Online monitoring method and system for geological parameters of coal mine tunnel

Abstract

The invention relates to the technical field of geological survey, and discloses a coal mine tunnel geological parameter on-line monitoring method and system, and the method comprises the steps: dynamically adjusting the deployment of nodes in a multi-parameter sensing network based on the surrounding rock stress distribution; dynamically calibrating the measurement parameters of the nodes to generate a standardized geological parameter set; time sequence dependence characteristics of the standardized geological parameter set and a pre-obtained historical accident data set are extracted in the two-channel input layer, key geological parameter mutation points in the time sequence dependence characteristics are focused based on a multi-head attention mechanism, the key geological parameter mutation points are input into the full connection layer, and a prediction probability is obtained; feeding back the prediction probability and the standardized geological parameter set to an optimization deployment scheme, obtaining the optimized prediction probability based on the optimized multi-parameter sensing network, and outputting a differentiated disaster response instruction; the effect of the on-line monitoring method for the geological parameters of the coal mine tunnel can be improved.

Description

Technical Field

[0001] This invention relates to the field of geological surveying technology, and in particular to a method and system for online monitoring of geological parameters in coal mine roadways. Background Technology

[0002] Existing monitoring methods are mostly based on fixed-deployment sensor networks, which are insufficient in responding to dynamic changes in the stress distribution of surrounding rock in coal mine roadways. During roadway excavation, the stress field of the surrounding rock changes significantly with the progress of mining. Fixed-deployment sensors cannot dynamically adjust the density and direction of monitoring nodes according to changes in stress distribution, resulting in missing or insufficient monitoring data in sensitive areas of stress concentration. This makes it difficult to accurately capture abrupt changes in geological parameters, thereby affecting the assessment of the stability of coal mine roadways.

[0003] In geological parameter analysis, existing technologies often lack the ability to effectively extract the time-series dependent characteristics of multiple parameters and focus on key abrupt changes. Traditional data analysis methods typically only perform threshold judgments on single parameters or simple parameter combinations, failing to fully explore the spatiotemporal correlations and temporal evolution patterns among different geological parameters. When potential safety hazards appear in coal mine roadways, multiple geological parameters will simultaneously undergo abnormal changes, but existing methods struggle to accurately identify key parameter abrupt changes and their interrelationships from complex time-series data. This leads to delayed or inaccurate assessments of geological conditions, hindering timely warnings of potential disaster risks. Summary of the Invention

[0004] This invention provides a method and system for online monitoring of geological parameters in coal mine roadways, the main purpose of which is to solve the problem of poor performance of online monitoring methods for geological parameters in coal mine roadways.

[0005] To achieve the above objectives, the present invention provides an online monitoring method for geological parameters in coal mine roadways, comprising:

[0006] S1. Dynamically adjust the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of surrounding rock in coal mine roadways;

[0007] S2. When the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes.

[0008] S3. Extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the stability of the coal mine roadway.

[0009] S4. Feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network, and obtain the optimized predicted probability based on the optimized multi-parameter sensor network;

[0010] S5. Based on the optimized predicted probabilities, classify the safety hazard levels and output differentiated disaster response instructions.

[0011] In a preferred embodiment, the dynamic adjustment of the deployment density and monitoring direction of nodes in the multi-parameter sensor network based on the surrounding rock stress distribution in coal mine roadways includes:

[0012] Initial stress distribution characteristics were extracted based on geological exploration reports and tunnel excavation trajectories.

[0013] Finite element numerical analysis was performed on the initial stress distribution characteristics to obtain the stress field cloud map of the coal mine roadway.

[0014] Stress gradient analysis is performed on the stress field cloud map, the region with the densest gradient division is identified as the sensitive region, and the energy difference threshold between the sensitive region and other regions is calculated.

[0015] The deployment density and monitoring direction of the nodes are dynamically determined based on the energy difference threshold.

[0016] In a preferred embodiment, when the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes, including:

[0017] An initial error distribution matrix is ​​established based on the reliable historical data of the nodes;

[0018] The initial error distribution matrix is ​​corrected based on the coupling effect of real-time ambient temperature and humidity on the sensor measurement parameters to obtain the target error distribution matrix of the node;

[0019] The importance of the node in the multi-parameter sensor network is quantified based on the target error distribution matrix to obtain the weight parameters of the node;

[0020] The weighted parameters are weighted and fused with the real-time geological data obtained by the multi-parameter sensor network from the geological parameter collection of the coal mine roadway to generate a standardized geological parameter set for the coal mine roadway.

[0021] In a preferred embodiment, the extraction of the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer includes:

[0022] A historical accident dataset for the coal mine roadways is compiled, which includes time-series data of geological parameters and corresponding roadway accident cases.

[0023] The standardized geological parameter set is input into the real-time geological parameter channel, and the historical accident dataset is input into the historical anomaly pattern channel to construct a dual-channel input layer.

[0024] The data of the dual-channel input layer is divided into a forward layer and a reverse layer. The forward layer and the reverse layer after time step expansion are concatenated into vectors to obtain the forward temporal dependency features and the reverse temporal dependency features of the coal mine roadway. The forward temporal dependency features and the reverse temporal dependency features are combined into temporal dependency features.

[0025] In a preferred embodiment, the step of focusing on abrupt changes in key geological parameters in the time-dependent features based on the multi-head attention mechanism includes:

[0026] Abrupt changes in parameters exceeding the normal range are used as the criteria for determining the multi-head attention mechanism.

[0027] At the spatial attention level, regions where the dependency weight of the temporal dependency feature exceeds the determination criteria are marked as high-risk regions;

[0028] At the time attention level, identify the mutation nodes of the time-dependent features, and construct a risk heatmap of the time-dependent features using the mutation nodes as timestamps;

[0029] By overlaying the high-risk area and the risk heatmap, the overlapping abrupt change points are taken as the key geological parameter abrupt change points in the time-dependent features.

[0030] In a preferred embodiment, the step of inputting the abrupt change points of the key geological parameters into the fully connected layer to obtain the predicted probability of the stability of the coal mine roadway includes:

[0031] The risk information and time-series information of the abrupt change points of the key geological parameters are compressed into risk heat map features and time-series features, respectively;

[0032] The spliced ​​risk heat map features and the time series features are input into a fully connected layer and then ReLU activation is performed to obtain the predicted probability of the stability of the coal mine roadway.

[0033] In a preferred embodiment, the predicted probability and the standardized geological parameter set are fed back to S1 to optimize the deployment scheme of the multi-parameter sensor network, including:

[0034] Spatiotemporal correlation detection is performed on the node and its neighboring nodes. When the detected value of the spatiotemporal correlation detection exceeds a preset normal threshold, the node is determined to be an abnormal node.

[0035] The missing data of the abnormal node is obtained by performing Bayesian network inference on the abnormal node, and the missing data is fed back to the multi-parameter sensor network.

[0036] When the abnormal node appears, the node position is calibrated and redundant node replacement is triggered based on the standardized geological parameter set. The data of the node position calibration and the data of the redundant node replacement are used as the optimized deployment scheme of the node, and the optimized deployment scheme is fed back to S1.

[0037] In a preferred embodiment, the deployment scheme includes:

[0038] When the predicted probability is dangerous, the node at the dangerous location will be adjusted, and redundant nodes will be activated for backup. The redundant nodes will be used as the optimized node positions in the deployment scheme.

[0039] When the predicted probability is no danger, the existing position of the node is maintained, and the multi-parameter sensor network is used to continue monitoring the coal mine roadway.

[0040] In a preferred embodiment, the step of classifying safety hazard levels based on the optimized predicted probabilities and outputting differentiated disaster response instructions includes:

[0041] When all parameters of the optimized predicted probability fluctuate within a preset safety threshold, the coal mine roadway operates normally.

[0042] When a single parameter of the optimized predicted probability exceeds a preset safety threshold, an early warning tracking is initiated for the coal mine roadway.

[0043] When several parameters of the optimized predicted probability exceed a preset safety threshold, the state with risk is confirmed and an early warning is issued.

[0044] When all parameters of the optimized predicted probability exceed the preset safety threshold, a compound disaster is confirmed, personnel are evacuated, and an alarm is issued.

[0045] To address the above problems, the present invention also provides a method and system for online monitoring of geological parameters in coal mine roadways, the system comprising:

[0046] A multi-parameter sensor network adjustment module is used to dynamically adjust the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of surrounding rock in coal mine roadways.

[0047] The geological parameter acquisition module is used to generate a standardized geological parameter set by dynamically calibrating the measurement parameters of the nodes when the multi-parameter sensor network acquires geological parameters of the coal mine roadway.

[0048] The roadway stability prediction module is used to extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the coal mine roadway stability.

[0049] An optimization module is used to feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network and obtain the optimized predicted probability based on the optimized multi-parameter sensor network.

[0050] The disaster early warning module is used to classify safety hazard levels based on the optimized predicted probabilities and output differentiated disaster response instructions.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. This invention dynamically adjusts the deployment density and monitoring direction of multi-parameter sensor network nodes based on the distribution of surrounding rock stress, enabling more targeted acquisition of geological parameters in key areas. By utilizing a dual-channel input layer to extract the temporal dependency features of standardized geological parameter sets and historical accident datasets, and by focusing on abrupt changes in key geological parameters through a multi-head attention mechanism, combined with the calculation of prediction probabilities using a fully connected layer, it can deeply explore the spatiotemporal correlations and temporal evolution patterns between parameters, achieving accurate assessment of coal mine roadway stability and improving the accuracy of geological condition judgment.

[0053] 2. This invention establishes an initial error distribution matrix and corrects it for real-time environmental factors to obtain a target error distribution matrix, dynamically calibrating the node measurement parameters. Simultaneously, it quantifies node weights based on the target error distribution matrix and weights them with real-time geological data to generate a standardized geological parameter set, effectively eliminating environmental interference and improving data reliability. Furthermore, by detecting the spatiotemporal correlation of nodes, using Bayesian network inference to repair missing data from abnormal nodes, and combining node position calibration and redundant node replacement optimization deployment schemes, the invention ensures the integrity and accuracy of the monitoring data. Attached Figure Description

[0054] Figure 1 A flowchart illustrating an online monitoring method for geological parameters in coal mine roadways according to an embodiment of the present invention;

[0055] Figure 2 This is a functional module diagram of an online monitoring method system for geological parameters in coal mine roadways provided in an embodiment of the present invention;

[0056] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0057] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0058] This application provides an online monitoring method for geological parameters in coal mine roadways. The executing entity of this online monitoring method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the online monitoring method for geological parameters in coal mine roadways can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0059] Reference Figure 1 The diagram shown is a flowchart illustrating an online monitoring method for geological parameters in coal mine roadways according to an embodiment of the present invention. In this embodiment, the online monitoring method for geological parameters in coal mine roadways includes:

[0060] S1. Dynamically adjust the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of surrounding rock in coal mine roadways;

[0061] In this embodiment of the invention, the dynamic adjustment of the deployment density and monitoring direction of nodes in the multi-parameter sensor network based on the surrounding rock stress distribution in coal mine roadways includes:

[0062] Initial stress distribution characteristics were extracted based on geological exploration reports and tunnel excavation trajectories.

[0063] Finite element numerical analysis was performed on the initial stress distribution characteristics to obtain the stress field cloud map of the coal mine roadway.

[0064] Stress gradient analysis is performed on the stress field cloud map, the region with the densest gradient division is identified as the sensitive region, and the energy difference threshold between the sensitive region and other regions is calculated.

[0065] The deployment density and monitoring direction of the nodes are dynamically determined based on the energy difference threshold.

[0066] Specifically, the geological exploration report is studied in detail to extract information such as the geological structure, rock strata characteristics, and burial depth of the coal mining area. At the same time, the tunnel excavation trajectory data is combined to determine the specific location and direction of the tunnel in the geological body. Through spatial analysis methods, the geological body is divided into multiple units according to certain rules. The stress state of each unit under the action of the gravity of the overlying rock strata and tectonic stress before tunnel excavation is analyzed to obtain the initial stress distribution characteristics.

[0067] Furthermore, the acquired initial stress distribution characteristic data are imported into finite element analysis software to establish a finite element model of the coal mine roadway and surrounding rock mass. Different material properties are assigned to the model according to the characteristics of the rock strata, and boundary conditions are set to simulate the actual geological environment. The model is solved using the finite element numerical calculation method to obtain the stress field cloud map of the coal mine roadway. The stress field cloud map intuitively presents the stress distribution of the roadway and surrounding rock mass.

[0068] Furthermore, the obtained stress field cloud map is analyzed, and the stress gradient at each point in the cloud map is calculated using the numerical difference method. That is, the cloud map is divided into multiple small regions, and the ratio of the difference in stress values ​​between adjacent regions to the distance between regions is calculated to obtain the stress gradient at each point. The stress gradient calculation results are visualized, and the region with the densest gradient division is identified as the sensitive region. Then, the difference in stress values ​​between the sensitive region and other regions is calculated using statistical analysis methods to determine the energy difference threshold.

[0069] Furthermore, based on the determined energy difference threshold, a node deployment strategy is formulated. For sensitive areas within the energy difference threshold range, the node deployment density is increased to ensure more detailed monitoring of the area. For other areas, the node deployment density is appropriately reduced. When determining the node monitoring direction, the node monitoring direction in sensitive areas is aligned with the direction of the greatest stress gradient change, while the node monitoring direction in other areas is rationally arranged according to the main stress direction. This dynamically divides the node deployment density and monitoring direction.

[0070] In summary, by dynamically adjusting the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of the surrounding rock in coal mine roadways, differentiated deployment of nodes can be achieved according to the actual stress conditions in different areas of the roadway.

[0071] In summary, by extracting the initial stress distribution characteristics from geological exploration reports and tunnel excavation trajectories and conducting finite element numerical analysis, sensitive areas with dense stress gradients can be accurately located. These areas are more prone to surrounding rock deformation or failure due to stress concentration and are therefore key monitoring areas.

[0072] In summary, dynamically adjusting the node deployment density based on the energy difference threshold between sensitive areas and other areas can increase the number of nodes in high-risk areas, improve the accuracy and density of monitoring data, and ensure timely capture of abnormal stress changes. At the same time, adjusting the node monitoring direction according to the principal stress direction in different areas ensures that the sensor measurement axis is consistent with the stress transmission direction, avoiding measurement errors caused by angular deviations, thereby more accurately reflecting the stress state of the surrounding rock.

[0073] In summary, this dynamic adjustment method changes the blindness of the traditional uniform deployment of nodes, which not only improves the monitoring system's ability to perceive the stability of the roadway, but also optimizes the allocation of node resources. While reducing hardware costs, it enhances the pertinence and effectiveness of monitoring, and provides more reliable basic data support for the safe operation of coal mine roadways.

[0074] S2. When the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes.

[0075] In this embodiment of the invention, when the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes, including:

[0076] An initial error distribution matrix is ​​established based on the reliable historical data of the nodes;

[0077] The initial error distribution matrix is ​​corrected based on the coupling effect of real-time ambient temperature and humidity on the sensor measurement parameters to obtain the target error distribution matrix of the node;

[0078] The importance of the node in the multi-parameter sensor network is quantified based on the target error distribution matrix to obtain the weight parameters of the node;

[0079] The weighted parameters are weighted and fused with the real-time geological data obtained by the multi-parameter sensor network from the geological parameter collection of the coal mine roadway to generate a standardized geological parameter set for the coal mine roadway.

[0080] Specifically, all reliable historical data collected by the collection node within a certain period of time is organized and arranged according to data type and collection time order.

[0081] Furthermore, a matrix is ​​constructed with nodes as rows and different data types or collection time points as columns.

[0082] Furthermore, each element in the matrix represents the error value of the corresponding node at a specific data type or time point. By analyzing the difference between historical data and the true value, the matrix elements are calculated and filled to establish the initial error distribution matrix.

[0083] Furthermore, real-time ambient temperature and humidity data are acquired to analyze the influence of temperature and humidity on the sensor measurement parameters.

[0084] Furthermore, a correction model is established that adjusts each element in the initial error distribution matrix based on the changes in temperature and humidity.

[0085] Furthermore, the specific method is to add or subtract corresponding correction values ​​to each element in the initial error distribution matrix based on the correspondence between temperature, humidity and error for different measurement parameters of each node, thereby obtaining the node target error distribution matrix after considering the influence of real-time environmental temperature and humidity.

[0086] Furthermore, an evaluation method is employed to quantify the importance of nodes in a multi-parameter sensor network. During the evaluation, the impact of node error distribution on the overall network data accuracy is comprehensively considered.

[0087] Furthermore, nodes with a wide error distribution range and large error values ​​have a greater impact on network data and a relatively lower importance assessment value; conversely, nodes with small errors and stable distributions have a higher importance assessment value.

[0088] Furthermore, this evaluation result is converted into a numerical form, which is the node weight parameter.

[0089] Furthermore, the weight parameters of each node are correlated with the real-time geological data collected by that node through a multi-parameter sensor network.

[0090] Furthermore, for each geological parameter, the parameter value collected by each node is multiplied by the weight parameter of the corresponding node, and then these products are added together and divided by the sum of all weight parameters. In this way, the real-time geological data is weighted and fused.

[0091] Furthermore, after weighted fusion processing, a standardized geological parameter set for coal mine roadways is generated, which includes multiple geological parameters and has a unified standard.

[0092] In summary, when collecting geological parameters of coal mine roadways using multi-parameter sensor networks, generating a standardized set of geological parameters by dynamically calibrating the node measurement parameters can effectively eliminate the interference of environmental factors on sensor data and improve the reliability and consistency of the data.

[0093] In summary, establishing an initial error distribution matrix based on reliable historical data of nodes can quantify the error performance of different nodes in historical operation and identify systematic errors caused by equipment aging, installation deviations, etc. Then, the initial error distribution matrix is ​​corrected according to the coupling effect of real-time ambient temperature and humidity on sensor measurement parameters to obtain the target error distribution matrix. This process can dynamically compensate for measurement drift caused by environmental variables (such as temperature and humidity changes).

[0094] For example, rising temperatures may cause some sensors to overestimate their readings. By correcting the model, the error value can be adjusted in real time to make the data closer to the actual geological parameters.

[0095] In summary, by further quantifying the importance of nodes based on the target error distribution matrix and generating weight parameters, different data credibility can be assigned according to the node error level. Nodes with small errors and high stability account for a higher proportion in data fusion, thus avoiding the distortion of the overall monitoring results due to the data of individual high-error nodes.

[0096] In summary, the weighted fusion method generates a standardized set of geological parameters, which can transform multi-source and heterogeneous raw measurement data into a unified standard dataset, eliminate data barriers caused by differences in sensor models and accuracy, and provide high-quality input data for subsequent tunnel stability analysis and prediction models.

[0097] In summary, this dynamic calibration mechanism significantly improves the accuracy and availability of monitoring data, ensuring that data-driven risk assessments and disaster early warnings are more scientific and reliable. At the same time, standardized processing enhances the compatibility and traceability of data across different systems, laying a solid data foundation for intelligent monitoring and safety management of coal mine roadways.

[0098] S3. Extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the stability of the coal mine roadway.

[0099] In this embodiment of the invention, the step of extracting the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer includes:

[0100] A historical accident dataset for the coal mine roadways is compiled, which includes time-series data of geological parameters and corresponding roadway accident cases.

[0101] The standardized geological parameter set is input into the real-time geological parameter channel, and the historical accident dataset is input into the historical anomaly pattern channel to construct a dual-channel input layer.

[0102] The data of the dual-channel input layer is divided into a forward layer and a reverse layer. The forward layer and the reverse layer after time step expansion are concatenated into vectors to obtain the forward temporal dependency features and the reverse temporal dependency features of the coal mine roadway. The forward temporal dependency features and the reverse temporal dependency features are combined into temporal dependency features.

[0103] The multi-head attention mechanism focuses on abrupt changes in key geological parameters within the temporal-dependent features, including:

[0104] Abrupt changes in parameters exceeding the normal range are used as the criteria for determining the multi-head attention mechanism.

[0105] At the spatial attention level, regions where the dependency weight of the temporal dependency feature exceeds the determination criteria are marked as high-risk regions;

[0106] At the time attention level, identify the mutation nodes of the time-dependent features, and construct a risk heatmap of the time-dependent features using the mutation nodes as timestamps;

[0107] By overlaying the high-risk area and the risk heatmap, the overlapping abrupt change points are taken as the key geological parameter abrupt change points in the time-dependent features.

[0108] The step of inputting the abrupt change points of the key geological parameters into the fully connected layer to obtain the predicted probability of the stability of the coal mine roadway includes:

[0109] The risk information and time-series information of the abrupt change points of the key geological parameters are compressed into risk heat map features and time-series features, respectively;

[0110] The spliced ​​risk heat map features and the time series features are input into a fully connected layer and then ReLU activation is performed to obtain the predicted probability of the stability of the coal mine roadway.

[0111] Specifically, various geological parameter data generated during the past operation of coal mine roadways are collected and arranged in chronological order to form geological parameter time series data. At the same time, roadway accident cases that have occurred are compiled and recorded. These geological parameter time series data and corresponding roadway accident cases are integrated together to form a historical accident dataset of coal mine roadways.

[0112] Furthermore, a data input structure is constructed, creating two independent data channels. One channel is dedicated to inputting the previously generated standardized geological parameter set, serving as the real-time geological parameter channel; the other channel is used to input the collected historical accident dataset, serving as the historical anomaly pattern channel. Together, these two channels form a dual-channel input layer.

[0113] Furthermore, the data received by the dual-channel input layer is processed by dividing the data into two parts along the time dimension: one part is the forward layer, in which the data is arranged sequentially according to time; the other part is the reverse layer, in which the data is arranged in reverse time order.

[0114] Furthermore, after the data of the forward and reverse layers are expanded at time steps, the data vectors at each time step are connected sequentially in a specific order to obtain the forward and reverse temporal dependency features of the coal mine roadway. Finally, these two features are merged and aggregated to obtain the complete temporal dependency features.

[0115] Furthermore, situations where geological parameters change beyond the pre-defined normal range are identified as abrupt changes, and these abrupt changes are used as the basis and condition for judgment by the multi-head attention mechanism.

[0116] Furthermore, at the spatial attention level, temporal dependency characteristics are analyzed, dependency weights for each region are calculated, and regions whose dependency weights reach or exceed the previously set criteria for determining abrupt changes are marked and identified as high-risk regions.

[0117] Furthermore, at the time attention level, nodes where geological parameters change abruptly are found in the time-dependent features. Once a change node is identified, the time corresponding to that node is used as a timestamp to visualize the risk level at different time points in the entire time-dependent features in the form of a heat map, thus forming a risk heat map.

[0118] Furthermore, the high-risk areas marked by the spatial attention level and the risk heatmap constructed by the temporal attention level are overlaid to find the abrupt change points corresponding to the overlapping parts of the two graphics. These abrupt change points are the key geological parameter abrupt change points in the temporal dependence features.

[0119] Furthermore, for key geological parameter mutation points, risk information and time series information contained therein are extracted respectively. Data compression technology is used to transform risk information into risk heat map features and time series information into time series features. Then, these two types of features are spliced ​​together in a certain order.

[0120] Furthermore, the stitched risk heat map features and time series features are input into the fully connected layer. The fully connected layer processes the input data and activates the processed data through the ReLU activation function. Through this operation, a value is finally obtained, which represents the predicted probability of the stability of the coal mine roadway.

[0121] Furthermore, for each data value output by the fully connected layer, if the data value is greater than zero, its original value is directly retained as the output; if the data value is less than or equal to zero, it is forcibly set to zero as the output.

[0122] Furthermore, the ReLU activation function can filter out positive valid signals and suppress negative invalid signals, thereby introducing nonlinear characteristics into the neural network, helping the model to better learn complex patterns in the data, and improving the accuracy of calculating the probability of predicting the stability of coal mine roadways.

[0123] In summary, by extracting the temporal dependency features of standardized geological parameter sets and historical accident datasets in the dual-channel input layer, focusing on key geological parameter mutation points based on the multi-head attention mechanism, and inputting them into the fully connected layer to obtain the predicted probability, we can deeply mine potential risk features related to roadway stability from the spatiotemporal dimensions.

[0124] In summary, by constructing a dual-channel input layer that includes a real-time data channel and a historical anomaly pattern channel, the temporal patterns of current geological parameters and historical accident data can be compared and analyzed. For example, the real-time stress change curve can be matched with the stress fluctuation pattern before the occurrence of historical accidents, thereby capturing the temporal dependence features that have the significance of accident precursors.

[0125] In summary, dividing dual-channel data into forward and reverse layers and concatenating the vectors can simultaneously capture the forward evolution patterns of time series (such as the trend of parameter increase over time) and reverse correlation features (such as the abnormal starting point when parameters are traced back after an accident), avoiding feature omissions caused by single-direction analysis.

[0126] In summary, the multi-head attention mechanism, by setting parameter mutation judgment conditions, can mark high-risk areas with abnormal stress gradient concentration at the spatial attention level and identify time nodes of parameter mutation at the temporal attention level. The key geological parameter mutation points determined by the superposition of the two can accurately locate high-risk states that simultaneously possess spatial stress concentration and temporal mutation characteristics.

[0127] For example, a region experiences a sudden increase in stress at a specific point in time, which closely matches the spatial stress distribution before historical accidents.

[0128] In summary, the risk and time-series information of key mutation points are compressed and input into the fully connected layer for ReLU activation. Through nonlinear transformation, multidimensional features can be mapped into a single predicted probability value, which intuitively reflects the stability state of the roadway.

[0129] In summary, this method, by integrating real-time data with historical experience and combining spatiotemporal attention mechanisms to focus on key risk points, can more comprehensively capture multi-factor coupled risks under complex geological conditions compared to traditional single-parameter threshold early warning. This improves the accuracy and timeliness of predictions and provides a data-driven intelligent analysis method for dynamic risk assessment of coal mine roadways. It effectively enhances the early identification capability of potential disasters and buys more response time for safety decisions.

[0130] S4. Feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network, and obtain the optimized predicted probability based on the optimized multi-parameter sensor network;

[0131] In this embodiment of the invention, the predicted probability and the standardized geological parameter set are fed back to S1 to optimize the deployment scheme of the multi-parameter sensor network, including:

[0132] Spatiotemporal correlation detection is performed on the node and its neighboring nodes. When the detected value of the spatiotemporal correlation detection exceeds a preset normal threshold, the node is determined to be an abnormal node.

[0133] The missing data of the abnormal node is obtained by performing Bayesian network inference on the abnormal node, and the missing data is fed back to the multi-parameter sensor network.

[0134] When the abnormal node appears, the node position is calibrated and redundant node replacement is triggered based on the standardized geological parameter set. The data of the node position calibration and the data of the redundant node replacement are used as the optimized deployment scheme of the node, and the optimized deployment scheme is fed back to S1.

[0135] The deployment scheme includes:

[0136] When the predicted probability is dangerous, the node at the dangerous location will be adjusted, and redundant nodes will be activated for backup. The redundant nodes will be used as the optimized node positions in the deployment scheme.

[0137] When the predicted probability is no danger, the existing position of the node is maintained, and the multi-parameter sensor network is used to continue monitoring the coal mine roadway.

[0138] Specifically, spatiotemporal correlation detection is performed on the data of each node and its neighboring nodes, which involves analyzing the data change patterns of each node in the time series and the consistency of its data with neighboring nodes in terms of spatial location.

[0139] Furthermore, in terms of the time dimension, we compare whether the trend of the current data of the node is consistent with that of the historical data; in terms of the spatial dimension, we compare whether the difference between the node data and the data of adjacent nodes at the same moment is within a reasonable range.

[0140] Furthermore, the detection results from the time and space dimensions are combined into a single detection value. When this detection value exceeds a pre-set normal threshold, the node is determined to be an abnormal node.

[0141] Furthermore, after identifying anomalous nodes, Bayesian network inference methods are used to process the missing data of these nodes. First, a Bayesian network model is constructed, using the data of anomalous nodes and related nodes as variables in the network. The network is trained using historical data to learn the probabilistic dependencies between variables.

[0142] Furthermore, based on the existing data of the abnormal node and the real-time data of the adjacent nodes, the most probable value of the missing data is calculated through probabilistic inference of the Bayesian network, and these inferred missing data are fed back to the multi-parameter sensor network to supplement the data integrity.

[0143] Furthermore, when an abnormal node is detected, the node's position is first calibrated. The actual physical location of the abnormal node is obtained using the Global Positioning System (GPS) or other positioning technologies, compared with a preset location recorded in the system, the position deviation is calculated, and the node's coordinates in the system are adjusted based on the deviation value to complete the position calibration.

[0144] Furthermore, based on a standardized set of geological parameters, suitable nodes are selected from pre-set redundant nodes for replacement.

[0145] Furthermore, during the replacement process, based on the monitoring needs of each area of ​​the roadway and the location of redundant nodes, redundant nodes are activated and deployed to the location of abnormal nodes or nearby areas that require enhanced monitoring. At the same time, the specific data of node location calibration and the detailed information of redundant node replacement are recorded. This data is then compiled into an optimized node deployment plan and fed back to step S1 (i.e., the initial node deployment stage) to optimize subsequent node deployment strategies.

[0146] Furthermore, when the predicted probability indicates a danger, adjustments are made to the nodes at the dangerous locations. First, the nodes corresponding to the dangerous locations are identified, and the deployment of these nodes is analyzed. Then, nodes with suitable distances from the dangerous locations and matching monitoring functions are selected from the redundant nodes as backup replacement nodes.

[0147] Furthermore, the redundant nodes to be prepared for replacement are activated, and their monitoring parameters and directions are adjusted so that they can cover the monitoring needs of dangerous locations. At the same time, the positions and parameter settings of these redundant nodes are used as the optimized node positions of the deployment plan and updated to the entire monitoring network to enhance the monitoring capability of dangerous areas.

[0148] Furthermore, when the predicted probability indicates no danger, the existing node positions remain unchanged, and no node adjustment or replacement operations are performed.

[0149] Furthermore, the multi-parameter sensor network continues to collect and transmit various geological parameters of the coal mine roadway in real time according to the current deployment plan and monitoring parameters, continuously monitoring the stability of the roadway and providing continuous data support for subsequent analysis and prediction.

[0150] In summary, by feeding the predicted probability and standardized geological parameter set back to S1 to optimize the deployment scheme of the multi-parameter sensor network, and obtaining the optimized predicted probability based on the optimized network, a closed-loop dynamic adjustment mechanism of "monitoring-analysis-optimization-re-monitoring" can be formed to continuously improve the adaptability and accuracy of the monitoring system.

[0151] In summary, by detecting the spatiotemporal correlation between nodes and their neighboring nodes, abnormal nodes caused by equipment failure, location offset, or environmental interference can be detected in a timely manner.

[0152] For example, if the spatial consistency of a node's data deviates significantly from that of its neighboring nodes and its time series fluctuations are abnormal, it is identified as an abnormal node and the missing data is supplemented through Bayesian network inference to avoid monitoring blind spots caused by missing data.

[0153] In summary, in handling abnormal nodes, by combining the standardized geological parameter set to trigger redundant node replacement and perform position calibration, the node layout can be dynamically adjusted according to the actual geological parameter distribution of the current roadway.

[0154] For example, redundant nodes can be added in areas of sudden stress change to increase monitoring density, while correcting the physical location deviation of nodes to ensure the accuracy of data spatial coordinates.

[0155] In summary, this optimized deployment scheme combines real-time monitoring data with prediction results. When the predicted probability indicates danger, it proactively adjusts the nodes at dangerous locations and activates redundant nodes for backup, enabling the monitoring network to adapt to changes in the risk status of the roadway.

[0156] For example, deploy more nodes in high-risk areas in advance to capture subtle parameter changes; maintain the existing layout when there is no danger predicted to avoid wasting resources.

[0157] In summary, the monitoring system can be continuously upgraded through feedback optimization mechanisms: on the one hand, the repair of abnormal nodes and the dynamic replacement of redundant nodes can continuously improve the integrity and reliability of the network and reduce false alarms or missed alarms caused by equipment problems; on the other hand, deployment adjustments based on predicted probabilities can ensure that nodes always focus on high-risk areas and key parameters.

[0158] For example, increasing node density and adjusting monitoring direction in areas with frequent historical accidents can improve the ability to warn of potential disasters.

[0159] In summary, the optimized multi-parameter sensor network can output more accurate prediction probabilities, forming a virtuous cycle and providing dynamic and intelligent infrastructure support for safety monitoring of coal mine roadways, effectively reducing safety risks caused by lagging monitoring systems or unreasonable layout.

[0160] S5. Based on the optimized predicted probabilities, classify the safety hazard levels and output differentiated disaster response instructions.

[0161] In this embodiment of the invention, the step of classifying safety hazard levels based on the optimized predicted probabilities and outputting differentiated disaster response instructions includes:

[0162] When all parameters of the optimized predicted probability fluctuate within a preset safety threshold, the coal mine roadway operates normally.

[0163] When a single parameter of the optimized predicted probability exceeds a preset safety threshold, an early warning tracking is initiated for the coal mine roadway.

[0164] When several parameters of the optimized predicted probability exceed a preset safety threshold, the state with risk is confirmed and an early warning is issued.

[0165] When all parameters of the optimized predicted probability exceed the preset safety threshold, a compound disaster is confirmed, personnel are evacuated, and an alarm is issued.

[0166] Specifically, the real-time fluctuations of all parameters in the optimized predicted probability are continuously monitored, and the current value of each parameter is compared with a pre-set safety threshold.

[0167] Furthermore, if the values ​​of all parameters fluctuate within the allowable range of the safety threshold, it indicates that the stability of the coal mine roadway is in a normal state, and the coal mine roadway is judged to be operating normally. The existing monitoring and production status should be maintained without additional intervention.

[0168] Furthermore, when the real-time value of a single parameter in the optimized predicted probability exceeds a preset safety threshold, an early warning and tracking mechanism for that parameter is immediately activated.

[0169] Furthermore, the types of parameters exceeding the threshold and their corresponding monitoring areas are first identified, and the monitoring frequency of the parameter in the area is increased. The parameter is then collected in real time with high density by a multi-parameter sensor network. At the same time, historical data is retrieved for comparative analysis to observe the parameter change trend and determine whether there is a possibility of continuous deterioration, so as to promptly identify potential risks and carry out targeted treatment.

[0170] Furthermore, if several parameters in the optimized predicted probabilities simultaneously exceed the preset safety threshold, the risk status corresponding to these parameters is first confirmed.

[0171] Furthermore, by cross-validating data, we can compare other relevant monitoring data and historical accident cases to rule out false alarms caused by sensor malfunctions or data transmission errors.

[0172] Furthermore, once the risk is confirmed to exist, the early warning system is immediately triggered to issue an alert to relevant departments and personnel. At the same time, the emergency response plan is activated, personnel are organized to investigate the risk area, and necessary preventive measures are taken to prevent the risk from escalating further.

[0173] Furthermore, when all parameters in the optimized predicted probability exceed the preset safety threshold, it is determined that a compound disaster may occur.

[0174] Furthermore, at this point, the highest level of emergency response is quickly activated. First, an emergency evacuation order is issued to all personnel in the coal mine through the broadcasting system and communication equipment, specifying the evacuation routes and assembly points, and organizing personnel to evacuate to a safe area in an orderly manner.

[0175] Furthermore, alerts should be issued to external rescue agencies, providing detailed information on the location and type of the disaster, and awaiting support from professional rescue forces. During the evacuation process, continuous monitoring of various parameters should be conducted to ensure the safe and orderly evacuation operation and minimize disaster losses.

[0176] In summary, by classifying safety hazard levels based on optimized predicted probabilities and outputting differentiated disaster response instructions, it is possible to implement graded responses according to the actual risk status of coal mine roadways, thereby achieving precise and efficient safety management.

[0177] In general, when all parameters fluctuate within the safety threshold, the roadway is considered to be operating normally, avoiding production interruptions caused by over-response and ensuring operational efficiency. When a single parameter exceeds the threshold, an early warning tracking is initiated. By increasing the monitoring frequency and comparing historical data, initial anomalies can be detected in a timely manner and targeted investigations can be carried out to prevent the risk from escalating. For example, intensive monitoring can be carried out in areas with abnormal single stress parameters to confirm whether there are signs of local loosening of the surrounding rock.

[0178] In general, if several parameters exceed the limits simultaneously, false alarms can be eliminated through cross-validation before triggering the warning, which can reduce misoperation caused by occasional sensor failures and ensure the reliability of the alarm.

[0179] For example, by comparing changes in multiple parameters such as stress, displacement, and humidity, an emergency procedure can be initiated only after confirming whether there is a real risk of surrounding rock instability. When all parameters exceed the limits, a compound disaster is confirmed, and the highest level of emergency response is immediately implemented, with personnel evacuation and external rescue being initiated simultaneously. This tiered response mechanism can take appropriate measures at different stages before a disaster occurs, avoiding insufficient or excessive response.

[0180] In summary, the core of differentiated response instructions is to handle situations in layers based on risk levels: maintain production under low-risk conditions, conduct precise monitoring under medium-risk conditions, issue decisive warnings under high-risk conditions, and evacuate quickly under extremely high-risk conditions. This not only conforms to the "tiered prevention and control" principle of coal mine safety management, but also enhances the pertinence of emergency response by dynamically adjusting resource allocation (such as manpower and equipment input).

[0181] In summary, this data-driven classification method avoids the limitations of the traditional "one-size-fits-all" early warning model. It can dynamically adjust response strategies based on real-time monitoring data, enhance the flexibility and scientific nature of safety management, minimize disaster losses, and ensure the continuity and safety of coal mine production.

[0182] like Figure 2 The diagram shown is a functional block diagram of an online monitoring system for geological parameters of coal mine roadways provided in an embodiment of the present invention.

[0183] The online monitoring method system 100 for geological parameters of coal mine roadways described in this invention can be installed in an electronic device. Depending on the functions implemented, the online monitoring method system 100 for geological parameters of coal mine roadways may include a multi-parameter sensor network adjustment module 101, a geological parameter acquisition module 102, a roadway stability prediction module 103, an optimization module 104, and a disaster early warning module 105. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0184] In this embodiment, the functions of each module / unit are as follows:

[0185] The multi-parameter sensor network adjustment module 101 is used to dynamically adjust the deployment density and monitoring direction of nodes in the multi-parameter sensor network based on the stress distribution of the surrounding rock in the coal mine roadway.

[0186] The geological parameter acquisition module 102 is used to generate a standardized geological parameter set by dynamically calibrating the measurement parameters of the nodes when the multi-parameter sensor network acquires geological parameters of the coal mine roadway.

[0187] The roadway stability prediction module 103 is used to extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the coal mine roadway stability.

[0188] The optimization module 104 is used to feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network and obtain the optimized predicted probability based on the optimized multi-parameter sensor network.

[0189] The disaster early warning module 105 is used to classify the safety hazard level based on the optimized predicted probability and output differentiated disaster response instructions.

[0190] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0191] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0192] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0193] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0194] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0195] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for online monitoring of geological parameters in coal mine roadways, characterized in that, The method includes: S1. Dynamically adjust the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of surrounding rock in coal mine roadways; S2. When the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes. S3. Extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the stability of the coal mine roadway. S4. Feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network, and obtain the optimized predicted probability based on the optimized multi-parameter sensor network; S5. Based on the optimized predicted probabilities, classify the safety hazard levels and output differentiated disaster response instructions.

2. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The deployment density and monitoring direction of nodes in the multi-parameter sensor network for dynamically adjusting the surrounding rock stress distribution in coal mine roadways include: Initial stress distribution characteristics were extracted based on geological exploration reports and tunnel excavation trajectories. Finite element numerical analysis was performed on the initial stress distribution characteristics to obtain the stress field cloud map of the coal mine roadway. Stress gradient analysis is performed on the stress field cloud map, the region with the densest gradient division is identified as the sensitive region, and the energy difference threshold between the sensitive region and other regions is calculated. The deployment density and monitoring direction of the nodes are dynamically determined based on the energy difference threshold.

3. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, When the multi-parameter sensor network collects geological parameters of the coal mine roadway, a standardized geological parameter set is generated by dynamically calibrating the measurement parameters of the nodes, including: An initial error distribution matrix is ​​established based on the reliable historical data of the nodes; The initial error distribution matrix is ​​corrected based on the coupling effect of real-time ambient temperature and humidity on the sensor measurement parameters to obtain the target error distribution matrix of the node; The importance of the node in the multi-parameter sensor network is quantified based on the target error distribution matrix to obtain the weight parameters of the node; The weighted parameters are weighted and fused with the real-time geological data obtained by the multi-parameter sensor network from the geological parameter collection of the coal mine roadway to generate a standardized geological parameter set for the coal mine roadway.

4. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The extraction of temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer includes: A historical accident dataset for the coal mine roadways is compiled, which includes time-series data of geological parameters and corresponding roadway accident cases. The standardized geological parameter set is input into the real-time geological parameter channel, and the historical accident dataset is input into the historical anomaly pattern channel to construct a dual-channel input layer. The data of the dual-channel input layer is divided into a forward layer and a reverse layer. The forward layer and the reverse layer after time step expansion are concatenated into vectors to obtain the forward temporal dependency features and the reverse temporal dependency features of the coal mine roadway. The forward temporal dependency features and the reverse temporal dependency features are combined into temporal dependency features.

5. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The multi-head attention mechanism focuses on abrupt changes in key geological parameters within the temporal-dependent features, including: Abrupt changes in parameters exceeding the normal range are used as the criteria for determining the multi-head attention mechanism. At the spatial attention level, regions where the dependency weight of the temporal dependency feature exceeds the determination criteria are marked as high-risk regions; At the time attention level, identify the mutation nodes of the time-dependent features, and construct a risk heatmap of the time-dependent features using the mutation nodes as timestamps; By overlaying the high-risk area and the risk heatmap, the overlapping abrupt change points are taken as the key geological parameter abrupt change points in the time-dependent features.

6. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The step of inputting the abrupt change points of the key geological parameters into the fully connected layer to obtain the predicted probability of the stability of the coal mine roadway includes: The risk information and time-series information of the abrupt change points of the key geological parameters are compressed into risk heat map features and time-series features, respectively; The spliced ​​risk heat map features and the time series features are input into a fully connected layer and then ReLU activation is performed to obtain the predicted probability of the stability of the coal mine roadway.

7. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The predicted probability and the standardized geological parameter set are fed back to S1 to optimize the deployment scheme of the multi-parameter sensor network, including: Spatiotemporal correlation detection is performed on the node and its neighboring nodes. When the detected value of the spatiotemporal correlation detection exceeds a preset normal threshold, the node is determined to be an abnormal node. The missing data of the abnormal node is obtained by performing Bayesian network inference on the abnormal node, and the missing data is fed back to the multi-parameter sensor network. When the abnormal node appears, the node position is calibrated and redundant node replacement is triggered based on the standardized geological parameter set. The data of the node position calibration and the data of the redundant node replacement are used as the optimized deployment scheme of the node, and the optimized deployment scheme is fed back to S1.

8. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The deployment scheme includes: When the predicted probability is dangerous, the node at the dangerous location will be adjusted, and redundant nodes will be activated for backup. The redundant nodes will be used as the optimized node positions in the deployment scheme. When the predicted probability is no danger, the existing position of the node is maintained, and the multi-parameter sensor network is used to continue monitoring the coal mine roadway.

9. The online monitoring method for geological parameters of coal mine roadways as described in claim 1, characterized in that, The process of classifying safety hazard levels based on the optimized predicted probabilities and outputting differentiated disaster response instructions includes: When all parameters of the optimized predicted probability fluctuate within a preset safety threshold, the coal mine roadway operates normally. When a single parameter of the optimized predicted probability exceeds a preset safety threshold, an early warning tracking is initiated for the coal mine roadway. When several parameters of the optimized predicted probability exceed a preset safety threshold, the state with risk is confirmed and an early warning is issued. When all parameters of the optimized predicted probability exceed the preset safety threshold, a compound disaster is confirmed, personnel are evacuated, and an alarm is issued.

10. A method and system for online monitoring of geological parameters in coal mine roadways, characterized in that, The system includes: A multi-parameter sensor network adjustment module is used to dynamically adjust the deployment density and monitoring direction of nodes in a multi-parameter sensor network based on the stress distribution of surrounding rock in coal mine roadways. The geological parameter acquisition module is used to generate a standardized geological parameter set by dynamically calibrating the measurement parameters of the nodes when the multi-parameter sensor network acquires geological parameters of the coal mine roadway. The roadway stability prediction module is used to extract the temporal dependency features of the standardized geological parameter set and the pre-acquired historical accident dataset in the dual-channel input layer, focus on the key geological parameter mutation points in the temporal dependency features based on the multi-head attention mechanism, input the key geological parameter mutation points into the fully connected layer, and obtain the predicted probability of the coal mine roadway stability. An optimization module is used to feed back the predicted probability and the standardized geological parameter set to S1 to optimize the deployment scheme of the multi-parameter sensor network and obtain the optimized predicted probability based on the optimized multi-parameter sensor network. The disaster early warning module is used to classify safety hazard levels based on the optimized predicted probabilities and output differentiated disaster response instructions.

Images