A mine rock burst prevention and control method, medium, device and product

Abstract

This application discloses a method, medium, equipment, and product for preventing and controlling rockbursts in mines. The method includes: acquiring multi-source data to characterize the geological conditions of the mine; constructing a three-dimensional voxel grid including multiple voxels along the three-dimensional space of the mine based on the multi-source data; inverting the structural characteristic index for characterizing coal body structural elements, the physical property characteristic index for characterizing coal body physical properties, and the stress characteristic index for characterizing stress field elements for each voxel based on the multi-source data, structural characteristic index, physical property characteristic index, and stress characteristic index; and generating a prevention and control strategy for rockbursts in mines based on the multi-source data, structural characteristic index, physical property characteristic index, and stress characteristic index. This application unifies multi-source data into the same space through a three-dimensional voxel grid, overcoming the deficiency of static zoning in reflecting detailed changes, and generates differentiated prevention and control strategies based on the collaborative identification of the three elements, providing a design basis for the prevention and control of rockbursts in mines.

Description

Technical Field

[0001] This application belongs to the field of coal mine safety mining and mine dynamic disaster prevention and control, specifically involving a method for preventing and controlling mine rockburst, a storage medium, electronic equipment and computer program products. Background Technology

[0002] Rockbursts are a typical type of dynamic disaster characterized by the instantaneous, violent, and non-steady release of elastic strain energy accumulated within the coal and rock mass during mining, induced by external disturbances. Their development and occurrence involve the coupling effects of multiple levels, including geological structure, rock mechanics, and engineering disturbances, exhibiting significant suddenness, nonlinearity, and strong destructiveness. As the focus of coal mining in my country shifts to deeper, ultra-deeper coalfields and structurally complex areas, the occurrence of rockbursts in newly built mines during coal seam exposure, initial mining, and even the excavation of infrastructure roadways is increasingly common, causing serious consequences such as roadway damage, equipment destruction, casualties, and production interruptions. How to accurately, proactively, and dynamically identify rockburst hazards in newly built mines, given the lack of long-term mining data, is a critical scientific and engineering problem urgently needing to be solved in the field of mine safety production.

[0003] From the perspective of the physical mechanism of rockburst disaster chain, its gestation and occurrence process is actually governed by three types of interrelated and mutually conditional factors. First, the structural elements of the coal seam include the geometric and structural features at both macroscopic and microscopic scales, such as bedding, joints, primary and tectonic fractures, interbedded rock, faults, folds, scour zones, abrupt changes in coal thickness, and the combination of roof and floor strata. These structural elements fundamentally determine the geometric boundaries of the energy reservoir, the dominant energy transfer paths, and the geometric morphology of the failure surface. Second, the physical properties of the coal seam include uniaxial compressive strength (reflecting the coal's ability to resist external forces), elastic modulus and Poisson's ratio (reflecting the coal's deformation capacity), elastic energy index (reflecting the coal's energy storage capacity), impact energy index (reflecting the coal's release capacity), dynamic failure time (reflecting the coal's failure rate), and related parameters such as brittleness index, acoustic wave velocity, and density. These physical properties determine the upper limit of elastic strain energy that the coal seam can accumulate under given loading conditions, and the rate and form in which this energy is released. Third, the stress field elements include the original rock stress determined by burial depth and geological history, as well as mining-induced stresses such as bearing pressure, transmitted stress, and dynamic load superposition induced by mining activities. The stress field provides the energy driving force and triggering conditions for the formation of rockbursts, and is the direct external factor that transforms rockbursts from potential risks into actual events.

[0004] The three factors mentioned above do not act in isolation, but rather work together in a highly coupled manner to determine the incubation path and probability of rockbursts. Structural factors control the boundary conditions for energy storage and release, physical properties determine the upper limit of energy storage and release capacity, and stress field factors provide the source and spatiotemporal gradient of energy. Any one factor, without the other two, cannot fully explain the phenomenon of rockbursts. For example, high-strength coal bodies are not easily impacted if located in low-stress areas, and similarly, coal bodies with high ductility and no obvious structural weaknesses are also unlikely to experience impacts in high-stress areas. Conversely, even if the physical properties of the coal body are prone to impact, without a suitable stress gradient and structural release surface, it is difficult to trigger severe instability.

[0005] However, in existing technologies, structural information is provided by geological exploration departments in the form of two-dimensional plan views, cross-sections, and borehole columnar sections; physical property information is provided by rock mechanics laboratories using the four national standard indicators obtained from a small number of samples; and stress information is obtained by in-situ rock stress measurement units through hydraulic fracturing or stress relief at a few measuring points. These three types of data typically use different coordinate systems, different precisions, and different sampling densities, and are stored separately in independent paper reports or professional databases, lacking a digital carrier that can integrate them under a unified space, time, and semantic framework. During the usage phase, engineering technicians often rely primarily on manual experience, overlaying the three types of information into maps, comparing tables, and making qualitative judgments to ultimately form a static rockburst hazard zoning map. Figure 1 Once completed, these drawings are rarely updated as mining progresses, making it even more difficult to reflect subtle heterogeneity at the working face scale. Consequently, mine rockburst prevention measures consistently lack a basis for differentiated and refined design for different regions, time periods, and working conditions.

[0006] Therefore, there is a lack of existing technologies that provide a basis for decision-making regarding the differentiated and refined prevention and control of mine rockbursts. Summary of the Invention

[0007] The purpose of this application is to provide a method, medium, equipment, and product for preventing and controlling rockbursts in mines. This can solve the problem in the prior art that the three types of information—structure, physical properties, and stress—cannot be uniformly represented and integrated in three-dimensional space, resulting in a lack of differentiated and refined design basis for mine rockburst prevention and control measures for different regions, time periods, and working conditions.

[0008] In a first aspect, embodiments of this application provide a method for preventing and controlling rockbursts in mines, the method comprising: Acquire multi-source data to characterize the geological conditions of the mine; Based on the multi-source data, a three-dimensional voxel mesh including multiple voxels is constructed along the three-dimensional space of the mine. Based on the multi-source data and the three-dimensional voxel grid, the structural characteristic index for characterizing coal body structural elements, the physical property characteristic index for characterizing coal body physical property elements, and the stress characteristic index for characterizing stress field elements of each voxel are respectively inverted. A strategy for preventing and controlling mine rockbursts is generated based on the multi-source data, the structural characteristic index, the physical property characteristic index, and the stress characteristic index.

[0009] Secondly, embodiments of this application provide a storage medium that stores computer instructions, which, when executed by a computer, are used to perform the steps of the mine rockburst prevention and control method as described in the first aspect.

[0010] Thirdly, embodiments of this application provide an electronic device, including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the mine rockburst prevention and control method as described in the first aspect.

[0011] Fourthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the mine rockburst prevention and control method as described in the first aspect.

[0012] In this embodiment, multi-source data for characterizing mine geological conditions is acquired, and a three-dimensional voxel mesh including multiple voxels is constructed along the three-dimensional space of the mine based on the multi-source data. This integrates the acquired multi-source data, originally stored in different coordinate systems, with different precisions and sampling densities, into a unified spatial coordinate system for high-resolution characterization. This provides a digital carrier for the fusion of multi-source data under a unified space, time, and semantic framework, effectively depicting subtle heterogeneity at the working face scale. It overcomes the deficiency of existing static zoning maps in reflecting detailed changes, and characterizes the data by inverting each voxel separately. The structural characteristic index of coal body structural elements, the physical property characteristic index for characterizing coal body physical properties, and the stress characteristic index for characterizing stress field elements can transform the originally unstructured three elements into three quantifiable and comparable characteristic indices. This upgrades the three elements from traditional graphical overlay and qualitative judgment based on human experience to quantitative indexation. Based on the above three characteristic indices, mine rockburst prevention and control strategies can be generated. Based on the synergistic identification of the three elements, differentiated prevention and control strategies can be generated according to the spatial distribution differences of the three indices under different regions, time periods, and working conditions, thereby providing a refined design basis for mine rockburst prevention and control measures. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the architecture of a digital system for mine rockburst control provided in an embodiment of this application; Figure 2 This is a flowchart illustrating the steps of a method for preventing and controlling rockbursts in mines, as provided in an embodiment of this application. Figure 3 This is an overall flowchart of a method for preventing and controlling rockbursts in mines, provided in an embodiment of this application; Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0016] The method for preventing and controlling mine rockbursts provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0017] Reference Figure 1 This is a schematic diagram of the architecture of a digital system for mine rockburst control provided in this application embodiment. Specifically, it includes a multi-source data input layer 101, a three-dimensional voxelized digital base 102, a three-element component inversion layer 103, a three-element coupling identification layer 104, a dynamic update loop layer 105, and a visualization and decision-making linkage layer 106. The digital system for mine rockburst control is used to implement a method for preventing and controlling mine rockbursts. The three-element component inversion layer 103 includes three modules arranged in parallel: a structural inversion module, a physical property inversion module, and a stress inversion module.

[0018] Reference Figure 2 This is a flowchart illustrating the steps of a method for preventing and controlling rockbursts in mines, as provided in an embodiment of this application. The method specifically includes the following steps: Step 201: Obtain multi-source data to characterize the geological conditions of the mine; In this embodiment, basic mine data is collected through a multi-source data input layer 101. This basic data is then classified and integrated to form multi-source data representing different dimensions of the mine's geological conditions. This multi-source data includes, but is not limited to, geological data representing the mine's geological occurrence and structural characteristics, test data representing the physical and mechanical properties of coal and rock, and monitoring data representing the response to mining disturbances. It should be noted that the geological body within the mine area includes coal seams and surrounding rock. Coal seams are the direct carriers of rockbursts, while surrounding rock, including the roof and floor, is a crucial medium for stress evolution and energy accumulation during mining. The geological data, test data, and monitoring data all originate from the coal seams, surrounding rock, and boreholes within the mine area, collectively representing different dimensions of the mine's geological conditions.

[0019] Geological data includes, but is not limited to, seismic exploration data, drilling data, and well logging data, used to provide information on the stratigraphic structure, geological morphology, and coal seam occurrence of the mine. Testing data includes, but is not limited to, in-situ stress testing data, acoustic monitoring data, and mechanical testing data, used to provide information on the mechanical parameters and physical response of the coal and rock mass. Monitoring data includes, but is not limited to, microseismic monitoring data, mine pressure monitoring data, and drill cuttings monitoring data, used to provide dynamic response information on coal and rock mass fracturing and stress adjustment during mining.

[0020] Step 202: Based on the multi-source data, construct a three-dimensional voxel mesh including multiple voxels along the three-dimensional space of the mine. In this embodiment, the multi-source data obtained in step 201 is used as input, and a three-dimensional voxel grid composed of several voxels is constructed using the three-dimensional space of the mine as the modeling scope. The three-dimensional voxel grid is a set of the smallest spatial units (i.e., voxels) formed by adjacent grid surfaces after discretizing the three-dimensional space of the mine in the strike, dip, and vertical directions at a set step size. Each voxel has unique spatial coordinates and index identifier, and serves as the smallest basic unit for carrying geological information, test information, and monitoring information in subsequent steps.

[0021] In the existing rockburst assessment process for newly built mines, seismic exploration results are presented in the form of structural maps and contour maps, drilling results are presented in the form of columnar charts, and test results are presented in the form of discrete measurement point data. The assessment conclusions are output in the form of two-dimensional drawings, paper reports, and stand-alone software analysis conclusions. This expression method is essentially a two-and-a-half-dimensional expression method with multiple layers superimposed. There is a lack of unified spatial benchmarks and semantic connections between the various information sources. It cannot be directly called by the mine integrated automation platform, safety monitoring network system, or digital twin mine platform, and it is also difficult to provide a data foundation for subsequent dynamic updates.

[0022] To address the aforementioned shortcomings, the three-dimensional voxel mesh constructed in this application adopts a full three-dimensional representation method with voxels as the smallest semantic units. By mapping and aggregating multi-source data into corresponding voxels, each voxel simultaneously carries three types of parameters—structure, physical properties, and stress—and their uncertainty information, while retaining the timestamps of the parameters. This unifies various types of information that were originally scattered in different coordinate systems, with different precisions, and different data formats under the same spatial framework and the same semantic system.

[0023] It should be noted that the three-dimensional voxel mesh in this embodiment serves as a digital base, providing the spatial foundation for all subsequent work. Subsequent structural characteristic index inversion, physical property characteristic index inversion, stress characteristic index inversion, three-element coupling and fusion, dynamic updates, and prevention and control strategy generation all use voxels as the operation objects for spatial positioning and data reading and writing. This transforms the evaluation results, which could only be statically presented in the form of drawings and reports, into structured digital products that can be directly called by the digital platform, providing a unified data carrier for the dynamic management of the entire life cycle of mine rockburst prevention and control.

[0024] Step 203: Based on the multi-source data and the three-dimensional voxel grid, respectively invert the structural characteristic index used to characterize the coal body structural elements, the physical property characteristic index used to characterize the coal body physical property elements, and the stress characteristic index used to characterize the stress field elements of each voxel. In this embodiment, the three-element sub-item inversion layer 103 uses a three-dimensional voxel grid as a unified spatial framework and different dimensions of data from multiple sources as inputs for each inversion process. It performs digital inversions on the three elements of rockburst, outputting three continuous fields: structural characteristic index, physical property characteristic index, and stress characteristic index for each voxel. Specifically, the structural inversion module, based on seismic exploration data, drilling data, and well logging data, inverts structural characteristic indices reflecting the integrity and tectonic development of the coal body on the three-dimensional voxel grid to characterize the coal body's structural elements; the physical property inversion module, based on acoustic monitoring data and mechanical test data, inverts physical property characteristic indices reflecting the mechanical properties of the coal body on the three-dimensional voxel grid to characterize the coal body's physical property elements; and the stress inversion module, based on in-situ rock stress test data and microseismic monitoring data, inverts stress characteristic indices reflecting the distribution of in-situ stress and the state of energy accumulation on the three-dimensional voxel grid to characterize the stress field elements. These three characteristic indices all use voxels as the basic spatial unit and are output in the form of continuous numerical values ​​between 0 and 1, together constituting a quantitative characterization of the three elements of mine rockburst at the full spatial and voxel levels.

[0025] Specifically, the purpose of the structural inversion module in digital inversion is to transform the geometric and structural information within the coal seam and adjacent strata of the roof and floor from discrete seismic exploration data, drilling data, and well logging data into a continuous three-dimensional structural feature field, and ultimately condense it into a structural feature index that can participate in coupling identification; the purpose of the physical property inversion module in digital inversion is to characterize the differences in the mechanical behavior of the coal body in space in the form of a continuous field, and ultimately condense it into a physical property feature index that reflects the physical property conditions of impact tendency; the purpose of the stress inversion module in digital inversion is to combine discrete original rock stress test data and microseismic monitoring data under the engineering reality of extremely limited original rock stress test data in newly built mines, reconstruct a continuous original rock stress field and mining-induced stress field covering the entire mining space, and ultimately condense it into a stress feature index that reflects the conditions driven by geostress.

[0026] Step 204: Generate a mine rockburst prevention and control strategy based on the multi-source data, the structural characteristic index, the physical property characteristic index, and the stress characteristic index.

[0027] In the existing rockburst assessment process for newly constructed mines, from an engineering practice perspective, it can be summarized into the following independent steps. First, the geological exploration department, relying on 3D seismic exploration and geological drilling data, identifies structural anomalies within the mining area, such as faults, folds, collapse columns, and coal thickness variations, and compiles a mine geological structure map and coal seam contour map as the structural baseline for rockburst assessment. Second, the rock mechanics laboratory processes a small number of coal cores obtained from the boreholes into standard specimens and conducts uniaxial compression tests according to national standards to measure four static indicators reflecting the coal seam's rockburst tendency. Combined with classification thresholds, the overall rockburst tendency level of the coal seam is given. Third, a professional geostress measurement unit selects a few measuring points within the mine area to conduct in-situ rock stress tests, obtaining the magnitude and direction of the principal stresses at several discrete points. Finally, engineering technicians, based on coal mine rockburst prevention and control regulations or industry experience, combine the above three types of results through manual overlay and qualitative analysis to form a rockburst hazard zoning map for the entire mine. Finally, after the working face enters the mining stage, microseismic monitoring, online stress monitoring, and electromagnetic radiation monitoring are gradually deployed to conduct on-site early warning. The above process has the advantages of standardization and operability in my country's current coal mine safety production supervision system. However, when facing newly built mines, especially deep mines with complex structures and variable geological conditions, it has exposed several deep-seated defects.

[0028] However, in existing technical solutions, the three types of information—structure, physical properties, and stress—are provided by three different disciplines: geology, rock mechanics, and geostress measurement. The results are stored separately in the form of graphs, tables, and reports, failing to achieve pixel-level coupling within a unified spatial carrier. The formation of rockbursts is actually the result of the synergistic effect of three factors: structure controlling the energy release path, physical properties determining energy storage capacity, and stress providing energy drive. Evaluating these three factors in isolation will mask the coupled amplification effect. In engineering practice, it is common to encounter situations where the physical property indicators of a certain area do not reach a strong rockburst tendency, but it is located within a fault influence zone and has a high stress concentration coefficient, making its actual rockburst risk significantly higher than the simple sum of the three factors evaluated independently. Existing technologies cannot systematically identify or quantitatively characterize such synergistic amplification phenomena. Furthermore, existing predictive models that rely on regression training based on historical microseismic and mine stress data fail in the early stages of coal seam exposure in newly built mines due to severe sample scarcity.

[0029] Therefore, in this embodiment, the structural characteristic index, physical property characteristic index, and stress characteristic index are first fused into a comprehensive risk index on a three-dimensional voxel mesh. This comprehensive risk index can be calculated at the voxel level, is hierarchical, and can be updated over time. A corresponding prevention and control strategy is then generated based on this comprehensive risk index. This comprehensive risk index reflects the combined effect of structure, physical properties, and stress. It is calculated point-by-point on a voxel basis and supports dynamic updates, thus overcoming the problem of isolated evaluation of the three factors in existing technologies. Furthermore, it does not rely on historical sample training and is suitable for the cold start scenario of newly built mines.

[0030] In practical applications, after a newly built mine enters the mining stage, the underground monitoring system continuously generates monitoring data such as micro-vibrations, mine pressure, drill cuttings volume, online stress monitoring, and electromagnetic radiation. This application embodiment efficiently integrates these monitoring data into the three-element field and comprehensive index to support dynamic early warning and generate differentiated prevention and control strategies. Specifically, based on the monitoring data continuously acquired during mining, the three-element characteristic index and risk comprehensive index of the affected area are incrementally updated to reflect the dynamic evolution of the three elements under mining disturbances. Then, spatial clustering analysis is performed based on the updated risk comprehensive index and characteristic index. Finally, based on the analysis results, corresponding differentiated prevention and control measures are matched to generate a mine rockburst prevention and control strategy. The prevention and control strategy refers to decision-making schemes used to reduce the risk of mine rockburst and inhibit its further evolution, including but not limited to grouting reinforcement, advanced support, large-diameter pressure relief boreholes, coal seam water injection, coal seam blasting pressure relief, and optimized mining deployment—all feasible on-site solutions.

[0031] This application embodiment acquires multi-source data for characterizing mine geological conditions and constructs a three-dimensional voxel mesh including multiple voxels along the three-dimensional space of the mine based on the multi-source data. This integrates the acquired multi-source data, originally scattered across different coordinate systems, with different precisions and sampling densities, into a unified spatial coordinate system for high-resolution characterization. This provides a digital carrier for the fusion of multi-source data under a unified space, time, and semantic framework, effectively depicting subtle heterogeneity at the working face scale. It overcomes the deficiency of existing static zoning maps in failing to reflect detailed changes, and uses the inversion of each voxel to characterize the coal body. The structural characteristic index of structural elements, the physical property index for characterizing coal body physical properties, and the stress characteristic index for characterizing stress field elements can transform the originally unstructured three elements into three quantifiable and comparable characteristic indices. This upgrades the three elements from traditional graphical overlay and qualitative judgment based on human experience to quantitative indexation. Based on the above three characteristic indices, mine rockburst prevention and control strategies can be generated. Based on the synergistic identification of the three elements, differentiated prevention and control strategies can be generated according to the spatial distribution differences of the three indices under different regions, time periods, and working conditions, thereby providing a refined design basis for mine rockburst prevention and control measures.

[0032] In one embodiment of this application, the multi-source data includes geological data, test data, and monitoring data. The step of constructing a three-dimensional voxel mesh comprising multiple voxels along the three-dimensional space of the mine based on the multi-source data includes: Based on the geological data, determine the spatial extent of the three-dimensional space and the initial grid step size; The initial grid step size is adaptively adjusted based on the geological data and the test data to obtain the adjusted grid step size, and a grid framework is generated based on the adjusted grid step size. The spatial range is discretized in three dimensions according to the grid framework to obtain multiple grid surfaces arranged along the three-dimensional space; A voxel is defined based on a spatial unit formed by the enclosing of adjacent mesh surfaces, and a three-dimensional voxel mesh composed of multiple voxels is generated within the spatial range. The geological data, test data, and monitoring data belonging to the same spatial range are aggregated to obtain the initial attribute value of each voxel; Each voxel is assigned an attribute based on its initial attribute value.

[0033] In this embodiment, the three-dimensional space of the mine is first discretized in the strike, dip, and vertical directions based on geological data to obtain the spatial range of the three-dimensional space. Specifically, the design range of the mine and the coal seam occurrence area can be determined based on the seismic exploration data, drilling data, and well logging data in the geological data. The total span of the grid is determined based on the design range of the mine and the coal seam occurrence area, so that the constructed three-dimensional voxel grid can cover the mining space that needs to be evaluated.

[0034] Secondly, the coal seam thickness can be determined based on drilling data from geological data. Then, based on preset accuracy requirements and the coal seam thickness, the grid scale can be adaptively selected. Specifically, an initial grid step size can be set according to the preset accuracy requirements, and then error checks are performed by comparing the attribute changes between adjacent voxels. For areas exceeding the preset change threshold, a local step size refinement method is applied to determine the grid scale. Here, the preset accuracy requirements refer to the minimum identifiable anomaly scale required by the project. The grid scale is determined by the grid step size used to characterize spatial discretization accuracy, including a horizontal step size for characterizing horizontal discretization accuracy and a vertical step size for characterizing vertical discretization accuracy. The horizontal step size is taken as the minimum identifiable scale in the preset accuracy requirements, and the vertical step size is generally taken as several equal parts of the coal seam thickness. Preferably, the horizontal step size is generally on the order of several meters to ten meters along the horizontal direction, and the vertical step size is no more than one-third of the coal seam thickness.

[0035] Based on this, the initial grid step size is adaptively adjusted to obtain the adjusted grid step size, and a mesh framework is generated based on the adjusted grid step size. Specifically, the attribute gradients of adjacent voxels and the calculation results at different scales are compared. When the relative change of attributes exceeds a preset change threshold, the grid step size is halved; the same halving operation is performed when the grid step size is near critical areas such as faults, tunnels, and working faces. The above process is iterated until the attribute change is lower than the preset change threshold or reaches the minimum allowable scale, resulting in the adjusted grid step size. Here, the minimum recognition scale is the highest spatial resolution required by the project, and the minimum allowable scale is the lowest limit that can be reached during the step size refinement process.

[0036] The spatial extent is discretized in three dimensions according to a mesh framework, resulting in multiple mesh faces arranged along the three-dimensional space. The spatial extent is further discretized along the orientation, dip, and vertical directions using an adjusted mesh step size, forming a three-dimensional mesh framework. At this point, voxels are defined based on the spatial units enclosed by adjacent mesh faces. Specifically, a voxel is the smallest spatial unit enclosed by adjacent mesh faces, and each voxel is enclosed by six adjacent mesh faces. An initial voxel mesh composed of multiple voxels is generated within the spatial extent. Each voxel is uniquely identified by its center coordinates and three-dimensional row and column numbers. Subsequent queries, assignments, interpolations, updates, and clustering all use this identifier as the primary key.

[0037] In one embodiment, defining voxels based on spatial units enclosed by adjacent mesh surfaces, and generating a three-dimensional voxel mesh composed of multiple voxels within the spatial range, includes: Voxels are defined based on spatial units formed by adjacent mesh surfaces, and an initial voxel mesh composed of multiple voxels is generated within the spatial range. The initial voxel grid is trimmed based on the geological data to obtain the three-dimensional voxel grid.

[0038] Specifically, based on seismic exploration data, drilling data, and well logging data in geological data, the top and bottom surfaces of the coal seam can be determined. The voxels located outside the scope defined by the top and bottom surfaces of the coal seam are deleted, and only the effective voxels located within the influence range of the coal seam and its top and bottom surfaces are retained. This ensures that the voxel mesh is distributed only within the effective space, avoiding the consumption of storage and computing resources in the absence of effective space.

[0039] After cropping, based on the spatial attributes of the geological, test, and monitoring data, data belonging to the same spatial range are aggregated to obtain the initial attribute value for each voxel in the 3D voxel grid. Specifically, multi-source data are mapped to the spatial range occupied by the corresponding voxel in the voxel grid according to their respective spatial coordinates. Data falling within the same voxel range are grouped together to form the initial attribute value for that voxel. Based on the initial attribute value of each voxel, attributes are assigned to each voxel in the 3D voxel grid.

[0040] Each voxel's initial attribute values ​​include five groups: geometric attributes, structural attributes, physical property attributes, stress attributes, and meta-attributes. Geometric attributes record the voxel's center coordinates and dimensions; structural attributes record fracture density, tectonic influence factors, and lithological abrupt change rate; physical property attributes record uniaxial compressive strength, elastic modulus, Poisson's ratio, longitudinal and transverse wave velocities, apparent density, and water content; stress attributes record the original rock stress components, mining-induced stress components, stress concentration factor, and elastic energy density; and meta-attributes record the confidence level and most recent update time for each attribute.

[0041] It should be noted that the grid is embedded within the existing coordinate reference system of the mine, using the mine's internal coordinate system as a benchmark. Spatial data is organized and managed hierarchically according to four levels: mine, coal seam, grid block, and voxel. Grid blocks are organized using an octree, and voxels employ a combined index of 3D row and column numbers and Morton coding, thus supporting rapid querying, slicing, and updating at the working face scale. The establishment of this coordinate reference system and hierarchical management structure allows multi-source data to be mapped, aggregated, and computed under a unified spatial benchmark. All subsequent algorithms operate around the same spatial carrier, avoiding coordinate transformation errors and spatial misalignment problems caused by different professional data using different coordinate benchmarks in traditional methods. It also avoids engineering problems caused by coordinate transformation, differences in sampling density, and version inconsistencies in traditional multi-layer methods. Furthermore, the confidence level and timestamp retained in the meta-attributes provide a foundation for subsequent uncertainty propagation and dynamic updates.

[0042] In one embodiment of this application, the geological data includes seismic exploration data, drilling data, and well logging data. Based on the multi-source data and the three-dimensional voxel grid, structural characteristic indices for each voxel to characterize coal body structural elements are inverted, including: Seismic attributes are extracted from the seismic exploration data, and samples are taken according to the spatial coordinates of each voxel in the three-dimensional voxel grid to obtain the multidimensional seismic attribute vector of each voxel. The multidimensional seismic attribute vector is processed to obtain the principal component attributes of each voxel; The spatial neighborhood relationship between each voxel is determined based on the spatial coordinates of each voxel in the three-dimensional voxel grid, and a three-dimensional data block is constructed based on the principal component attributes and the spatial neighborhood relationship. The structural anomaly probability of each voxel is determined based on the three-dimensional data block and the drilling data. Based on the drilling data and the logging data, the borehole structure constraints of each borehole in the mine are determined; Based on the borehole structure constraints and the multidimensional seismic attribute vector, collaborative interpolation is performed in the three-dimensional voxel grid to obtain a continuous structural parameter field. Based on the structural anomaly probability and the continuous structural parameter field, the structural characteristic index of each voxel is determined.

[0043] In one embodiment of this application, the structural characteristic index used to characterize the coal body structural elements is inverted based on the pre-constructed three-dimensional voxel grid, mainly relying on seismic exploration data, drilling data and well logging data in geological data.

[0044] First, various seismic attributes are extracted from the post-stack amplitude volume and its derived attribute volume of seismic exploration data. These seismic attributes include, but are not limited to, root mean square amplitude reflecting amplitude anomalies, coherence attributes reflecting stratigraphic discontinuities, curvature attributes characterizing local bending and deflection, and instantaneous frequency attributes reflecting frequency attenuation anomalies. After extraction, each seismic attribute is sampled according to the spatial coordinates of each voxel in a three-dimensional voxel grid. Specifically, each seismic attribute is trilinearly sampled according to the coordinates of a unified voxel center, and the seismic attribute value is assigned to the corresponding voxel, thereby obtaining a multidimensional seismic attribute vector for each voxel.

[0045] Subsequently, the multidimensional seismic attribute vector is processed to obtain the principal component attributes of each voxel. Preferably, the multidimensional seismic attribute vector is subjected to dimensionality reduction and orthogonalization to obtain the principal component attributes of each voxel. Specifically, each attribute is normalized within the well field, and then principal component analysis is used to reduce the multidimensional attributes to mutually orthogonal principal components. Principal component analysis outputs the score and corresponding contribution rate of each principal component as the principal component attributes. Preferably, principal components with a cumulative contribution rate of not less than 85% are selected as inputs for subsequent processing to reduce redundancy and collinearity among attributes.

[0046] Based on this, the spatial neighborhood relationships between voxels are determined according to their spatial coordinates in the 3D voxel grid, and a 3D data block is constructed based on the aforementioned principal component attributes and these spatial neighborhood relationships. Specifically, during construction, with the target voxel as the center, the principal component attributes of all voxels within the set neighborhood window size are extracted from the 3D voxel grid. These principal component attributes are arranged according to their relative spatial positions within the window, forming a 3D data block with spatial structure. This 3D data block simultaneously contains the principal component attribute information of the central voxel and its surrounding voxels, providing input with spatial context information for the subsequent 3D convolutional neural network. Subsequently, using the voxel attribute block in the 3D voxel grid as input, the 3D data block is processed in conjunction with drilling data. The borehole structure information of the exposed area is used as supervision to identify voxel-level structural anomalies, and finally, the structural anomaly probability of each voxel is determined, that is, the probability value of each voxel belonging to structural anomaly categories such as faults, collapse columns, coal thickness abrupt change zones, or coal seam tear zones.

[0047] Meanwhile, based on drilling and logging data, the borehole structure constraints of each borehole in the mine are determined. Specifically, all drilling and logging data entering the coal seam are processed uniformly. The original logging data, such as electrical imaging, natural gamma, density, resistivity, and sonic transit time, are sequentially subjected to depth correction, outlier removal, missing segment interpolation, and equal-interval resampling. Curves of different dimensions are robustly standardized and then spliced ​​into a multivariate sequence according to the same borehole depth. Drilling data, such as electrical imaging interpretation results and core descriptions, are converted into discrete labels, and the multivariate sequence and discrete labels are synchronously aligned. Based on this, three parameters are calculated: borehole fracture density, borehole lithological abrupt change rate, and borehole coal integrity. Among them, borehole fracture density is determined by the number of electrical imaging fractures or the cumulative trace length per unit borehole depth; borehole lithological abrupt change rate is determined by the comprehensive gradient of the standardized curve within the sliding window; and borehole coal integrity is jointly calibrated by the core integrity and the acoustic and resistivity responses. The borehole fracture density, borehole lithological abrupt change rate, borehole coal integrity, and their spatial coordinates and confidence levels obtained along the depth of each borehole together constitute the borehole structural constraints.

[0048] Subsequently, using the borehole structure constraint as a hard constraint and the multidimensional seismic attribute vector as a soft constraint, collaborative interpolation is performed in a three-dimensional voxel grid to obtain a continuous structural parameter field. The borehole structure constraint, as a hard constraint, means that the interpolation results must strictly meet the measured values ​​at the borehole location, while the multidimensional seismic attribute vector, as a soft constraint, controls the spatial variation trend and allows deviations within the error range. Preferably, the roof and floor surfaces of the coal seam can also be used as geometric boundaries to limit the effective domain of the interpolation. Finally, a geostatistical collaborative interpolation method is used to extend the borehole fracture density, borehole lithological abrupt change rate, and borehole coal integrity at the borehole scale into a three-dimensional continuous structural parameter field covering the entire space.

[0049] Furthermore, the target structural influence factors for each voxel are determined based on the structural anomaly probability and drilling data, and the target fracture density and target lithological abrupt change rate for each voxel are determined based on the continuous structural parameter field. Specifically, the target structural influence factors are calculated jointly from the structural anomaly probability, borehole fracture density, and the normalized distance of the voxel from structures such as faults, fold axes, or collapse columns. The characteristic length used for distance attenuation is determined based on fault displacement and coal seam depth. The target fracture density and target lithological abrupt change rate are directly extracted from the continuous structural parameter field.

[0050] After normalizing the target fracture density, target structural influence factor, and target lithological abrupt change rate of each voxel, they are linearly combined according to the calibration weights, giving each voxel a continuous value between 0 and 1. Preferably, the calibration weights are determined using exposed faults, folds, fracture zones, and impact manifestation locations as samples, through constrained logistic regression or grid search, with the criteria being the maximum cross-validation AUC, non-negative weights, and a sum of weights equal to one. When the sample is insufficient, initial values ​​are set using expert priors and are continuously corrected as new exposed data is released. The closer the structural characteristic index is to 1, the more favorable the structural conditions in the area where the voxel is located are for the incubation and occurrence of rockbursts.

[0051] It should be noted that the embodiments of this application involve structural parameters at multiple levels. Among them, the borehole fracture density and borehole lithological mutation rate are raw parameters directly calculated at the borehole location based on logging data and core description. These two parameters exist only at discrete borehole locations, representing the measured values ​​at that point. The data volume is sparse but the accuracy is high. After co-interpolation, these parameters are expanded into a three-dimensional continuous structural parameter field covering the entire space. At this time, the fracture density value and lithological mutation rate value obtained by each voxel from this continuous structural parameter field are recorded as the target fracture density and target lithological mutation rate, representing the optimal estimated values ​​at the voxel location after multi-source data constraints. The fracture density and tectonic influence factor in the structural attributes of the three-dimensional voxel grid are initial values ​​obtained by preliminary aggregation of multi-source data when the three-dimensional voxel grid is constructed. Their spatial coverage depends on the spatial distribution of the raw data. The target fracture density and target lithological mutation rate are data obtained by incorporating more refined borehole logging processing and seismic attribute constraints on the initial structural attributes of the three-dimensional voxel grid.

[0052] This application upgrades the structural information from the traditional discrete expression of two-dimensional line or surface elements to a voxel-level continuous field expression. The influence range of structural elements such as faults and folds is explicitly quantified as a continuous field that decays with distance, thereby enabling voxel-level coupling calculations with subsequent physical property fields and stress fields.

[0053] In one embodiment of this application, the test data includes acoustic monitoring data and mechanical test data. Based on the multi-source data and the three-dimensional voxel grid, the physical property index used to characterize the physical properties of coal is inverted, including: Based on the acoustic monitoring data, the wave velocity parameters of each voxel are obtained by inversion in the three-dimensional voxel grid. Based on the drilling data, the mechanical test data, the wave velocity parameters, and the structural characteristic index, the set of mechanical parameter indices for each voxel is inverted; By fusing the mechanical parameter indices in the set of mechanical parameter indices, the physical property index of each voxel is obtained.

[0054] In this embodiment, the main reliance is on acoustic monitoring data and mechanical test data from the test data, combined with drilling data from the geological data and the determined structural characteristic index. Based on the constructed three-dimensional voxel grid, the physical property characteristic index used to characterize the physical property elements of the coal body is inverted. Specifically, the differences in the mechanical behavior of the coal body in space are characterized in the form of a continuous field and finally condensed into a physical property characteristic index that reflects the physical property conditions of impact tendency.

[0055] First, wave velocity parameters for each voxel are obtained by inverting acoustic monitoring data in a three-dimensional voxel grid. Specifically, an underground active acoustic monitoring system is used as the primary means. Transmitting and receiving probe arrays are arranged at certain intervals along the roadway to perform three-dimensional acoustic tomography on the coal seam. The acoustic tomography employs a travel-time-based iterative imaging algorithm, controlling the convergence speed and stability through iterative relaxation factors. The acoustic monitoring data is inverted into P-wave and S-wave velocity fields covering the working face scale. The inversion results are then written into the corresponding voxels in the three-dimensional voxel grid according to their spatial coordinates, thus obtaining wave velocity parameters for each voxel.

[0056] Simultaneously, sample labels for model training were obtained based on drilling and mechanical test data. Specifically, representative locations in the exposed area were selected for core sampling and fabricated into standard specimens. Uniaxial compression tests were conducted according to national standards to obtain four national standard indicators: uniaxial compressive strength, elastic energy index, impact energy index, and dynamic failure time. Drilling data provided information such as apparent density and coal seam depth. Apparent density was obtained from density logging, and coal seam depth was calculated from the elevations of the top and bottom plates of the coal seam in the borehole and the surface elevation. Mechanical test data provided measured values ​​for the four national standard indicators. Based on this, a multi-task regression model was constructed, using the wave velocity parameters of each voxel, the apparent density and coal seam depth from the drilling data, and the determined structural characteristic indices as input features, and the aforementioned four national standard indicators as multi-task output targets.

[0057] The multi-task regression model is trained and cross-validated on samples from the exposed area. It employs a compact network structure and physical consistency constraints to avoid overfitting on small samples, ensure smooth results in structurally homogeneous regions, and exhibit reasonable gradient abrupt changes when crossing structurally anomalous zones. After training, the multi-task regression model is applied to a three-dimensional voxel mesh. The input parameters of each voxel are assigned according to its spatial location, and data between boreholes are completed using spatial interpolation constrained by formation boundaries. This allows the multi-task regression model to invert and obtain the set of mechanical parameter indices for each voxel, namely, estimates of uniaxial compressive strength, elastic energy index, impact energy index, and dynamic failure time for each voxel.

[0058] After obtaining the set of mechanical parameter indices for each voxel, the indices are fused to obtain the physical property index for each voxel. During the fusion process, referring to the grading thresholds of each indicator in the national standard, the four indicators for each voxel are piecewise linearly normalized to map them to a unified numerical range. Then, they are weighted and fused according to pre-defined weights. The weights are determined using the impact tendency classification of historical coal samples as labels, through non-negative constraint multiple regression or entropy weighting combined with expert weighting. The weights are normalized so that the sum of the weights is 1, with the weight that minimizes the cross-validation error used as the initial value. After the above fusion, each voxel obtains a physical property index between 0 and 1. This index comprehensively reflects the inherent impact tendency of the coal body at the physical property level; the higher the value, the more physical conditions the voxel possesses for impact generation.

[0059] This application extends the traditional method of inverting only a single parameter to simultaneously inverting four national standard indicators. This ensures that the physical property index strictly corresponds to the current industry evaluation standards in terms of physical meaning, facilitating engineering review and regulatory verification within the industry. Simultaneously, by using the structural characteristic index as one of the input constraints for physical property inversion, the physical mechanism of structural damage to coal body continuity and subsequent deterioration of physical properties is explicitly modeled. This allows for the automatic capture of physical property deterioration in structurally anomalous regions such as faults and fracture zones, thereby avoiding the overestimation of physical properties in structurally anomalous areas caused by sample scarcity in existing technologies.

[0060] In one embodiment of this application, the geological data includes tectonic background information and mining spatial location, the test data includes in-situ rock stress test data, and the monitoring data includes microseismic monitoring data. Based on the multi-source data and the three-dimensional voxel grid, the stress characteristic index of each voxel used to characterize the stress field element is inverted, including: Based on the original rock stress test data and the tectonic background information, the original rock stress field of the mine is constructed in the three-dimensional voxel grid; the original rock stress field is used to characterize the spatial distribution of in-situ stress in the mine when it is not affected by mining activities; Based on the mining location of the mine and the original rock stress field, the mining-induced stress field of the mine is determined; the mining-induced stress field is used to characterize the spatial distribution of ground stress in the mine under the influence of mining activities. The mining-induced stress field is corrected based on the microseismic monitoring data, and the stress concentration factor and elastic energy density of each voxel are determined based on the corrected mining-induced stress field. The stress characteristic index of each voxel is determined based on the stress concentration factor and the elastic energy density.

[0061] In this embodiment, the system primarily relies on tectonic background information and mining spatial location from geological data, original rock stress test data from test data, and microseismic monitoring data from monitoring data. Based on a pre-constructed three-dimensional voxel grid, stress characteristic indices used to characterize stress field elements are inverted. The stress field is the direct driving source of rockbursts and is also the scarcest data element in newly constructed mines. In this embodiment, under the condition of a limited number of hydraulic fracturing or stress relief measurement points, by combining tectonic background and microseismic data, a continuous stress field covering the mining area can be obtained and ultimately condensed into stress characteristic indices.

[0062] Based on in-situ stress test data and tectonic background information, an in-situ stress field for the mine is constructed in a three-dimensional voxel grid. This in-situ stress field characterizes the spatial distribution of in-situ stress in the mine before mining activities. Specifically, using measurement point data obtained through hydraulic fracturing or stress relief as hard constraints, and combining the direction of the maximum horizontal principal stress determined by the tectonic background of the mine field, the stress tensor of each measurement point is transformed into a unified mine coordinate system and quality-classified. Based on this, the burial depth trend terms of the maximum principal stress, intermediate principal stress, and minimum principal stress are fitted respectively. The trend residuals are spatially interpolated using a Kriging interpolation method considering tectonic zoning. Finally, the trend terms are superimposed with the residual field, and conditional corrections are applied at the borehole measurement points to obtain a continuous in-situ stress field covering the entire mine field and its estimated variance.

[0063] It should be noted that the maximum, intermediate, and minimum principal stresses mentioned above are the three principal components obtained after eigenvalue decomposition of the stress tensor. Their directions are given by the corresponding eigenvectors and can be further converted into components in the mine coordinate system, rather than fixed coordinate axis direction components in the spatial coordinate system. After the above processing, each voxel in the three-dimensional voxel mesh obtains the magnitude and direction information of the principal stresses of the original rock stress state.

[0064] Secondly, based on the mining location and the original rock stress field, the mining-induced stress field of the mine is determined. This mining-induced stress field is used to characterize the spatial distribution of ground stress under the influence of mining activities. In one embodiment, the boundaries of the working face, roadway, and coal pillar are determined based on the mining location. These boundaries are used as geometric inputs, and the pre-constructed original rock stress field is used as the background field. Based on the elastic half-space theory and the coal pillar load transfer relationship, response kernel functions for the advance support pressure, lateral support pressure, and pressure transmitted from adjacent goaf areas are established respectively. Furthermore, based on the mining location, the mining height, burial depth, working face length, advance distance, and roof parameters can also be determined. The peak coefficient and influence range of each response kernel are determined based on the mining height, burial depth, working face length, advance distance, and roof parameters. The three types of response kernels are superimposed to obtain the mining-induced additional stress field. This mining-induced additional stress field is then superimposed on the original rock stress field to obtain the mining-induced stress field. In this embodiment, the response kernel and parameter lookup table for unit load are pre-calculated in the offline stage. In the online stage, only scaling, translation and voxel convolution of the response kernel need to be completed according to the current mining geometry. Therefore, this embodiment does not require repeated meshing or iterative solution of large finite element equations, and the stress field can be updated quickly after each working face advance.

[0065] To further improve the accuracy and reliability of the stress field, it is necessary to further correct the mining-induced stress field based on microseismic monitoring data. Specifically, moment tensor decomposition is performed on the microseismic events, retaining three eigenvalues ​​and their principal axes: the maximum compression axis, the maximum tension axis, and the intermediate principal axis. The maximum compression axis and the maximum tension axis are used to characterize the main force mechanism, while the intermediate principal axis is used to maintain the integrity of the tensor and participate in stress correction.

[0066] Using the prior stress vector *s* as the state estimate of the current mining-induced stress field, and the principal axis directions and relative intensities obtained from the microseismic moment tensor decomposition as the observation vector *d*, an observation operator *H* is established to map the stress state to the microseismic observation space. This operator characterizes the uncertainty of the two types of information. Based on this, the maximum a posteriori estimate is calculated using the covariance matrix *Cs* of the prior stress error and the covariance matrix *Cd* of the microseismic observation error. Where s0 is the prior stress vector before fusion, s MAP This represents the maximum a posteriori stress estimation vector obtained after fusing microseismic observations; T denotes matrix transpose. 1 indicates matrix inversion; the smaller Cs is, the stronger the prior stress constraint; the smaller Cd is, the greater the weight of microseismic observations in the correction.

[0067] During the correction process, observation residuals are constructed based on the principal axis direction and relative intensity of the microseismic events. These residuals are then distributed to nearby voxels according to distance attenuation within the event's influence radius. Maximum a posteriori (MAP) estimates of s are then used. MAPThe stress components are adjusted, while the original mining-induced stress field results are maintained for voxels outside the event's influence range. The magnitude of each correction is limited by the covariance matrix. After the above corrections, the stress field of each voxel in the three-dimensional voxel grid is constrained and adjusted by the microseismic monitoring data.

[0068] The stress concentration factor of each voxel is obtained by calculating the ratio of the corrected mining-induced stress field to the original rock stress field, reflecting the degree of stress increase caused by mining activities. The elastic energy density of each voxel is calculated based on the principal stress values ​​of the corrected mining-induced stress field and the elastic parameters of the coal and rock mass, reflecting the ability of a unit volume of coal and rock mass to store elastic strain energy. It should be noted that both parameters are calculated voxel-by-voxel in a three-dimensional voxel grid, ensuring that each voxel obtains a corresponding stress concentration factor and elastic energy density value.

[0069] Finally, the stress concentration factor and elastic energy density of each voxel are piecewise linearly normalized, and then weighted and combined according to pre-defined weights. The weights of each component are determined through non-negative constraint regression and cross-validation, based on the discriminative power of the maximum principal stress, stress concentration factor, stress gradient, and dynamic correction in historical impact data. When data is insufficient, prior weights are set based on mechanistic importance. After this fusion, each voxel obtains a stress characteristic index between 0 and 1; a larger stress characteristic index indicates that the stress conditions of that voxel are more favorable for rockburst occurrence.

[0070] This application upgrades the original rock stress inversion from the traditional extensive extrapolation of a single point representing the entire well field to a multi-source fusion inversion based on hard constraints, tectonic background, and soft constraints, ensuring consistency between the stress field and tectonic units. Simultaneously, it automatically links the mining-induced stress inversion with the working face advancement progress, enabling continuous updates of the stress field as production progresses. Furthermore, it upgrades the stress index from the traditional single-dimensional discrimination of high stress equaling high risk to a dual-driven discrimination based on stress concentration coefficient and elastic energy density, effectively avoiding the problem of high-stress but low-energy-storage areas or high-energy-storage but low-stress areas being incorrectly identified as high-risk areas.

[0071] Preferably, in structural inversion, the 3D convolutional neural network can be replaced by other 3D segmentation network structures, or voxel-level classification can be performed using machine learning methods such as random forests and gradient boosting based on seismic attributes. In property inversion, neural network regression can be replaced by methods such as Gaussian process regression, random forests, and gradient boosting; in cases with very few samples, it can degenerate into multiple linear regression supplemented with physical consistency constraints. In stress inversion, the analytical solution of elasticity can be replaced by a fast calculation scheme based on a finite element offline surrogate model, or a surrogate learning scheme based on a discrete element model.

[0072] In one embodiment of this application, the strategy for preventing and controlling mine rockburst based on the multi-source data, the structural characteristic index, the physical property characteristic index, and the stress characteristic index includes: The comprehensive risk index of each voxel is determined based on the structural characteristic index, the physical property characteristic index, and the stress characteristic index; Based on the monitoring data, the structural characteristic index, physical property index, stress characteristic index, and comprehensive risk index of the voxels in the affected area of ​​the three-dimensional voxel grid are incrementally updated to obtain the index set of each voxel. Based on the index set of each voxel, voxels that meet the preset clustering requirements are clustered to form candidate risk regions; When the candidate risk area reaches the preset triggering condition, an early warning event is generated to characterize the existence of rockburst risk in the candidate risk area; The dominant element of the warning event is determined based on the index set of voxels contained in the candidate risk area; Based on the dominant factors of the warning event, a prevention and control strategy for reducing the risk of rockburst in the candidate risk areas is generated.

[0073] In this embodiment, the three-element coupling identification layer 104 generates a mine rockburst prevention and control strategy based on multi-source data and the determined structural characteristic index, physical property characteristic index, and stress characteristic index. Specifically, the visualization and decision-making linkage layer 106, based on the three-element sub-indices, presents the relevant early warning results in a three-dimensional visualization on the mine digital twin platform, providing multiple query modes such as by coal seam slice, by working face profile, by time series, and by element components. This ensures that the early warning results are no longer isolated hazard zoning maps, but rather decision support products that are linked to on-site executable prevention and control processes.

[0074] First, the comprehensive risk index of each voxel is determined based on the structural characteristic index, physical property characteristic index, and stress characteristic index. The comprehensive risk index also outputs the dominant component identifier and is divided into four levels according to the preset threshold: strong impact risk, moderate impact risk, weak impact risk, and no obvious impact risk, which are visually identified by red, orange, yellow, and green, respectively.

[0075] After a newly built mine enters the mining stage, the underground monitoring system continuously generates monitoring data such as micro-vibrations, mine pressure, drill cuttings volume, online stress monitoring, and electromagnetic radiation. In this embodiment, after determining the comprehensive risk index of each voxel, this monitoring data is efficiently integrated into the three-element field and the comprehensive index to support dynamic early warning. Specifically, based on the monitoring data, the structural characteristic index, physical property index, stress characteristic index, and comprehensive risk index of the voxels in the affected area of ​​the three-dimensional voxel grid are incrementally updated to obtain the index set of each voxel.

[0076] In one embodiment, during exponential updates, an extended state vector is defined for each voxel i at time k. x i，k =[I s I p I σ ,I,ΔI s ΔI p ΔI σ ] T This vector sequentially contains the structural characteristic index, physical property characteristic index, stress characteristic index, and risk comprehensive index, as well as the increments of these three indices relative to the previous period, and is stored together with the voxel number, timestamp, and covariance matrix. State equation X i，k =F i，k X i，k-1 +B i，k u i，k +w i，k The observation equation z is used to describe the evolution of each index over time. i，k =H i，k X i，k +v i，k It is used to map state variables to corresponding monitoring quantities.

[0077] Where i is the voxel number, k is the time, and x i，k Let I be the extended state vector of voxel i in period k. s I p I σ The structure, physical properties, and stress exponent at time k are respectively ΔI s ΔI p ΔI σ These represent the increments of the structure, physical properties, and stress exponent at time k relative to the previous time k-1, respectively; the superscript T indicates vector transpose; F i，k B is the state transition matrix, describing the evolution of each index over time. i，k To control the input matrix; u i，k The control vector is composed of external forces such as the working face advance distance and dynamic load intensity; z i，k For monitoring and observation equations; H i，k This is the observation matrix, used to map state variables to corresponding monitored quantities; w i，k With v i，k These are process noise and observation noise, respectively, and are usually assumed to be zero-mean random vectors.

[0078] Specifically, the working face advance distance can be determined based on production scheduling records, coal mining machine positioning, or working face measurement data from monitoring data. Dynamic load intensity can be comprehensively calculated based on microseismic event energy, magnitude, event rate, and the distance from the voxel to the seismic source.

[0079] Different monitoring data are mapped to corresponding index categories based on their physical meaning. In one embodiment, the structural characteristic index, physical property index, stress characteristic index, and comprehensive risk index of voxels in the affected region of the three-dimensional voxel mesh are incrementally updated to obtain the index set for each voxel, including: Based on the density of microseismic events in the monitoring data and according to the Gutenberg-Richter law log 10 The value of b is obtained by fitting N=a-bM, and the structural feature index is updated based on the value of b. The microseismic energy release rate is obtained by dividing the total energy of events within the sliding time window in the monitoring data by the window length. It is then allocated to voxels within the affected area according to the event location to characterize coal damage accumulation. The physical property index is updated based on the microseismic energy release rate. The abnormality of drill cuttings is determined based on the amount of drill cuttings in the monitoring data. When the amount of drill cuttings is abnormal, the stress characteristic index of the corresponding voxel is updated according to the preset stress index correction rule based on the abnormal value of drill cuttings and its spatial location.

[0080] Specifically, the dynamically updated loop layer 105 normalizes and weights the microseismic event density, energy release rate, and b-value decrease by confidence level, thus forming microseismic structural observations. When event clustering increases or the b-value decreases, the structural characteristic index increases accordingly. It should be noted that each type of observation carries both accuracy level and effective spatiotemporal range to determine its update weight.

[0081] Preferably, the dynamic update employs a strategy combining a sliding window and Kalman filtering. The system establishes a circular queue of length N update cycles for each type of monitoring quantity. New data is written to the tail of the queue after time alignment and quality control, while the oldest cycle data is deleted. Simultaneously, the number of valid samples, mean, variance, and missing data markers are maintained. During filtering, the Kalman gain is calculated based on the predicted covariance and observation noise. Then, according to the Kalman filter x=x - +K(z-Hx) - The system integrates predicted and measured values. When the measured error is small, R is decreased and K is increased to enhance the weight of the measured data. When the measured error is large, more model predictions are retained. After each update, the three-element index and comprehensive index of the voxel are written back to the three-dimensional voxel grid, and the timestamp is updated, so that the prevention and control strategy can be continuously updated with the mining progress.

[0082] Among them, P - R is the state prediction error covariance matrix before the update; H is the observation noise covariance matrix; T is the observation matrix, where the superscript T indicates matrix transpose; K is the Kalman gain matrix, used to reconcile the weights of model predictions and measured data; x -... 1 indicates matrix inversion. When R decreases, the reliability of the observed data increases, and K increases accordingly; when P... - When the value is reduced, the uncertainty of the model prediction decreases, and the updated results are more likely to retain the predicted values.

[0083] After incremental updates are completed, voxels that meet the preset clustering requirements are clustered based on the index set of each voxel to form candidate risk areas. Specifically, all voxels within a specified range in front of the scanning work area are scanned, and spatially connected voxels exceeding the risk threshold are clustered to form candidate risk areas. Each candidate risk area records its spatial boundary, center location, volume, maximum and average comprehensive index, dominant element, duration, and confidence level, and is matched with the clustering results of the previous period according to spatial overlap rate and center distance. When a candidate risk area reaches the preset triggering condition, an early warning event is generated to characterize the presence of rockburst risk in that candidate risk area.

[0084] Preferably, the preset triggering conditions include the risk level, spatial area, duration period or growth rate of the candidate risk area reaching the corresponding triggering threshold. When the triggering conditions are met, the system generates an early warning event. The early warning event inherits information such as the location, range, level, dominant elements and confidence level of the cluster area and pushes it to the mine monitoring and production scheduling system.

[0085] Subsequently, the dominant factor of the early warning event is determined based on the set of indices of the voxels contained in the candidate risk area. Specifically, the confidence-weighted average of the structural index, physical property index, and stress index in the medium and higher risk clusters needs to be calculated first. When the largest sub-index exceeds a set difference relative to the second largest sub-index, the component is determined to be a single dominant factor; otherwise, it is determined to be dominated by multiple factors.

[0086] Based on the dominant factors of the early warning event, prevention and control strategies for reducing the risk of rockbursts in candidate risk areas can be generated. Specifically, when the structural component is dominant (i.e., the structural characteristic index is significantly higher than the physical property index and stress characteristic index within the risk clustering area), measures such as grouting reinforcement, advanced support, and avoidance structures are recommended. When the physical property component is dominant (i.e., the physical property index is significantly higher than the structural characteristic index and stress characteristic index), measures to reduce the energy storage capacity of the coal seam, such as large-diameter pressure relief boreholes and coal seam water injection, are recommended. When the stress component is dominant (i.e., the stress characteristic index is significantly higher than the structural characteristic index and physical property index), measures to reduce stress concentration, such as coal seam blasting pressure relief, adjusting the advance speed, and optimizing mining deployment, are recommended. When multiple factors are dominant, corresponding measures are recommended in combination.

[0087] It should be noted that after each monitoring cycle, the system automatically provides short-term and medium-term risk trend forecasts for the comprehensive index profile within a certain advance distance in front of the work area, providing quantitative basis for the production organization and prevention and control measures deployment in the next cycle, thus making the early warning results a decision support product that is connected with the on-site executable prevention and control processes.

[0088] Preferably, regarding the coupling form of the three elements, the weighted geometric average plus short-board gain term can be replaced by non-additive fusion based on Choquet fuzzy integrals, or by the joint probability over-threshold method based on Copula functions. Regarding dynamic updates, extended Kalman filtering can be replaced by unscented Kalman filtering, particle filtering, Bayesian model averaging, etc. Regarding early warning triggering, density-based spatial clustering can be replaced by topological clustering based on topological persistence analysis. Regarding visualization and decision-making linkage, hierarchical thresholds, color mapping, and the prevention and control measure library can all be parameterized according to the actual situation of the mine.

[0089] In one embodiment of this application, to ensure the engineering reliability of the method, key parameters and uncertainties are systematically managed during implementation. Specifically, in the model training and index calibration stages, historical impact events and high-stress manifestation areas are used as positive samples, and stable production areas are used as negative samples, with the validation set divided according to time sequence. First, the normalized upper and lower limits of individual indicators are determined, and then the weights of the three elements, the weakest link gain coefficient, and the risk classification threshold are calibrated through constrained search to optimize the comprehensive target composed of early warning recall rate, false alarm rate, and lead time. Finally, rolling verification is performed using newly disclosed and monitored data.

[0090] Regarding the expression of uncertainty, the dynamic update module simultaneously calculates the expected values ​​and confidence intervals of the three-factor index and the comprehensive index, and writes the results into the three-dimensional voxelized digital base 102 for use by the visualization and decision-making linkage layer 106. The expected value is used to determine the risk level, and the confidence interval is used to adjust the credibility of the warning and the transparency of the display, triggering manual review when uncertainty is high. The confidence interval is estimated using a combination of parametric error propagation and Monte Carlo simulation, and is visually displayed through color transparency or shading.

[0091] In terms of data governance, the system establishes a traceability chain for input data sources, recording the processors, parameters, and timestamps from the original observations, intermediate processing, to the final fusion, so that each warning can be traced back to the underlying data source.

[0092] In one embodiment of this application, determining the comprehensive risk index of each voxel based on the structural characteristic index, the physical property characteristic index, and the stress characteristic index includes: Extract the confidence level of each voxel from its attributes; The weights of the structural characteristic index, physical property index and stress characteristic index of each voxel are determined based on the confidence level of each voxel. Determine whether the structural characteristic index, physical property characteristic index, and stress characteristic index of each voxel are all greater than the preset gain activation threshold; If the structural characteristic index, physical property index, and stress characteristic index of each voxel are all greater than the preset gain activation threshold, then the structural characteristic index, physical property index, and stress characteristic index of each voxel are fused based on the preset gain coefficient and the weight to obtain the comprehensive risk index of each voxel. If at least one of the structural characteristic index, physical property index, and stress characteristic index of each voxel is less than or equal to the preset gain activation threshold, then the structural characteristic index, physical property index, and stress characteristic index of each voxel are fused based on the weights to obtain the comprehensive risk index of each voxel.

[0093] When all three factors simultaneously enter the danger zone, the overall risk is higher than that of simple linear superposition; however, if any one factor is significantly low, the overall risk will be suppressed. Therefore, in this embodiment, the three factors are integrated from two aspects: synergistic amplification and bottleneck constraint, to determine the comprehensive risk index. Specifically, the comprehensive risk index is calculated by combining a weighted geometric mean with a bottleneck gain term.

[0094] In one embodiment, before fusion, the confidence level of each voxel is extracted from its meta-attributes. This confidence level reflects the reliability of the data source and the uncertainty of spatial interpolation for each voxel during the three-element inversion process. Based on the confidence level of each voxel, the weights corresponding to the structural characteristic index, physical property index, and stress characteristic index of each voxel are determined. The higher the confidence level, the more sufficient the data support for that voxel in that element is, and the higher the weight is assigned in the fusion. It can be understood that the basic weights of the three elements are determined by historical cases in the initial stage, and are dynamically corrected and normalized based on the confidence level in the online stage.

[0095] Furthermore, it is determined whether the structural characteristic index, physical property characteristic index, and stress characteristic index of each voxel are all greater than the preset gain activation threshold. If all three indices of each voxel are greater than the preset gain activation threshold, it means that the voxel has entered the danger zone in all three aspects. The synergistic amplification effect among the three needs to be explicitly characterized. Therefore, the three indices are fused based on the preset gain coefficient and weights. Specifically, the fusion is carried out in the form of a weighted geometric average, and the synergistic gain amplification determined by the preset gain coefficient is applied to the fusion result to obtain the comprehensive risk index of each voxel.

[0096] If at least one of the three indices of each voxel is less than or equal to the preset gain activation threshold, it means that the voxel has a significant weakness. The overall risk is suppressed by the weakness and will not be amplified by the synergy. Therefore, the three indices are weighted and geometrically averaged, but no synergy gain amplification is applied to obtain the overall risk index of each voxel.

[0097] Specifically, the comprehensive risk index I can be expressed by the following formula:

[0098] Among them, I s I p I σ These are the structural characteristic index, physical property characteristic index, and stress characteristic index, respectively. s w p w σ These represent the weights corresponding to the structural characteristic index, physical property characteristic index, and stress characteristic index, respectively. λ is a preset gain coefficient used to adjust the amplification degree of the synergistic effect of the three elements. min Indicate I s I p I σ The minimum value in, I0 is the preset gain start threshold, max(0, I min -I0) means only if I s I p I σ Gain is activated only when all three indices exceed the preset gain activation threshold. The power of each sub-indice represents its relative contribution to the overall risk.

[0099] In one embodiment, while outputting the comprehensive risk index, the dominant component identifier is also output, and divided into four levels according to preset thresholds: strong impact risk, moderate impact risk, weak impact risk, and no significant impact risk, which are visually identified by red, orange, yellow, and green, respectively. This fusion method avoids false high risk caused by high values ​​of a single factor, and strengthens the risk expression when all three factors simultaneously enter the danger zone.

[0100] This application's embodiments determine a comprehensive risk index based on two fundamental physical characteristics of rockbursts: synergistic amplification and the weakest link effect. This avoids the limitations of the linear weighting method in terms of physical meaning. At the same time, the three sub-indices and the comprehensive index are output simultaneously, so that the high risk of any voxel can be decomposed into three components: structure, physical properties, and stress, providing a precise attribution basis for subsequent differentiated prevention and control measures.

[0101] To enable those skilled in the art to better understand the method for preventing and controlling rockbursts in mines provided in the embodiments of this application, the complete implementation process of the embodiments of this application is presented here in chronological order.

[0102] Reference Figure 3 This is an overall flowchart of a method for preventing and controlling rockbursts in mines, provided in an embodiment of this application, which specifically includes the following steps: Step 301: Mine basic data collection and initialization of 3D voxelized digital base.

[0103] Multi-source data for characterizing mine geological conditions is acquired. Based on the acquired multi-source data, a three-dimensional voxel mesh including multiple voxels is constructed along the three-dimensional space of the mine. The multi-source data is mapped and aggregated into the corresponding voxels according to their respective spatial coordinates, so that each voxel contains multi-dimensional attributes such as geometric information, structural information, physical property information, stress information and meta-attributes, and the initialization of the three-dimensional voxelized digital base 102 is completed.

[0104] Step 302: Perform structural inversion, physical property inversion, and stress inversion along the three parallel branches respectively.

[0105] Based on three-dimensional voxel grids and different data subsets from multi-source data, digital inversion is performed for the three elements of rockburst. The structural inversion branch inverts structural characteristic indices based on seismic exploration data, drilling data, and well logging data; the physical property inversion branch inverts physical property characteristic indices based on acoustic monitoring data and mechanical test data; and the stress inversion branch inverts stress characteristic indices based on in-situ stress test data and microseismic monitoring data. The three branches operate independently and in parallel, outputting structural characteristic index fields, physical property characteristic index fields, and stress characteristic index fields expressed in continuous numerical form using voxels as the basic spatial unit.

[0106] Step 303: Three-element coupling identification, outputting the initial version (Comprehensive Risk Index) CIBI voxel field and preliminary rockburst hazard zoning.

[0107] The outputs from the three branches are fused onto a 3D voxel grid, and the comprehensive risk index is calculated independently for each voxel. The fusion process combines a weighted geometric mean with a synergistic gain term. When all three indices of a voxel exceed a preset gain threshold, a synergistic gain is applied to the fusion result; when any element's index is low, the comprehensive index is suppressed through the multiplier effect of the weighted geometric mean. Based on the comprehensive risk index of each voxel, risk levels are classified according to preset thresholds, and an initial version of the comprehensive shock risk index voxel field and preliminary rockburst hazard zoning are output.

[0108] Step 304: After entering the mining stage, access the multi-source monitoring data in the well for dynamic updates, and trigger graded early warning and differentiated prevention and control suggestions.

[0109] After a newly built mine enters the mining phase, the microseismic monitoring data, mine pressure monitoring data, drill cuttings monitoring data, electromagnetic radiation monitoring data, and online stress monitoring data continuously generated by the underground monitoring system are categorized into corresponding index categories according to the element-specific mapping rules. After time alignment and quality control, these data are written into the extended state vector as the observation input for Kalman filtering. Based on the extended state vector and the Kalman filtering algorithm, the three-element index and comprehensive risk index of voxels in the affected area are incrementally updated. After the update is completed, it is written back to the digital base and the timestamp is updated. Subsequently, voxels within a specified range in front of the working face are scanned. Voxels that exceed the risk threshold and are spatially connected are clustered to form candidate risk areas. When a candidate risk area reaches the preset trigger condition, an early warning event is generated. The early warning is classified into four levels—red, orange, yellow, and green—according to the comprehensive risk index threshold for graded early warning. The dominant factors are determined based on the index set of voxels contained in the candidate risk area. Differential prevention and control measures are then matched according to the type of dominant factor. When structure is dominant, measures such as grouting reinforcement, advanced support, and avoidance structures are recommended. When physical properties are dominant, measures to reduce the energy storage capacity of the coal seam, such as large-diameter pressure-relief boreholes and coal seam water injection, are recommended. When stress is dominant, measures to reduce stress concentration, such as coal seam blasting pressure relief, adjusting the advance speed, and optimizing mining deployment, are recommended. The closed-loop system feeds back observation information upstream after each monitoring cycle to allow for rolling calibration of the inversion model and weights.

[0110] It should be noted that the three inversion branches in step 302 above are executed in parallel, not sequentially. After the initialization of the 3D voxelized digital base 102, the three branches of structural inversion, physical property inversion, and stress inversion run independently and in parallel, without interdependence, and output the structural characteristic index field, physical property characteristic index field, and stress characteristic index field respectively before converging at the three-element coupling identification node. Step 304 constitutes a dynamic update closed loop, which runs continuously during the mining phase. After each loop, observation information is fed back upstream to perform rolling calibration of the inversion model and weights.

[0111] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0112] This application also provides a storage medium that stores computer instructions. When the computer executes the computer instructions, it is used to perform various processes of the above-described method for preventing and controlling rockbursts in mines, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0113] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0114] This application also provides an electronic device, including a processor 4010, a memory 409, and a program or instructions stored in the memory 409 and executable on the processor 4010. When the program or instructions are executed by the processor 4010, they implement the various processes of the above-described mine rockburst prevention and control method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0115] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0116] Figure 4 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of this application.

[0117] The electronic device 400 includes, but is not limited to, components such as: radio frequency unit 401, network module 402, audio output unit 403, input unit 404, sensor 405, display unit 406, user input unit 407, interface unit 408, memory 409, and processor 4010.

[0118] Those skilled in the art will understand that the electronic device 400 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 4010 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0119] This application also provides a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, it implements the various processes of the above-described method for preventing and controlling rockbursts in mines and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0120] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0121] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0122] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for preventing and controlling rockbursts in mines, characterized in that, include: Acquire multi-source data to characterize the geological conditions of the mine; Based on the multi-source data, a three-dimensional voxel mesh including multiple voxels is constructed along the three-dimensional space of the mine. Based on the multi-source data and the three-dimensional voxel grid, the structural characteristic index for characterizing coal body structural elements, the physical property characteristic index for characterizing coal body physical property elements, and the stress characteristic index for characterizing stress field elements of each voxel are respectively inverted. A strategy for preventing and controlling mine rockbursts is generated based on the multi-source data, the structural characteristic index, the physical property characteristic index, and the stress characteristic index.

2. The prevention and control method according to claim 1, characterized in that, The multi-source data includes geological data, test data, and monitoring data. Based on the multi-source data, constructing a three-dimensional voxel mesh comprising multiple voxels along the three-dimensional space of the mine includes: Based on the geological data, determine the spatial extent of the three-dimensional space and the initial grid step size; The initial grid step size is adaptively adjusted based on the geological data and the test data to obtain the adjusted grid step size, and a grid framework is generated based on the adjusted grid step size. The spatial range is discretized in three dimensions according to the grid framework to obtain multiple grid surfaces arranged along the three-dimensional space; Voxels are defined based on spatial units formed by adjacent mesh surfaces, and a three-dimensional voxel mesh composed of multiple voxels is generated within the spatial range. The geological data, test data, and monitoring data belonging to the same spatial range are aggregated to obtain the initial attribute value of each voxel; Each voxel is assigned an attribute based on its initial attribute value.

3. The prevention and control method according to claim 2, characterized in that, The geological data includes seismic exploration data, drilling data, and well logging data. Based on the multi-source data and the three-dimensional voxel grid, structural characteristic indices for each voxel to characterize coal body structural elements are inverted, including: Seismic attributes are extracted from the seismic exploration data, and samples are taken according to the spatial coordinates of each voxel in the three-dimensional voxel grid to obtain the multidimensional seismic attribute vector of each voxel. The multidimensional seismic attribute vector is processed to obtain the principal component attributes of each voxel; The spatial neighborhood relationship between each voxel is determined based on the spatial coordinates of each voxel in the three-dimensional voxel grid, and a three-dimensional data block is constructed based on the principal component attributes and the spatial neighborhood relationship. The structural anomaly probability of each voxel is determined based on the three-dimensional data block and the drilling data. Based on the drilling data and the logging data, the borehole structure constraints of each borehole in the mine are determined; Based on the borehole structure constraints and the multidimensional seismic attribute vector, collaborative interpolation is performed in the three-dimensional voxel grid to obtain a continuous structural parameter field. Based on the structural anomaly probability and the continuous structural parameter field, the structural characteristic index of each voxel is determined.

4. The prevention and control method according to claim 3, characterized in that, The test data includes acoustic monitoring data and mechanical test data. Based on the multi-source data and the three-dimensional voxel grid, the physical property indexes used to characterize the physical properties of coal are inverted for each voxel, including: Based on the acoustic monitoring data, the wave velocity parameters of each voxel are obtained by inversion in the three-dimensional voxel grid. Based on the drilling data, the mechanical test data, the wave velocity parameters, and the structural characteristic index, the set of mechanical parameter indices for each voxel is inverted; By fusing the mechanical parameter indices in the set of mechanical parameter indices, the physical property index of each voxel is obtained.

5. The prevention and control method according to claim 2, characterized in that, The geological data includes tectonic background information and mining spatial location; the test data includes in-situ rock stress test data; and the monitoring data includes microseismic monitoring data. Based on the multi-source data and the three-dimensional voxel grid, stress characteristic indices for characterizing stress field elements are inverted for each voxel, including: Based on the original rock stress test data and the tectonic background information, the original rock stress field of the mine is constructed in the three-dimensional voxel grid; the original rock stress field is used to characterize the spatial distribution of in-situ stress in the mine when it is not affected by mining activities; Based on the mining location of the mine and the original rock stress field, the mining-induced stress field of the mine is determined; the mining-induced stress field is used to characterize the spatial distribution of ground stress in the mine under the influence of mining activities. The mining-induced stress field is corrected based on the microseismic monitoring data, and the stress concentration factor and elastic energy density of each voxel are determined based on the corrected mining-induced stress field. The stress characteristic index of each voxel is determined based on the stress concentration factor and the elastic energy density.

6. The prevention and control method according to claim 2, characterized in that, The strategy for preventing and controlling mine rockbursts based on the multi-source data, the structural characteristic index, the physical property characteristic index, and the stress characteristic index includes: The comprehensive risk index of each voxel is determined based on the structural characteristic index, the physical property characteristic index, and the stress characteristic index; Based on the monitoring data, the structural characteristic index, physical property index, stress characteristic index, and comprehensive risk index of the voxels in the affected area of ​​the three-dimensional voxel grid are incrementally updated to obtain the index set of each voxel. Based on the index set of each voxel, voxels that meet the preset clustering requirements are clustered to form candidate risk regions; When the candidate risk area reaches the preset triggering condition, an early warning event is generated to characterize the presence of rockburst risk in the candidate risk area; The dominant element of the warning event is determined based on the index set of voxels contained in the candidate risk area; Based on the dominant factors of the warning event, a prevention and control strategy for reducing the risk of rockburst in the candidate risk areas is generated.

7. The prevention and control method according to claim 6, characterized in that, The determination of the comprehensive risk index for each voxel based on the structural characteristic index, the physical property characteristic index, and the stress characteristic index includes: Extract the confidence level of each voxel from its attributes; The weights of the structural characteristic index, physical property index and stress characteristic index of each voxel are determined based on the confidence level of each voxel. Determine whether the structural characteristic index, physical property characteristic index, and stress characteristic index of each voxel are all greater than the preset gain activation threshold; If the structural characteristic index, physical property index, and stress characteristic index of each voxel are all greater than the preset gain activation threshold, then the structural characteristic index, physical property index, and stress characteristic index of each voxel are fused based on the preset gain coefficient and the weight to obtain the comprehensive risk index of each voxel. If at least one of the structural characteristic index, physical property index, and stress characteristic index of each voxel is less than or equal to the preset gain activation threshold, then the structural characteristic index, physical property index, and stress characteristic index of each voxel are fused based on the weights to obtain the comprehensive risk index of each voxel.

8. A storage medium, characterized in that, The storage medium stores computer instructions, which, when executed by the computer, are used to perform a method for preventing and controlling rockbursts in mines as described in any one of claims 1-7.

9. An electronic device, characterized in that, Includes at least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a method for preventing and controlling rockbursts in mines as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements a method for preventing and controlling rockbursts in mines as described in any one of claims 1-7.

Images