A rock burst prediction method and system driven by earthquake source inversion data

Abstract

The present invention discloses a rock burst prediction method and system driven by source inversion data. The prediction method includes: quantitatively inverting and calculating the core parameters of the coal rock fracture focal mechanism through the original acoustic wave data collected on site. A rock burst source data prediction model is constructed based on the Informer algorithm to accurately predict the time and location of rock burst. Combined with the predicted time, location and core parameters of the focal mechanism of future rupture sources, a rock burst risk assessment model is constructed to predict the magnitude and nature of rock burst, and a dual-driven rock burst intelligent model of source data prediction and rock burst risk assessment is established to achieve multi-dimensional intelligent prediction of the temporal response, spatial distribution and evolutionary attributes of the rock burst disaster process. This method combines source data prediction and risk assessment, which can not only predict the risk of rock burst disasters, but also intuitively reproduce the rock burst evolution process and disaster-causing mechanism.

Description

Technical Field

[0001] The present invention relates to the field of rock burst prediction, and in particular to a rock burst prediction method and system driven by earthquake source inversion data. Background Art

[0002] Rock burst is one of the most serious dynamic hazards in underground projects such as coal mines. Its prediction and prevention are crucial for ensuring the safety and efficiency of coal mine production. As coal mining depths continue to increase, the incidence and severity of rock burst disasters are also increasing. The complex mechanism of rock burst disasters involves multiple factors, such as the physical properties of the rock mass, the ground stress state, and mining disturbances, which hinders the prediction of rock burst disasters. However, with the development of monitoring technology, monitoring methods such as microseismic monitoring, geoacoustic monitoring, and electromagnetic radiation technology have become more popular, providing faster and more accurate early warning information for real-time monitoring of coal mine rock burst disasters.

[0003] The combination of big data technology with monitoring methods such as microseismic monitoring, ground sound monitoring and electromagnetic radiation technology has become an important technical method for predicting rock burst disasters. CN118584542A proposed to build a prediction model through TFT neural network, and use b value, energy index, cumulative apparent volume and Schmidt number multivariate to predict rock burst, thereby improving the accuracy of rock burst prediction. CN117708705A provides a rock burst warning method based on the comparison of long-term and short-term characteristics of microseismic data, and introduces prior knowledge into the algorithm design to automatically extract the rock burst warning implicit warning indicators. This method improves the generalization ability of handling rock burst disasters in different mines. CN117849855A proposed a rock burst prediction method based on the comparison of monitoring data with different hazards. Changing the algorithm structure solves the problem of poor algorithm generalization ability caused by data imbalance, and effectively improves the warning effect of rock burst events. These prediction methods effectively improve the quality of data prediction by improving the algorithm structure, but do not extract data information directly related to rock burst disasters. In order to more accurately predict rock burst disasters, a data and physics dual-driven model is constructed. Rock burst disasters are predicted by directly analyzing the source mechanism of the original monitoring data. This model can better capture the precursor information of rock burst and improve the generalization ability of rock burst disaster prediction. Summary of the Invention

[0004] In response to the problems and needs raised above, this solution proposes a rock burst prediction method and system driven by earthquake source inversion data. By adopting the following technical features, it can achieve the above technical objectives and bring about many other technical effects.

[0005] An object of the present invention is to provide a rock burst prediction method driven by earthquake source inversion data, comprising the following steps:

[0006] S10: Arrange a microseismic monitoring system according to the actual monitoring range of the mine site, use the microseismic monitoring system to collect raw acoustic wave data from the mine, perform positioning calculation and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors arranged in the coal seam at the working face;

[0007] S20: Use the standardized processing method to preprocess the core parameters of the rupture source, and divide the preprocessed data set into training set, test set and validation set;

[0008] S30: Construct a rockburst source data prediction model based on the Informer algorithm. Input the training set into the rockburst source data prediction model. Use the optimized loss function backpropagation to train the rockburst source data prediction model. Use the validation set to test and evaluate the trained rockburst source data prediction model. Continuously adjust the training parameters until the requirements are met. Use the rockburst source data prediction model to predict the time, direction, magnitude, and core parameters of future rupture sources.

[0009] S40: Construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as the input set, and establish a risk rating index by calculating the cumulative difference index of the core parameters of each rupture source. Combined with the rock burst time, direction and risk rating index, the rock burst disaster prediction risk is evaluated.

[0010] S50: First, use the predicted time, direction, magnitude and core parameters of the rupture source to reproduce the cross-scale gradual changes in the rock burst disaster process from the fracture surface to the fracture field; then use the risk rating index to predict the evolution degree of the rock burst disaster and further identify the development trend of the rock burst.

[0011] In addition, the earthquake source inversion data-driven rock burst prediction method according to the present invention may also have the following technical features:

[0012] In one example of the present invention, in step S10, positioning calculation and focal mechanism inversion are performed on the original acoustic wave data to obtain the core parameters of the focal mechanism directly related to the rupture, including the following steps:

[0013] S11: Post-process the collected acoustic wave data and extract the first wave arrival time and amplitude information using the red delay information criterion method;

[0014] S12: Using the arrival time data of the first acoustic wave at sensors at different locations of the microseismic monitoring system, the simplex algorithm is used to calculate the spatial location of the earthquake source;

[0015] S13: Based on the constraints of the tensile-shear rupture source model, the six components M of the source rupture moment tensor are solved using the acoustic first wave amplitude data collected by sensors at different locations. pq ;

[0016] S14: Based on the calculated moment tensor components, quantitative inversion is performed to calculate the core parameters of the coal rock fracture source, wherein the core parameters of the coal rock fracture source include fracture area, orientation, tension-shear properties, moment magnitude, and released energy.

[0017] In one example of the present invention, in step S14, the specific calculation formula of the core parameters of the coal rock fracture source is as follows:

[0018] Broken area:

[0019]

[0020] Where ΔV is the rupture area, M1 and M3 are the moment tensors M kl The characteristic value of , μ is the Lame constant, ξ is the proportional coefficient of the rupture area and displacement;

[0021] Orientation:

[0022]

[0023] n=(cosβ,0,±sinβ)

[0024]

[0025] Where β is the angle between the direction of movement of the earthquake source rupture surface and the normal direction, M1, M2, and M3 are the moment tensors M kl The eigenvalues ​​of , μ and λ are the Lame constants, n is the spatial orientation, and b is the direction of motion;

[0026] Shearing properties:

[0027]

[0028]

[0029]

[0030] Where V t is the tension volume, V s is the shear volume, K is the tension-shear ratio: K>1, the tensile rupture property is dominant; K<1, the shear rupture property is dominant;

[0031] Magnitude:

[0032]

[0033] Where M1, M2, and M3 are moment tensors M kl The characteristic value of

[0034] Release Energy:

[0035]

[0036] Where: σ t is the tensile strength, σ s For shear strength.

[0037] In one example of the present invention, in step S20, a standardized processing method is used to pre-process the core parameters of the rupture source, including:

[0038] For preprocessing parameters x1, x2, ..., x n-1 , x n The normalization formula is as follows:

[0039]

[0040] Where, is the minimum value of the preprocessing parameter, is the maximum value of the preprocessing parameter, y i ∈[0,1].

[0041] In one example of the present invention, in step S30, a rock burst source data prediction model is constructed based on the Informer algorithm. The rock burst source data prediction model includes: an encoder, a decoder, and a fully connected layer.

[0042] The decoder includes a masked multi-head probabilistic sparse self-attention mechanism module and a multi-head attention mechanism module, wherein the masked multi-head probabilistic sparse self-attention mechanism layer is configured to ensure the autoregressive nature of the model, prevent future information leakage, and improve computational efficiency; the multi-head attention mechanism layer is configured to capture long-range dependencies in the sequence and process information in different subspaces in parallel through a multi-head mechanism, thereby enhancing the model's ability to understand complex time series data;

[0043] The encoder includes a multi-head probabilistic sparse self-attention module and a self-attention distillation module. The multi-head probabilistic sparse self-attention module is configured to reduce the computational complexity of the model by randomly selecting sequence blocks and calculating attention scores only within these blocks, while enhancing the model's ability to capture local temporal dependencies. The self-attention distillation module is configured to improve the model's ability to understand long-range dependencies of long-sequence inputs by performing feature compression through the distillation step, enabling the model to better capture and utilize long-term dependency information in long-sequence data.

[0044] The fully connected layer is configured to map the input into a higher dimensional space and then transform it back to the original dimension after filtering.

[0045] In one example of the present invention, in step S30, using the optimized loss function back propagation to train the rock burst source data prediction model, and using the validation set to test and evaluate the trained rock burst source data prediction model includes the following steps:

[0046] S301: Calculate the data sample error and construct a loss function. The expression of the loss function is as follows:

[0047]

[0048] Where n is the number of samples; l is the true value; is the predicted value;

[0049] S302: Back propagation training model: Calculate the gradient of the loss function with respect to the weights and biases of each layer of the output layer in reverse order from the output layer.

[0050] S303: Gradient descent optimization loss function: Calculate the loss function L with respect to a i The partial derivative of And update the core parameters of each rupture source: Where α is the learning rate that controls the step size of each update; the gradient is repeatedly calculated and the weights and biases of the prediction model are updated until the stopping condition is met;

[0051] S304: Repeat S301 to S303 to determine whether the model meets the requirements. If not, adjust the parameters accordingly and continue training until the requirements are met and use the test set for prediction.

[0052] In an example of the present invention, step S40 specifically includes the following steps:

[0053] S401: Assume that the core parameter set of the focal mechanism for dynamic prediction of rock burst is P = (m, K, ΔV, bV t …);

[0054] S402: Calculate the cumulative difference index, assuming that the data sequence of the parameters in the time window is q1, q2, ..., q n-1 ,q n (n>3) where q n is the parameter when the time series is n; the subtraction of the previous and next terms of the sequence, Δq2=q2-q1,…,Δq n-1 =q n-1 -q n-2 , Δq n =q n -q n-1 , let Gj =min{Δq n ,0},H j =max{Δq n , 0}, then

[0055]

[0056]

[0057] Let the cumulative difference R w =F1 / (F1+F2), R w ∈[0,1], calculate the cumulative difference values ​​of each parameter in the core parameter P of the rock burst dynamic prediction, and obtain the cumulative difference set of rock burst risk assessment parameters [R m , R K , R ΔV , R bVt …R w ];

[0058] S403: Determine the risk level evaluation index expression:

[0059] F p＝ R m a1+R K a2+R ΔV a3+R bVt a4+…+R w a w

[0060] Where a1, a2, a3, a4, ... a w The rock burst risk level is divided into the following levels: p Evaluate rock burst risk and predict the level of rock burst risk in the future.

[0061] In an example of the present invention, step S50 specifically includes the following steps:

[0062] The time, orientation and focal mechanism parameters obtained by the rock burst source data prediction model are used to characterize the gradual evolution process of the source fracture surface from generation, expansion, penetration to the cross-scale fracture field.

[0063] The risk rating index F obtained by the rock burst risk assessment model p Predict the evolution degree of rock burst disaster using R K , R ΔV , R bVt Judge the nature of rock bursts that will occur in the future, and further identify the development trend of rock bursts, to achieve multi-dimensional intelligent prediction of the temporal response, spatial distribution, and evolutionary properties of rock burst disaster processes.

[0064] Another object of the present invention is to provide an intelligent rock burst prediction system driven by earthquake source inversion data, comprising:

[0065] a parameter acquisition module configured to deploy a microseismic monitoring system based on the actual monitoring range of the mine site, collect raw acoustic wave data from the mine using the microseismic monitoring system, perform location calculations and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors deployed within the coal seam at the working face;

[0066] a data partitioning module configured to preprocess the core parameters of the rupture source using a standardized processing method, and to divide the preprocessed data set into a training set, a test set, and a validation set;

[0067] The prediction model module is configured to construct a rock burst source data prediction model based on the Informer algorithm, input the training set into the rock burst source data prediction model, train the rock burst source data prediction model using the optimized loss function back propagation, test and evaluate the trained rock burst source data prediction model using the validation set, and continuously adjust the training parameters until the requirements are met; and predict the time, direction, magnitude and core parameters of the future rupture source through the rock burst source data prediction model;

[0068] A risk model module is configured to construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as an input set, establishing a risk rating index by calculating the cumulative difference index of the core parameters of each rupture source, and evaluating the rock burst disaster prediction risk by combining the rock burst time, direction and risk rating index;

[0069] The comprehensive prediction module is configured to first use the predicted time, direction, magnitude and core parameters of the rupture source to reproduce the cross-scale gradual changes in the rock burst disaster process from the crack surface to the crack field; then use the risk rating indicators to predict the degree of evolution of the rock burst disaster and further identify the development trend of the rock burst.

[0070] In one example of the present invention, the risk model module includes:

[0071] The core parameter setting unit is configured to set the core parameter set of the focal mechanism for dynamic prediction of rock burst pressure P = (m, K, ΔV, bV t …);

[0072] The cumulative difference indicator unit is configured to set the data sequence of the parameters in the time window to be q1, q2, ..., q n-1 ,q n (n>3) where qn is the parameter when the time series is n; the subtraction of the previous and next terms of the sequence, Δq2=q2-q1,…,Δq n-1 =q n-1 -q n-2 , Δq n =q n -q n-1 , let G j =min{Δq n ,0},H j =max{Δq n , 0}, then

[0073]

[0074]

[0075] Let the cumulative difference R w =F1 / (F1+F2), R w ∈[0,1], calculate the cumulative difference values ​​of each parameter in the core parameter P of the rock burst dynamic prediction, and obtain the cumulative difference set of rock burst risk assessment parameters [R m , R K , R ΔV , R bVt …R w ];

[0076] The risk level evaluation unit is configured to determine the risk level evaluation index expression as follows:

[0077] F p＝ R m a1+R K a2+R ΔV a3+R bVt a4+…+R w a w

[0078] Where a1, a2, a3, a4, ... a w The rock burst risk level is divided into the following levels: p Evaluate rock burst risk and predict the level of rock burst risk in the future.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] Compared to traditional rockburst prediction models that typically use raw acoustic time-series signals as a data set, this method uses focal mechanism inversion to analyze the core parameters of the focal mechanism directly related to rockburst disasters. This method can provide more accurate and direct data reflecting the characteristics of the disaster.

[0081] This method uses the core parameters of the earthquake focal mechanism as the data set. Due to the direct correlation between the core parameters and the nature of the disaster, the evolution process of rock burst disasters in different mines is essentially the same, which enables the prediction model to be better generalized to various mine environments, improving the applicability and robustness of the model.

[0082] This method builds a rockburst source data prediction model based on the Informer algorithm, which can efficiently handle long time series prediction problems. By optimizing the loss function and backpropagation training, combined with verification and evaluation on a validation set, the parameters are continuously adjusted until the model meets the requirements. This ensures the effectiveness of the prediction model in actual rockburst prediction applications and can more accurately predict the time and location of future rupture sources.

[0083] This method constructs a dual-driven rock burst intelligent prediction model for earthquake source data prediction and rock burst risk assessment. It can not only accurately identify and predict the time and location of rock burst disasters, but also further judge the development trend of rock burst disaster evolution. It forms a multi-dimensional intelligent prediction method for rock burst that integrates source parameter prediction and disaster evolution rating, providing more reliable technical support for mine safety.

[0084] Hereinafter, the best embodiment of the present invention will be described in more detail with reference to the accompanying drawings so that the features and advantages of the present invention can be easily understood. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings of the embodiments of the present invention. The drawings are only used to illustrate some embodiments of the present invention, but not to limit all embodiments of the present invention thereto.

[0086] Figure 1 Flow chart of a rock burst prediction method driven by earthquake source inversion data according to an embodiment of the present invention;

[0087] Figure 2 A schematic diagram of the structure of a rock burst source data prediction model based on the Informer algorithm according to an embodiment of the present invention;

[0088] Figure 3 Schematic diagram of focal mechanism inversion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to make the purpose, technical solution and advantages of the technical solution of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of specific embodiments of the present invention. The same figure marks in the drawings represent the same parts. It should be noted that the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the described 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.

[0090] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by persons of ordinary skill in the field to which the invention belongs. The words "first", "second" and similar terms used in the patent application specification and claims of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "a" or "an" do not necessarily indicate a quantity limitation. Words such as "include" or "comprising" mean that the elements or objects preceding the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0091] According to the first aspect of the present invention, a rock burst prediction method driven by earthquake source inversion data is as follows: Figure 1 and Figure 3 As shown, the following steps are included:

[0092] S10: Arrange a microseismic monitoring system according to the actual monitoring range of the mine site, use the microseismic monitoring system to collect raw acoustic wave data from the mine, perform positioning calculation and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors arranged in the coal seam at the working face;

[0093] S20: Use the standardized processing method to preprocess the core parameters of the rupture source, and divide the preprocessed data set into training set, test set and validation set;

[0094] S30: Construct a rockburst source data prediction model based on the Informer algorithm. Input the training set into the rockburst source data prediction model. Use the optimized loss function backpropagation to train the rockburst source data prediction model. Use the validation set to test and evaluate the trained rockburst source data prediction model. Continuously adjust the training parameters until the requirements are met. Use the rockburst source data prediction model to predict the time, direction, magnitude, and core parameters of future rupture sources.

[0095] S40: Construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as the input set, and establish a risk rating index by calculating the cumulative difference index of the core parameters of each rupture source. Combined with the rock burst time, direction and risk rating index, the rock burst disaster prediction risk is evaluated.

[0096] S50: First, the predicted occurrence time, orientation, magnitude and core parameters of the rupture source are used to reproduce the cross-scale gradual changes of the rock burst disaster process from the fracture surface to the fracture field; then the risk rating indicators are used to predict the evolution degree of the rock burst disaster, further identify the development trend of the rock burst, realize the dual-driven intelligent prediction of rock burst by earthquake source data prediction and rock burst risk assessment, and comprehensively predict the evolution process of rock burst disaster risk.

[0097] Compared to traditional rockburst prediction models that typically use raw acoustic time-series signals as a data set, this method uses focal mechanism inversion to analyze the core parameters of the focal mechanism directly related to rockburst disasters. This method can provide more accurate and direct data reflecting the characteristics of the disaster.

[0098] This method uses the core parameters of the earthquake focal mechanism as the data set. Due to the direct correlation between the core parameters and the nature of the disaster, the evolution process of rock burst disasters in different mines is essentially the same, which enables the prediction model to be better generalized to various mine environments, improving the applicability and robustness of the model.

[0099] This method builds a rockburst source data prediction model based on the Informer algorithm, which can efficiently handle long time series prediction problems. By optimizing the loss function and backpropagation training, combined with verification and evaluation on a validation set, the parameters are continuously adjusted until the model meets the requirements. This ensures the effectiveness of the prediction model in actual rockburst prediction applications and can more accurately predict the time and location of future rupture sources.

[0100] This method constructs a dual-driven rock burst intelligent prediction model for earthquake source data prediction and rock burst risk assessment. It can not only accurately identify and predict the time and location of rock burst disasters, but also further judge the development trend of rock burst disaster evolution. It forms a multi-dimensional intelligent prediction method for rock burst that integrates source parameter prediction and disaster evolution rating, providing more reliable technical support for mine safety.

[0101] In one example of the present invention, in step S10, positioning calculation and focal mechanism inversion are performed on the original acoustic wave data to obtain the core parameters of the focal mechanism directly related to the rupture, including the following steps:

[0102] S11: Post-process the collected acoustic wave data and extract the first wave arrival time and amplitude information using the red delay information criterion method;

[0103] S12: Using the arrival time data of the first acoustic wave at sensors at different locations of the microseismic monitoring system, the simplex algorithm is used to calculate the spatial location of the earthquake source;

[0104] S13: Based on the constraints of the tensile-shear rupture source model, the six components M of the source rupture moment tensor are solved using the acoustic first wave amplitude data collected by sensors at different locations. pq ;

[0105] S14: Based on the calculated moment tensor components, quantitative inversion is performed to calculate the core parameters of the coal rock fracture source, wherein the core parameters of the coal rock fracture source include fracture area, orientation, tension-shear properties, moment magnitude, and released energy.

[0106] In one example of the present invention, in step S14, the specific calculation formula of the core parameters of the coal rock fracture source is as follows:

[0107] Broken area:

[0108]

[0109] Where ΔV is the rupture area, M1 and M3 are the moment tensors M kl The characteristic value of , μ is the Lame constant, ξ is the proportional coefficient of the rupture area and displacement;

[0110] Orientation:

[0111]

[0112] n=(cosβ,0,±sinβ)

[0113]

[0114] Where β is the angle between the direction of movement of the earthquake source rupture surface and the normal direction, M1, M2, and M3 are the moment tensors Mkl The eigenvalues ​​of , μ and λ are the Lame constants, n is the spatial orientation, and b is the direction of motion;

[0115] Shearing properties:

[0116]

[0117]

[0118]

[0119] Where V t is the tension volume, V s is the shear volume, K is the tension-shear ratio: K>1, the tensile rupture property is dominant; K<1, the shear rupture property is dominant;

[0120] Magnitude:

[0121]

[0122] Where M1, M2, and M3 are moment tensors M kl The characteristic value of

[0123] Release Energy:

[0124]

[0125] Where: σ t is the tensile strength, σ s For shear strength.

[0126] In one example of the present invention, in step S20, a standardized processing method is used to pre-process the core parameters of the rupture source, including:

[0127] For preprocessing parameters x1, x2, ..., x n-1 , x n The normalization formula is as follows:

[0128]

[0129] Where, is the minimum value of the preprocessing parameter, is the maximum value of the preprocessing parameter, y i ∈[0,1].

[0130] In one example of the present invention, Figure 2 As shown, in step S30, a rock burst source data prediction model is constructed based on the Informer algorithm. The rock burst source data prediction model includes: an encoder, a decoder, and a fully connected layer.

[0131] The decoder includes a masked multi-head probabilistic sparse self-attention mechanism module and a multi-head attention mechanism module, wherein the masked multi-head probabilistic sparse self-attention mechanism layer is configured to ensure the autoregressive nature of the model, prevent future information leakage, and improve computational efficiency; the multi-head attention mechanism layer is configured to capture long-range dependencies in the sequence and process information in different subspaces in parallel through a multi-head mechanism, thereby enhancing the model's ability to understand complex time series data;

[0132] The encoder includes a multi-head probabilistic sparse self-attention module and a self-attention distillation module. The multi-head probabilistic sparse self-attention module is configured to reduce the computational complexity of the model by randomly selecting sequence blocks and calculating attention scores only within these blocks, while enhancing the model's ability to capture local temporal dependencies. The self-attention distillation module is configured to improve the model's ability to understand long-range dependencies of long-sequence inputs by performing feature compression through the distillation step, enabling the model to better capture and utilize long-term dependency information in long-sequence data.

[0133] The fully connected layer is configured to map the input into a higher dimensional space and then transform it back to the original dimension after filtering.

[0134] The detailed process of data processing by the rock burst source data prediction model is as follows:

[0135] The original data set is input into the encoder and decoder respectively. In the encoder, the multi-head probabilistic sparse self-attention module reduces the computational complexity of the model by randomly selecting sequence blocks and calculating the attention scores only within these blocks, while enhancing the model's ability to capture local temporal dependencies. The self-attention distillation module improves the model's ability to understand long-range dependencies of long sequence inputs. Feature compression is performed through the distillation step, enabling the model to better capture and utilize long-term dependency information in long sequence data. The original data processed by the encoder is input into the decoder and processed by the decoder together with the original data set initially input into the decoder. That is, the masked multi-head probabilistic sparse self-attention mechanism layer ensures the autoregressive nature of the model, prevents future information leakage, and improves computational efficiency. By adding a mask to the attention block, it ensures that the model can only access current and past information, but not future information, during prediction. The multi-head attention mechanism layer captures long-range dependencies in the sequence and processes information in different subspaces in parallel through a multi-head mechanism, enhancing the model's ability to understand complex time series data. Finally, the fully connected layer maps the input to a higher-dimensional space and, after filtering, returns it to the original dimension. This step uses a fully connected network (replaced by CNN in Informer) to enhance the model's nonlinear transformation capabilities of features, further improving the model's predictive performance.

[0136] In one example of the present invention, in step S30, using the optimized loss function back propagation to train the rock burst source data prediction model, and using the validation set to test and evaluate the trained rock burst source data prediction model includes the following steps:

[0137] S301: Calculate the data sample error and construct a loss function. The expression of the loss function is as follows:

[0138]

[0139] Where n is the number of samples; l is the true value; is the predicted value;

[0140] S302: Back propagation training model: Calculate the gradient of the loss function with respect to the weights and biases of each layer of the output layer in reverse order from the output layer.

[0141] S303: Gradient descent optimization loss function: Calculate the loss function L with respect to a i The partial derivative of And update the core parameters of each rupture source: Where α is the learning rate that controls the step size of each update; the gradient is repeatedly calculated and the weights and biases of the prediction model are updated until the stopping condition is met;

[0142] S304: Repeat S301 to S303 to determine whether the model meets the requirements. If not, adjust the parameters accordingly and continue training until the requirements are met and use the test set for prediction.

[0143] In an example of the present invention, step S40 specifically includes the following steps:

[0144] S401: Assume that the core parameter set of the focal mechanism for dynamic prediction of rock burst is P = (m, K, ΔV, bV t …);

[0145] S402: Calculate the cumulative difference index, assuming that the data sequence of the parameters in the time window is q1, q2, ..., q n-1 ,q n (n>3) where q n is the parameter when the time series is n; the subtraction of the previous and next terms in the sequence, Δq2=q2-q1,…,Δq n-1 =q n-1 -q n-2 , Δq n =q n -q n-1 , let G j =min{Δq n ,0},H j =max{Δq n, 0}, then

[0146]

[0147]

[0148] Let the cumulative difference R w =F1 / (F1+F2), R w ∈[0,1], calculate the cumulative difference values ​​of each parameter in the core parameter P of the rock burst dynamic prediction, and obtain the cumulative difference set of rock burst risk assessment parameters [R m , R K , R ΔV , R bVt …R w ];

[0149] S403: Determine the risk level evaluation index expression:

[0150] F p＝ R m a1+R K a2+R ΔV a3+R bVt a4+…+R w a w

[0151] Where a1, a2, a3, a4, ... a w The rock burst risk level is divided into the following levels: p Evaluate rock burst risk and predict the level of rock burst risk in the future.

[0152] In an example of the present invention, step S50 specifically includes the following steps:

[0153] The time, orientation and focal mechanism parameters obtained by the rock burst source data prediction model are used to characterize the gradual evolution process of the source fracture surface from generation, expansion, penetration to the cross-scale fracture field.

[0154] The risk rating index F obtained by the rock burst risk assessment model p Predict the evolution degree of rock burst disaster using R K , R ΔV , R bVt Judge the nature of rock bursts that will occur in the future, and further identify the development trend of rock bursts, to achieve multi-dimensional intelligent prediction of the temporal response, spatial distribution, and evolutionary properties of rock burst disaster processes.

[0155] According to a second aspect of the present invention, a rock burst intelligent prediction system driven by earthquake source inversion data comprises:

[0156] a parameter acquisition module configured to deploy a microseismic monitoring system based on the actual monitoring range of the mine site, collect raw acoustic wave data from the mine using the microseismic monitoring system, perform location calculations and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors deployed within the coal seam at the working face;

[0157] a data partitioning module configured to preprocess the core parameters of the rupture source using a standardized processing method, and to divide the preprocessed data set into a training set, a test set, and a validation set;

[0158] The prediction model module is configured to construct a rock burst source data prediction model based on the Informer algorithm, input the training set into the rock burst source data prediction model, train the rock burst source data prediction model using the optimized loss function back propagation, test and evaluate the trained rock burst source data prediction model using the validation set, and continuously adjust the training parameters until the requirements are met; and predict the time, direction, magnitude and core parameters of the future rupture source through the rock burst source data prediction model;

[0159] A risk model module is configured to construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as an input set, establishing a risk rating index by calculating the cumulative difference index of the core parameters of each rupture source, and evaluating the rock burst disaster prediction risk by combining the rock burst time, direction and risk rating index;

[0160] The comprehensive prediction module is configured to first use the predicted time, direction, magnitude and core parameters of the rupture source to reproduce the cross-scale gradual changes of the rock burst disaster process from the crack surface to the crack field; then use the risk rating indicators to predict the evolution degree of the rock burst disaster, further identify the development trend of the rock burst, realize the dual-driven intelligent prediction of rock burst by earthquake source data prediction and rock burst risk assessment, and comprehensively predict the evolution process of rock burst disaster risk.

[0161] Compared to traditional rockburst prediction models that typically use raw acoustic time-series signals as a data set, this system uses focal mechanism inversion to analyze the core parameters of the focal mechanism directly related to rockburst disasters. This method provides more accurate and direct data reflecting the characteristics of the disaster.

[0162] The system uses the core parameters of the earthquake source mechanism as the data set. Due to the direct correlation between the core parameters and the nature of the disaster, the evolution process of rock burst disasters in different mines is essentially the same, which enables the prediction model to be better generalized to various mine environments, improving the applicability and robustness of the model.

[0163] The system builds a rock burst source data prediction model based on the Informer algorithm, which can efficiently handle long time series prediction problems. By optimizing the loss function and backpropagation training, combined with verification and evaluation of the validation set, the parameters are continuously adjusted until the model meets the requirements. This ensures the effectiveness of the prediction model in actual rock burst prediction applications and can more accurately predict the time and location of future rupture sources.

[0164] The system constructs a dual-driven rock burst intelligent prediction model for earthquake source data prediction and rock burst risk assessment. It can not only accurately identify and predict the time and location of rock burst disasters, but also further judge the development trend of rock burst disaster evolution. It forms a multi-dimensional intelligent prediction method for rock burst that integrates source parameter prediction and disaster evolution rating, providing more reliable technical support for mine safety.

[0165] In one example of the present invention, the prediction model module includes:

[0166] The loss function unit is configured to construct a loss function and calculate the data sample error. The expression of the loss function is as follows:

[0167]

[0168] Where n is the number of samples; l is the true value; is the predicted value;

[0169] A back-propagation training unit configured to calculate the gradient of the loss function with respect to the weights and biases of each layer of the output layer in reverse order from the output layer;

[0170] Loss function optimization unit: configured to calculate the loss function L with respect to a i The partial derivative of And update the core parameters of each rupture source: Where α is the learning rate that controls the step size of each update; the gradient is repeatedly calculated and the weights and biases of the prediction model are updated until the stopping condition is met;

[0171] The model judgment unit is configured to determine whether the model meets the requirements. If not, the parameters are adjusted accordingly and training is continued until the requirements are met and predictions are made using the test set.

[0172] In one example of the present invention, the risk model module includes:

[0173] The core parameter setting unit is configured to set the core parameter set of the focal mechanism for dynamic prediction of rock burst pressure P = (m, K, ΔV, bV t …);

[0174] The cumulative difference indicator unit is configured to set the data sequence of the parameters in the time window to be q1, q2, ..., q n-1 ,q n (n>3) where q n is the parameter when the time series is n; the subtraction of the previous and next terms in the sequence, Δq2=q2-q1,…,Δq n-1 =q n-1 -q n-2 , Δq n =q n -q n-1 , let G j =min{Δq n ,0},H j =max{Δq n , 0}, then

[0175]

[0176]

[0177] Let the cumulative difference R w =F1 / (F1+F2), R w ∈[0,1], calculate the cumulative difference values ​​of each parameter in the core parameter P of the rock burst dynamic prediction, and obtain the cumulative difference set of rock burst risk assessment parameters [R m , R K , R ΔV , R bVt …R w ];

[0178] The risk level evaluation unit is configured to determine the risk level evaluation index expression as follows:

[0179] F p＝ R m a1+R K a2+R ΔV a3+R bVt a4+…+R w a w

[0180] Where a1, a2, a3, a4, ... a w The rock burst risk level is divided into the following levels: p Evaluate rock burst risk and predict the level of rock burst risk in the future.

[0181] The above describes in detail an exemplary implementation of the earthquake source inversion data-driven rock burst prediction method and system proposed in the present invention with reference to the preferred embodiments. However, those skilled in the art will understand that, without departing from the concept of the present invention, various modifications and variations can be made to the above-mentioned specific embodiments, and various technical features and structures proposed in the present invention can be combined in various ways without exceeding the scope of protection of the present invention, which is determined by the appended claims.

Claims

1. A rock burst prediction method driven by earthquake source inversion data, characterized in that: The steps include: S10: Arrange a microseismic monitoring system according to the actual monitoring range of the mine site, use the microseismic monitoring system to collect raw acoustic wave data from the mine, perform positioning calculation and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors arranged in the coal seam at the working face; S20: Use standardized processing methods to preprocess the core parameters of coal rock fracture sources, and divide the preprocessed data set into training set, test set and validation set; S30: Construct a rock burst source data prediction model based on the Informer algorithm. Input the training set into the rock burst source data prediction model. Use the optimized loss function backpropagation to train the rock burst source data prediction model. Use the validation set to test and evaluate the trained rock burst source data prediction model. Continuously adjust the parameters of the rock burst source data prediction model for training until the requirements are met. Use the rock burst source data prediction model to predict the time, direction, magnitude, and core parameters of the coal and rock fracture sources in the future. S40: Construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as an input set, and establish a risk rating index by calculating the cumulative difference index of each core parameter of the focal mechanism. The rock burst disaster prediction risk is evaluated by combining the rock burst time, direction and risk rating index; S50: First, the predicted time, direction, magnitude of the rupture source and the core parameters of the coal and rock rupture source are used to reproduce the cross-scale gradual changes in the rock burst disaster process from the fracture surface to the fracture field; then the risk rating indicators are used to predict the evolution degree of the rock burst disaster and further identify the development trend of the rock burst.

2. The earthquake source inversion data-driven rock burst prediction method according to claim 1, characterized in that: In step S10, positioning calculation and focal mechanism inversion are performed on the original acoustic wave data to obtain core parameters of the coal rock fracture source directly related to the fracture, including the following steps: S11: Post-process the collected acoustic wave data and extract the first wave arrival time and amplitude information using the red delay information criterion method; S12: Using the arrival time data of the first acoustic wave at sensors at different locations of the microseismic monitoring system, the simplex algorithm is used to calculate the spatial location of the earthquake source; S13: Based on the constraints of the tensile-shear rupture source model, the six components of the source rupture moment tensor are solved using the acoustic first wave amplitude data collected by sensors at different locations. M pq ; S14: Based on the calculated moment tensor components, quantitative inversion is performed to calculate the core parameters of the coal rock fracture source, wherein the core parameters of the coal rock fracture source include fracture area, orientation, tension-shear properties, moment magnitude, and released energy.

3. The rock burst prediction method driven by earthquake source inversion data according to claim 2, characterized in that: In step S14, the specific calculation formula of the core parameters of the coal rock fracture source is as follows: Broken area: ； Where, is the broken area, M 1. M 3 is the moment tensor M kl The characteristic value of , μ is the second Lame constant, ξ is the proportional coefficient of the fracture area and displacement; Orientation: ； Where, is the angle between the movement direction of the earthquake source rupture surface and the normal direction, M 1. M 2. M 3 is the moment tensor M kl The eigenvalue of , λ is the first Lame constant, k is the spatial orientation, b is the direction of movement; Shearing properties: ； Where, V t is the tension volume, V s is the shear volume, K is the tension-shear ratio: K>1, the tensile rupture property is dominant; K<1, the shear rupture property is dominant; Moment magnitude: ； Where, M 1. M 2. M 3 is the moment tensor M kl The characteristic value of Release Energy: ； Where: σ t is the tensile strength, σ s For shear strength.

4. The earthquake source inversion data-driven rock burst prediction method according to claim 1, characterized in that: In step S20, a standardized processing method is used to pre-process the core parameters of the coal and rock fracture source, including: The preprocessing parameters are x 1 ，x 2 ，…，x n-1 ，x n For normalization processing, the formula is as follows: ； Where, is the minimum value of the preprocessing parameter, is the maximum value of the preprocessing parameter, y i ∈[0,1].

5. The earthquake source inversion data-driven rock burst prediction method according to claim 1, characterized in that: In step S30, a rock burst source data prediction model is constructed based on the Informer algorithm. The rock burst source data prediction model includes an encoder, a decoder, and a fully connected layer. The decoder includes a masked multi-head probabilistic sparse self-attention mechanism module and a multi-head attention mechanism module, wherein the masked multi-head probabilistic sparse self-attention mechanism module is configured to ensure the autoregressive nature of the model, prevent future information leakage, and improve computational efficiency; the multi-head attention mechanism module is configured to capture long-range dependencies in the sequence and process information in different subspaces in parallel through a multi-head mechanism, thereby enhancing the model's ability to understand complex time series data; The encoder includes a multi-head probabilistic sparse self-attention module and a self-attention distillation module. The multi-head probabilistic sparse self-attention module is configured to reduce the computational complexity of the model by randomly selecting sequence blocks and calculating attention scores only within these blocks, while enhancing the model's ability to capture local temporal dependencies. The self-attention distillation module is configured to improve the model's ability to understand long-range dependencies of long-sequence inputs by performing feature compression through the distillation step, enabling the model to better capture and utilize long-term dependency information in long-sequence data. The fully connected layer is configured to map the input into a higher dimensional space and then transform it back to the original dimension after filtering.

6. The earthquake source inversion data-driven rock burst prediction method according to claim 1, characterized in that: In step S30, the rock burst source data prediction model is trained using back propagation of the optimized loss function, and the trained rock burst source data prediction model is tested and evaluated using the validation set, including the following steps: S301: Calculate the data sample error and construct a loss function. The expression of the loss function is as follows: ； Where N is the number of samples; is the true value; is the predicted value; S302: Back propagation training of the rock burst source data prediction model: calculating the gradient of the loss function with respect to the weights and biases of each layer of the output layer in reverse order from the output layer; S303: Gradient descent optimization loss function: Calculate the loss function L about The partial derivative of ; And update the core parameters of each coal rock fracture source: ,in, α It is the learning rate that controls the step size of each update; it repeatedly calculates the gradient and updates the weights and biases of the impact ground pressure source data prediction model until the stopping condition is met; S304: Repeat S301 to S303 to determine whether the rock burst source data prediction model meets the requirements. If not, adjust the rock burst source data prediction model parameters accordingly and continue training until the requirements are met and prediction is performed using the test set.

7. An intelligent rock burst prediction system driven by earthquake source inversion data, characterized in that: include: a parameter acquisition module configured to deploy a microseismic monitoring system based on the actual monitoring range of the mine site, collect raw acoustic wave data from the mine using the microseismic monitoring system, perform location calculations and focal mechanism inversion on the raw acoustic wave data, and obtain core parameters of the coal and rock fracture source directly related to the fracture; wherein the microseismic monitoring system includes microseismic sensors deployed within the coal seam at the working face; a data partitioning module configured to preprocess the core parameters of the coal and rock fracture source using a standardized processing method, and to divide the preprocessed data set into a training set, a test set, and a validation set; The prediction model module is configured to construct a rock burst source data prediction model based on the Informer algorithm, input the training set into the rock burst source data prediction model, train the rock burst source data prediction model using the optimized loss function back propagation, test and evaluate the trained rock burst source data prediction model using the validation set, and continuously adjust the parameters of the rock burst source data prediction model for training until the requirements are met; and predict the time, direction, magnitude and core parameters of the coal and rock fracture source in the future through the rock burst source data prediction model; A risk model module is configured to construct a rock burst risk assessment model, using the predicted values ​​of the core parameters of the focal mechanism obtained by the rock burst source data prediction model as an input set, establishing a risk rating index by calculating the cumulative difference index of each core parameter of the focal mechanism, and evaluating the rock burst disaster prediction risk by combining the rock burst time, direction and risk rating index; The comprehensive prediction module is configured to first use the predicted time, direction, magnitude of the rupture source and the core parameters of the coal and rock rupture source to reproduce the cross-scale gradual changes in the rock burst disaster process from the fracture surface to the fracture field; then use the risk rating indicators to predict the degree of evolution of the rock burst disaster and further identify the development trend of the rock burst.

Images