: Compare the disaster propagation evolution behavior, personnel emergency evacuation behavior, ventilation control behavior, and equipment interlocking triggering behavior between the virtual competition exercise result space and the underground mine digital twin to obtain subject response bias. All identified four types of bias are uniformly classified and numbered to obtain a virtual-real bias feature set, providing comprehensive and traceable bias base data for subsequent calculation of multiple twin credibility attenuation factors and reverse correction of the twin.

[0059] Step A420: Based on the aforementioned virtual-real deviation feature set, calculate multiple twin credibility attenuation factors. These multiple twin credibility attenuation factors include a data source credibility attenuation factor, a spatial region credibility attenuation factor, a time lag credibility attenuation factor, and a subject response credibility attenuation factor. The formula for calculating the data source credibility attenuation factor is as follows: ; in, As a data source credibility decay factor, For the predetermined deviation threshold, This is due to data source deviation.

[0060] A larger data source credibility decay factor indicates higher credibility of the data source dimension. A smaller data source credibility decay factor indicates more severe credibility decay of the data source dimension. Similarly, the spatial region credibility decay factor = (predetermined deviation threshold - spatial region deviation) / predetermined deviation threshold. The time lag credibility decay factor = (predetermined deviation threshold - time lag deviation) / predetermined deviation threshold. The subject response credibility decay factor = (predetermined deviation threshold - subject response deviation) / predetermined deviation threshold. Multiple twin credibility decay factors can intuitively quantify the degree of credibility decay of each dimension of the underground digital twin in the mine, providing a unified, standardized, and feasible quantitative basis for subsequent model decay correction and virtual-real linkage reverse correction.

[0061] Step A430: The underground digital twin of the mine is attenuated and corrected according to the multiple twin credibility attenuation factors to obtain the first optimized underground digital twin. For the data source dimension, the original data deviation of the underground digital twin is compensated according to the data source credibility attenuation factor, and the initial disaster parameters, environmental parameters, and equipment parameters are recalibrated to offset simulation errors caused by data source deviations. For the spatial region dimension, the disaster spread range is scaled proportionally according to the spatial region credibility attenuation factor, the risk node positions are fine-tuned, and the spatial distribution of personnel and equipment is corrected to compensate for spatial matching deviations. For the time lag dimension, the simulation timing step size is adjusted according to the time lag credibility attenuation factor, and the action delay time is compensated to correct the speed deviations in disaster propagation, personnel evacuation, and equipment interlocking. For the subject response dimension, the disaster evolution speed, personnel evacuation sensitivity, ventilation control intensity, and equipment interlocking response threshold are adjusted according to the subject response credibility attenuation factor to correct the subject behavior distortion problem. Finally, the first optimized underground digital twin is obtained. It achieves quantitative and precise correction, where the larger the deviation, the stronger the correction, and the smaller the deviation, the weaker the fine-tuning. This solves the problems of inconsistency, distortion, offset, and lag in the simulation of digital twins in underground mines, and significantly improves the overall stability of digital twins in underground mines.

[0062] Step A440: Perform reverse correction on the underground digital twin of the mine according to the virtual-real deviation feature set to obtain a second optimized underground twin. Step A450: Merge the first optimized underground twin and the second optimized underground twin to generate the corrected underground twin.

[0063] By utilizing the data source deviation, spatial region deviation, time lag deviation, and subject response deviation from the virtual-real deviation feature set, a root-cause reverse correction is performed on the underground digital twin to obtain a second optimized underground twin. Specifically, for data source deviation, if the simulated initial gas concentration is found to be higher than the actual site, or if there is a fixed offset in the environmental temperature and humidity parameters, the underlying parameter initialization script of the underground digital twin is directly modified, replacing the model's default initial values ​​with actual on-site monitoring values ​​to completely eliminate data input deviation. For spatial region deviation, if the virtual disaster spread range is found to be larger than the actual coverage area of ​​the real roadway, or if the risk node coordinates are offset, or if the personnel distribution area is misaligned, the underground 3D roadway grid coordinates are recalibrated, the disaster spread spatial boundary is reset, and the spatial point mapping relationship between equipment and personnel is corrected to eliminate spatial modeling errors. For time lag deviation, if the disaster arrival time is found to be lagging behind the actual on-site time, or if the equipment interlock response is delayed, the simulation timing step size is modified, the system response delay is compressed, and the event trigger timing is aligned to resolve the inconsistency in the overall simulation speed. To address the deviation in the main response, if it is found that the personnel evacuation speed is too fast, the ventilation control sensitivity is insufficient, the disaster evolution rate is distorted, or the equipment interlocking trigger is too sensitive, the multi-subject behavior parameters are recalibrated. The systemic defects of unrealistic main responses and mismatched evolution patterns are repaired from the bottom layer. Finally, the full-dimensional bottom-layer reverse calibration is completed, and a second optimized downhole twin is obtained that eliminates system errors and the simulation pattern fits the real working conditions.

[0064] Furthermore, a corrected downhole twin is generated by fusing the first and second optimized downhole twins. Specifically, the first optimized downhole twin uses multiple twin reliability decay factors to perform proportional fine-tuning, enabling it to adjust more based on the magnitude of the deviation, ensuring stable twin operation without severe oscillations. The second optimized downhole twin completes fundamental rectification, thoroughly fixing inherent defects in data, space, timing, and the subject. During the fusion process, for non-systematic deviations such as minor data fluctuations and small spatial offsets, the proportional correction results of the first optimized downhole twin are uniformly retained to maintain the twin's operational stability. For systemic issues such as inaccurate timing, distorted subject behavior, spatial topology errors, and inherent parameter offsets, all are replaced with the underlying correction results of the second optimized downhole twin. Simultaneously, the simulation step size and spatial coordinates of the first and second optimized downhole twins are unified, ultimately resulting in a corrected downhole twin that is free from system deviations and operational fluctuations, and possesses both high precision and high stability. This enables the matching of virtual drills with real-time downhole conditions, improving the realism and adaptability of virtual drill scenarios.

[0065] Step A420 includes: Steps A421 and A500: The multi-hazard spatiotemporal coupled risk propagation map is updated and reinforced based on the corrected underground twin to obtain dynamic risk warning results. First, the corrected underground twin and the multi-hazard spatiotemporal coupled risk propagation map are connected via a standardized data interface to extract the real-time risk intensity (e.g., gas concentration of 0.8% and roof displacement of 12mm) of multiple underground risk nodes in the corrected underground twin, the actual transmission rate of multiple risk propagation edges (e.g., gas diffusion velocity of 0.6m / s in the north wing return airway), and the disaster arrival time of multiple key time-series nodes. These data are compared item by item with the corresponding parameters of the multi-hazard spatiotemporal coupled risk propagation map, and parameters with deviations are updated using feedback. That is, for risk nodes, the original risk value is replaced with the real-time risk value in the corrected underground twin; for example, the risk value of the No. 3 tunneling face is updated from 0.3 to 0.7. For the risk propagation edges, the propagation weights are recalculated based on the corrected disaster transmission rate; for example, the propagation weight of the south wing transport roadway is adjusted from 0.4 to 0.6. Finally, based on the updated multi-hazard spatiotemporal coupled risk propagation map, dynamic risk warning results containing risk location and risk level are generated. This completely solves the problems of static solidification, parameter dependence on experience, insufficient consideration of coupling effects, and delayed warnings in traditional multi-hazard risk propagation maps. It realizes dynamic self-iteration and continuous optimization of the risk map, improves the accuracy of multi-hazard coupled risk identification underground, and provides timely and reliable decision-making basis for dynamic risk prevention and emergency response in mines. Based on the multiple twin credibility decay factors, twin credibility decay warning signals are generated. The twin credibility decay warning signals are alarm instructions and prompts triggered by each twin credibility decay factor to indicate abnormal credibility status of the digital twin. First, warning judgment thresholds are pre-set for multiple twin credibility decay factors, and two levels, normal and warning, are divided according to the requirements of mine simulation operation. For example, when each twin's credibility decay factor is greater than or equal to a certain value, it is considered a normal range; when each twin's credibility decay factor is less than 0, it is considered a warning range. By comparing each twin's credibility decay factor with the warning range, a twin credibility decay warning signal is generated when any twin's credibility decay factor falls into the warning range. Through multiple twin credibility decay factors, the credibility changes of each dimension of the digital twin can be monitored in real time, allowing for early identification of abnormal states such as model distortion and excessive deviations, thus preventing the failure of risk simulation and virtual exercise results due to insufficient twin credibility.

[0066] Step A500: Based on the corrected underground twin, the multi-hazard spatiotemporal coupled risk propagation map is updated with feedback reinforcement to obtain dynamic risk warning results. First, the corrected underground twin and the multi-hazard spatiotemporal coupled risk propagation map are connected via a standardized data interface. The real-time risk intensity of multiple underground risk nodes in the corrected underground twin (e.g., gas concentration of 0.8% and roof displacement of 12mm in the No. 2 coal face), the actual transmission rate of multiple risk propagation edges (e.g., gas diffusion velocity of 0.6m / s in the north wing return airway), and the disaster arrival time of multiple key time-series nodes are extracted. These data are compared item by item with the corresponding parameters in the multi-hazard spatiotemporal coupled risk propagation map, and parameters with deviations are updated with feedback. That is, for risk nodes, the original risk value is replaced with the real-time risk value in the corrected underground twin; for example, the risk value of the No. 3 tunneling face is updated from 0.3 to 0.7. For each risk propagation edge, the propagation weight is recalculated based on the corrected disaster transmission rate; for example, the propagation weight of the south wing transport roadway is adjusted from 0.4 to 0.6. Finally, based on the updated multi-hazard spatiotemporal coupled risk propagation map, dynamic risk warning results containing risk location and risk level are generated. This completely solves the problems of static fixation, parameter dependence on experience, insufficient consideration of coupling effects, and delayed warnings in traditional multi-hazard risk propagation maps. It achieves dynamic self-iteration and continuous optimization of the risk map, improves the accuracy of multi-hazard coupled risk identification underground, and provides timely and reliable decision-making basis for dynamic risk prevention and emergency response in mines.

[0067] Step A500 includes: Step A510: Based on the dynamic risk warning results, configure an underground risk suppression mechanism and execute the underground risk pre-suppression according to the underground risk suppression mechanism. The underground risk suppression mechanism is a set of risk handling measures that match different types of disasters such as underground gas, roof collapse, water hazards, and fires, including automated equipment actions, personnel dispatch instructions, ventilation control strategies, and equipment interlocking rules. Risk pre-suppression is a proactive safety management process that takes preventive measures to block the spread of risks, reduce the intensity of risks, and prevent the escalation of risks before a disaster occurs or in its early stages. The underground risk suppression mechanism includes: if the dynamic risk warning results include the risk of increased gas concentration, configuring the main ventilation fan speed to increase parameters and closing the air doors leading to the risk area, etc. If the dynamic risk warning results include the risk of excessive roof displacement, configuring the roof stress monitoring equipment to increase sampling, activating temporary support devices, and notifying the personnel working in the area to evacuate, etc. If the dynamic risk warning results include the risk of water hazard warning, configuring the closure of waterproof gates, activation of drainage pumps, and evacuation of personnel in low-lying areas, etc. This has enabled a shift from "passive response" to "proactive prevention" of underground mine risks, advancing the timing of risk management from after a disaster to the nascent stage, shortening the risk response time, effectively preventing the spread and escalation of risks, reducing the probability of accidents and disaster losses, and providing proactive and intelligent safety assurance for safe production in underground mines.

[0068] In summary, the method for dynamic risk early warning and virtual drills of mines based on digital twins provided in this application has the following technical effects: This method analyzes the multi-dimensional disaster coupling and propagation relationships using a digital twin of the mine underground, establishing a multi-hazard spatiotemporal coupled risk propagation map. It then uses an underground risk propagation induction matrix to dynamically collapse and reconstruct the risk propagation chain of this map, obtaining a dynamic risk collapse skeleton network. Based on this network, multiple counterfactual virtual exercise branches are constructed, and multi-agent competitive simulations are performed using these branches to obtain a virtual competitive exercise result space. Based on this result space, the credibility attenuation of the underground digital twin is calculated, and reverse correction based on virtual-real linkage is applied to obtain a corrected underground twin. Finally, the corrected underground twin is used to reinforce and update the multi-hazard spatiotemporal coupled risk propagation map, yielding dynamic risk warning results. This approach solves the technical problems of existing technologies where underground mine risk warnings rely on single-point threshold judgments, making it difficult to characterize multi-hazard coupling and propagation relationships, and causing a disconnect between virtual exercises and real-time underground conditions, resulting in poor quality underground mine safety management. It has achieved the technical effects of improving the accuracy of underground risk propagation prediction, realizing dynamic linkage early warning of multiple disaster risks, enhancing the realism and adaptability of virtual exercise scenarios, and improving the quality of underground mine safety management.

[0069] Example 2: Based on the same inventive concept as the method for dynamic risk early warning and virtual drills of digital twin mines described in the previous examples, this invention also provides a system for dynamic risk early warning and virtual drills of digital twin mines. Please refer to the appendix. Figure 6 The system includes: The disaster coupling propagation module 1 is used to construct a digital twin of the mine underground based on the multi-source monitoring dataset of the mine underground, and to perform multi-dimensional disaster coupling propagation relationship analysis based on the digital twin of the mine underground to establish a multi-hazard spatiotemporal coupling risk propagation map; The dynamic collapse reconstruction module 2 is used to construct the downhole risk propagation induction matrix and perform dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map according to the downhole risk propagation induction matrix to obtain the dynamic risk collapse skeleton network. The multi-agent competition simulation module 3 is used to construct multiple counterfactual virtual exercise branches based on the dynamic risk collapse skeleton network, and to perform multi-agent competition simulation based on the multiple counterfactual virtual exercise branches to obtain the virtual competition exercise result space. The virtual-real linkage reverse correction module 4 is used to perform credibility attenuation calculation and virtual-real linkage reverse correction on the underground digital twin of the mine based on the virtual competition exercise result space, and obtain the corrected underground twin; The feedback reinforcement update module 5 is used to perform feedback reinforcement updates on the multi-hazard spatiotemporal coupling risk propagation map based on the corrected downhole twin, and obtain dynamic risk warning results.

[0070] Furthermore, the disaster coupling propagation module 1 is also used to: perform spatiotemporal mapping of multidimensional disaster factors based on the underground digital twin of the mine to obtain a spatiotemporal distribution map of multidimensional disaster factors; perform semantic extraction of underground risk nodes based on the spatiotemporal distribution map of multidimensional disaster factors to obtain a set of underground risk nodes; and perform cross-hazard coupling propagation relationship identification based on the set of underground risk nodes to obtain a multi-hazard spatiotemporal coupling risk propagation map.

[0071] Furthermore, the dynamic collapse reconstruction module 2 is also used for: performing propagation inertia analysis based on the multi-hazard spatiotemporal coupled risk propagation map to construct an underground risk propagation inertial field; predicting the migration of risk hotspots based on the underground risk propagation inertial field to obtain the underground dynamic risk migration trajectory; configuring criticality based on the underground dynamic risk migration trajectory to obtain the risk migration critical attention trajectory; and performing dynamic parameter configuration for induction intervention based on the risk migration critical attention trajectory to obtain the underground risk propagation induction matrix.

[0072] Furthermore, the dynamic collapse reconstruction module 2 is also used for: performing propagation blocking analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtaining risk propagation blocking analysis results; performing propagation intensity attenuation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtaining risk propagation intensity attenuation analysis results; performing propagation direction reorientation analysis on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix, and obtaining risk propagation direction reorientation analysis results; and performing dynamic collapse reconstruction on the multi-hazard spatiotemporal coupled risk propagation map based on the risk propagation blocking analysis results, the risk propagation intensity attenuation analysis results, and the risk propagation direction reorientation analysis results to generate the dynamic risk collapse skeleton network.

[0073] Furthermore, the multi-agent competitive simulation module 3 is also used for: identifying high-risk propagation main pathways based on the dynamic risk collapse skeleton network to obtain multiple baseline accident evolution chains; performing multi-node counterfactual perturbations on each baseline accident evolution chain to generate multiple counterfactual perturbation condition sets; evaluating and filtering the simulation value of each counterfactual perturbation condition set to obtain multiple preferred counterfactual perturbation conditions; and mapping and modeling the multiple baseline accident evolution chains with the multiple preferred counterfactual perturbation conditions to obtain the multiple counterfactual virtual simulation branches.

[0074] Furthermore, the multi-entity competitive simulation module 3 is also used for: performing personnel evacuation simulations based on each counterfactual virtual exercise branch, and obtaining the simulation results for each personnel evacuation; performing disaster spread simulations based on each counterfactual virtual exercise branch, and obtaining the simulation results for each disaster spread; performing ventilation adjustment simulations based on each counterfactual virtual exercise branch, and obtaining the simulation results for each ventilation adjustment; performing equipment interlocking simulations based on each counterfactual virtual exercise branch, and obtaining the simulation results for each equipment interlocking; and performing competitive correlation based on the simulation results for each personnel evacuation, the simulation results for each disaster spread, the simulation results for each ventilation adjustment, and the simulation results for each equipment interlocking to obtain the virtual competitive exercise result space.

[0075] Furthermore, the virtual-real linkage reverse correction module 4 is also used to: compare the virtual competition exercise result space with the underground digital twin of the mine to obtain a virtual-real deviation feature set; calculate multiple twin credibility attenuation factors based on the virtual-real deviation feature set; perform attenuation correction on the underground digital twin of the mine based on the multiple twin credibility attenuation factors to obtain a first optimized underground twin; perform reverse correction on the underground digital twin of the mine based on the virtual-real deviation feature set to obtain a second optimized underground twin; and fuse the first optimized underground twin and the second optimized underground twin to generate the corrected underground twin.

[0076] The multi-source monitoring dataset includes environmental monitoring data, equipment operation data, personnel monitoring data, ventilation operation data, and geological monitoring data.

[0077] Furthermore, the feedback reinforcement update module 5 is also used to: configure an underground risk suppression mechanism based on the dynamic risk warning result, and execute the underground risk pre-suppression in the mine based on the underground risk suppression mechanism.

[0078] Furthermore, the virtual-real linkage reverse correction module 4 is also used to generate a twin credibility decay warning signal based on the multiple twin credibility decay factors.

[0079] The dynamic risk warning and virtual drill system for digital twin mines provided in this embodiment of the invention can execute the dynamic risk warning and virtual drill system for digital twin mines provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0080] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0081] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for dynamic risk early warning and virtual drills in digital twin mines, characterized in that: The method includes: Based on the multi-source monitoring dataset of underground mines, a digital twin of underground mines is constructed, and the multi-dimensional disaster coupling and propagation relationship is analyzed based on the digital twin of underground mines to establish a multi-hazard spatiotemporal coupling risk propagation map; Construct a downhole risk propagation induction matrix, and perform dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map based on the downhole risk propagation induction matrix to obtain a dynamic risk collapse skeleton network; Multiple counterfactual virtual exercise branches are constructed based on the dynamic risk collapse skeleton network, and multi-agent competitive simulation is performed based on the multiple counterfactual virtual exercise branches to obtain the virtual competitive exercise result space; Based on the virtual competition exercise result space, the credibility attenuation calculation and virtual-real linkage reverse correction are performed on the underground digital twin of the mine to obtain the corrected underground twin; The multi-hazard spatiotemporal coupled risk propagation map is updated and reinforced based on the corrected downhole twin to obtain dynamic risk warning results; The construction of the downhole risk propagation induction matrix includes: Based on the spatiotemporal coupled risk propagation map of the multiple hazards, propagation inertia analysis is performed to construct the underground risk propagation inertial field in the mine. Based on the inertial field of underground risk propagation, risk hot zone migration is predicted, and the underground dynamic risk migration trajectory is obtained. Based on the downhole dynamic risk migration trajectory, criticality is configured to obtain the risk migration critical attention trajectory; Based on the risk migration critical attention trajectory, the dynamic parameter configuration for induction intervention is executed to obtain the downhole risk propagation induction matrix. The dynamic parameter configuration for induction intervention refers to the parameters of the corresponding intervention measures matched for the risk migration critical attention trajectory.

2. The method as described in claim 1, characterized in that, Based on the aforementioned underground digital twin of the mine, multi-dimensional disaster coupling and propagation relationships are analyzed, and a multi-hazard spatiotemporal coupled risk propagation map is established, including: Based on the underground digital twin of the mine, a multidimensional disaster factor spatiotemporal mapping is performed to obtain a multidimensional disaster factor spatiotemporal distribution map; Based on the spatiotemporal distribution map of the multidimensional disaster factors, semantic extraction of downhole risk nodes is performed to obtain a set of downhole risk nodes; Based on the set of downhole risk nodes, cross-hazard coupling propagation relationships are identified to obtain the spatiotemporal coupled risk propagation map of the multi-hazards.

3. The method as described in claim 1, characterized in that, Based on the downhole risk propagation induction matrix, the multi-hazard spatiotemporal coupled risk propagation map is dynamically reconstructed by risk propagation chain collapse to obtain a dynamic risk collapse skeleton network, including: Based on the downhole risk propagation induction matrix, a propagation blocking analysis is performed on the spatiotemporal coupled risk propagation map of the multiple hazards to obtain the risk propagation blocking analysis results. Based on the downhole risk propagation induction matrix, the propagation intensity attenuation analysis is performed on the multi-hazard spatiotemporal coupled risk propagation map to obtain the risk propagation intensity attenuation analysis results. Based on the downhole risk propagation induction matrix, the propagation direction redefinition analysis is performed on the multi-hazard spatiotemporal coupled risk propagation map to obtain the risk propagation direction redefinition analysis results. Based on the risk propagation blocking analysis results, the risk propagation intensity attenuation analysis results, and the risk propagation direction reorientation analysis results, the spatiotemporal coupled multi-hazard risk propagation map is dynamically collapsed and reconstructed to generate the dynamic risk collapse skeleton network.

4. The method as described in claim 1, characterized in that, Multiple counterfactual virtual exercise branches are constructed based on the dynamic risk collapse skeleton network, including: Based on the dynamic risk collapse skeleton network, high-risk propagation main pathways are identified, and multiple baseline accident evolution chains are obtained. For each baseline incident evolution chain, perform multi-node counterfactual perturbations to generate multiple counterfactual perturbation condition sets; For each set of counterfactual perturbation conditions, a value evaluation and screening process is performed to obtain multiple preferred counterfactual perturbation conditions; The multiple baseline incident evolution chains are mapped and modeled with the multiple preferred counterfactual perturbation conditions to obtain the multiple counterfactual virtual exercise branches.

5. The method as described in claim 1, characterized in that, Based on the multiple counterfactual virtual exercise branches, a multi-agent competitive simulation is performed to obtain the virtual competitive exercise result space, including: Based on the personnel evacuation simulation of each counterfactual virtual exercise branch, obtain the evacuation simulation results for each person; Based on each counterfactual virtual exercise branch, a disaster diffusion subject simulation is performed to obtain the results of each disaster diffusion simulation. Perform ventilation adjustment simulations based on each counterfactual virtual exercise branch and obtain the simulation results for each ventilation adjustment. Based on each counterfactual virtual exercise branch, perform the main simulation of equipment interlocking and obtain the simulation results of each equipment interlocking. The virtual competitive exercise result space is obtained by competitively associating the results of the personnel evacuation simulation, the disaster spread simulation, the ventilation adjustment simulation, and the equipment interlocking simulation.

6. The method as described in claim 1, characterized in that, Based on the virtual competition exercise result space, the credibility attenuation of the underground digital twin is calculated and the virtual-real linkage reverse correction is performed to obtain the corrected underground twin, including: The virtual competition exercise result space is compared with the underground digital twin of the mine to obtain a set of virtual-real deviation features; Based on the aforementioned set of virtual-real deviation features, multiple twin reliability attenuation factors are calculated; The underground digital twin of the mine is attenuated and corrected according to the multiple twin credibility attenuation factors to obtain the first optimized underground twin; The underground digital twin of the mine is reverse-corrected based on the set of virtual-real deviation features to obtain a second optimized underground twin; The corrected downhole twin is generated by fusing the first optimized downhole twin and the second optimized downhole twin.

7. The method as described in claim 1, characterized in that, The multi-source monitoring dataset includes environmental monitoring data, equipment operation data, personnel monitoring data, ventilation operation data, and geological monitoring data; Obtain dynamic risk warning results, including: Based on the dynamic risk warning results, configure an underground risk suppression mechanism, and implement the underground risk pre-suppression in the mine according to the underground risk suppression mechanism.

8. The method as described in claim 6, characterized in that, Calculate multiple twin confidence decay factors, including: Based on the multiple twin credibility decay factors, a twin credibility decay warning signal is generated.

9. A dynamic risk early warning and virtual drill system for digital twin mines, characterized in that: The system is used to implement the method according to any one of claims 1 to 8, the system comprising: The disaster coupling propagation module is used to construct a digital twin of the mine underground based on the multi-source monitoring dataset of the mine underground, and to analyze the multi-dimensional disaster coupling propagation relationship based on the digital twin of the mine underground, and to establish a multi-hazard spatiotemporal coupling risk propagation map; The dynamic collapse reconstruction module is used to construct a downhole risk propagation induction matrix and perform dynamic collapse reconstruction of the risk propagation chain on the multi-hazard spatiotemporal coupled risk propagation map according to the downhole risk propagation induction matrix to obtain a dynamic risk collapse skeleton network. The multi-agent competition simulation module is used to construct multiple counterfactual virtual exercise branches based on the dynamic risk collapse skeleton network, and to perform multi-agent competition simulation based on the multiple counterfactual virtual exercise branches to obtain the virtual competition exercise result space. The virtual-real linkage reverse correction module is used to perform credibility attenuation calculation and virtual-real linkage reverse correction on the underground digital twin of the mine based on the virtual competitive exercise result space, and obtain the corrected underground twin; the feedback reinforcement update module is used to perform feedback reinforcement update on the multi-hazard spatiotemporal coupling risk propagation map based on the corrected underground twin, and obtain dynamic risk warning results.

Images