Mine earthquake complex network multi-scale sensing and overlying strata self-adaptive regulation and control method

Abstract

The invention relates to the technical field of mine safety, and provides a mine earthquake complex network multi-scale sensing and overlying strata adaptive regulation and control method. Constructing a dynamic complex network model of the coal pillar-key layer structure by performing spatio-temporal scale fusion on micro-seismic monitoring data of the mine coal pillar-key layer mechanism and a numerical simulation result; and dynamically updating the dynamic complex network model of the coal pillar-key layer structure by adopting a bidirectional collaborative optimization strategy so as to generate stress judgment and energy release indexes of collaborative instability induced shock of the coal pillar-key layer structure, and refining the process and mode of energy release in combination with micro-shock monitoring data and numerical simulation so as to realize the energy release of the coal pillar-key layer structure. Complete mine earthquake monitoring and control from macroscopic stress transfer to mesoscopic micro-earthquake activity and then to microscopic crack extension are realized, and powerful support is provided for mine earthquake early warning and prevention and control in a mine.

Description

Technical Field

[0001] This application relates to the field of mine safety technology, and in particular to a method for multi-scale sensing and overburden adaptive control of complex mine seismic networks. Background Technology

[0002] Mine tremors are a common hazard in coal mine production, primarily caused by mining disturbances that trigger rock mass stress instability, energy accumulation, and sudden release. As mining depth increases, the frequency and intensity of mine tremors also gradually increase, especially under conditions of high gas content, thick coal seams, and complex geological formations, where the risks are more pronounced. Mine tremors not only pose a serious threat to safe mine production but can also trigger secondary disasters such as roof collapse and gas explosions. Summary of the Invention

[0003] The purpose of this application is to provide a multi-scale sensing and overburden adaptive control method for complex mining seismic networks, so as to solve or alleviate the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, this application provides the following technical solution: This application provides a method for multi-scale sensing and overburden adaptive control of complex mining seismic networks, including: The microseismic monitoring data and numerical simulation results of the coal pillar-key layer structure in the mine are fused in a spatiotemporal scale to construct a dynamic complex network model of the coal pillar-key layer structure. The key areas for mine seismic propagation are identified by the node centrality of the dynamic complex network model. A bidirectional collaborative optimization strategy combining graph theory-based centrality analysis and machine learning prediction is adopted. By using historical data on microseismic energy release in key areas, the dynamic complex network model of the coal pillar-key layer structure is dynamically updated. Based on the optimization update results, stress criteria and energy release indicators are dynamically generated to provide early warning of the collaborative instability state of the coal pillar-key layer structure. In response to the stress criterion for coordinated instability reaching the early warning threshold, seismic suppression and control are implemented in key areas. Based on the microseismic monitoring data after seismic suppression and control, the dynamic complex network model and the stress criterion for coordinated instability are dynamically updated to verify the effect of seismic suppression and control and optimize subsequent seismic suppression and control schemes.

[0005] Preferably, the dynamic complex network model of the coal pillar-key layer structure includes: using multiple numerical simulation grid cells of the mine spatial region as network nodes, and connecting multiple numerical simulation grid cells based on stress-energy transfer inside the mine to construct a macro-scale model of the mine; Wherein, according to the formula: Determine the grid cells in the macroscale model Its adjacent grid cells Stress-energy transfer efficiency ; In the formula, For grid cells in macro-scale models With grid cells stress gradient, This represents the maximum stress gradient within the mesh elements of the macroscopic model. They are grid cells With grid cells strain energy density, This represents the maximum strain energy density within the grid cells of the macroscopic-scale model. To adjust the exponent, the nonlinear effect of the strain energy density of the grid cells on the energy transfer efficiency is characterized; The attenuation factor characterizes the grid cell. With grid cells Impact on energy transfer efficiency; For grid cells With grid cells The physical distance between them The characteristic length represents the scale of the influence range or transmission distance.

[0006] Preferably, the dynamic complex network model of the coal pillar-key layer structure further includes: projecting the source points of mine microseismic events onto network nodes that divide the mine seismic activity area network into grids, and connecting the source points of microseismic events through the time interval and spatial interval of microseismic events to construct a mesoscale model of the mine seismic activity area network; Wherein, according to the formula: Determining microseismic events in mesoscale models Microseismic events Energy correlation ; In the formula, Microseismic events Microseismic events The time interval, Microseismic events Microseismic events The spatial distance of the epicenter, Microseismic events Microseismic events The energy of microseismic events; These are the time decay constant and the spatial decay constant, respectively. This represents the maximum microseismic event energy in the mesoscale model. This is the energy decay index, used to adjust the effect of energy on connection strength.

[0007] Preferably, the dynamic complex network model of the coal pillar-key layer structure further includes: a microscale model of the local fracture network in the mine, which uses particle contact points or fracture breakpoints as network nodes, the contact force between particle contact points or the fracture opening between fracture breakpoints as connection weights, and connects particle contact points or fracture breakpoints through contact force chains between particle contact points or fracture expansion paths between fracture breakpoints.

[0008] Preferably, the degree centrality and / or betweenness centrality of network nodes in a dynamic complex network model are calculated to determine the propagation path and key areas of mine tremors.

[0009] Preferably, the dynamic complex network model of the coal pillar-critical layer structure is dynamically updated, including: The dynamic complex network model of the coal pillar-key layer structure is corrected by mesh update, random perturbation parameters, and inverse optimization. Based on the microseismic activity of mine faults, the parameters of a dynamic complex network model are dynamically adjusted using a time-series graph attention network.

[0010] Preferably, the dynamic complex network model of the coal pillar-key layer structure is mesh corrected through mesh updates, random perturbation parameters, and inverse optimization, including: In response to the stress gradient of an element in a dynamic complex network model of a coal pillar-key layer structure exceeding a preset gradient threshold, an adaptive finite element method based on error estimation is used to insert new network nodes into the dynamic complex network model. In the dynamic complex network model of coal pillar-critical layer structure, adjacent network nodes in the low-stress variation region are merged. Based on historical microseismic event catalogs and stress monitoring data, the node positions and connections of network nodes in the dynamic complex network model of the coal pillar-key layer structure are adjusted using gradient descent or genetic algorithms.

[0011] Preferably, based on the microseismic activity of the mine fault, the parameters of the dynamic complex network model are dynamically adjusted using a time-series graph attention network, including: Dynamic updates of node embeddings in dynamic complex network models are achieved through structural attention and temporal attention. In fault simulation, the connection weights of network nodes in a dynamic complex network model are dynamically adjusted using a time-series graph attention network based on changes in velocity distribution. The mesh partitioning of a dynamic complex network model is optimized in reverse based on the variational method, and the transition zone position of the mesh partitioning of the dynamic complex network model is adjusted by gradient descent. The grid density distribution of a dynamic complex network model is optimized by using the attention weight feedback of a temporal graph attention network. The boundary nodes of the macroscopic network in the dynamic complex network model are associated with the breakpoints of the microscopic network in the dynamic complex network model, and the multi-scale network is mapped to the same feature space through graph embedding operation to generate a unified network topology.

[0012] Preferably, the stress criteria and energy release indicators dynamically generated for early warning of the coal pillar-critical layer structure entering a state of coordinated instability include: Based on the macro-scale model in the dynamic complex network model, the stress-energy transfer relationship between coal pillars, key layers and fault zones in the mine is determined. Based on the mesoscale model in the dynamic complex network model, the spatiotemporal distribution characteristics of mine microseismic events are extracted, and the stress concentration path is determined by the connection relationship of network nodes in the mesoscale model. Based on the microscale model in the dynamic complex network model, and combined with numerical simulation and microseismic monitoring data of the mine, the fracture propagation path and its energy release characteristics are predicted.

[0013] In response to the stress criterion of coordinated instability reaching the early warning threshold, a seismic control scheme is generated and implemented for the key area, including at least one of hydraulic fracturing, pressure relief drilling, or adjustment of the mining sequence plan in the key area.

[0014] Beneficial effects: The multi-scale sensing and overburden adaptive control method for complex mine seismic networks provided in this application integrates microseismic monitoring data and numerical simulation results of the coal pillar-key layer structure at spatiotemporal scales to construct a dynamic complex network model of the coal pillar-key layer structure. The key areas for mine seismic propagation are identified through the node centrality of the dynamic complex network model. A bidirectional collaborative optimization strategy combining graph theory-based centrality analysis and machine learning prediction is employed. Historical data on microseismic energy release in the key areas is used to dynamically update the dynamic complex network model of the coal pillar-key layer structure. Based on the optimization update results, stress criteria and energy release indicators are dynamically generated to provide early warning of the collaborative instability state of the coal pillar-key layer structure. When the stress criteria for collaborative instability reach the early warning threshold, seismic suppression control is implemented in the key areas. Based on the microseismic monitoring data after seismic suppression control, the dynamic complex network model and the stress criteria for collaborative instability are dynamically updated to verify the seismic suppression control effect and optimize subsequent seismic suppression control schemes.

[0015] Therefore, by using a dynamic and complex network model of the coal pillar-key layer structure, the energy transfer paths and stress distribution patterns between various regions in the mine (such as coal pillars, key layers, and fault zones) are revealed, providing a physical basis for the spatiotemporal distribution of microseismic events and the extraction of stress concentration paths. This allows for the identification of stress concentration areas in the mine through the spatiotemporal distribution of microseismic events, thereby determining the hotspots of mine earthquakes and providing direction for fracture propagation path analysis. By combining microseismic monitoring data and numerical simulations, the process and mode of energy release are refined, achieving complete mine earthquake monitoring from macroscopic stress transfer to mesoscopic microseismic activity and then to microscopic fracture propagation. After implementing seismic suppression and control, monitoring feedback is conducted through the microseismic and stress monitoring system to achieve continuous iterative optimization of the complex network structure, risk criteria, and control strategies, forming an adaptive closed loop of "sensing-diagnosis-control-verification," providing strong support for mine earthquake early warning and prevention. Attached Figure Description

[0016] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. Wherein: Figure 1 This is a flowchart illustrating a multi-scale sensing and overburden adaptive control method for complex mining seismic networks according to some embodiments of this application; Figure 2 This is a logical view of multi-scale sensing and overburden adaptive control of complex mining seismic networks provided according to embodiments of this application; Figure 3 This is a schematic diagram of a scenario for multi-scale sensing and overburden adaptive control of complex mining seismic networks according to embodiments of this application; Figure 4 This is a schematic diagram of the technical route of a multi-scale sensing and overburden adaptive control method for complex mining seismic networks according to some embodiments of this application; Figure 5 This is a schematic diagram of adaptive control logic provided according to some embodiments of this application. Detailed Implementation

[0017] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Various examples are provided by way of explanation and not by way of limitation. In fact, those skilled in the art will understand that modifications and variations can be made to the present application without departing from the scope or spirit of the present application. For example, a feature shown or described as part of one embodiment may be used in another embodiment to produce yet another embodiment. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0018] Mine seismic events are generally classified into overburden type, fault type, and coal pillar type. Chain seismic events, on the other hand, are a new type of instability mode that is different from the traditional single rock layer instability or local stress concentration failure caused by specific disturbances in the "coal pillar-key layer" system under deep mining conditions. They are mainly affected by factors such as mining depth and stress, mining speed and disturbance intensity, geological structure and rock mass properties.

[0019] Among them, the mining speed during the mining process Rockbursts occur relatively frequently, but when At this time, the frequency of rockbursts increases significantly. Under deep, high-stress conditions, excessively high mining speeds may cause the rock mass strain energy to accumulate beyond a critical value, leading to chain-like fractures. Simultaneously, with increasing mining depth (e.g., greater than 1000 meters), the original rock stress increases significantly, the rock mass becomes more brittle, and the intensity of seismic energy release increases exponentially. Furthermore, near faults, folds, and other tectonic zones, the rock mass integrity is poor, and fractures are well-developed, providing pathways for energy transfer. For example, fault ignition leads to stress field reconstruction, easily forming a "stress shell imbalance" and triggering seismic events.

[0020] Under deep mining conditions, chain tremors (i.e., continuous or cascading tremor events triggered by mining activities) have become a significant threat to coal mine safety. Their occurrence mechanism is complex, involving the coupling of multiple processes such as stress redistribution in the rock mass, critical stratum failure, and energy accumulation and release. Instability and failure caused by chain tremors are often not independent failures of single components, but rather a dynamic coupling of coal pillar instability and critical stratum failure. Furthermore, the failure is characterized by significant suddenness, often accompanied by tremors and rockbursts, involving the multi-field coordinated evolution of stress, fractures, and energy. After failure, the overburden structure exhibits chain-like instability and energy release.

[0021] Traditional mine seismic monitoring mainly relies on methods such as microseismic signal monitoring, stress field analysis, and source location. However, when faced with complex geological conditions and the coupling effects of multiple factors, it suffers from problems such as low prediction accuracy and delayed early warning response. Therefore, this embodiment proposes a multi-scale sensing method for complex mine seismic networks and adaptive overburden control. Based on a complex network model, it senses mine seismic activity in multiple dimensions and at multiple scales, enabling dynamic prediction of mine seismic development. Simultaneously, based on the causes and mechanisms of mine seismic events, an adaptive overburden control strategy is adopted to achieve proactive prevention and control of mine seismic events.

[0022] like Figures 1 to 5 As shown, the multi-scale sensing and overburden adaptive control method for complex seismic networks includes: Step S101: The microseismic monitoring data and numerical simulation results of the coal pillar-key layer structure in the mine are fused at the spatiotemporal scale to construct a dynamic complex network model of the coal pillar-key layer structure, and the key areas for mine seismic propagation are identified by the node centrality of the dynamic complex network model.

[0023] In this embodiment, by collecting microseismic events, stress monitoring data, and numerical simulation results of the coal pillar-key layer structure, the energy accumulation, transfer, and release laws in the coal pillar-key layer coupled system are clarified, and a dynamic complex network model of the coal pillar-key layer structure is established. This is based on broadband microseismic monitoring technology (frequency... The microseismic waveforms and energy evolution of the entire process of rock strata instability at different scales were collected to clarify the dynamic instability process of the coal pillar-key layer structure under mining disturbance; and the mine geological conditions and coal pillar-key layer structural characteristics were obtained through fine geophysical exploration methods such as borehole columnar diagrams, fiber optic stress monitoring and ground-penetrating radar, so as to determine the spatial layout and property parameters of the coal pillar-key layer structure.

[0024] Numerical simulations of the coal pillar-key layer structure (such as macro- and micro-scale coupled simulations of the coal pillar-key layer using FLAC3D and PFC) are used to reproduce the fracture propagation and macro-destruction processes of the coal pillar-key layer structure under different mining conditions. The spatial and temporal characteristics of coordinated instability are presented intuitively, and the dynamic coupled instability of the coal pillar-key layer under different stress paths is explored. The stress-strain curves, energy release laws, and acoustic emission characteristics of rock samples are obtained. The entire process of fracture initiation, propagation, and rock mass fracturing during the coordinated instability of the coal pillar-key layer is analyzed, and the stress distribution characteristics, fracture propagation rate, and energy release laws are obtained.

[0025] Furthermore, microseismic monitoring data (event location, waveform characteristics, energy, moment tensor, etc.) are fused with numerical simulation results (stress distribution, fracture propagation path, energy release characteristics) at spatiotemporal scales. Specifically, through network node and attribute definition, key areas of the mine (coal pillars, key layers, support structures) are used as network nodes to construct a propagation path network. The dynamic evolution law of the coal pillar-key layer system and the chain effect of node state transformation are analyzed. A stress-fracture-energy multi-field dynamic complex network model of the coal pillar-key layer structure is established to reveal the inherent physical law of the coordinated instability chain effect, laying a data foundation for multi-scale dynamic perception and real-time early warning of mine earthquakes.

[0026] In this embodiment, the dynamic complex network model of the coal pillar-key layer structure includes a macro-scale overall mine network, a meso-scale key node network, and a micro-scale local fracture network. Thus, the dynamic complex network model of the coal pillar-key layer structure enables the accurate identification of mine seismic energy and stress propagation paths.

[0027] Specifically, a macro-scale model of the coal pillar-key layer structure is constructed based on the overall mine topology network to analyze the stress-energy transfer relationship between the coal pillar, key layer, and fault zone; a meso-scale model of the coal pillar-key layer structure is constructed based on the mine seismic activity area network to extract the spatiotemporal distribution characteristics of microseismic events and analyze stress concentration paths using network node connection relationships; and a micro-scale model of the coal pillar-key layer is constructed based on the local fracture network of the mine to evolve the fracture propagation path and energy release in the coal pillar-key layer structure.

[0028] In this embodiment, the macroscopic scale is the overall topological network of the mine. During the construction of the macroscopic scale model, the mine spatial region is discretized into multiple numerical simulation grid cells as network nodes of the model. These multiple numerical simulation grid cells are then connected based on the stress-energy transfer within the mine to construct the macroscopic scale model of the mine. In other words, the discretized numerical simulation grid cells of the mine spatial region are used as network nodes, and the stress-energy transfer efficiency between adjacent grid cells is used as the connection weight to connect multiple numerical simulation grid cells, thus constructing the macroscopic scale model of the mine. The model is constructed according to the following formula: Determine the grid cells in the macroscale model Its adjacent grid cells Stress-energy transfer efficiency In the formula, For grid cells in macro-scale models With grid cells stress gradient, This represents the maximum stress gradient within the mesh elements of the macroscopic model. They are grid cells With grid cells strain energy density, This represents the maximum strain energy density within the grid cells of the macroscopic-scale model. To adjust the exponent, the nonlinear effect of the strain energy density of the grid cells on the energy transfer efficiency is characterized; The attenuation factor characterizes the grid cell. With grid cells Impact on energy transfer efficiency; For grid cells With grid cells The physical distance between them The characteristic length is a scale that characterizes the range or transmission distance of the impact of seismic propagation (such as the width of the fracture zone).

[0029] Therefore, by discretizing the spatial region of the mine into multiple numerical simulation grid cells, and constructing a macroscopic mine model based on stress-energy transfer within the mine, the physical interactions between different regions are simulated through the connection between each grid cell. This enables accurate simulation and analysis of the dynamic changes in energy transfer and stress distribution of complex physical processes within the mine (such as the occurrence and evolution of mine tremors), effectively improving the research on the occurrence, development, and propagation mechanisms of mine tremors.

[0030] In this embodiment, the mesoscale model is a network of seismic activity regions, and the network nodes of the mesoscale model are the epicenters of microseismic events. Edges represent spatiotemporal correlations (connected by specified time intervals and spatial intervals, such as edges connecting areas with a time interval of less than 5 minutes and a distance of less than 20 meters). The connection weights are determined by the energy correlation of the microseismic events. Specifically, when constructing the mesoscale model, the seismic activity region network is divided into grids, and the epicenters of microseismic events are projected onto the grid model. The attributes of the corresponding grid nodes in this grid model are related to the frequency and energy of the microseismic event. The frequency of the microseismic event reflects the intensity of seismic activity in the corresponding area of ​​the grid node, and the energy of the microseismic event reflects the potential risk level of the corresponding area.

[0031] Specifically, the source points of mine microseismic events are projected onto the network nodes of a gridded network representing the mine's seismic activity area. The source points of microseismic events are then connected using the time and spatial intervals between the events to construct a mesoscale model of the mine's seismic activity area network. In other words, the source points of mine microseismic events are projected onto the gridded network nodes of the mine's seismic activity area network, and the corresponding network nodes are connected using the energy correlation of microseismic events at specified time and spatial intervals as connection weights, forming a mesoscale model of the mine's seismic activity area network within the coal pillar-key layer structure. The formula is as follows: Determining microseismic events in mesoscale models Microseismic events Energy correlation In the formula, Microseismic events Microseismic events The time interval, Microseismic events Microseismic events The spatial distance of the epicenter, Microseismic events Microseismic events The energy of microseismic events; These are the time decay constant and the spatial decay constant, respectively. This represents the maximum microseismic event energy in the mesoscale model.

[0032] The energy decay index is used to adjust the effect of energy on connection strength. In the dynamic complex network model of the coal pillar-critical layer structure, the energy decay index... This quantitatively characterizes the rate attenuation of seismic energy (in the form of elastic waves) as it propagates through rock mass media, with increasing distance. This rate directly relates to the physical reality of dynamic complex networks and the reliability of calculation results. Energy attenuation index. The larger the value, the faster the energy decay, indicating that the rock mass absorbs vibration energy strongly or has a complex propagation path (such as well-developed fractures); Energy decay index The smaller the value, the slower the energy decay, indicating that the energy can propagate further, the rock mass has good integrity, or it is in a dominant stress transmission channel.

[0033] The energy released by the point source of the mine earthquake travels a certain distance. The seismic energy at the receiving point (network node) for: In the formula, The source energy of the seismic point source in the mine; These are the inversion coefficients.

[0034] In a specific example, the energy decay index is determined using a calibration method based on multi-source data fusion. Specifically, this is achieved by acquiring historical microseismic monitoring data from known mines (with known source energy). and receiving point Energy received at the location ), for the energy decay index Inversion coefficients Inversion is performed to obtain the energy attenuation index in a specific propagation direction of the mine earthquake. Inversion coefficients The actual value.

[0035] Therefore, by projecting the frequency and energy of microseismic events into a grid model and dynamically adjusting the grid division according to the magnitude of mine seismic risk, a high-resolution network model of mine seismic activity area is established to accurately describe the spatial characteristics of mine seismic activity and provide effective decision-making for mine seismic prevention and control.

[0036] In this embodiment, the microscale is a local fracture network. In the mine, the fracture opening is affected by factors such as stress disturbance, pressure changes, and the influence of surrounding rock strata during the mining process. The stronger the fracture activity, the greater the fracture opening. Contact force chains exist within granular materials (such as rocks, coal blocks, and fillers in the mine) and are strong force paths formed by a large number of particles transmitting force through contact points. In the microscale model, particle contact points or fracture breakpoints are used as network nodes, and the connection weights between network nodes reflect the magnitude of the contact force or the degree of fracture opening between network nodes. The contact force between particle contact points or the fracture opening between fracture breakpoints can be calculated using the finite element method, discrete element method, or other numerical simulation methods.

[0037] Specifically, in the construction of the microscale model, particle contact points or fracture points are used as network nodes, and the contact force between particle contact points or the fracture opening between fracture points are used as connection weights. Particle contact points or fracture points are connected through contact force chains between particle contact points or fracture propagation paths between fracture points to construct a microscale model of the local fracture network in the mine. The model is constructed according to the formula: Determine network nodes in the microscale model With network nodes Connection weights between In the formula, These are preset weighting coefficients used to control the relative importance of contact forces between particle contact points or crack opening between crack fracture points. For network nodes With network nodes Contact force, This represents the maximum contact force between network nodes in the microscale model. For network nodes With network nodes Crack opening between This represents the maximum gap opening between network nodes in the microscale model.

[0038] Therefore, by constructing a multi-field, multi-scale dynamic complex network model for sensing mine seismic activity, multiple factors such as microseismic activity, stress, and crack propagation are integrated into the association of nodes and edges in the dynamic complex network model. Then, through multi-scale coupling analysis, a technical foundation and support are provided for real-time sensing of the evolution process of mine seismic activity.

[0039] In this embodiment, the spatiotemporal distribution characteristics of microseismic events are extracted using a mesoscale model within a dynamic complex network model. Centrality analysis is then performed on the network nodes of the dynamic complex network model. Based on nodes whose centrality index exceeds a preset threshold, key areas for seismic propagation are identified. Specifically, degree centrality analysis determines areas with frequent seismic activity and easy energy accumulation; betweenness centrality analysis identifies the main channels of seismic energy propagation, thus pinpointing areas controlling seismic expansion.

[0040] Wherein, according to the formula: Determine network nodes Degree centrality In the formula, To connect with network nodes Connected network nodes, To connect with network nodes The combination of connected network nodes, For network nodes With network nodes The weight of energy transfer between them. Specifically, according to the formula: Determine the receiving point (network node). and receiving point (network node) Energy transfer weight between In the formula, For receiving points (network nodes) The energy of the mine earthquake, For receiving points (network nodes) The energy of the mine earthquake, For receiving points (network nodes) and receiving point (network node) Distance between For receiving points (network nodes) To the receiving point (network node) The energy decay index in the direction of .

[0041] In this embodiment, the highly central region is the area where mine seismic events occur frequently (high-risk points). A Long Short-Time Memory (LSTM) network is used to analyze the historical data of microseismic energy release (mine seismic time series data) at the high-risk points to predict the time window in which mine seismic events may occur. Firstly, based on the network nodes... and network nodes Energy transfer weight between The model is dynamically adjusted according to the weights: For network nodes and network nodes Energy transfer weight between Make corrections to obtain network nodes. and network nodes Energy transfer correction weight between In the formula, For network nodes To network node The energy decay index in the direction of , The energy of the mine earthquake is generated by network nodes. Transmitted to network nodes The time interval.

[0042] Then, construct the network connection matrix of the dynamic complex network model. (matrix The elements are the energy transfer correction weights between network nodes. Next, historical data (mine seismic time series data) and the network connectivity matrix are used... The LSTM network is trained by inputting the time series data of mine tremors within the current time window of each network node. The trained LSTM network is then used to predict mine tremor events and the time window in which the mine tremor event may occur.

[0043] In this embodiment, betweenness centrality is used to analyze the main channels of seismic energy propagation, identify the regions controlling seismic expansion, and determine that regions or channels with high betweenness centrality are key hubs for seismic energy transfer. Specifically, according to the formula: In the formula, For nodes betweenness centrality, For network nodes To network node The total number of shortest paths, For network nodes To network node And through network nodes The number of shortest paths.

[0044] In this embodiment, the main occurrence area of ​​the seismic event is determined by degree centrality, and the propagation channel of the seismic energy is determined by betweenness centrality. Then, numerical simulation is used to simulate the seismic event, and the obtained simulated data (stress gradient) is compared with the real data collected by the seismic monitoring system to determine the simulation error. The model is then optimized and adjusted based on the simulation error. Finally, a convolutional neural network (CNN) model is used to extract the spectral features of the seismic event. This involves automatically extracting features (network topology features, temporal features of each node, etc.) from the data (historical microseismic data (time, location, energy), stress monitoring time-series data, mining progress data) through multiple convolutional layers, reducing data dimensionality through pooling layers, and performing regression prediction on the possible time window of the seismic event through fully connected layers.

[0045] In this embodiment, degree centrality and betweenness centrality are used to quantify the importance of network nodes. Based on the analysis of key pivots (percolation probability) in the percolation network model, hidden high-risk nodes in the network are identified, and the critical conditions for the occurrence of mine-induced earthquakes are analyzed based on network evolution. Specifically, the rock mass is abstracted as a network, with "nodes" representing rock blocks or regions and "edges" representing fractures or weak surfaces, thus constructing a percolation network model. Under mine-induced earthquake conditions, as stress loading or damage accumulation occurs (the probability of an "edge" being occupied increases), the percolation probability increases. (Increase), in percolation network models, connected clusters may form. When the probability... Reaching the percolation threshold When the permeation network model is in progress, a huge connected cluster will appear, indicating that the rock mass has undergone macroscopic through-through failure. Through permeation network model simulation, the key edge or node combination that contributes the most to the overall connectivity of the rock mass can be identified when it is close to the adjacent state (macroscopic through-through failure).

[0046] Step S102: Adopting a bidirectional collaborative optimization strategy that combines graph theory-based centrality analysis with machine learning prediction, the dynamic complex network model of the coal pillar-key layer structure is dynamically updated using historical data on microseismic energy release in key areas. Based on the optimization update results, stress criteria and energy release indicators are dynamically generated to provide early warning of the collaborative instability state of the coal pillar-key layer structure.

[0047] Mine network models are affected by variational grids and discontinuous media (faults). For example, the stress transfer efficiency changes due to grid geometric deformation caused by dynamic changes in the grid during simulation due to mining activities, rock mass deformation, or numerical optimization requirements; geological structures such as faults and fractures cause abrupt changes in the mechanical properties of the medium. To address this, this embodiment summarizes and employs a bidirectional collaborative optimization strategy combining graph theory-based centrality analysis and machine learning prediction to dynamically update the dynamic complex network model of the coal pillar-key layer structure.

[0048] On the one hand, mesh correction is performed on the dynamic complex network model of the coal pillar-key layer structure through mesh updates, random perturbation parameters, and reverse optimization. Based on changes in the physical field (e.g., stress or strain fields), the mesh is dynamically adjusted using numerical simulation methods to achieve mesh updates. For example, the mesh can be further refined in stress concentration areas of the coal pillar or key layer to more accurately capture local phenomena. By refining or coarsening the mesh of the dynamic complex network model of the coal pillar-key layer structure, the model's mesh resolution can be changed to adapt to different scale requirements. Simultaneously, random perturbation parameters are introduced during mesh generation to simulate the impact of uncertainties on the mine structure, such as the inhomogeneity of rock strata and differences in coal seam hardness, to further enhance the adaptability and robustness of the simulation.

[0049] Furthermore, a reverse analysis method based on numerical solutions is used to analyze the gap between the results and the target under the current model, and then adjust the mesh division to meet the requirements of higher accuracy. In this way, by reverse optimizing the existing model and adjusting the mesh division method, more accurate dynamic changes can be obtained in different working faces, coal pillars, or key layer areas.

[0050] Specifically, when the stress gradient of a unit in the dynamic complex network model of a coal pillar-key layer structure exceeds a preset gradient threshold, an adaptive finite element method based on error estimation inserts new network nodes into the dynamic complex network model. In other words, when the stress gradient of a mesh unit in the dynamic complex network model of a coal pillar-key layer structure exceeds the preset gradient threshold, it indicates that there is significant stress concentration or variation in these areas. The adaptive finite element method based on error estimation identifies these high-stress variation areas and inserts new network nodes within these areas. That is, by calculating the stress gradient of the mesh unit, the stress concentration areas are accurately located, and the mesh in the stress concentration areas is refined (by inserting new network nodes) based on error estimation, improving the accuracy and stability of the model; by inserting new network nodes, the accuracy of high-stress areas is further improved, enhancing the simulation of mine seismic activity.

[0051] In the dynamic complex network model of the coal pillar-key layer structure, some regions exhibit low stress due to small stress changes. By merging adjacent network nodes in the low stress variation region of the dynamic complex network model of the coal pillar-key layer structure, the computational complexity is reduced, the computational efficiency is effectively improved, and redundant calculations are avoided. At the same time, it is ensured that the model can effectively reflect the dynamic behavior of the coal pillar-key layer structure without affecting the accuracy.

[0052] Specifically, in the dynamic complex network model of the coal pillar-critical layer structure, a stress change threshold is set (e.g., Stress variation areas exceeding the stress variation threshold are classified as high-stress areas, while stress variation areas below the threshold are classified as low-stress areas. In a specific example, stress field distribution can be extracted based on FLAC3D / PFC simulation results, and stress variation areas can be divided into high-stress and low-stress areas; alternatively, multi-source monitoring data can be obtained through microseismic system monitoring, and clustering algorithms (such as K-means, DBSCAN) can be used to spatially distance the multi-source monitoring data to automatically identify stress zones.

[0053] During the merging of adjacent network nodes in low stress change regions, nodes with stress change rates less than the stress change threshold and distances less than a preset value are selected. Network nodes with similar attributes (such as similar energy and frequency domain characteristics) are merged using graph shrinkage or hierarchical clustering. The network topology before and after the merger is compared to ensure that the critical path and central nodes remain unchanged.

[0054] To further optimize the dynamic complex network model of the coal pillar-critical layer structure, based on historical microseismic event catalogs and stress monitoring data, the node positions and connections in the dynamic complex network model of the coal pillar-critical layer structure are adjusted using gradient descent or genetic algorithms. Historical microseismic events provide a reference for model adjustment (especially when identifying and optimizing high-risk seismic areas); real-time stress monitoring data from the mine provides an understanding of the current stress state of the coal pillar-critical layer structure and a basis for optimizing stress concentration areas.

[0055] Gradient descent or genetic algorithms are used to enable the model to continuously adjust node positions and connections during iteration, allowing it to accurately capture the dynamic changes in actual stress and mine seismic events during its evolution. Gradient descent is used to find the optimal node positions and connections to reduce model output errors; genetic algorithms are used to simulate natural selection processes to select the optimal solution from multiple possible network structures, enhancing the model's adaptability and accuracy.

[0056] On the other hand, based on the microseismic activity of mine faults, the parameters of a dynamic complex network model are dynamically adjusted using a time-series graph attention network. By introducing an attention mechanism into the time-series graph attention network, the changes in the dynamic complex network model in the time dimension are dynamically captured. By assigning different weights to different nodes (coal pillars or key layer parameters) at different times, the focus on time nodes is strengthened, thereby optimizing the dynamic characteristics of the model.

[0057] The microseismic activity in a mine reflects the activity state of the mine fault. By inputting the microseismic activity in the mine as time series data into a time series graph attention network, and combining the adaptive learning capability of the time series graph attention network, the network can automatically learn the potential evolution law of the mine structure through the analysis of the time series data, and optimize the model parameters (such as mesh accuracy, mesh generation, perturbation parameters, etc.) in real time. Furthermore, by analyzing the microseismic signals at the mine fault, the occurrence law and source characteristics of mine earthquakes can be explored.

[0058] Data on microseismic activity in mines includes: mine microseismic data, mine stress state and energy distribution data, time-series data and vibration modes within time windows, mine geological and engineering parameters, and mine field monitoring feedback data. Among these, mine microseismic data includes source frequency, source location, magnitude, and energy release. By monitoring the frequency of microseismic events in the mine (especially in areas near coal pillars or critical layers), i.e., the frequency of occurrence of microseismic events (source frequency), stress changes or structural instability risks in the mine can be effectively predicted. Source location data obtained through microseismic location technology allows understanding of stress distribution and seismic activity in different areas within the mine, effectively identifying potential seismic induction zones and facilitating further adjustment of model parameters. The magnitude and energy of microseismic events determine regional stress changes and stability in the mine, effectively predicting the risk of local structural instability or stress concentration.

[0059] Mine stress state and energy distribution data include stress field data, energy accumulation, and release. By monitoring the stress state of the mine (e.g., stress distribution in coal pillars and key layers), information about the stability of coal pillars and key layers is provided to the model to effectively capture potential instability areas, thereby dynamically adjusting model parameters such as mesh generation and mesh accuracy. By recording and analyzing the transmission and release processes of stress and energy in mine seismic activity, attention mechanisms are used to dynamically capture and optimize the conditions for the occurrence of seismic events, and model parameters are adjusted to more accurately predict seismic risks.

[0060] Time-series data is used by the time-series graph attention network to capture the temporal evolution of mine microseismic events. For example, microseismic activity in certain areas may exhibit periodicity or fluctuation, and time-series data can effectively reveal the dynamic changes in mine stress. By analyzing vibration patterns within an event window (e.g., the fluctuation of source activity within a characteristic time period), the time-series graph attention network can learn the occurrence patterns of mine seismic events, further optimizing and adjusting the dynamic parameters of the model.

[0061] The geological and engineering parameters of a mine include coal pillar width, critical layer thickness, and surrounding rock properties. Among these, coal pillar width and critical layer thickness directly affect mine stability. Monitoring these geological and engineering parameters can effectively determine the likelihood of mine-induced seismic events. By incorporating surrounding rock properties (such as strength and elastic modulus) into the model and using an attention mechanism to adjust the model's regional divisions and parameters, the influence of the physical properties of the surrounding rock on the occurrence of mine-induced seismic events can be further demonstrated.

[0062] The on-site monitoring feedback data mainly consists of sensor data deployed in the mine, including real-time feedback data from various mine sensors (such as stress sensors, accelerometers, and micro-seismic sensors). The attention mechanism in the time-series graph attention network adaptively adjusts based on the on-site monitoring feedback data to effectively enhance the model's dynamic response capability.

[0063] Based on the data of microseismic activity, the model parameters of the time series graph attention network (such as mesh generation parameters, structural parameters, etc.) are adjusted in real time. For example, based on the characteristics of vibration intensity, source location, source frequency and other features in different regions, the calculation parameters in the model (such as mesh refinement degree, disturbance parameters, etc.) are adjusted to reflect the stress changes and dynamic effects of the internal structure of the mine, and to optimize the multi-scale coupling of the coal pillar-key layer structure in a timely manner.

[0064] In the process of dynamically adjusting the parameters of a dynamic complex network model based on a time-series graph attention network, the node embedding of the dynamic complex network model is dynamically updated through structural attention and temporal attention. Specifically, the embedding vector of each network node is updated by calculating the relative importance of each network node relative to its neighboring network nodes; different weights are assigned to data at different time steps based on temporal information (the time interval of microseismic events), and the attention mechanism is used to identify the correlation importance between each time step and the seismic prediction.

[0065] In the process of dynamically adjusting the parameters of a dynamic complex network model, during fault simulation, the connection weights of network nodes in the model are dynamically adjusted using a time-series graph attention network based on changes in velocity distribution. Specifically, a fault model incorporating velocity variation information is constructed based on the geological fault characteristics of the mine. As time progresses, the stress and velocity of the fault in the mine change, affecting the induction and propagation path of mine tremors. Based on changes in velocity distribution, the time-series graph attention network dynamically adjusts the connection weights, updating the stress, energy, or vibration transmission intensity between different nodes in the mine, thereby improving the accuracy of mine tremor prediction.

[0066] In the process of dynamically adjusting the parameters of a dynamic complex network model, a variational method is used to perform inverse optimization of the mesh generation, and gradient descent is used to adjust the transition zone positions of the mesh. In dynamic complex network models, mesh generation accuracy determines the model's analytical capability. Transition zones are typically areas with significant stress gradient changes in mines, requiring higher mesh density. By using a variational method to inversely optimize the mesh generation, adjusting the mesh size and distribution, computational resources can be concentrated on areas requiring fine analysis, especially in seismically active zones and near faults. Gradient descent is then used to gradually adjust the boundary positions of the transition zones, enabling the model to more accurately describe the stress field and vibration propagation paths.

[0067] During the dynamic adjustment of parameters of the dynamic complex network model, the grid density distribution of the model is optimized through the attention weight feedback of the time-series graph attention network. Specifically, by introducing the time-series graph attention mechanism, the model can identify the importance of different time periods and regions for mine seismic prediction, and dynamically adjust the grid density distribution based on the attention weight feedback. This ensures better resolution in important areas (such as areas with frequent mine seismic events or stress concentration areas), thereby balancing computational accuracy and resources and effectively reducing computational overhead.

[0068] In the process of dynamically adjusting the parameters of the dynamic complex network model, the boundary nodes of the macroscopic network in the dynamic complex network model are associated with the fracture discontinuities of the microscopic network in the dynamic complex network model. Then, graph embedding is used to map the multi-scale networks to the same feature space, generating a unified network topology. Specifically, after identifying the boundary nodes in the macroscopic network (e.g., the main structural nodes of a mine) and the fracture discontinuities in the microscopic network (e.g., cracks or faults in a mine), a pre-defined mapping rule is used to establish the association between the macroscopic and microscopic networks. Graph embedding is then used to preserve the topological relationships in the network structure, mapping the macroscopic and microscopic networks to the same feature space. This allows the model to comprehensively consider the characteristics of the mine at different scales, reflecting both the macroscopic and microscopic features of the mine, providing a global perspective for mine tremor prediction and control.

[0069] Therefore, the dynamic updates of node embedding provide the model with a more accurate mine network structure, and the dynamic adjustment of connection weights further strengthens the interrelationships between nodes in the mine, making the model more adaptable to the actual occurrence patterns of mine tremors. Mesh optimization and transition zone adjustment effectively ensure the accurate representation of complex mine structures, improving the computational efficiency and accuracy of the model. Graph embedding operations integrate networks of different scales to achieve a unified model perspective and improve the overall performance of dynamic complex network models.

[0070] By combining time-series graph attention networks and variational methods, a dynamic complex network model of the coal pillar-critical layer structure can be effectively optimized. The model parameters, grid division, and node connections can be dynamically adjusted to improve the accuracy and efficiency of mine tremor prediction and prevention. Through precise grid adjustment, optimized stress and energy transfer, and unified network topology, the complex dynamic characteristics of the mine can be better reflected, providing technical support for mine safety.

[0071] In this embodiment, the fault zone is a natural crack or fracture zone in the rock strata. Its stress distribution and transmission are unique, playing a decisive role in the occurrence of mine tremors and becoming the path for energy release and tremor propagation. Especially under mining disturbances, fault zone activity often causes strong mine tremors. The interaction between coal pillars and key layers directly affects mine stability, and stress transmission relationships are reflected through mechanisms such as coal pillar support and key layer rupture. As a critical path for stress transmission, the existence and activity of the fault zone directly affect the propagation and energy release of mine tremors. In this embodiment, in the dynamic complex network model, the stress concentration path is the transmission path of stress and energy generated within the mine due to mining activities, coal pillar instability, and rock strata deformation. Stress concentration is the path through which stress accumulated, concentrated, and propagated along channels in space due to the effects of coal pillars, key layers, or fault zones.

[0072] In a mine, the primary function of a coal pillar is to support the surrounding rock and withstand the pressure during mining. Mining activities can lead to the fracturing or local instability of the coal pillar, thereby altering the stress distribution between the coal pillar and the overlying strata (such as critical layers). In this case, stress will concentrate along the contact surface or fracture surface between the coal pillar and the critical layer. This path represents the interaction region between the coal pillar and the critical layer, or the stress transmission path after the coal pillar fracture. Common fracture zones in mines are naturally occurring cracks in the rock strata. When mining activities affect these fracture zones, local stress will propagate along them. Especially when the fracture zone is active, it may become a critical channel for energy release. In this case, the stress concentration path is the channel from the coal seam (or critical layer) to the fracture zone. Therefore, based on the macroscopic scale model in a dynamic complex network model, the stress-energy transmission relationship between coal pillars, critical layers, and fracture zones in a mine is determined, mainly including: the stress concentration path between the coal pillar and the critical layer, and the stress concentration path between the coal seam and the fracture zone.

[0073] In this embodiment, under the background of mine earthquake, the source point is usually the stress release point on the coal seam or fault zone. When a local area (such as a coal seam or key layer) ruptures or slips, energy will spread rapidly to the surrounding area and may form stress concentration in adjacent areas. This path is the stress transfer from the source point of the mine earthquake to other areas of the mine (such as nearby coal seams, faults, etc.).

[0074] In the overall outcome of a mine, stress concentration paths involve not only coal pillars, key layers, and fracture zones, but also the interaction of surrounding rock and support structures. Mine mining leads to stress transfer and concentration between different areas. These paths connect multiple important structures in the mine through its structural hierarchy. The specific locations of stress concentration may be fracture zones at the mine depth, the contact interface between the ore layer and the surrounding rock, and stress transfer channels between coal pillars and key layers.

[0075] To address this, a mesoscale model within a dynamic complex network model is used to extract the spatiotemporal distribution characteristics of mine microseismic events. Stress concentration paths are then determined through the connectivity of network nodes within the mesoscale model. Numerical simulation methods, such as the finite element method and the discrete element method, can be employed to simulate the stress distribution in structures like coal pillars, key layers, and fault zones within the mine. This allows for the identification of stress concentration regions and specific stress concentration paths. Furthermore, the analysis of stress gradients and local deformations identifies the most significant stress concentration pathways.

[0076] In addition, actual stress data and microseismic event data can be obtained through mine microseismic monitoring systems, stress sensors, and geological surveys. By analyzing the stress release information at different depths and in different areas of the mine, spatiotemporal distribution analysis can be performed to identify stress concentration areas and extract stress concentration paths. Furthermore, the different mechanical properties of different rock strata and structures in the mine, such as the differences in the mechanical properties of coal seams, fault zones, and key layers, can affect the stress distribution in these areas, allowing for the prediction of stress concentration paths based on these differences in mechanical properties.

[0077] In this way, stress concentration paths are characterized to represent the path of stress transmission from one region (such as a coal seam or fault zone) to another region (such as a coal pillar, surrounding rock, or other faults). Stress concentration paths are determined by methods such as numerical simulation, monitoring data analysis, and rock strata mechanical properties. The interaction between different structures in the mine, such as the contact surface between the coal pillar and the key layer, and the crack surface between the coal seam and the fault zone, are described.

[0078] In this embodiment, the microscale model primarily focuses on the microstructure within the mine, such as coal seams, fractures, and particle contact points. It analyzes the impact of changes in this microstructure on the occurrence and propagation of mine tremors. Furthermore, it analyzes the microscale model through numerical simulation and microseismic monitoring to determine the occurrence and evolution of mine tremor events. Specifically, based on the microscale model within a dynamic complex network model, and combined with numerical simulation and mine microseismic monitoring data, it predicts fracture propagation paths and their energy release characteristics.

[0079] On the one hand, in the microscale model, the discrete element method (DEM) is used to simulate particle contact points, fracture propagation, and mechanical behavior in the mine. By simulating the contact force chains between particles and the propagation path of fractures, the changes in microstructure under mining activities or external disturbances are analyzed. The finite element method is used to calculate the impact of microscale behaviors such as stress concentration and fracture propagation on overall mine seismic activity. In this way, numerical simulations can map microscale mechanical behaviors (such as fracture opening, contact force, and stress distribution) onto a macroscale model, predicting the overall stress state of the mine and potential seismic zones.

[0080] On the other hand, microseismic monitoring systems (such as microseismic sensors deployed in the mine) collect microseismic event data in the mine. By analyzing the temporal information and spatial distribution of microseismic events, the inducing source, propagation path, and focal intensity of mine tremors can be identified. Combining microseismic monitoring data with mechanical simulations, the influence of key microstructures in the mine (such as cracks and contact points) on microseismic events can be identified. By analyzing the location and frequency of microseismic events, changes in the microstructure within the mine, such as crack opening and coal pillar rupture, can be inferred.

[0081] In addition, numerical simulations can be calibrated using microseismic monitoring data to ensure that the simulation results match actual mine seismic events. By comparing the simulated stress changes, crack propagation, and the occurrence of microseismic events, the parameters in the numerical simulation can be adjusted to improve the model's prediction accuracy.

[0082] When combining numerical simulation and microseismic monitoring data in a microscale model, firstly, a microstructural model of the mine is established through data simulation, including the mechanical properties of the coal seam, the distribution of fractures, and the interaction between particles. Numerical simulation is then performed using the microstructural model to predict the stress, strain, and mechanical behavior of each micro-unit in the mine. Simultaneously, microseismic sensors are deployed in the mine to collect spatiotemporal distribution data of microseismic events, in order to determine the spatial location, magnitude, and time series of microseismic sources, reflecting information such as the dynamic changes within the mine.

[0083] Then, the collected microseismic monitoring data is compared and calibrated with the numerical simulation results. By analyzing the spatiotemporal distribution information of microseismic events, the relationship between microseismic events and specific microstructures (such as cracks and contact forces) is inferred. The prediction accuracy of the numerical simulation is verified using microseismic monitoring data; for example, deviations between the numerical simulation results and the microseismic monitoring data are analyzed. Relevant parameters in the numerical simulation (such as mechanical parameters, material parameters, and contact forces) are adjusted to improve the model's accuracy and practicality. Finally, real-time monitoring of mine microseismic activity is conducted, and the microseismic monitoring data and numerical model are dynamically updated interactively to predict changes in the mine's microstructure, achieving single-unit early warning and prevention of mine seismic risks.

[0084] This microscale model combines microseismic monitoring data with numerical simulation. Microseismic monitoring data provides dynamic characteristics of stress release in the mine, while numerical simulation predicts potential stress concentration areas, fracture propagation paths, and other mechanical behaviors, leading to a comprehensive understanding of the microscopic changes within the mine. The spatiotemporal distribution characteristics of microseismic events reveal the evolution of the mine's microstructure; for example, frequent microseismic events indicate fracture propagation or coal seam rupture in certain areas, while numerical simulation can predict stress and deformation in these areas. After predicting the mine's microstructure through numerical simulation, the prediction results are validated using microseismic monitoring data, and the numerical model is corrected and optimized, enabling real-time monitoring and risk prediction of mine microseismic activity.

[0085] Therefore, by combining microseismic monitoring data with numerical simulation in a microscale model, the microstructure of the mine can be effectively monitored and analyzed dynamically. The spatiotemporal distribution of actual mine seismic activity can be revealed by using microseismic monitoring data, and the stress, deformation and crack propagation of the microstructure in the mine can be predicted and analyzed by using numerical simulation. Through this combination, the mechanism of mine seismic occurrence can be revealed, providing real-time monitoring and early warning support for safe production in the mine.

[0086] In this embodiment, the stress state of the entire mine is captured by a macroscopic model, especially the stress transmission between coal pillars, key layers and major fault zones. By utilizing the stress and energy exchange between different parts of the mine (such as coal pillars and fault zones), potential risk areas in the mine are obtained, and the stress-energy transmission relationship between coal pillars, key layers and major fault zones is determined. In particular, when coal pillars or key layers encounter stress concentration, physical basis is provided for the spatiotemporal distribution characteristics of microseismic events and the extraction of stress concentration paths.

[0087] By analyzing the temporal and spatial characteristics of microseismic events in mines using mesoscale models, particularly the location, frequency, and evolution of these events, and utilizing the connectivity of network nodes, stress concentration paths within the mines can be deduced. This reveals areas in the mines that are prone to stress release (such as specific cracks or faults), providing guidance for refining crack propagation paths and identifying energy release characteristics.

[0088] By combining numerical simulations and microseismic monitoring data in a microscale model, the propagation behavior of fractures and their impact on mine structures are simulated to predict the occurrence of mine tremors. Ultimately, a complete mine tremor prediction and control model is formed, encompassing macroscopic stress transmission, mesoscopic microseismic activity, and microscopic fracture propagation. Through iterative updates, early warning and prevention strategies for mine tremors are continuously optimized.

[0089] Step S103: In response to the stress criterion of coordinated instability reaching the warning threshold, seismic suppression control is carried out in key areas. Based on the microseismic monitoring data after seismic suppression control, the dynamic complex network model and the stress criterion of coordinated instability are dynamically updated to verify the effect of seismic suppression control and optimize the subsequent seismic suppression control scheme.

[0090] In this embodiment, the propagation path of mine vibration is analyzed using the network centrality (degree centrality, betweenness centrality) of a dynamic complex network model, high-risk nodes of mine vibration are identified, and a network topology for mine vibration energy transfer is established. Combined with numerical simulation results, the core areas of mine vibration energy accumulation and propagation are analyzed. Numerical simulation analysis is used to identify high-stress areas under different mining conditions, and the dynamic complex network model is dynamically updated and optimized to improve the identification accuracy of key nodes. Furthermore, by combining mine vibration monitoring, numerical simulation, and the network model, the stress and microseismic event change characteristics of mine vibration risk areas are extracted. Automatic risk warnings are then provided for the coordinated instability state of the coal pillar-key layer structure using stress criteria and energy release indicators.

[0091] When the stress criterion for coordinated instability reaches the warning threshold, a seismic control scheme is generated and implemented for key areas. This includes measures such as hydraulic fracturing, pressure relief drilling, blasting for pressure relief, artificial backfilling, and adjustments to mining sequence plans in key areas. The impact of different measures on the propagation path of mine tremors and the control effect on the stress field are compared, and the mechanisms by which different control measures affect the stress field, fracture propagation, and energy release are analyzed.

[0092] By combining numerical simulation and field monitoring, this study verifies the stress release effect in key areas of the mine by implementing seismic control measures. It analyzes the impact of key nodes—stress release (pressure relief drilling), fracture guidance (active fracturing), and mining parameters (mining speed)—on the seismic propagation path, thereby optimizing seismic risk control strategies. Furthermore, by adjusting microseismic activity changes after implementing different control measures, the study compares and analyzes the seismic suppression effects under different conditions. Based on seismic monitoring data, a quantitative evaluation system of "key node energy accumulation - stress control - seismic damage suppression" is established, resulting in precise seismic suppression.

[0093] The implementation of various seismic control measures affects high-risk critical nodes in a dynamic complex network model, altering seismic propagation paths, weakening energy accumulation areas, and optimizing stress distribution in the mine. By analyzing the scope and magnitude of the impact of each seismic control measure and its quantitative effect on the network structure, a quantitative relationship is established with existing microseismic monitoring results. Specifically, after the implementation of pressure relief drilling, the adjacency matrix in the dynamic complex network model is analyzed. The stress weights of high betweenness centrality nodes are updated; after hydraulic fracturing is performed, attenuation coefficients are applied to the connecting edges of the hydraulic fracturing region based on the degree of stress release after fracture guidance.

[0094] In this embodiment, a control and optimization architecture based on monitoring data is established, driven by real-time feedback from microseismic monitoring data. This enables dynamic adjustment of mine tremor prevention and control measures, forming a three-in-one dynamic early warning system for mine tremors: monitoring, analysis, and feedback. Dynamic updates of a dynamic complex network model are used to optimize the control of key nodes, improving the intelligence level of mine tremor disaster prevention and control. Ultimately, based on a dynamic complex network model of multi-scale mine tremors, adaptive control of mine seismic suppression is achieved. Based on field-measured stress and microseismic monitoring data, the applicability of different control measures is analyzed, and the seismic suppression control scheme is optimized. Furthermore, combining field mine tremor monitoring, numerical simulation, and the dynamic complex network model, the changes in stress and microseismic events in mine tremor risk areas are analyzed, updating the mine tremor risk assessment, optimizing the seismic suppression control strategy, and establishing a real-time early warning and adaptive control feedback system for mine tremor risk. This forms a closed-loop system from monitoring data to control strategy to feedback optimization, achieving proactive mine tremor prevention and control.

[0095] In the description of this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0096] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for multi-scale sensing and overburden adaptive control of complex mining and seismic networks, characterized in that, include: The microseismic monitoring data and numerical simulation results of the coal pillar-key layer structure in the mine are fused in a spatiotemporal scale to construct a dynamic complex network model of the coal pillar-key layer structure. The key areas for mine seismic propagation are identified by the node centrality of the dynamic complex network model. A bidirectional collaborative optimization strategy combining graph theory-based centrality analysis and machine learning prediction is adopted. By using historical data on microseismic energy release in key areas, the dynamic complex network model of the coal pillar-key layer structure is dynamically updated. Based on the optimization update results, stress criteria and energy release indicators are dynamically generated to provide early warning of the collaborative instability state of the coal pillar-key layer structure. In response to the stress criterion for coordinated instability reaching the early warning threshold, seismic suppression and control are implemented in key areas. Based on the microseismic monitoring data after seismic suppression and control, the dynamic complex network model and the stress criterion for coordinated instability are dynamically updated to verify the effect of seismic suppression and control and optimize subsequent seismic suppression and control schemes.

2. The method according to claim 1, characterized in that, The dynamic complex network model of the coal pillar-critical layer structure includes: A macroscopic model of the mine is constructed by using multiple numerical simulation grid cells, which are discretized into a spatial region of the mine, as network nodes and connecting them based on the stress-energy transfer within the mine. Wherein, according to the formula: Determine the grid cells in the macroscale model Its adjacent grid cells Stress-energy transfer efficiency ; In the formula, For grid cells in macro-scale models With grid cells stress gradient, This represents the maximum stress gradient within the mesh elements of the macroscopic model. They are grid cells With grid cells strain energy density, This represents the maximum strain energy density within the grid cells of the macroscopic model. To adjust the exponent, the nonlinear effect of the strain energy density of the grid cells on the energy transfer efficiency is characterized; The attenuation factor characterizes the grid cell. With grid cells Impact on energy transfer efficiency; For grid cells With grid cells The physical distance between them The characteristic length represents the scale of the influence range or transmission distance.

3. The method according to claim 1, characterized in that, The dynamic complex network model of the coal pillar-critical layer structure also includes: The source points of mine microseismic events are projected onto the network nodes of the gridded network of mine seismic activity areas, and the source points of microseismic events are connected by the time interval and spatial interval of the microseismic events to construct a mesoscale model of the mine seismic activity area network. Wherein, according to the formula: Determining microseismic events in mesoscale models Microseismic events Energy correlation ; In the formula, Microseismic events Microseismic events The time interval, Microseismic events Microseismic events The spatial distance of the epicenter, Microseismic events Microseismic events The energy of microseismic events; These are the time decay constant and the spatial decay constant, respectively. This represents the maximum microseismic event energy in the mesoscale model. This is the energy decay index, used to adjust the effect of energy on connection strength.

4. The method according to claim 1, characterized in that, The dynamic complex network model of the coal pillar-critical layer structure also includes: A microscale model of a local fracture network in a mine is constructed by using particle contact points or fracture breakpoints as network nodes, and the contact force between particle contact points or the fracture opening between fracture breakpoints as connection weights. The particle contact points or fracture breakpoints are connected through contact force chains between particle contact points or fracture expansion paths between fracture breakpoints.

5. The method according to claim 1, characterized in that, Calculate the degree centrality and / or betweenness centrality of network nodes in a dynamic complex network model to determine the propagation path and key areas of mine tremors.

6. The method according to claim 1, characterized in that, Dynamic updates are performed on the dynamic complex network model of the coal pillar-critical layer structure, including: The dynamic complex network model of the coal pillar-key layer structure is corrected by mesh update, random perturbation parameters, and inverse optimization. Based on the microseismic activity of mine faults, the parameters of a dynamic complex network model are dynamically adjusted using a time-series graph attention network.

7. The method according to claim 6, characterized in that, The dynamic complex network model of the coal pillar-key layer structure is mesh-corrected through mesh updates, random perturbation parameters, and inverse optimization, including: In response to the stress gradient of an element in a dynamic complex network model of a coal pillar-key layer structure exceeding a preset gradient threshold, an adaptive finite element method based on error estimation is used to insert new network nodes into the dynamic complex network model. In the dynamic complex network model of coal pillar-critical layer structure, adjacent network nodes in the low-stress variation region are merged. Based on historical microseismic event catalogs and stress monitoring data, the node positions and connections of network nodes in the dynamic complex network model of the coal pillar-key layer structure are adjusted using gradient descent or genetic algorithms.

8. The method according to claim 6, characterized in that, Based on the microseismic activity of mine faults, the parameters of a dynamic complex network model are dynamically adjusted using a time-series graph attention network, including: Dynamic updates of node embeddings in dynamic complex network models are achieved through structural attention and temporal attention. In fault simulation, the connection weights of network nodes in a dynamic complex network model are dynamically adjusted based on changes in velocity distribution using a time-series graph attention network. The mesh partitioning of a dynamic complex network model is optimized in reverse based on the variational method, and the transition zone position of the mesh partitioning of the dynamic complex network model is adjusted by gradient descent. The grid density distribution of a dynamic complex network model is optimized by using the attention weight feedback of a temporal graph attention network. The boundary nodes of the macroscopic network in the dynamic complex network model are associated with the breakpoints of the microscopic network in the dynamic complex network model, and the multi-scale network is mapped to the same feature space through graph embedding operation to generate a unified network topology.

9. The method according to claim 1, characterized in that, Dynamically generate stress criteria and energy release indicators for early warning of the coal pillar-critical layer structure entering a coordinated instability state, including: Based on the macro-scale model in the dynamic complex network model, the stress-energy transfer relationship between coal pillars, key layers and fault zones in the mine is determined; Based on the mesoscale model in the dynamic complex network model, the spatiotemporal distribution characteristics of mine microseismic events are extracted, and the stress concentration path is determined by the connection relationship of network nodes in the mesoscale model. Based on the microscale model in the dynamic complex network model, and combined with numerical simulation and microseismic monitoring data of the mine, the fracture propagation path and its energy release characteristics are predicted.

10. The method according to claim 1, characterized in that, In response to the stress criterion of coordinated instability reaching the early warning threshold, a seismic control scheme is generated and implemented for the key area, including at least one of hydraulic fracturing, pressure relief drilling, or adjustment of the mining sequence plan in the key area.

Images