Coal mining rock burst prevention and early warning system

Abstract

The invention discloses a coal mining rock burst prevention and early warning system which comprises a data acquisition system, and the data acquisition system comprises a sensor network, a self-adaptive sensor layout module, a signal sending module and a power supply module. The sensor network comprises a stress sensor, a displacement sensor and a vibration sensor which are arranged in a coal mine tunnel; the signal sending module is used for sending data acquired by the sensor; the power module is used for supplying power to the sensor network and the signal sending module. The coal mining rock burst prevention and early warning system is compatible with three kinds of sensing data of a stress sensor, a displacement sensor and a vibration sensor, and the positions and density of the sensors in a high-risk area are dynamically adjusted. And establishing a coal seam rock burst risk evaluation expert system by using the monitored result, and evaluating the early warning result. The sensor layout strategy of the coal mining rock burst prevention and early warning system is optimized, the monitoring capability of a high-risk area is improved, and the rock burst danger early warning level can be improved.

Description

Technical Field

[0001] The present invention relates to the field of coal mine safety technology, and in particular to coal mine rock burst early warning technology. Background Art

[0002] With the continuous increase in the depth and intensity of coal mining, rock burst has become one of the major threats to coal mine safety production. In the past, the monitoring and prevention of rock burst in coal mines mainly relied on traditional methods, such as empirical judgment and borehole stress monitoring. However, these methods have problems such as low monitoring accuracy, untimely response, and inaccurate prediction, which are difficult to meet the needs of modern coal mine safety production.

[0003] In recent years, with the rapid development of sensor technology, information technology and artificial intelligence technology, new ideas and methods have been provided for the prevention and control of rock burst in coal mines. For example, high-precision pressure, displacement and vibration sensors can collect more accurate ground pressure data in real time; big data analysis and machine learning algorithms can mine potential laws and features from massive amounts of data to improve the accuracy of rock burst prediction; advanced communication technology can achieve rapid data transmission and real-time processing, providing a guarantee for timely warning.

[0004] However, the existing rock burst prevention and control technologies and systems still have some shortcomings. Monitoring sensors are generally evenly distributed in various areas, but the rock burst risks in various areas of the mine are different. Too many sensors in areas with low rock burst risk will cause waste of resources, while insufficient sensors in areas with high rock burst risk may cause risk assessment errors due to insufficient and incomplete data.

[0005] At the same time, the existing impact ground pressure prevention and control technologies and systems still have technical problems such as insufficient data integration and analysis capabilities, and incomplete and incomplete early warning systems. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a coal mining rock burst prevention and control early warning system to solve technical problems such as poor layout strategy of coal mining rock burst monitoring sensors, insufficient data integration and analysis capabilities, and an imperfect early warning system.

[0007] The coal mine impact rock pressure prevention and control early warning system of the present invention comprises a data acquisition system, wherein the data acquisition system comprises a sensor network, an adaptive sensor layout module, a signal transmission module and a power supply module; the sensor network comprises a stress sensor, a displacement sensor and a vibration sensor arranged in a coal mine tunnel, wherein the three sensors are respectively used to monitor the stress, displacement and seismic wave data of the surrounding rock and the working face of the coal mine tunnel; the signal transmission module is used to transmit the data collected by the sensor; the power supply module is used to supply power to the sensor network and the signal transmission module; the adaptive sensor layout module dynamically optimizes the layout positions and densities of the three sensors by the following method:

[0008] 1) defining the initial position coordinates of each sensor and the rock burst risk level corresponding to the sensor, wherein the risk level includes high risk, medium risk and low risk, and storing the initial position coordinates of each sensor and the corresponding rock burst risk level in a list;

[0009] 2) The adaptive sensor deployment module receives stress, displacement, and vibration data collected by the sensor network, and then performs risk level assessment based on the received sensor data:

[0010] If the data collected by the sensor meets any one of the following conditions: stress data>60, displacement data>0.6, vibration data>7, the rock burst risk level is: high risk;

[0011] If the data collected by the sensor meets any one of the following conditions: 55<stress data ≤60, 0.55<displacement data ≤0.6, 6<vibration data ≤7, and does not meet the high risk condition, the geological conditions are assessed as medium risk, and the rock burst risk level is: medium risk;

[0012] If the data collected by the sensor simultaneously meets the following conditions: stress data ≤ 55, displacement data ≤ 0.55, and vibration data ≤ 6, then the rock burst risk level is: low risk;

[0013] Update the rock burst risk level corresponding to the sensor in the list according to the assessment results;

[0014] 3) Obtain the number of sensors corresponding to the high risk level in the updated list and the location coordinates of the sensors, obtain the boundary of the high risk area of ​​rock burst according to the location coordinates, and then calculate the area of ​​the high risk area; then calculate the density of sensors in the high risk area; if it is calculated that the sensor layout density in the high risk area is less than the set threshold, the number of sensors in the high risk area will be increased so that the sensor density in the high risk area is greater than the set threshold, thereby strengthening the monitoring of the high risk area; if it is calculated that the sensor layout density in the high risk area is greater than the set threshold, the layout position and number of sensors will not be adjusted.

[0015] Furthermore, when the number of sensors arranged in the high-risk area is increased in step 3), the midpoint between two adjacent sensors in the high-risk area is used as the arrangement point of the newly added sensor, and the coordinates of the arrangement point are updated in the list.

[0016] Furthermore, the rock burst risk levels corresponding to the sensors in the list are updated according to a set period, and the sensor deployment positions are adjusted.

[0017] Furthermore, when the number of sensors arranged in the high-risk area is increased in the step 3), the added sensors are from the low-risk area.

[0018] The coal mining rock burst prevention and early warning system further includes a data analysis system and an early warning system;

[0019] The data analysis system includes a data receiving module, a data conversion and storage module, a data preprocessing module, a risk analysis module, a model training module, a model evaluation module and a result output module; the signal receiving module is used to receive data sent by the signal sending module; the data conversion and storage module converts the analog signal received by the signal receiving module into a digital signal, and stores the converted data in a database; the data preprocessing module is used to perform preprocessing operations on the data converted into digital signals; the risk analysis module is used to process the preprocessed data to give a risk prediction result; the model training module divides the data processed by the data preprocessing module into a training set, a verification set and a test set to train, verify and test the risk analysis module respectively; the model evaluation module uses the test set to evaluate the finally trained risk analysis module; the result output module outputs the result of the risk analysis in a clear and intuitive form;

[0020] The early warning system includes a signal receiving and processing module, an early warning level setting module, a risk judgment module, an early warning information sending module, and a power supply and backup module; the signal receiving and processing module is responsible for receiving the risk analysis results output by the data analysis system; the early warning level setting module is used to divide the early warning level according to the set threshold, and set specific early warning methods and response requirements according to the early warning level; the risk judgment module determines the early warning level to which the risk analysis result belongs based on the preset risk threshold; the early warning information sending module is used to send early warning information in different ways; the power supply and backup module supplies power to each module of the early warning system and backs up the early warning data.

[0021] Furthermore, the risk analysis module includes an LSTM layer and an MLP layer connected to the LSTM layer. The LSTM layer takes the data of three dimensions collected by the stress sensor, the displacement sensor and the vibration sensor as input and extracts data features. The MLP layer processes the features extracted by the LSTM layer and outputs the risk analysis results.

[0022] Further, the risk analysis module is trained in the following manner:

[0023] Parameter initialization: The weights of the LSTM layer and the MLP layer are initialized using the Xavier method, the bias is initialized to 0, the Adam optimizer is selected, the learning rate is 0.001, and the loss function uses the mean square error;

[0024] 32 samples are extracted from the training set as a batch at a time, and the batch data is input into the LSTM layer. The LSTM layer processes the sample information and outputs the hidden state. The hidden state is passed to the MLP layer for further processing and prediction. The mean square error loss of the batch is calculated, the gradient is calculated through backpropagation, and the weights and biases of the model are updated using the Adam optimizer.

[0025] L2 regularization was applied with a regularization parameter of 0.001. The number of LSTM units, the number of MLP layers and neurons, and the learning rate were adjusted by experimenting on the validation set. Training was stopped if the loss on the validation set did not decrease after 10 consecutive training rounds.

[0026] Furthermore, the model evaluation module calculates the mean square error and the mean absolute error, and determines whether the performance of the risk analysis module meets the requirements based on the calculation results. If not, the results are fed back to the model training module for adjustment.

[0027] Furthermore, the risk judgment module judges the warning level of the risk analysis result according to the preset risk threshold, including: when the stress value monitored by the stress sensor exceeds 70MPa, a mild assessment is triggered; if the strain value detected by the displacement sensor increases by more than 3% within 30 minutes, and the stress value monitored by the stress sensor exceeds 70MPa, a moderate risk warning is triggered; when the vibration frequency detected by the vibration sensor exceeds 8Hz and the amplitude exceeds 4mm, and the stress value monitored by the stress sensor exceeds 75MPa, and the strain value detected by the displacement sensor increases by more than 4% within 1 hour, a severe risk warning is triggered.

[0028] Furthermore, the coal mining rock burst prevention and early warning system includes a coal seam rock burst hazard assessment expert system;

[0029] The coal seam rock burst hazard assessment expert system includes a knowledge base module, an inference engine module, an explanation engine module and a human-machine interface module; the knowledge base module stores typical rock burst cases, rock burst theoretical research results, rock burst prevention and control laws and regulations, and rock burst prevention and control technologies as a weight judgment basis for the inference engine module; the inference engine module performs inference evaluation on the early warning results of the early warning system; the explanation engine module outputs the inference results intuitively to the user human-machine interface; the human-machine interface module provides interactive functions for users;

[0030] The inference engine module uses the hierarchical analysis method to perform reasoning in the following way:

[0031] A hierarchical model was established: target layer: impact rock pressure risk warning result evaluation; criterion layer: warning accuracy, warning timeliness, and warning consequence severity; sub-criteria layer: warning accuracy is divided into false alarm rate and missed alarm rate; warning timeliness is divided into warning issuance time and response time; warning consequence severity refers to casualties and economic losses caused by impact after warning;

[0032] Construct judgment matrix and consistency test: For the criterion layer, judge the relative importance of warning accuracy, warning timeliness, and warning consequence severity based on the expert experience in the knowledge base, and construct a judgment matrix. For the sub-criterion layer, judge the importance of each sub-criterion layer factor relative to the criterion layer factor to which it belongs, and construct the corresponding judgment matrix. Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, and perform a consistency test. If the consistency ratio is less than 0.1, the judgment matrix is ​​considered to have satisfactory consistency;

[0033] Hierarchical single sorting and total sorting: Hierarchical single sorting: Calculate the maximum eigenvalue of each judgment matrix and its corresponding eigenvector, and obtain the relative importance weight of each level element to the previous level element after normalization; Hierarchical total sorting: Synthesize the weights of each level to obtain the total weight of each factor of the sub-criteria layer to the target layer;

[0034] Analyze the warning results: multiply the scores of each sub-criteria layer by the corresponding weight and sum them up to obtain the comprehensive score of the warning result; judge the quality of the warning result based on the comprehensive score to provide a basis for subsequent decision-making.

[0035] Beneficial effects of the present invention:

[0036] 1. The coal mining rock burst prevention and early warning system of the present invention is compatible with three types of sensor data, namely, stress sensors, displacement sensors, and vibration sensors. By dynamically adjusting the position and density of sensors in high-risk areas, the sensor deployment strategy of the coal mining rock burst prevention and early warning system is optimized, and the monitoring capability of high-risk areas is improved, which is beneficial to improving the accuracy of rock burst early warning.

[0037] 2. The coal mine impact ground pressure prevention and control early warning system of the present invention organically integrates data monitoring, data analysis and early warning information transmission to form a complete early warning mechanism and process system, which can effectively warn of impact ground pressure in advance, reduce the probability of accidents, and reduce the huge economic losses caused by tunnel repairs, equipment replacement, production stagnation, etc. due to impact ground pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a structural diagram of the coal mining rock burst prevention and early warning system.

[0039] Figure 2 Schematic diagram of the sensor layout.

[0040] Figure 3 Schematic diagram of the model training module.

[0041] Figure 4 This is the structural system diagram of the expert system for coal seam rock burst hazard assessment. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0043] like Figure 1 As shown, in this embodiment, the coal mining rock burst prevention and control early warning system includes a data acquisition system, and the data acquisition system includes a sensor network, an adaptive sensor layout module, a signal sending module and a power supply module; the sensor network includes a stress sensor 1, a displacement sensor 2 and a vibration sensor 3 arranged in the coal mine tunnel, and the three sensors are used to monitor the stress, displacement and seismic wave data of the surrounding rock and the working face of the coal mine tunnel respectively; the signal sending module is used to send the data collected by the sensor; the power supply module is used to power the sensor network and the signal sending module; the adaptive sensor layout module dynamically optimizes the layout position and density of the three sensors by the following method:

[0044] 1) Define the initial position coordinates of each sensor and the rock burst risk level corresponding to the sensor, wherein the risk level includes high risk, medium risk and low risk, and store the initial position coordinates of each sensor and the corresponding rock burst risk level in a list.

[0045] 2) The adaptive sensor deployment module receives stress, displacement, and vibration data collected by the sensor network, and then performs risk level assessment based on the received sensor data:

[0046] If the data collected by the sensor meets any one of the following conditions: stress data>60, displacement data>0.6, vibration data>7, the rock burst risk level is high risk.

[0047] If the data collected by the sensor meets any one of the following conditions: 55<stress data ≤60, 0.55<displacement data ≤0.6, 6<vibration data ≤7, and does not meet the high-risk conditions, the geological conditions are assessed as medium risk, and the impact rock pressure risk level is: medium risk.

[0048] If the data collected by the sensor simultaneously meets the following conditions: stress data ≤55, displacement data ≤0.55, and vibration data ≤6, the rock burst risk level is: low risk.

[0049] The rock burst risk level corresponding to the sensor in the list is updated according to the assessment results.

[0050] 3) Obtain the number of sensors corresponding to the high risk level in the updated list and the location coordinates of the sensors, obtain the boundary of the high risk area of ​​rock burst according to the location coordinates, and then calculate the area of ​​the high risk area; then calculate the density of sensors in the high risk area; if it is calculated that the sensor layout density in the high risk area is less than the set threshold, the number of sensors in the high risk area will be increased so that the sensor density in the high risk area is greater than the set threshold, thereby strengthening the monitoring of the high risk area; if it is calculated that the sensor layout density in the high risk area is greater than the set threshold, the layout position and number of sensors will not be adjusted.

[0051] The coal mining rock burst prevention and control early warning system in this embodiment is compatible with three types of sensor data, namely, stress sensors, displacement sensors, and vibration sensors. By dynamically adjusting the position and density of sensors in high-risk areas, the sensor deployment strategy of the coal mining rock burst prevention and control early warning system is optimized, thereby improving the monitoring capability of high-risk areas, which in turn helps to improve the accuracy of rock burst early warning.

[0052] As an improvement to the above embodiment, when the number of sensors arranged in the high-risk area is increased in step 3), the midpoint between two adjacent sensors in the high-risk area is used as the arrangement point of the newly added sensor, and the coordinates of the arrangement point are updated in the list. The midpoint between two adjacent sensors in the high-risk area is used as the arrangement point of the newly added sensor, so that the newly added sensors can be evenly distributed in the high-risk area, and the position coordinates of the newly added sensors can also be easily obtained.

[0053] As an improvement to the above embodiment, the rock burst risk level corresponding to the sensor in the list is updated according to a set period, and the sensor layout position is adjusted. As the working face advances, the rock burst risk in the tunnel surrounding rock may change, so regular updates can timely adjust the monitoring strategy, which is conducive to better prevention of rock burst risk.

[0054] As an improvement to the above embodiment, when the number of sensors arranged in the high-risk area is increased in step 3), the added sensors are from the low-risk area. This improvement can reduce sensor waste and help save costs.

[0055] The coal mining rock burst prevention and early warning system described in this embodiment includes a data analysis system and an early warning system.

[0056] The data analysis system includes a data receiving module, a data conversion and storage module, a data preprocessing module, a risk analysis module, a model training module, a model evaluation module and a result output module. The signal receiving module is used to receive data sent by the signal sending module. The data conversion and storage module converts the analog signal received by the signal receiving module into a digital signal and stores the converted data in a database.

[0057] The data preprocessing module is used to perform preprocessing operations on the data after being converted into digital signals; the preprocessing operations include processing missing values ​​and abnormal values ​​of the data after being converted into digital signals, and performing data cleaning and standardization. First, process the missing values. It may be found that the mining stress, displacement or vibration data at certain moments are missing, and the average value of the adjacent data is used to fill the missing values. Next, identify the abnormal values ​​and determine whether the data is abnormal by setting a reasonable threshold range; for abnormal values, corrections can be made according to the data distribution. Then perform data cleaning to remove duplicate data records or delete data that is obviously inconsistent with the actual situation. Finally, standardization is performed to standardize all data so that its mean is 0 and its standard deviation is 1.

[0058] The risk analysis module is used to process the data after preprocessing operations to give risk prediction results. The model training module divides the data processed by the data preprocessing module into a training set, a validation set and a test set, specifically 70% for training, 15% for validation, and 15% for testing, so as to train, validate and test the risk analysis module respectively. The model evaluation module uses the test set to evaluate the finally trained risk analysis module. The result output module outputs the results of the risk analysis in a clear and intuitive form, including outputting in the form of generating reports and data tables, and delivering the analysis results in a timely manner.

[0059] In this embodiment, the risk analysis module includes an LSTM layer and an MLP layer connected to the LSTM layer. The LSTM layer takes the three-dimensional data collected by the stress sensor, displacement sensor and vibration sensor as input and extracts data features. The MLP layer processes the features extracted by the LSTM layer and outputs the risk analysis results. The risk analysis module is trained in the following way:

[0060] Parameter initialization: The weights of the LSTM layer and the MLP layer are initialized using the Xavier method, the bias is initialized to 0, the Adam optimizer is selected, the learning rate is 0.001, and the loss function uses the mean square error;

[0061] 32 samples are extracted from the training set as a batch at a time, and the batch data is input into the LSTM layer. The LSTM layer processes the sample information and outputs the hidden state. The hidden state is passed to the MLP layer for further processing and prediction. The mean square error loss of the batch is calculated, the gradient is calculated through backpropagation, and the weights and biases of the model are updated using the Adam optimizer.

[0062] L2 regularization was applied with a regularization parameter of 0.001. The number of LSTM units, the number of MLP layers and neurons, and the learning rate were adjusted by experimenting on the validation set. Training was stopped if the loss on the validation set did not decrease after 10 consecutive training rounds.

[0063] The model evaluation module calculates the mean square error and the mean absolute error, and determines whether the performance of the risk analysis module meets the requirements based on the calculation results. If not, it is fed back to the model training module for adjustment.

[0064] In this embodiment, the early warning system includes a signal receiving and processing module, an early warning level setting module, a risk judgment module, an early warning information sending module, and a power supply and backup module; the signal receiving and processing module is responsible for receiving the risk analysis results output by the data analysis system; the early warning level setting module is used to divide the early warning level according to the set threshold, and set specific early warning methods and response requirements according to the early warning level; the risk judgment module determines the early warning level to which the risk analysis result belongs based on the preset risk threshold; the early warning information sending module is used to send early warning information in different ways; the power supply and backup module supplies power to each module of the early warning system and backs up the early warning data.

[0065] In this embodiment, the risk judgment module judges the warning level of the risk analysis result according to the preset risk threshold, including: when the stress value monitored by the stress sensor exceeds 70MPa, a mild assessment is triggered; if the strain value detected by the displacement sensor increases by more than 3% within 30 minutes, and the stress value monitored by the stress sensor exceeds 70MPa, a moderate risk warning is triggered; when the vibration frequency detected by the vibration sensor exceeds 8Hz and the amplitude exceeds 4mm, and the stress value monitored by the stress sensor exceeds 75MPa, and the strain value detected by the displacement sensor increases by more than 4% within 1 hour, a severe risk warning is triggered.

[0066] The coal mining rock burst prevention and early warning system in this embodiment also includes a coal seam rock burst hazard assessment expert system.

[0067] The knowledge base module is the source of the inference engine data weight. The knowledge base module contains a large number of typical rock burst cases, rock burst theoretical research results, rock burst prevention and control laws and regulations, and rock burst prevention and control technologies. Users can judge the relative importance of early warning accuracy, early warning timeliness, and severity of early warning consequences, as well as quantitative scores at the sub-criteria level based on the expert experience in the knowledge base module.

[0068] The inference engine module adopts the hierarchical analysis method to analyze the rationality of the rock burst risk warning assessment of the early warning system. The interpretation engine module outputs the inference results intuitively to the user human-machine interface; the human-machine interface module provides users with interactive functions to facilitate users to view the results.

[0069] In this embodiment, the coal seam rock burst hazard assessment expert system inference engine module, the inference engine module includes a target layer: rock burst warning result evaluation.

[0070] Criteria level: warning accuracy, warning timeliness, and severity of warning consequences.

[0071] Sub-criteria level: Warning accuracy: false alarm rate, missed alarm rate. Warning timeliness: warning issuance time, response time. Warning consequence severity: casualty estimate, economic loss estimate.

[0072] Construct the criterion layer judgment matrix and consistency test, calculate the consistency ratio CR, and judge that the matrix has satisfactory consistency. For the sub-criterion layer judgment matrix, perform consistency test and confirm that it has satisfactory consistency.

[0073] Hierarchical single ranking and total ranking: Calculate the weight of each factor in the criterion layer relative to the target layer based on the knowledge base. For example, the result is that the weight of warning accuracy is 0.5, the weight of warning timeliness is 0.3, and the weight of warning consequence severity is 0.2.

[0074] The weights of each factor in the sub-criteria layer relative to the factors in the corresponding criterion layer are calculated based on the knowledge base, such as the weight of false alarm rate is 0.4, and the weight of missed alarm rate is 0.6. The weight of warning issuance time is 0.4, and the response time is 0.6. The estimated casualties are 0.7, and the estimated economic losses are 0.3. The weights of each level are synthesized to obtain the total weight of each factor in the sub-criteria layer relative to the target layer.

[0075] Analyze the warning results. Assume that a warning is a mild risk warning. The specific situation is as follows: the false alarm rate is 5%, and the missed alarm rate is 3%. The warning is issued in 10 minutes and the response time is 20 minutes. There are no casualties and no economic losses. These specific situations are quantified and scored according to the knowledge base module.

[0076] After quantitative scoring, for example, the false alarm rate score is 80 points, the missed alarm rate score is 85 points, the warning issuance time score is 75 points, the response time score is 70 points, the estimated casualties score is 80 points, and the estimated economic loss score is 85 points.

[0077] The scores of each sub-criteria layer are multiplied by their total weight relative to the target layer and the sum is calculated to obtain the comprehensive score of the warning result. According to the pre-set evaluation criteria, the warning result can be judged as reliable (100 points-90 points), relatively reliable (90 points-75 points), or unreliable (below 75 points).

[0078] The warning information sending module includes an audio-visual alarm module, which is used to send different sound and light warning signals according to different warning levels. In this embodiment, the audio-visual alarm module sends audio-visual signals of different colors according to three different warning levels of mild, moderate and severe. The warning level changes from mild to severe, the color sent changes from yellow to orange and finally to red, and the alarm sound changes from an interval of 3s to an interval of 1s to no interval, so as to attract the attention of on-site personnel.

[0079] In this embodiment, the warning information sending module includes a text message notification module, which is connected to the mobile communication network to send warning text messages to preset relevant personnel to ensure that the relevant personnel can receive the warning notification in time even if they are not on site.

[0080] In this embodiment, the warning information sending module also includes a broadcast notification module, which issues corresponding warning information and instructions to the entire mining area through the broadcast system inside the coal mine. The broadcast can cover a larger range, allowing more people to receive emergency notifications.

[0081] In this embodiment, the warning information sending module also includes a display module, including a large-screen display and an electronic billboard. The large-screen display is placed in the bottom chamber of the well and the parking lot, and the electronic billboard is placed in the tunnel. The display module is used to intuitively display the warning information, so that the staff can check the location of the rock burst and the rock burst warning level at any time.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A coal mining rock burst prevention and early warning system, characterized by: The invention comprises a data acquisition system, wherein the data acquisition system comprises a sensor network, an adaptive sensor layout module, a signal transmission module and a power supply module; the sensor network comprises a stress sensor, a displacement sensor and a vibration sensor arranged in a coal mine tunnel, wherein the three sensors are respectively used to monitor the stress, displacement and seismic wave data of the surrounding rock and the working face of the coal mine tunnel; the signal transmission module is used to transmit the data collected by the sensor; the power supply module is used to supply power to the sensor network and the signal transmission module; the adaptive sensor layout module dynamically optimizes the layout position and density of the three sensors by the following method: 1) defining the initial position coordinates of each sensor and the rock burst risk level corresponding to the sensor, wherein the risk level includes high risk, medium risk and low risk, and storing the initial position coordinates of each sensor and the corresponding rock burst risk level in a list; 2) The adaptive sensor deployment module receives stress, displacement, and vibration data collected by the sensor network, and then performs risk level assessment based on the received sensor data: If the data collected by the sensor meets any one of the following conditions: stress data>60, displacement data>0.6, vibration data>7, the rock burst risk level is: high risk; If the data collected by the sensor meets any one of the following conditions: 55<stress data ≤60, 0.55<displacement data ≤0.6, 6<vibration data ≤7, and does not meet the high risk condition, the geological conditions are assessed as medium risk, and the rock burst risk level is: medium risk; If the data collected by the sensor simultaneously meets the following conditions: stress data ≤ 55, displacement data ≤ 0.55, and vibration data ≤ 6, then the rock burst risk level is: low risk; Update the rock burst risk level corresponding to the sensor in the list according to the assessment results; 3) Obtain the number of sensors corresponding to the high risk level in the updated list and the location coordinates of the sensors, obtain the boundary of the high risk area of ​​rock burst according to the location coordinates, and then calculate the area of ​​the high risk area; then calculate the density of sensors in the high risk area; if it is calculated that the density of sensor layout in the high risk area is less than the set threshold, the number of sensors arranged in the high risk area will be increased so that the sensor density in the high risk area is greater than the set threshold, thereby strengthening the monitoring of the high risk area; If the calculated sensor layout density in the high-risk area is greater than the set threshold, the layout position and number of sensors will not be adjusted.

2. The coal mining rock burst prevention and early warning system according to claim 1 is characterized by: When the number of sensors deployed in the high-risk area is increased in step 3), the midpoint between two adjacent sensors in the high-risk area is used as the deployment point of the newly added sensor, and the coordinates of the deployment point are updated in the list.

3. The coal mining rock burst prevention and early warning system according to claim 1 is characterized by: Update the rock burst risk level corresponding to the sensors in the list according to the set period, and adjust the sensor layout position.

4. The coal mining rock burst prevention and early warning system according to claim 1 is characterized by: When the number of sensors arranged in the high-risk area is increased in the step 3), the added sensors are from the low-risk area.

5. The coal mining rock burst prevention and early warning system according to claim 1 is characterized by: Including data analysis system and early warning system; The data analysis system includes a data receiving module, a data conversion and storage module, a data preprocessing module, a risk analysis module, a model training module, a model evaluation module and a result output module; the signal receiving module is used to receive data sent by the signal sending module; the data conversion and storage module converts the analog signal received by the signal receiving module into a digital signal, and stores the converted data in a database; the data preprocessing module is used to perform preprocessing operations on the data converted into digital signals; the risk analysis module is used to process the preprocessed data to give a risk prediction result; the model training module divides the data processed by the data preprocessing module into a training set, a verification set and a test set to train, verify and test the risk analysis module respectively; the model evaluation module uses the test set to evaluate the finally trained risk analysis module; the result output module outputs the result of the risk analysis in a clear and intuitive form; The early warning system includes a signal receiving and processing module, an early warning level setting module, a risk judgment module, an early warning information sending module, and a power supply and backup module; the signal receiving and processing module is responsible for receiving the risk analysis results output by the data analysis system; the early warning level setting module is used to divide the early warning level according to the set threshold, and set a specific early warning method and response requirements according to the early warning level; the risk judgment module judges the early warning level to which the risk analysis result belongs according to the preset risk threshold; The warning information sending module is used to send warning information in different ways; The power supply and backup module supplies power to each module of the early warning system and backs up the early warning data.

6. The coal mining rock burst prevention and early warning system according to claim 5 is characterized by: The risk analysis module includes an LSTM layer and an MLP layer connected to the LSTM layer. The LSTM layer takes the data of three dimensions collected by the stress sensor, the displacement sensor and the vibration sensor as input and extracts data features. The MLP layer processes the features extracted by the LSTM layer and outputs the risk analysis results.

7. The coal mining rock burst prevention and early warning system according to claim 6 is characterized by: The risk analysis module is trained in the following manner: Parameter initialization: The weights of the LSTM layer and the MLP layer are initialized using the Xavier method, the bias is initialized to 0, the Adam optimizer is selected, the learning rate is 0.001, and the loss function uses the mean square error; 32 samples are extracted from the training set as a batch at a time, and the batch data is input into the LSTM layer. The LSTM layer processes the sample information and outputs the hidden state. The hidden state is passed to the MLP layer for further processing and prediction. The mean square error loss of the batch is calculated, the gradient is calculated through backpropagation, and the weights and biases of the model are updated using the Adam optimizer. L2 regularization was applied with a regularization parameter of 0.

001. The number of LSTM units, the number of MLP layers and neurons, and the learning rate were adjusted by experimenting on the validation set. Training was stopped if the loss on the validation set did not decrease after 10 consecutive training rounds.

8. The coal mining rock burst prevention and early warning system according to claim 7 is characterized by: The model evaluation module calculates the mean square error and the mean absolute error, and determines whether the performance of the risk analysis module meets the requirements based on the calculation results. If not, it is fed back to the model training module for adjustment.

9. The coal mining rock burst prevention and early warning system according to claim 8 is characterized by: The risk judgment module judges the warning level of the risk analysis result according to the preset risk threshold value, including: when the stress value monitored by the stress sensor exceeds 70MPa, a mild risk warning is triggered; if the strain value detected by the displacement sensor increases by more than 3% within 30 minutes, and the stress value monitored by the stress sensor exceeds 70MPa, a moderate risk warning is triggered; when the vibration frequency detected by the vibration sensor exceeds 8Hz and the amplitude exceeds 4mm, and the stress value monitored by the stress sensor exceeds 75MPa, and the strain value detected by the displacement sensor increases by more than 4% within 1 hour, a severe risk warning is triggered.

10. The coal mining rock burst prevention and early warning system according to claim 1, characterized in that: Including coal seam rock burst hazard assessment expert system; The coal seam rock burst hazard assessment expert system includes a knowledge base module, an inference engine module, an explanation engine module and a human-machine interface module; the knowledge base module stores typical rock burst cases, rock burst theoretical research results, rock burst prevention and control laws and regulations, and rock burst prevention and control technologies as a weight judgment basis for the inference engine module; the inference engine module performs inference evaluation on the early warning results of the early warning system; the explanation engine module outputs the inference results intuitively to the user human-machine interface; the human-machine interface module provides interactive functions for users; The inference engine module uses the hierarchical analysis method to perform reasoning in the following way: Establish a hierarchical model: target layer: impact rock pressure risk warning result evaluation; criterion layer: warning accuracy, warning timeliness, and warning consequence severity; sub-criterion layer: warning accuracy is divided into false alarm rate and missed alarm rate; Warning timeliness is divided into warning issuance time and response time; The severity of the warning consequence refers to the casualties and economic losses caused when the impact occurs after the warning; Construct judgment matrix and consistency test: For the criterion layer, judge the relative importance of warning accuracy, warning timeliness, and warning consequence severity based on the expert experience in the knowledge base, and construct a judgment matrix. For the sub-criterion layer, judge the importance of each sub-criterion layer factor relative to the criterion layer factor to which it belongs, and construct the corresponding judgment matrix. Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, and perform a consistency test. If the consistency ratio is less than 0.1, the judgment matrix is ​​considered to have satisfactory consistency; Hierarchical single sorting and total sorting: Hierarchical single sorting: Calculate the maximum eigenvalue of each judgment matrix and its corresponding eigenvector, and obtain the relative importance weight of each level element to the previous level element after normalization; Hierarchical total sorting: Synthesize the weights of each level to obtain the total weight of each factor of the sub-criteria layer to the target layer; Analyze the warning results: multiply the scores of each sub-criteria layer by the corresponding weight and sum them up to obtain the comprehensive score of the warning result; The quality of the early warning results is judged based on the comprehensive score, providing a basis for subsequent decision-making.

Images