Rock burst risk identification method for coal pillar

Abstract

The invention discloses a rock burst risk identification method for a coal pillar, and the method comprises the steps: constructing a multi-source monitoring data set, extracting a plurality of dynamic characteristic parameters, such as micro-seismic energy and stress concentration ratio, constructing a comprehensive precursor score in combination with a coal pillar simulation critical value, and introducing a classification model to carry out the risk grade identification. The problem that a traditional method depends on a single index and early warning staticization is solved, meanwhile, through fusion of finite element simulation and real-time monitoring, the expression ability of the stability evolution process of the coal pillar is enhanced, the systematic recognition process from feature extraction, score evaluation to early warning output is achieved, the accuracy and adaptability of rock burst risk recognition are enhanced, and the risk recognition efficiency is improved. And the capabilities of high-discrimination risk discrimination and field application adaptability are realized.

Description

Technical Field

[0001] This invention relates to the field of rockburst risk prevention technology for coal pillars, and more specifically, to a method for identifying rockburst risks in coal pillars. Background Technology

[0002] As mining depth increases, the level of ground stress rises significantly, and strong stress concentration zones easily form around the coal pillar, further inducing dynamic disasters such as rockbursts. Rockbursts, characterized by their sudden occurrence, violent energy release, and wide-ranging impact, are one of the major types of disasters restricting safe production in mines.

[0003] The incubation and triggering process of rockbursts often manifests as a complex dynamic process involving the coupling and evolution of multiple physical fields, exhibiting significant nonlinear characteristics and spatiotemporal heterogeneity. Traditional monitoring and identification methods mostly focus on trend judgment of single indicators or static early warning based on thresholds, lacking the ability to collaboratively process multi-source monitoring data and dynamically model the mechanical behavior of coal pillars, making it difficult to accurately reflect their stability evolution process. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying the risk of rockburst in coal pillars, so as to solve the problems mentioned in the background art.

[0005] According to one aspect of this application, a method for identifying rockburst risk in coal pillars is provided, the method comprising the following steps: S1. Collect monitoring information of the area where the coal pillar is located. The monitoring information includes microseismic signals, rock strain values, geostress values ​​and surrounding rock pressure values. At the same time, acquire the geometric dimensions of the coal pillar and the geological parameters of the rock strata to construct a multi-source monitoring dataset for the coal pillar. S2. The multi-source monitoring data of the coal pillar is filtered, denoised, standardized and spatiotemporally aligned to extract four types of monitoring characteristic parameters: local stress concentration, microseismic energy concentration, strain rate and pressure change rate of the coal pillar, and a set of monitoring characteristic parameter vectors is constructed. S3. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a stress field evolution model of the coal pillar is established using the finite element method, the simulation results are determined, and the critical stress value σ_cr and critical strain threshold ε_cr of the coal pillar are determined according to the simulation results. S4. The monitoring feature parameter vector, the instability critical stress value σ_cr, and the critical strain threshold ε_cr are used as input features, and weighted based on preset weight coefficients to determine the comprehensive precursor score. S5. Input the comprehensive precursor score into the pre-trained classification model and output the risk level of coal pillar rockburst; S6. When the risk level reaches the warning threshold, generate and output coal pillar rock burst warning information.

[0006] Preferably, in step S3, extracting the instability critical stress value σ_cr and the critical strain threshold ε_cr includes the following steps: S31. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a three-dimensional finite element model is constructed, boundary conditions and load conditions are set, and stress-strain process simulation is performed. S32. Obtain the principal stress-strain curves on the central axis of the coal pillar or in the high stress concentration area; S33. Identify the first stress peak value σ_max that appears in the curve and set it as σ_cr, and at the same time extract the strain value corresponding to this point and set it as ε_cr. S34. When the stress-strain curve has multiple peaks or multiple plastic yield zones, select the peak point that first enters the descending segment as the corresponding position of σ_cr.

[0007] Preferably, in step S31, constructing the three-dimensional finite element model includes: Obtain precise size data of the coal pillar and collect geological information on the thickness, dip angle, and bedding distribution of the top and bottom strata of the coal pillar and input it into the model, where the model includes the coal pillar and the surrounding rock area of ​​at least 3 times the size of the coal pillar; Set the coal body parameters, including elastic modulus, Poisson's ratio, cohesion, internal friction angle, tensile strength, density, etc., and input the elastic modulus, strength parameters and joint surface characteristics of the top and bottom rock strata in layers; Boundary constraints are set, with the bottom boundary set as a fixed constraint to simulate the solid support effect between the coal seam floor and the deep bedrock, the four sides set as horizontal displacement constraints to simulate the lateral pressure of the surrounding rock on the coal pillar, and the bottom boundary set to apply vertical stress to simulate the self-weight of the overlying strata. The static load is set to calculate the vertical stress based on the burial depth, σ_v = ρgh, where ρ is the rock density and h is the burial depth; High-density grids are used to generate coal pillar areas, and gradually sparse grids are used to generate surrounding rock areas. Local grid densification is applied to high stress concentration areas. Dynamic loads are applied to the top or side of the coal pillar, and the transient response is solved by time history analysis.

[0008] Preferably, applying dynamic loads to the top or side of the coal pillar includes: Treating the coal pillar under blasting dynamic load as a linear time-invariant system, the blasting vibration waveform is decomposed into a superposition of intensity nonstationary and frequency nonstationary. The Gamma function is used to describe the variation of blasting vibration intensity with time.

[0009] Where A0 is the peak acceleration, t p α represents the peak time, and β represents the fitting parameters.

[0010] Preferably, the training process of the classification model includes the following steps: S51. Collect coal mine monitoring samples with historical rockburst records, the samples including monitoring characteristic parameter vectors, scoring values ​​and historical risk level labels; S52. Perform data cleaning, outlier removal, and standardization on the sample data; S53. Select a supervised classification model for training; S54. Input the score value and monitoring feature parameter vector, and use the historically labeled risk level as the output label. Construct the training set and validation set to complete the model training and parameter tuning. S55. Use the trained model to predict the risk level of new data.

[0011] Preferably, the formula for the comprehensive prodromal symptoms score S is set as follows: ; in, For microseismic energy concentration, For strain rate, For stress concentration, For the rate of pressure change, , , , For the corresponding weighting coefficients, the following conditions must be met: .

[0012] Preferably, the formula for calculating the stress intensity is:

[0013] This represents the measured value of ground stress at a specific monitoring point. The time derivative of the geostress represents the slope of the stress change. This is the historical average. The It characterizes the rate of change under unit stress and sensitively reflects the risk of sudden stress increase or decrease.

[0014] The preferred formula for calculating microseismic energy concentration is:

[0015] , representing the energy of microseismic events within the i-th spatial unit. The proportion of total energy, To monitor the total energy of microseismic events falling into the i-th spatial grid within the monitoring period, The total number of units in the spatial division. The The smaller the value, the more concentrated the energy; the larger the value, the more dispersed the energy distribution.

[0016] Preferably, the formula for calculating the strain rate is:

[0017] This represents the strain value at the current moment. The time interval for strain sampling. The It represents the acceleration of strain change per unit time. A positive value indicates that the deformation is accelerating, while a negative value indicates that the deformation trend is slowing down.

[0018] Preferably, the formula for calculating the rate of pressure change is:

[0019] This is the overall score for the current moment. The score from the previous moment. A time-weighted coefficient is used to control the inertia and sensitivity of the scoring. The Used to smooth out score fluctuations.

[0020] This invention constructs a multi-source monitoring dataset, extracts multiple dynamic characteristic parameters such as microseismic energy and stress concentration, combines them with coal pillar simulation critical values ​​to construct a comprehensive precursor score, and introduces a classification model for risk level identification. This overcomes the problems of traditional methods relying on single indicators and static early warning. At the same time, by integrating finite element simulation and real-time monitoring, it enhances the ability to express the evolution process of coal pillar stability, realizes a systematic identification process from feature extraction, scoring evaluation to early warning output, enhances the accuracy and adaptability of rockburst risk identification, and achieves high-discrimination risk judgment and adaptability for field application. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method for identifying the risk of rockburst in a coal pillar according to an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 According to an embodiment of the present invention, a method for identifying rockburst risk in coal pillars is provided, the method comprising the following steps.

[0024] S1. Collect monitoring information of the area where the coal pillar is located. The monitoring information includes microseismic signals, rock strain values, geostress values ​​and surrounding rock pressure values. At the same time, acquire the geometric dimensions of the coal pillar and the geological parameters of the rock strata to construct a multi-source monitoring dataset for the coal pillar.

[0025] According to the above technical solution, by collecting multi-source information such as microseismic signals, strain values, and geostress values, a dynamic monitoring system covering the stress state and deformation behavior of the coal pillar and its surrounding rock is established. The geometric dimensions of the coal pillar and the geological parameters of the rock strata serve as the input basic data, providing the basis for boundary and material setting for subsequent simulation analysis.

[0026] Furthermore, the geometric dimensions of the coal pillar include its width, thickness, and height, while the geological parameters of the rock strata include rock mass density, elastic modulus, and Poisson's ratio.

[0027] S2. The multi-source monitoring data of the coal pillar is filtered, denoised, standardized and spatiotemporally aligned to extract four types of monitoring characteristic parameters: local stress concentration, microseismic energy concentration, strain rate and pressure change rate of the coal pillar, and a set of monitoring characteristic parameter vectors is constructed.

[0028] Based on the above technical solution, by using filtering, denoising and standardization, the interference caused by sensing errors or environmental factors in the monitoring data can be reduced, and the uniformity of the data in time and space can be improved. Based on this, the four types of feature parameters extracted are used to characterize the local stress and deformation response characteristics of the coal pillar, which is convenient for subsequent model quantification.

[0029] S3. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a stress field evolution model of the coal pillar is established using the finite element method. The simulation results are determined, and the critical stress value σ_cr and critical strain threshold ε_cr of the coal pillar are determined based on the simulation results.

[0030] According to the above technical solution, a finite element model is established based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata. Under controlled boundary and loading conditions, the stress-strain evolution process inside the coal pillar is numerically solved.

[0031] Furthermore, during the simulation, the central axis of the coal pillar or the stress concentration area is selected as the observation path. The principal stress-principal strain relationship curve is extracted, and the position of the first stress peak is identified. The stress value and strain value corresponding to the peak are defined as the instability critical stress σ_cr and critical strain ε_cr, respectively. When there are multiple peaks in the curve, the peak point that enters the falling segment earliest is selected as the instability characteristic point to characterize the simulation results of the coal pillar approaching the ultimate bearing state.

[0032] S4. Using the monitoring feature parameter vector, the instability critical stress value σ_cr, and the critical strain threshold ε_cr as input features, perform weighted processing based on preset weight coefficients to determine the comprehensive precursor score S.

[0033] According to the above technical solution, the monitoring parameters and simulated critical values ​​are used as inputs and combined by weight to form a comprehensive score. This score is used to express the degree of deviation of the current state of the coal pillar from the instability boundary, and realizes a normalized expression of the stability state from the parameter level.

[0034] S5. Input the comprehensive precursor score into the pre-trained classification model and output the risk level of coal pillar rockburst.

[0035] According to the above technical solution, the pre-trained classification model takes the scoring and monitoring features as input features and uses the established classification boundary to classify the current coal pillar status. Its principle is based on the mapping relationship established by the pattern recognition and supervised learning algorithms between samples.

[0036] S6. When the risk level reaches the warning threshold, generate and output early warning information for coal pillar rock burst.

[0037] According to the above technical solution, the set early warning threshold is used as a control trigger condition. When the system identification result meets the condition, an early warning instruction is generated through a preset logic process to drive the relevant system to respond.

[0038] In one embodiment, step S3, extracting the instability critical stress value σ_cr and the critical strain threshold ε_cr, includes the following steps.

[0039] S31. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a three-dimensional finite element model is constructed, boundary conditions and load conditions are set, and stress-strain process simulation is performed. Based on the above technical solution, a three-dimensional finite element model is constructed to approximate the complex structure and stress boundary of the coal pillar. The stress state of the coal pillar under actual working conditions is simulated by applying boundary conditions such as ground stress and mining disturbance.

[0040] In one embodiment, constructing a three-dimensional finite element model includes the following implementation steps: Geometric modeling and parameter preparation: Obtain the precise dimensions (height, width, length, shape) of the coal pillar based on borehole data or mining design drawings and input them. Collect geological information such as the thickness, dip angle, and bedding distribution of the top and bottom strata of the coal pillar. The model includes the coal pillar and the surrounding rock area of ​​at least 3 times the size of the coal pillar to eliminate boundary effects.

[0041] Material properties and constitutive model selection: Set coal body parameters, including elastic modulus, Poisson's ratio, cohesion, internal friction angle, tensile strength, density, etc. Input the elastic modulus, strength parameters and joint surface characteristics of the top and bottom rock strata in layers. The coal body adopts the elastoplastic model Mohr-Coulomb, and the surrounding rock adopts isotropic elasticity.

[0042] In one embodiment, if a seepage field (such as an aquifer) exists, the fluid-structure interaction module is activated to set the permeability coefficient and pore pressure boundaries. If the coal body has a fractured structure, a discrete fracture network (DFN) model needs to be introduced.

[0043] Boundary conditions and load application: Set boundary constraints, where the bottom boundary is set as a fixed constraint (displacement U=0) to simulate the solid support effect between the coal seam floor and the deep bedrock; the four sides are set as horizontal displacement constraints (Ux=0, Uy=0) to simulate the lateral pressure of the surrounding rock on the coal pillar; the middle boundary is set to apply vertical stress to simulate the self-weight or tectonic stress of the overlying strata.

[0044] The static load is set to calculate the vertical stress based on the burial depth, σ_v = ρgh, where ρ is the rock density and h is the burial depth.

[0045] In one embodiment, mining disturbances or seismic waves are considered, and dynamic loads such as sinusoidal or shock wave time histories are applied. If a tectonic residual stress field exists, measured horizontal stress components must be input.

[0046] Mesh generation and element type: High-density meshes (element size ≤ 1 / 10 of the minimum coal pillar size) are used in the coal pillar area, while gradually sparse meshes are used in the surrounding rock area. Local mesh refinement is applied to high stress concentration areas (such as the coal pillar edge). 8-node hexahedral elements (C3D8) or higher-order tetrahedral elements (C3D10) are used for coal and surrounding rock. Joints or faults are simulated using cohesive elements or surface-to-surface contact pairs to simulate slippage and separation.

[0047] In one embodiment, stress calculation results under different mesh densities are compared to ensure that the error is less than 5%.

[0048] Dynamic perturbation and geological structure modeling: Dynamic loads, such as blasting vibration waveforms, are applied to the top or sides of the coal pillar, and the transient response is solved through time history analysis. Based on geological exploration data, fault planes or joint groups are inserted into the model, and their mechanical parameters, including friction coefficient and normal / tangential stiffness, are set. If the fault may be activated, a slip criterion (such as Coulomb's friction law) is used.

[0049] In one embodiment, the theoretical model for constructing the blasting vibration waveform includes treating the coal pillar under blasting dynamic load as a linear time-invariant system, and decomposing the blasting vibration waveform into a superposition of intensity nonstationary and frequency nonstationary states. The Gamma function is used to describe the variation of blasting vibration intensity with time.

[0050] Where A0 is the peak acceleration, t p α represents the peak time, and β represents the fitting parameters.

[0051] In one embodiment, the specific parameter settings in the above model are shown in the table below: parameter coal body Top sandstone Bottom mudstone elastic modulus 2.5GPa 15GPa 8GPa Poisson's ratio 0.25 0.22 0.28 Cohesion 1.8MPa 3.5MPa 2MPa internal friction angle 35° 42° 28° tensile strength 0.5MPa 1.2MPa 0.8MPa S32. Obtain the principal stress-strain curves on the central axis of the coal pillar or in the high stress concentration area; Based on the above technical solution, extracting the stress-strain curves of the central axis of the coal pillar or the local high-stress zone, and identifying the local strength change trend and overall deformation response process of the material, is the basic data source for identifying the instability initiation point.

[0052] S33. Identify the first stress peak value σ_max in the curve and set it as σ_cr. At the same time, extract the strain value corresponding to this point and set it as ε_cr. According to the above technical solution, the first stress peak in the stress-strain curve usually corresponds to the limit point of the coal pillar's bearing capacity. The stress and strain values ​​at this location are extracted to define the instability index of the coal pillar under simulated conditions.

[0053] S34. When the stress-strain curve has multiple peaks or multiple plastic yield zones, select the peak point that first enters the descending segment as the corresponding position of σ_cr.

[0054] According to the above technical solution, when there are multiple stress peaks or non-monotonic change segments, selecting the peak corresponding to the first descending segment can avoid overestimating the stability of the coal pillar and determine its earliest instability trend from the initial turning point of the structural response.

[0055] In one embodiment, the training process of a classification model specifically includes the following steps: S51. Collect coal mine monitoring samples with historical rockburst records. The samples include monitoring characteristic parameter vectors, score values, and historical risk level labels.

[0056] According to the above technical solution, the collection of sample data with historical event labels provides a labeled training basis for the classification model, which facilitates the model to establish a mapping relationship between input features and output risk levels during the learning process.

[0057] S52. Perform data cleaning, outlier removal, and standardization on the sample data.

[0058] According to the above technical solutions, data cleaning and standardization can reduce the bias caused by differences in data distribution and outliers in model training, thereby improving the stability and convergence efficiency of the model learning process.

[0059] S53. Select a supervised classification model for training.

[0060] According to the above technical solution, the supervised classification model trains the classification boundary based on existing labeled data, and in principle utilizes the algorithm structure to mine the discriminative ability of input features.

[0061] Furthermore, the classification model can be any one of the following: support vector machine, random forest, gradient boosting decision tree, or deep neural network model.

[0062] S54. Input the score value and monitoring feature parameter vector, and use the historically labeled risk level as the output label. Construct the training set and validation set to complete the model training and parameter tuning.

[0063] According to the above technical solution, training and validation sets are constructed for model parameter adjustment and performance evaluation. The process is based on the principle of data partitioning and loss function minimization to calibrate and optimize the model's predictive ability.

[0064] S55. Use the trained model to predict the risk level of new data.

[0065] According to the above technical solution, when the trained model is used for new data identification, its role is to classify and judge unknown samples using existing decision functions, thereby achieving rapid estimation of risk level.

[0066] In one embodiment, specifically, the formula for the comprehensive aura score S is: ; in, For microseismic energy concentration, For strain rate, For stress concentration, For the rate of pressure change, , , , For the corresponding weighting coefficients, the following conditions must be met: .

[0067] According to the above technical solution, the process of constructing a comprehensive precursor score is based on the normalized representation of multiple monitoring parameters, and the importance of different indicators is relatively measured by setting weight coefficients. In this scoring function, various physical quantities reflect different response mechanisms of the coal pillar's stable state: microseismic energy concentration is used to describe the spatial concentration of energy release, strain rate characterizes the local structural deformation evolution trend, stress concentration characterizes the accumulation of local stress anomalies, and pressure change rate reflects the degree of dynamic disturbance of surrounding rock pressure.

[0068] The weighting coefficients can be determined based on historical data statistical analysis, expert experience, or parameter optimization algorithms, so that the scoring results numerically reflect the comprehensive trend of the coal pillar state approaching the instability boundary, thereby providing a unified input scale for subsequent classification models.

[0069] In one embodiment, specifically, the formula for calculating the stress intensity is:

[0070] This represents the measured value of ground stress at a specific monitoring point. The time derivative of the geostress represents the slope of the stress change. This is the historical average. It characterizes the rate of change under unit stress and sensitively reflects the risk of sudden stress increase or decrease.

[0071] According to the above technical solution, stress concentration is defined as the difference between the measured ground stress and its historical mean, divided by the time derivative. This expression reflects the fluctuation of local stress state from two dimensions: the degree of deviation and the rate of change. By introducing the time derivative as a normalization factor, the model's response sensitivity to short-term stress abrupt changes can be enhanced, and the expression of indicators under different working conditions can be standardized.

[0072] In one embodiment, the specific formula for calculating the microseismic energy concentration is as follows:

[0073] , representing the energy of microseismic events within the i-th spatial unit. The proportion of total energy, To monitor the total energy of microseismic events falling into the i-th spatial grid within the monitoring period, The total number of units in the spatial division. The smaller the value, the more concentrated the energy; the larger the value, the more dispersed the energy distribution.

[0074] According to the above technical solution, the microseismic energy concentration is constructed based on the normalized sum of squares of spatial energy distribution, reflecting the energy concentration of microseismic events in the spatial domain per unit time. The parameters can be used as a structured representation of the spatial distribution of energy disturbances during the modeling process, which is beneficial for the subsequent model to sensitively model the activity of the seismic source.

[0075] In one embodiment, the specific formula for calculating the strain rate is:

[0076] This represents the strain value at the current moment. The time interval for strain sampling. It represents the acceleration of strain change per unit time. A positive value indicates that the deformation is accelerating, while a negative value indicates that the deformation trend is slowing down.

[0077] According to the above technical solution, the strain rate is expressed as the difference between the current strain value and the previous strain value divided by the time interval, which is a numerical approximation of the first derivative. It is used to describe the deformation intensity of the structure per unit time, and combined with its sign information, it can further infer the local deformation trend, making it suitable for dynamic input expression in real-time continuous monitoring scenarios.

[0078] In one embodiment, the specific formula for calculating the rate of pressure change is:

[0079] This is the overall score for the current moment. The score from the previous moment. The time-weighted coefficient controls the inertia and sensitivity of the scoring. Used to smooth out score fluctuations.

[0080] According to the above scheme, the rate of pressure change is calculated by the difference between the current and previous comprehensive score values, and a weighting coefficient is introduced to adjust the influence of this difference on the continuous sequence. This structure is essentially a score change response within a time-sliding window, reflecting the rate of change of the pressure signal within the score domain, and balancing the hysteresis and sensitivity of the score response through parameter settings.

[0081] All parts not described in this invention are the same as or can be implemented using existing technology. Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identifying the risk of rockburst in coal pillars, characterized in that, The method includes the following steps: S1. Collect monitoring information of the area where the coal pillar is located. The monitoring information includes microseismic signals, rock strain values, geostress values ​​and surrounding rock pressure values. At the same time, acquire the geometric dimensions of the coal pillar and the geological parameters of the rock strata to construct a multi-source monitoring dataset for the coal pillar. S2. The multi-source monitoring data of the coal pillar is filtered, denoised, standardized and spatiotemporally aligned to extract four types of monitoring characteristic parameters: local stress concentration of the coal pillar, microseismic energy concentration, strain rate and pressure change rate, and a set of monitoring characteristic parameter vectors is constructed. S3. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a stress field evolution model of the coal pillar is established using the finite element method, the simulation results are determined, and the critical stress value σ_cr and critical strain threshold ε_cr of the coal pillar are determined according to the simulation results. S4. The monitoring feature parameter vector, the instability critical stress value σ_cr, and the critical strain threshold ε_cr are used as input features, and weighted based on preset weight coefficients to determine the comprehensive precursor score S. S5. Input the comprehensive precursor score into the pre-trained classification model and output the risk level of coal pillar rockburst; S6. When the risk level reaches the warning threshold, generate and output coal pillar rock burst warning information.

2. The method according to claim 1, characterized in that, In step S3, determining the critical stress value σ_cr and the critical strain threshold ε_cr for instability includes the following steps: S31. Based on the geometric dimensions of the coal pillar and the geological parameters of the rock strata, a three-dimensional finite element model is constructed, boundary conditions and load conditions are set, and stress-strain process simulation is performed. S32. Obtain the principal stress-strain curves on the central axis of the coal pillar or in the high stress concentration area; S33. Identify the first stress peak value σ_max that appears in the curve and set it as σ_cr, and at the same time extract the strain value corresponding to this point and set it as ε_cr. S34. When the stress-strain curve has multiple peaks or multiple plastic yield zones, select the peak point that first enters the descending segment as the corresponding position of σ_cr.

3. The method according to claim 2, characterized in that, In step S31, constructing the three-dimensional finite element model includes: Obtain precise size data of the coal pillar and collect geological information on the thickness, dip angle, and bedding distribution of the top and bottom strata of the coal pillar and input it into the model, where the model includes the coal pillar and the surrounding rock area of ​​at least 3 times the size of the coal pillar; Set the coal body parameters, including elastic modulus, Poisson's ratio, cohesion, internal friction angle, tensile strength, and density. Input the elastic modulus, strength parameters, and joint surface characteristics of the top and bottom rock strata in layers. Boundary constraints are set, with the bottom boundary set as a fixed constraint to simulate the solid support effect between the coal seam floor and the deep bedrock, the four sides set as horizontal displacement constraints to simulate the lateral pressure of the surrounding rock on the coal pillar, and the bottom boundary set to apply vertical stress to simulate the self-weight of the overlying strata. The static load is set to calculate the vertical stress based on the burial depth, σ_v = ρgh, where ρ is the rock density and h is the burial depth; High-density grids are used to generate coal pillar areas, and gradually sparse grids are used to generate surrounding rock areas. Local grid densification is applied to high stress concentration areas. Dynamic loads are applied to the top or side of the coal pillar, and the transient response is solved by time history analysis.

4. The method according to claim 3, characterized in that, Applying dynamic loads to the top or sides of a coal pillar includes: treating the coal pillar under blasting dynamic load as a linear time-invariant system, decomposing the blasting vibration waveform into a superposition of intensity non-stationary and frequency non-stationary components, and using the Gamma function to describe the variation of blasting vibration intensity with time. Where A0 is the peak acceleration, t p α represents the peak time, and β represents the fitting parameters.

5. The method according to claim 1, characterized in that, The training process of the classification model includes the following steps: S51. Collect coal mine monitoring samples with historical rockburst records, the samples including monitoring characteristic parameter vectors, score values ​​and historical risk level labels; S52. Perform data cleaning, outlier removal, and standardization on the sample data; S53. Select a supervised classification model for training; S54. Input the score value and monitoring feature parameter vector, and use the historically labeled risk level as the output label. Construct the training set and validation set to complete the model training and parameter tuning. S55. Use the trained model to predict the risk level of new data.

6. The method according to claim 1, characterized in that, in, The comprehensive prodromal score S is set as follows: ； in, For microseismic energy concentration, For strain rate, For stress concentration, For the rate of pressure change, , , , For the corresponding weighting coefficients, the following conditions must be met: .

7. The method according to claim 5, characterized in that, in, The formula for calculating stress concentration is: This represents the measured value of ground stress at a specific monitoring point. The time derivative of the geostress represents the slope of the stress change. The historical average, the It characterizes the rate of change under unit stress and sensitively reflects the risk of sudden stress increase or decrease.

8. The method according to claim 6, characterized in that, The formula for calculating the energy concentration of microseismic events is: , representing the energy of microseismic events within the i-th spatial unit. The proportion of total energy, To monitor the total energy of microseismic events falling into the i-th spatial grid within the monitoring period, The total number of units in the spatial division, the The smaller the value, the more concentrated the energy; the larger the value, the more dispersed the energy distribution.

9. The method according to claim 6, characterized in that, The formula for calculating strain rate is: This represents the strain value at the current moment. The strain sampling time interval, the It represents the acceleration of strain change per unit time. A positive value indicates that the deformation is accelerating, while a negative value indicates that the deformation trend is slowing down.

10. The method according to claim 6, characterized in that, The formula for calculating the rate of pressure change is: This is the overall score for the current moment. The score from the previous moment. The time-weighted coefficient controls the inertia and sensitivity of the scoring. Used to smooth out score fluctuations.

Images