Coal mine pressure risk assessment method based on complex system dynamics

Abstract

The invention discloses a coal mine pressure risk assessment method based on complex system dynamics, belongs to the technical field of mine pressure risk assessment, is used for mine pressure risk assessment, and constructs a complex system dynamics model, combines various sensor data, and carries out real-time monitoring and risk assessment of mine pressure changes. The method comprises the steps of geological data acquisition, sensor data real-time monitoring, data preprocessing, mine pressure prediction model construction, model training and optimization, real-time risk assessment and the like, data of a stress sensor, a displacement sensor and an acoustic emission sensor are comprehensively utilized, and data preprocessing is performed by adopting a Kalman filtering algorithm; and optimizing the mine pressure prediction model by using a graph convolutional network and an attention mechanism, and performing real-time risk assessment through an adaptive graph neural network. According to the method, the adaptive graph neural network is adopted for real-time risk assessment, potential risks in mine pressure changes can be found in time, risk scores are generated, and scientific decision support is provided for coal mine managers.

Description

Technical Field

[0001] The invention discloses a coal mine pressure risk assessment method based on complex system dynamics, belonging to the technical field of coal mine pressure risk assessment. Background Art

[0002] During the entire coal mining process, accurate assessment of mine pressure risk is an important part of ensuring safe production in coal mines. The change of mine pressure is affected by many factors, including geological structure, mining intensity, mining process, etc. These factors interact with each other to form a complex dynamic system. Traditional mine pressure monitoring methods usually rely on experience judgment and single sensor data, which cannot fully and real-time reflect the change of mine pressure, and easily lead to lag and inaccuracy in mine pressure risk assessment.

[0003] In the prior art, stress sensors, displacement sensors and acoustic emission sensors are mainly used to monitor the changes in mine pressure in coal mining areas. These sensors work independently, and the data obtained are isolated and unrelated, and cannot form comprehensive information on mine pressure changes. In addition, traditional mine pressure monitoring systems have limitations in data processing and analysis. They usually use simple statistical analysis methods, which cannot fully utilize the complex information in the sensor data, resulting in insufficient accuracy and reliability of mine pressure risk assessment results.

[0004] For data processing, the existing technologies mostly use methods based on empirical models. These models rely on historical data and manually set parameters, lack dynamic adaptability, and are difficult to cope with the complex and changeable geological conditions and mine pressure changes during coal mining. Traditional model training methods have the problem of poor generalization ability and cannot obtain consistent prediction results under different geological conditions and mining processes. At the same time, the data preprocessing methods in the existing technologies are relatively simple and it is difficult to effectively remove noise and outliers in sensor data, resulting in low quality of input data for the prediction model, thereby affecting the accuracy of mine pressure risk assessment.

[0005] In terms of the construction of the mine pressure risk assessment model, the existing technologies mostly use simple linear models or empirical formulas. These methods cannot fully capture the nonlinear dynamic characteristics of mine pressure changes and are difficult to accurately predict mine pressure risks. In addition, the existing technologies lack the comprehensive use of different sensor data and fail to fully consider the complex relationship between geological data and mine pressure change information, resulting in greater difficulty in the construction of prediction models and parameter optimization, and low prediction accuracy and reliability. Summary of the invention

[0006] The purpose of the present invention is to provide a coal mine pressure risk assessment method based on complex system dynamics to solve the problem of inaccurate coal mine pressure risk assessment in the prior art.

[0007] A coal mine pressure risk assessment method based on complex system dynamics, including:

[0008] S1. Collect geological data of the coal mining area, including rock layer thickness, rock layer hardness, fault distribution and physical property data of coal seam roof and floor;

[0009] S2. Stress sensors, displacement sensors and acoustic emission sensors are deployed in the coal mining area to monitor the mine pressure data in the coal mining area in real time to generate mine pressure change information, and upload D(t) to the central database. D(t) represents the set of mine pressure change information collected by all sensors at time t;

[0010] S3. Preprocess the collected geological data and mine pressure change information, use the Kalman filter algorithm to remove noise and outliers, and obtain a clean coal mining area data set D clean ;

[0011] S4. Based on the complex system dynamics model and combined with the coal mining area data set, a coal mine pressure prediction model is constructed;

[0012] S5. Use historical data to train the coal mine pressure prediction model, and use a method combining graph convolutional network and attention mechanism to optimize the parameters of the coal mine pressure prediction model to obtain the optimized coal mine pressure prediction model M. final ;

[0013] S6. Input the latest mine pressure change information into the trained coal mine pressure prediction model, use the adaptive graph neural network to perform real-time mine pressure risk assessment, and generate a risk score;

[0014] S7. Based on the results of the mine pressure risk assessment, the mine pressure changes during the coal mining process are dynamically predicted, and an adaptive filtering algorithm is used to generate a mine pressure risk warning signal. The warning signal includes three risk levels: low, medium and high. The mine pressure risk assessment results and warning signals are presented to coal mine managers in the form of mine pressure change graphs, risk level graphs and historical data comparison graphs to assist coal mine managers in making decisions.

[0015] S1 includes, S1.1, conducting borehole measurement in the coal mining area, arranging multiple borehole measurement points, using a multi-frequency laser rangefinder and a nonlinear sonic logging instrument to measure the thickness of rock formations at different depths, and obtaining rock formation thickness data H i , i represents the number of the rock layer, and the average value and standard deviation are calculated by repeated measurements:

[0016]

[0017] Where N is the number of measurements, H in represents the thickness of rock layer i measured at the nth time, represents the average thickness of rock layer i;

[0018] S1.2. Collect samples of various rock formations in the coal mining area, use compression test equipment to measure the hardness and compression test of the samples, obtain the hardness and compressive strength of the rock formation respectively, merge and calibrate the field measured data with the laboratory data, and determine the hardness and compressive strength data:

[0019]

[0020] In the formula, B i represents the hardness of rock layer i, M represents the number of measurements, B im represents the hardness of rock layer i measured for the mth time, ΔH represents the measurement deviation, H i represents the thickness of rock layer i, σ ci represents the compressive strength of rock layer i, σ cim represents the compressive strength of rock layer i measured for the mth time, Δ∈ represents the strain deviation, ∈ i represents the strain of rock layer i, α and β are adaptive data fusion weight factors.

[0021] S1 includes, S1.3, using geological radar and multi-frequency seismic wave reflection method to detect the distribution of faults in the coal mining area, and setting adaptive geological radar survey lines and multi-mode source receiving points to record the location of the faults F j , fault length L j and fault dip angle θ j , where j represents the fault number:

[0022]

[0023] In the formula, F j represents the location of fault j, K represents the number of radar lines, T k represents the measurement time of the kth measurement line, R(t) represents the radar reflection signal, θ j represents the dip angle of fault j, L j represents the length of fault j, S(t) represents the source signal, and t represents the time;

[0024] S1.4. Sampling the roof and floor of the coal seam to obtain measured density Elastic modulus and Poisson's ratio The microstructure of the samples was analyzed using laser interferometry and electron microscopy, and a multi-scale mechanical model of the physical properties of the top and bottom plates was established;

[0025]

[0026] In the formula, Indicates top and bottom plate k1 The volume of a point, ρ(x,y,z) represents the density distribution at the position (x,y,z), and (u,v,w) represent the displacement in the x,y,z directions respectively.

[0027] S2 includes, S2.1, arranging stress sensors, displacement sensors and acoustic emission sensors in the coal mining area, and the sensor arrangement plan is based on the geological characteristics of the coal mine and the mining process. i :

[0028] P i =(x i ,y i ,z i );

[0029] Where P i represents the position coordinates of sensor i, x i ,y i ,z i They represent the coordinates of the sensor in three-dimensional space respectively;

[0030] S2.2. Use stress sensors to monitor the changes in rock stress in the coal mining area in real time and obtain the stress value σ measured by sensor i at time t i (t):

[0031]

[0032] In the formula, σ 0 represents the initial stress value, Δσ i (t) represents the change in stress at time t, t 0 represents the initial time, It means the partial derivative of stress value with respect to the horizontal axis, It represents the partial derivative of stress value with respect to the ordinate, θ i is the sensor angle, κ n is the nth influence coefficient, τ n is the nth stress impact strength, E i is the elastic modulus of the rock formation, λ is the attenuation coefficient, t n is the time when the nth stress impact occurs, N is the total number of impacts, and t represents the moment.

[0033] S2 includes: S2.3, using displacement sensors to monitor the displacement changes of rock formations, and obtaining the displacement value u measured by sensor i at time t i (t):

[0034]

[0035] In the formula, u 0 represents the initial displacement value, Δu i(t) represents the change in displacement at time t, α i is the displacement amplitude, ω i is the angular frequency, is the initial phase angle, β i is the damping coefficient, γ j To affect the amplitude, η j is the vibration frequency, δ j is the attenuation coefficient, M is the total number of influencing factors, and j is the jth influencing factor;

[0036] S2.4. Use acoustic emission sensors to monitor the acoustic emission signals generated during coal mining, and obtain the acoustic emission signal intensity A received by sensor i at time t. i (t):

[0037] A i (t) = A 0 +ΔA i (t);

[0038] In the formula, A 0 Represents the background acoustic emission signal intensity, ΔA i (t) represents the change in the intensity of the acoustic emission signal at time t;

[0039] S2.5. Use the data synchronization algorithm to upload the collected stress data, displacement data and acoustic emission data to the central database via wireless transmission:

[0040]

[0041] In the formula, D(t) represents the set of all sensor data collected at time t, N 1 Indicates the total number of sensors.

[0042] S3 includes, S3.1, preliminary screening of the geological data processed by S1 and the mine pressure change information processed by S2, eliminating obviously erroneous data points, and forming a preliminary data set;

[0043] S3.2, apply Kalman filter algorithm to preprocess the preliminary data set to remove noise and outliers;

[0044] S3.3. Storing clean coal mining area data sets in a central database, the data sets containing geological data and mine pressure change information:

[0045]

[0046] Where D clean represents the clean coal mining area dataset, D raw represents the preliminary data set, including geological data and mine pressure change information, I is the unit matrix, K kThe Kalman gain matrix of the kth step, H is the observation matrix, z i is the observed value of the i-th step, K i H is the Kalman gain matrix at the i-th step.

[0047] S4 includes: S4.1. Based on the theory of complex system dynamics, the basic framework of the coal mine pressure prediction model is established, and the system state variable x(t) and system input u(t) are defined. t represents the time t. The clean coal mining area dataset D clean As model input, a coal mine pressure prediction model based on complex system dynamics is constructed:

[0048] x(t+1)=Ax(t)+Bu(t)+w(t); y(t)=Cx(t)+v(t);

[0049] In the formula, the system state variable x(t) represents the rock state at the current moment, and x(t) includes geological data {B i ,σ ci ,F j ,ρ k ,E k ,v k} and mine pressure change information D(t), the system input u(t) represents the mining intensity and coal mining progress, the process noise w(t) represents the unmeasured random disturbance, the observed output y(t) represents the measured mine pressure data, the measurement noise v(t) represents the error in the measurement process, and A, B, and C are the parameters to be determined for the model;

[0050] According to the system state transfer equation and observation equation, the coal mining area data set D clean Integrate the geological data and mine pressure change information into the coal mine pressure prediction model;

[0051] S4.2. Extract features from geological data and mine pressure change information and define feature vector f(t):

[0052] f(t)=[B i ,σc i ,F j ,ρ k ,E k ,v k ,D(t)];

[0053] Based on the theory of complex system dynamics, the mapping function φ is used to map the feature vector f(t) to the state variable x(t):

[0054] x(t)=φ(f(t));

[0055] Apply the state transfer equation to perform time series analysis on the eigenvector and calculate the state variable x(t+1) at the next moment:

[0056] x(t+1)=Aφ(f(t))+Bu(t)+w(t);

[0057] Apply the observation equation to transform the state variable f(t) into the observed output y(t):

[0058] y(t)=Cx(t)+v(t);

[0059] Correct A, B, C, w(t) and v(t) to ensure that the model accurately reflects the actual mine pressure change process;

[0060] S4.3. Compare and verify the corrected model with the actual monitoring data, adjust the model parameters to improve the prediction accuracy, and save the constructed mine pressure prediction model to the central database.

[0061] S5 includes, S5.1, collecting historical data of coal mines, including geological data and mine pressure change information, to form a historical data set D hist , preprocess the historical data set, input the processed data into the mine pressure prediction model, and construct the initial training data set D train ;

[0062] S5.2. Train the initial mine pressure prediction model based on the graph convolutional network and construct a graph structure. The nodes represent geological data and mine pressure change information, and the edges represent the relationship between the data. Define the node feature matrix H and the adjacency matrix A:

[0063] H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) W (l) );

[0064] Among them, H (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the activation function, and D is the degree matrix.

[0065] S5 includes, S5.3, introducing the attention mechanism, optimizing the attention weights between nodes, and calculating the attention weight α between node i and node j ij :

[0066]

[0067] In the formula, α ij is the attention weight between node i and node j, α is the weight vector of the attention mechanism, || represents the vector concatenation operation, N(i) is the set of neighbor nodes of node i, exp represents the experience function, LeakyReLU represents the activation function, and Wh i represents the weight matrix of node i, Whj represents the weight matrix of node j, Wh k represents the weight matrix of node k;

[0068] S5.4. Through multiple iterative training, adjust A, B, and C, minimize the prediction error function L, and optimize the coal mine pressure prediction model:

[0069]

[0070] Where N is the number of training samples, y i is the actual observed value, is the model prediction value, λ is the regularization parameter, and are the L2 norms of the weight matrix and adjacency matrix respectively;

[0071] S5.5. Verify and evaluate the trained mine pressure prediction model to ensure that the model has generalization ability under different geological conditions and mining processes, and obtain the optimized mine pressure prediction model M final .

[0072] S6 includes, S6.1, obtaining the latest mine pressure change information from the real-time monitoring system D real-time , including real-time stress data σ i (t), displacement data u i (t) and acoustic emission data A i (t);

[0073] S6.2, D real-time Input to the optimized coal mine pressure prediction model M final In the process of real-time mine pressure risk assessment, the input D real-time Processing is performed to conduct risk assessment on the output of the adaptive graph neural network to generate a mine pressure risk score R(t). The risk score is calculated based on the changing trend of real-time monitoring data:

[0074]

[0075] Where f is the risk assessment function, is the output feature matrix of the adaptive graph neural network, and L represents the number of network layers.

[0076] Compared with the prior art, the present invention has the following beneficial effects: The present invention comprehensively utilizes the data of stress sensors, displacement sensors and acoustic emission sensors to construct a coal mine pressure prediction model based on complex system dynamics. The Kalman filter algorithm is used to preprocess the data, which can effectively remove noise and outliers and ensure the quality of input data. The model parameters are optimized through graph convolutional networks and attention mechanisms, which improves the training effect and generalization ability of the model. In particular, the method of combining graph convolutional networks and attention mechanisms can fully mine the complex relationships in sensor data, improve the model's prediction accuracy for mine pressure changes, and significantly improve the accuracy and reliability of mine pressure risk assessment;

[0077] The present invention uses an adaptive graph neural network to perform real-time mine pressure risk assessment and generate a risk score. By acquiring mine pressure change information in real time, inputting the trained coal mine pressure prediction model, and using an adaptive graph neural network to process real-time monitoring data, a mine pressure risk score is generated. The method can timely discover potential risks in mine pressure changes, generate early warning signals including three risk levels of low, medium and high, and can achieve real-time monitoring and early warning of mine pressure risks. The assessment results and early warning signals are displayed to coal mine managers through a visual interface to assist them in making scientific decisions, thereby improving the safety production level of coal mines and reducing the occurrence of mine pressure accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 This is a flow chart of a coal mine pressure risk assessment method based on complex system dynamics proposed by the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] A coal mine pressure risk assessment method based on complex system dynamics, including:

[0081] S1. Collect geological data of the coal mining area, including rock layer thickness, rock layer hardness, fault distribution and physical property data of coal seam roof and floor;

[0082] S2. Stress sensors, displacement sensors and acoustic emission sensors are deployed in the coal mining area to monitor the mine pressure data in the coal mining area in real time to generate mine pressure change information, and upload D(t) to the central database. D(t) represents the set of mine pressure change information collected by all sensors at time t;

[0083] S3. Preprocess the collected geological data and mine pressure change information, use the Kalman filter algorithm to remove noise and outliers, and obtain a clean coal mining area data set D clean ;

[0084] S4. Based on the complex system dynamics model and combined with the coal mining area data set, a coal mine pressure prediction model is constructed;

[0085] S5. Use historical data to train the coal mine pressure prediction model, and use a method combining graph convolutional network and attention mechanism to optimize the parameters of the coal mine pressure prediction model to obtain the optimized coal mine pressure prediction model M. final ;

[0086] S6. Input the latest mine pressure change information into the trained coal mine pressure prediction model, use the adaptive graph neural network to perform real-time mine pressure risk assessment, and generate a risk score;

[0087] S7. Based on the results of the mine pressure risk assessment, the mine pressure changes during the coal mining process are dynamically predicted, and an adaptive filtering algorithm is used to generate a mine pressure risk warning signal. The warning signal includes three risk levels: low, medium and high. The mine pressure risk assessment results and warning signals are presented to coal mine managers in the form of mine pressure change graphs, risk level graphs and historical data comparison graphs to assist coal mine managers in making decisions.

[0088] S1 includes, S1.1, conducting borehole measurement in the coal mining area, arranging multiple borehole measurement points, using a multi-frequency laser rangefinder and a nonlinear sonic logging instrument to measure the thickness of rock formations at different depths, and obtaining rock formation thickness data H i , i represents the number of the rock layer, and the average value and standard deviation are calculated by repeated measurements:

[0089]

[0090] Where N is the number of measurements, H in represents the thickness of rock layer i measured at the nth time, represents the average thickness of rock layer i;

[0091] S1.2. Collect samples of various rock formations in the coal mining area, use compression test equipment to measure the hardness and compression test of the samples, obtain the hardness and compressive strength of the rock formation respectively, merge and calibrate the field measured data with the laboratory data, and determine the hardness and compressive strength data:

[0092]

[0093] In the formula, B i represents the hardness of rock layer i, M represents the number of measurements, B imrepresents the hardness of rock layer i measured for the mth time, ΔH represents the measurement deviation, H i represents the thickness of rock layer i, σ ci represents the compressive strength of rock layer i, σ cim represents the compressive strength of rock layer i measured for the mth time, Δ∈ represents the strain deviation, ∈ i represents the strain of rock layer i, α and β are adaptive data fusion weight factors.

[0094] S1 includes, S1.3, using geological radar and multi-frequency seismic wave reflection method to detect the distribution of faults in the coal mining area, and setting adaptive geological radar survey lines and multi-mode source receiving points to record the location of the faults F j , fault length L j and fault dip angle θ j , where j represents the fault number:

[0095]

[0096] In the formula, F j represents the location of fault j, K represents the number of radar lines, T k represents the measurement time of the kth measurement line, R(t) represents the radar reflection signal, θ j represents the dip angle of fault j, L j represents the length of fault j, S(t) represents the source signal, and t represents the time;

[0097] S1.4. Sampling the roof and floor of the coal seam to obtain measured density Elastic modulus and Poisson's ratio The microstructure of the samples was analyzed using laser interferometry and electron microscopy, and a multi-scale mechanical model of the physical properties of the top and bottom plates was established;

[0098]

[0099] In the formula, Indicates top and bottom plate k 1 The volume of a point, ρ(x,y,z) represents the density distribution at the position (x,y,z), and (u,v,w) represent the displacement in the x,y,z directions respectively.

[0100] S2 includes, S2.1, arranging stress sensors, displacement sensors and acoustic emission sensors in the coal mining area, and the sensor arrangement plan is based on the geological characteristics of the coal mine and the mining process. i :

[0101] P i =(x i ,y i ,z i );

[0102] Where P i represents the position coordinates of sensor i, x i ,y i ,z i They represent the coordinates of the sensor in three-dimensional space respectively;

[0103] S2.2. Use stress sensors to monitor the changes in rock stress in the coal mining area in real time and obtain the stress value σ measured by sensor i at time t i (t):

[0104]

[0105] In the formula, σ 0 represents the initial stress value, Δσ i (t) represents the change in stress at time t, t 0 represents the initial time, It means the partial derivative of stress value with respect to the horizontal axis, It represents the partial derivative of stress value with respect to the ordinate, θ i is the sensor angle, κ n is the nth influence coefficient, τ n is the nth stress impact strength, E i is the elastic modulus of the rock formation, λ is the attenuation coefficient, t n is the time when the nth stress impact occurs, N is the total number of impacts, and t represents the moment.

[0106] S2 includes: S2.3, using displacement sensors to monitor the displacement changes of rock formations, and obtaining the displacement value u measured by sensor i at time t i (t):

[0107]

[0108] In the formula, u 0 represents the initial displacement value, Δu i (t) represents the change in displacement at time t, α i is the displacement amplitude, ω i is the angular frequency, is the initial phase angle, β i is the damping coefficient, γ j To affect the amplitude, η j is the vibration frequency, δ j is the attenuation coefficient, M is the total number of influencing factors, and j is the jth influencing factor;

[0109] S2.4. Use acoustic emission sensors to monitor the acoustic emission signals generated during coal mining, and obtain the acoustic emission signal intensity A received by sensor i at time t. i (t):

[0110] A i (t) = A 0 +ΔA i (t);

[0111] In the formula, A 0 Represents the background acoustic emission signal intensity, ΔA i (t) represents the change in the intensity of the acoustic emission signal at time t;

[0112] S2.5. Use the data synchronization algorithm to upload the collected stress data, displacement data and acoustic emission data to the central database via wireless transmission:

[0113]

[0114] In the formula, D(t) represents the set of all sensor data collected at time t, N 1 Indicates the total number of sensors.

[0115] S3 includes, S3.1, preliminary screening of the geological data processed by S1 and the mine pressure change information processed by S2, eliminating obviously erroneous data points, and forming a preliminary data set;

[0116] S3.2, apply Kalman filter algorithm to preprocess the preliminary data set to remove noise and outliers;

[0117] S3.3. Storing clean coal mining area data sets in a central database, the data sets containing geological data and mine pressure change information:

[0118]

[0119] Where D clean represents the clean coal mining area dataset, D raw represents the preliminary data set, including geological data and mine pressure change information, I is the unit matrix, K k The Kalman gain matrix of the kth step, H is the observation matrix, z i is the observed value of step i, K i H is the Kalman gain matrix at the i-th step.

[0120] S4 includes: S4.1. Based on the theory of complex system dynamics, the basic framework of the coal mine pressure prediction model is established, and the system state variable x(t) and system input u(t) are defined. t represents the time t. The clean coal mining area dataset D clean As model input, a coal mine pressure prediction model based on complex system dynamics is constructed:

[0121] x(t+1)=Ax(t)+Bu(t)+w(t); y(t)=Cx(t)+v(t);

[0122] In the formula, the system state variable x(t) represents the rock state at the current moment, and x(t) includes geological data {B i ,σ ci ,F j ,ρ k ,E k ,v k} and mine pressure change information D(t), the system input u(t) represents the mining intensity and coal mining progress, the process noise w(t) represents the unmeasured random disturbance, the observed output y(t) represents the measured mine pressure data, the measurement noise v(t) represents the error in the measurement process, and A, B, and C are the parameters to be determined for the model;

[0123] According to the system state transfer equation and observation equation, the coal mining area data set D clean Integrate the geological data and mine pressure change information into the coal mine pressure prediction model;

[0124] S4.2. Extract features from geological data and mine pressure change information and define feature vector f(t):

[0125] f(t)=[B i ,σ ci ,F j ,ρ k ,E k ,v k ,D(t)];

[0126] Based on the theory of complex system dynamics, the mapping function φ is used to map the feature vector f(t) to the state variable x(t):

[0127] x(t)=φ(f(t));

[0128] Apply the state transfer equation to perform time series analysis on the eigenvector and calculate the state variable x(t+1) at the next moment:

[0129] x(t+1)=Δφ(f(t))+Bu(t)+w(t);

[0130] Apply the observation equation to transform the state variable f(t) into the observed output y(t):

[0131] y(t)=Cx(t)+v(t);

[0132] Correct A, B, C, w(t) and v(t) to ensure that the model accurately reflects the actual mine pressure change process;

[0133] S4.3. Compare and verify the corrected model with the actual monitoring data, adjust the model parameters to improve the prediction accuracy, and save the constructed mine pressure prediction model to the central database.

[0134] S5 includes, S5.1, collecting historical data of coal mines, including geological data and mine pressure change information, to form a historical data set D hist , preprocess the historical data set, input the processed data into the mine pressure prediction model, and construct the initial training data set D train ;

[0135] S5.2. Train the initial mine pressure prediction model based on the graph convolutional network and construct a graph structure. The nodes represent geological data and mine pressure change information, and the edges represent the relationship between the data. Define the node feature matrix H and the adjacency matrix A:

[0136] H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) W (l) );

[0137] Among them, H (l) is the node feature matrix of the lth layer, w (l) is the weight matrix of the lth layer, σ is the activation function, and D is the degree matrix.

[0138] S5 includes, S5.3, introducing the attention mechanism, optimizing the attention weights between nodes, and calculating the attention weight α between node i and node j ij :

[0139]

[0140] In the formula, α ij is the attention weight between node i and node j, α is the weight vector of the attention mechanism, || represents the vector concatenation operation, N(i) is the set of neighbor nodes of node i, exp represents the experience function, LeakyReLU represents the activation function, and Wh i represents the weight matrix of node i, Wh j represents the weight matrix of node j, Wh k represents the weight matrix of node k;

[0141] S5.4. Through multiple iterative training, adjust A, B, and C, minimize the prediction error function L, and optimize the coal mine pressure prediction model:

[0142]

[0143] Where N is the number of training samples, y i is the actual observed value, is the model prediction value, λ is the regularization parameter, and are the L2 norms of the weight matrix and adjacency matrix respectively;

[0144] S5.5. Verify and evaluate the trained mine pressure prediction model to ensure that the model has generalization ability under different geological conditions and mining processes, and obtain the optimized mine pressure prediction model M final .

[0145] S6 includes, S6.1, obtaining the latest mine pressure change information from the real-time monitoring system D real-time , including real-time stress data σ i (t), displacement data u i (t) and acoustic emission data A i (t);

[0146] S6.2, D real-time Input to the optimized coal mine pressure prediction model M final In the process of real-time mine pressure risk assessment, the input D real-time Processing is performed to conduct risk assessment on the output of the adaptive graph neural network to generate a mine pressure risk score R(t). The risk score is calculated based on the changing trend of real-time monitoring data:

[0147]

[0148] Where f is the risk assessment function, is the output feature matrix of the adaptive graph neural network, and L represents the number of network layers.

[0149] The technical process of the present invention is as follows Figure 1 As shown, the technical solution of the present invention constructs a mine pressure prediction model based on complex system dynamics by comprehensively utilizing multiple sensor data, which can accurately reflect the dynamic characteristics of mine pressure changes and improve the accuracy and reliability of mine pressure risk assessment. The Kalman filter algorithm is used for data preprocessing to effectively remove noise and outliers and ensure the quality of input data. The model parameters are optimized through graph convolutional networks and attention mechanisms, which improves the training effect and generalization ability of the model.

[0150] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features may be replaced by equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A coal mine pressure risk assessment method based on complex system dynamics, characterized in that: include: S1. Collect geological data of the coal mining area, including rock layer thickness, rock layer hardness, fault distribution and physical property data of coal seam roof and floor; S2. Stress sensors, displacement sensors and acoustic emission sensors are deployed in the coal mining area to monitor the mine pressure data in the coal mining area in real time to generate mine pressure change information, and upload D(t) to the central database. D(t) represents the set of mine pressure change information collected by all sensors at time t; S3. Preprocess the collected geological data and mine pressure change information, use the Kalman filter complex system dynamics model algorithm to remove noise and outliers, and obtain a clean coal mining area data set D clean ; S4. Based on the complex system dynamics model and combined with the coal mining area data set, a coal mine pressure prediction model is constructed; S5. Use historical data to train the coal mine pressure prediction model, and use a method combining graph convolutional network and attention mechanism to optimize the parameters of the coal mine pressure prediction model to obtain the optimized coal mine pressure prediction model M. final ; S6. Input the latest mine pressure change information into the trained coal mine pressure prediction model, use the adaptive graph neural network to perform real-time mine pressure risk assessment, and generate a risk score; S7. Based on the results of the mine pressure risk assessment, the mine pressure changes during the coal mining process are dynamically predicted, and an adaptive filtering algorithm is used to generate a mine pressure risk warning signal. The warning signal includes three risk levels: low, medium and high. The mine pressure risk assessment results and warning signals are presented to coal mine managers in the form of mine pressure change graphs, risk level graphs and historical data comparison graphs to assist coal mine managers in making decisions.

2. A method for coal mine pressure risk assessment based on complex system dynamics according to claim 1, characterized in that: S1 includes, S1.1, conducting borehole measurement in the coal mining area, arranging multiple borehole measurement points, using a multi-frequency laser rangefinder and a nonlinear sonic logging instrument to measure the thickness of rock formations at different depths, and obtaining rock formation thickness data H i , i represents the number of the rock layer, and the average value and standard deviation are calculated by repeated measurements: Where N is the number of measurements, H in represents the thickness of rock layer i measured at the nth time, represents the average thickness of rock layer i; S1.

2. Collect samples of various rock formations in the coal mining area, use compression test equipment to measure the hardness and compression test of the samples, obtain the hardness and compressive strength of the rock formation respectively, merge and calibrate the field measured data with the laboratory data, and determine the hardness and compressive strength data: In the formula, B i represents the hardness of rock layer i, M represents the number of measurements, B im represents the hardness of rock layer i measured for the mth time, ΔH represents the measurement deviation, H i represents the thickness of rock layer i, σ ci represents the compressive strength of rock layer i, σ cim represents the compressive strength of rock layer i measured for the mth time, Δ∈ represents the strain deviation, ∈ i represents the strain of rock layer i, α and β are adaptive data fusion weight factors.

3. A method for coal mine pressure risk assessment based on complex system dynamics according to claim 2, characterized in that: S1 includes, S1.3, using geological radar and multi-frequency seismic wave reflection method to detect the distribution of faults in the coal mining area, and setting adaptive geological radar survey lines and multi-mode source receiving points to record the location of the faults F j , fault length L j and fault dip angle θ j , where j represents the fault number: In the formula, F j represents the location of fault j, K represents the number of radar lines, T k represents the measurement time of the kth measurement line, R(t) represents the radar reflection signal, θ j represents the dip angle of fault j, L j represents the length of fault j, S(t) represents the source signal, and t represents the time; S1.

4. Sampling the roof and floor of the coal seam to obtain measured density Elastic modulus and Poisson's ratio The microstructure of the samples was analyzed using laser interferometry and electron microscopy, and a multi-scale mechanical model of the physical properties of the top and bottom plates was established; In the formula, Indicates top and bottom plate k 1 The volume of a point, ρ(x,y,z) represents the density distribution at the position (x,y,z), and (u,v,w) represent the displacement in the x,y,z directions respectively.

4. The method for coal mine pressure risk assessment based on complex system dynamics according to claim 1 is characterized in that: S2 includes, S2.1, arranging stress sensors, displacement sensors and acoustic emission sensors in the coal mining area, and the sensor arrangement plan is based on the geological characteristics of the coal mine and the mining process. i : P i =(x i ,y i ,z i ); Where P i represents the position coordinates of sensor i, x i ,y i ,z i They represent the coordinates of the sensor in three-dimensional space respectively; S2.

2. Use stress sensors to monitor the changes in rock stress in the coal mining area in real time and obtain the stress value σ measured by sensor i at time t i (t): In the formula, σ0 represents the initial stress value, Δσ i (t) represents the stress change at time t, t0 represents the initial time, It means the partial derivative of stress value with respect to the horizontal axis, It represents the partial derivative of stress value with respect to the ordinate, θ i is the sensor angle, κ n is the nth influence coefficient, τ n is the nth stress impact strength, E i is the elastic modulus of the rock formation, λ is the attenuation coefficient, t n is the time when the nth stress impact occurs, N is the total number of impacts, and t represents the moment.

5. A method for coal mine pressure risk assessment based on complex system dynamics according to claim 4, characterized in that: S2 includes: S2.3, using displacement sensors to monitor the displacement changes of rock formations, and obtaining the displacement value u measured by sensor i at time t i (t): Where u0 represents the initial displacement value, Δu i (t) represents the change in displacement at time t, α i is the displacement amplitude, ω i is the angular frequency, is the initial phase angle, β i is the damping coefficient, γ j To affect the amplitude, η j is the vibration frequency, δ j is the attenuation coefficient, M is the total number of influencing factors, and j is the jth influencing factor; S2.

4. Use acoustic emission sensors to monitor the acoustic emission signals generated during coal mining, and obtain the acoustic emission signal intensity A received by sensor i at time t. i (t): A i (t)=A0+ΔA i (t); Where A0 represents the background acoustic emission signal intensity, ΔA i (t) represents the change in the intensity of the acoustic emission signal at time t; S2.

5. Use the data synchronization algorithm to upload the collected stress data, displacement data and acoustic emission data to the central database via wireless transmission: In the formula, D(t) represents the set of all sensor data collected at time t, N 1 Indicates the total number of sensors.

6. The method for coal mine pressure risk assessment based on complex system dynamics according to claim 1 is characterized in that: S3 includes, S3.1, preliminary screening of the geological data processed by S1 and the mine pressure change information processed by S2, eliminating obviously erroneous data points, and forming a preliminary data set; S3.2, apply Kalman filter algorithm to preprocess the preliminary data set to remove noise and outliers; S3.

3. Storing clean coal mining area data sets in a central database, the data sets containing geological data and mine pressure change information: Where D clean represents the clean coal mining area dataset, D raw represents the preliminary data set, including geological data and mine pressure change information, I is the unit matrix, K k The Kalman gain matrix of the kth step, H is the observation matrix, z i is the observed value of the i-th step, K i H is the Kalman gain matrix at the i-th step.

7. The method for coal mine pressure risk assessment based on complex system dynamics according to claim 1 is characterized in that: S4 includes: S4.

1. Based on the theory of complex system dynamics, the basic framework of the coal mine pressure prediction model is established, and the system state variable x(t) and system input u(t) are defined. t represents the time t. The clean coal mining area dataset D clean As model input, a coal mine pressure prediction model based on complex system dynamics is constructed: x(t+1)=Ax(t)+Bu(t)+w(t); y(t)=Cx(t)+v(t); In the formula, the system state variable x(t) represents the rock state at the current moment, and x(t) includes geological data {B i ,σ ci ,F j ,ρ k ,E k ,v k } and mine pressure change information D(t), the system input u(t) represents the mining intensity and coal mining progress, the process noise w(t) represents the unmeasured random disturbance, the observed output y(t) represents the measured mine pressure data, the measurement noise v(t) represents the error in the measurement process, and A, B, and C are the parameters to be determined for the model; According to the system state transfer equation and observation equation, the coal mining area data set D clean Integrate the geological data and mine pressure change information into the coal mine pressure prediction model; S4.

2. Extract features from geological data and mine pressure change information and define feature vector f(t): f(t)=[B i ,σ ci ,F j ,ρ k ,E k ,v k ,D(t)]; Based on the theory of complex system dynamics, the mapping function φ is used to map the feature vector f(t) to the state variable x(t): x(t)=φ(f(t)); Apply the state transfer equation to perform time series analysis on the eigenvector and calculate the state variable x(t+1) at the next moment: x(t+1)=Aφ(f(t))+Bu(t)+w(t); Apply the observation equation to transform the state variable f(t) into the observed output y(t): y(t)=Cx(t)+v(t); Correct A, B, C, w(t) and v(t) to ensure that the model accurately reflects the actual mine pressure change process; S4.

3. Compare and verify the corrected model with the actual monitoring data, adjust the model parameters to improve the prediction accuracy, and save the constructed mine pressure prediction model to the central database.

8. A method for coal mine pressure risk assessment based on complex system dynamics according to claim 7, characterized in that: S5 includes, S5.1, collecting historical data of coal mines, including geological data and mine pressure change information, to form a historical data set D hist , preprocess the historical data set, input the processed data into the mine pressure prediction model, and construct the initial training data set D train ; S5.

2. Train the initial mine pressure prediction model based on the graph convolutional network and construct a graph structure. The nodes represent geological data and mine pressure change information, and the edges represent the relationship between the data. Define the node feature matrix H and the adjacency matrix A: H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) W (l) ); Among them, H (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the activation function, and D is the degree matrix.

9. A method for coal mine pressure risk assessment based on complex system dynamics according to claim 8, characterized in that: S5 includes, S5.3, introducing the attention mechanism, optimizing the attention weights between nodes, and calculating the attention weight α between node i and node j ij : In the formula, α ij is the attention weight between node i and node j, α is the weight vector of the attention mechanism, || represents the vector concatenation operation, N(i) is the set of neighbor nodes of node i, exp represents the experience function, LeakyReLU represents the activation function, and Wh i represents the weight matrix of node i, Wh j represents the weight matrix of node j, Wh k represents the weight matrix of node k; S5.

4. Through multiple iterative training, adjust A, B, and C, minimize the prediction error function L, and optimize the coal mine pressure prediction model: In the formula, N is the number of training samples, y i is the actual observed value, is the model prediction value, λ is the regularization parameter, and are the L2 norms of the weight matrix and adjacency matrix respectively; S5.

5. Verify and evaluate the trained mine pressure prediction model to ensure that the model has generalization ability under different geological conditions and mining processes, and obtain the optimized mine pressure prediction model M final .

10. The method for coal mine pressure risk assessment based on complex system dynamics according to claim 1, characterized in that: S6 includes, S6.1, obtaining the latest mine pressure change information from the real-time monitoring system D real-time , including real-time stress data σ i (t), displacement data u i (t) and acoustic emission data A i (t); S6.2, D real-time Input to the optimized coal mine pressure prediction model M final In the process of real-time mine pressure risk assessment, the input D real-time Processing is performed to conduct risk assessment on the output of the adaptive graph neural network to generate a mine pressure risk score R(t). The risk score is calculated based on the changing trend of real-time monitoring data: Where f is the risk assessment function, is the output feature matrix of the adaptive graph neural network, and L represents the number of network layers.

Images