Mine water inrush early warning method based on distributed optical fiber monitoring

Abstract

The application discloses a mine water inrush early warning method based on distributed optical fiber monitoring and belongs to the technical field of mine safety monitoring. In view of the technical problems that the spatial continuous perception of groundwater seepage state is absent in complex geological structure of mines and the dynamic evolution process of water inrush channels is difficult to capture, microseismic monitoring signals, distributed optical fiber strain signals and groundwater chemical ion concentration data are acquired, a random forest algorithm is used to extract rock mass microcrack expansion features, rock stratum deformation features and water enrichment features, a three-dimensional space grid is constructed to obtain an initial seepage state, a long short-term memory network is used to extract time sequence features to obtain a continuous seepage evolution sequence, the seepage gradient change value of adjacent time node data is calculated to determine a water-rock coupling abnormal area, a water inrush channel dynamic evolution track is generated by tracking adjacent abnormal areas, the spatial distance between the track extension direction and the safety boundary of a mining working face is calculated, and a water inrush early warning instruction is triggered when the distance is less than a preset threshold. The application is used for early warning of water inrush disasters in the mining process.

Description

Technical Field

[0001] This invention belongs to the field of mine safety monitoring technology, and in particular relates to a mine water inrush early warning method based on distributed optical fiber monitoring. Background Technology

[0002] Groundwater hazard prevention during mining operations is directly related to miners' safety and the continuity of mine production, representing one of the most severe challenges in mine safety management. Current mine water inrush monitoring primarily relies on point-based monitoring methods such as borehole water level observation and flow measurement. While these methods can obtain hydrological data for specific locations, they struggle to construct a complete dynamic picture of groundwater in mining environments with complex groundwater conditions and uneven aquifer distribution. Furthermore, due to numerous blind spots between monitoring points, the expansion of seepage channels or the development of fissures in concealed areas far from monitoring points often go undetected, leading to missed signs of water inrush.

[0003] More importantly, mine water inrushes undergo a dynamic evolution from gestation to eruption. The seepage of groundwater into the rock strata triggers a redistribution of stress in the surrounding rock, thereby altering the opening and connectivity of fractures. This water-rock coupling effect gives the water inrush channel a time-varying characteristic. Traditional monitoring methods, with their discrete data points, struggle to capture this continuous process, especially when groundwater pressure gradually accumulates in a certain section and fracture networks quietly connect. Without real-time tracking of this evolutionary trajectory, it's impossible to determine the direction of the water inrush risk and its rate of evolution. For example, ahead of the tunneling face, the water-bearing fault zone may be separated from the coal seam by only a layer of pressure rock. As mining disturbances occur, micro-fractures within the rock mass gradually expand and connect, continuously changing the groundwater seepage path. However, due to the lack of continuous monitoring of this dynamic process, early warning systems often only issue an alarm after a sudden surge in water volume, leaving extremely limited time for emergency response. Therefore, how to achieve continuous spatial perception of groundwater seepage in the complex geological environment of mines and accurately capture the dynamic evolution characteristics during the formation of water inrush channels, so as to provide reliable early warnings during the stage of water inrush risk accumulation, has become a key issue that urgently needs to be addressed in mine water hazard prevention and control. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a mine water inrush early warning method based on distributed optical fiber monitoring, thereby resolving the issues present in the prior art.

[0005] In a first aspect, to achieve the above objectives, the present invention provides a mine water inrush early warning method based on distributed optical fiber monitoring, comprising the following steps:

[0006] Acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area;

[0007] The random forest algorithm was used to process the microseismic monitoring signal, the distributed optical fiber strain signal, and the groundwater chemical ion concentration data to extract the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment.

[0008] A three-dimensional spatial grid of the mine geological body is constructed by the characteristics of the microfracture propagation of the rock mass, the characteristics of the deformation of the rock strata, and the characteristics of water enrichment, and the initial seepage state corresponding to each three-dimensional spatial grid is obtained.

[0009] The initial seepage state is input into a long short-term memory network for temporal feature extraction to obtain a continuous seepage evolution sequence;

[0010] Based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, a three-dimensional spatial grid with water-rock coupling anomalies is determined, and the water-rock coupling anomaly region is identified.

[0011] Based on the spatial distribution coordinates of the water-rock coupling anomaly region, adjacent water-rock coupling anomaly regions are tracked to generate the dynamic evolution trajectory of the water inflow channel;

[0012] Obtain the extension direction vector of the dynamic evolution trajectory of the water inflow channel, and calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face;

[0013] A water inrush warning command is triggered based on the comparison result between the spatial distance and the preset distance threshold.

[0014] Secondly, the present invention also provides a mine water inrush early warning system based on distributed optical fiber monitoring, for implementing a mine water inrush early warning method based on distributed optical fiber monitoring, the system comprising:

[0015] The data acquisition module is used to acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area.

[0016] The feature extraction module is used to process the microseismic monitoring signal, the distributed optical fiber strain signal and the groundwater chemical ion concentration data using the random forest algorithm to extract the characteristics of rock mass microfracture propagation, rock strata deformation and water enrichment.

[0017] The spatial grid modeling module is used to construct a three-dimensional spatial grid of the mine geological body through the characteristics of the rock mass microfracture propagation, the characteristics of the rock strata deformation, and the characteristics of water enrichment, and to obtain the initial seepage state corresponding to each three-dimensional spatial grid.

[0018] The temporal evolution analysis module is used to input the initial seepage state into a long short-term memory network for temporal feature extraction, obtain a continuous seepage evolution sequence, and determine the three-dimensional spatial grid with water-rock coupling anomalies based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, thereby identifying the water-rock coupling anomaly region.

[0019] The water inflow channel tracking module is used to track adjacent water-rock coupling anomaly areas based on the spatial distribution coordinates of the water-rock coupling anomaly area, and generate a dynamic evolution trajectory of the water inflow channel.

[0020] The early warning decision module is used to obtain the extension direction vector of the dynamic evolution trajectory of the water inrush channel, calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face, and trigger a water inrush early warning command based on the comparison result of the spatial distance and the preset distance threshold.

[0021] Thirdly, the present invention also provides a computer terminal device, comprising:

[0022] One or more processors;

[0023] A memory, coupled to the processor, for storing one or more programs;

[0024] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the mine water inrush early warning method based on distributed optical fiber monitoring in the first aspect described above.

[0025] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the mine water inrush early warning method based on distributed optical fiber monitoring in the first aspect described above.

[0026] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the mine water inrush early warning method based on distributed optical fiber monitoring described in the first aspect above.

[0027] Compared with the prior art, the present invention has the following advantages and technical effects:

[0028] This invention provides a mine water inrush early warning method based on distributed optical fiber monitoring. By integrating multi-source data such as distributed optical fiber strain monitoring and microseismic monitoring, this invention achieves continuous perception of the seepage state in the underground space of the mine, overcoming the shortcomings of traditional point-based monitoring methods that have monitoring blind spots. Based on the characteristics of rock mass micro-fracture propagation, rock stratum deformation, and water enrichment extracted by the random forest algorithm, combined with three-dimensional spatial grid modeling and time-series analysis of long short-term memory networks, it can accurately capture the dynamic evolution law of water-rock coupling during the formation of water inrush channels, solving the problem that discrete data points cannot reflect the continuous change process of the seepage field. By tracking the abnormal water-rock coupling area and generating the dynamic evolution trajectory of the water inrush channel, and calculating the spatial distance between the trajectory extension direction and the safety boundary of the mining face, an early warning command can be triggered at the stage of water inrush risk accumulation. This changes the passive situation of traditional methods that can only issue alarms after a sudden increase in water volume, providing reliable technical support for early warning and emergency response to mine water inrush disasters. Attached Figure Description

[0029] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0030] Figure 1 This is an overall flowchart of the mine water inrush early warning method based on distributed optical fiber monitoring according to an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the three-dimensional spatial grid and seepage field distribution of the mine geological body according to an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the clustering of water-rock coupling anomaly regions and the dynamic evolution trajectory of water inflow channels in an embodiment of the present invention. Detailed Implementation

[0033] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0034] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0035] Example 1

[0036] like Figure 1 As shown, this embodiment provides a mine water inrush early warning method based on distributed optical fiber monitoring, including:

[0037] Acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area;

[0038] The random forest algorithm was used to process the microseismic monitoring signal, the distributed optical fiber strain signal, and the groundwater chemical ion concentration data to extract the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment.

[0039] A three-dimensional spatial grid of the mine geological body is constructed by the characteristics of the microfracture propagation of the rock mass, the characteristics of the deformation of the rock strata, and the characteristics of water enrichment, and the initial seepage state corresponding to each three-dimensional spatial grid is obtained.

[0040] The initial seepage state is input into a long short-term memory network for temporal feature extraction to obtain a continuous seepage evolution sequence;

[0041] Based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, a three-dimensional spatial grid with water-rock coupling anomalies is determined, and the water-rock coupling anomaly region is identified.

[0042] Based on the spatial distribution coordinates of the water-rock coupling anomaly region, adjacent water-rock coupling anomaly regions are tracked to generate the dynamic evolution trajectory of the water inflow channel;

[0043] Obtain the extension direction vector of the dynamic evolution trajectory of the water inflow channel, and calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face;

[0044] A water inrush warning command is triggered based on the comparison result between the spatial distance and the preset distance threshold.

[0045] The specific process of the entire invention is as follows:

[0046] Step S101: Obtain microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data of the mining area.

[0047] Step S102: Based on the microseismic monitoring signal, the distributed optical fiber strain signal, and the groundwater chemical ion concentration data, the random forest algorithm is used to extract the characteristics of rock mass microfracture propagation, rock stratum deformation, and water enrichment.

[0048] Step S103: Construct a three-dimensional spatial grid of the mine geological body by using the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment to obtain the initial seepage state corresponding to each three-dimensional spatial grid.

[0049] Step S104: Input the initial seepage state into the long short-term memory network to extract temporal features and obtain a continuous seepage evolution sequence.

[0050] Step S105: Obtain data from adjacent time nodes in the continuous seepage evolution sequence and calculate the seepage gradient change value between adjacent time node data.

[0051] Step S106: If the seepage gradient change value is greater than the preset gradient threshold, it is determined that there is a water-rock coupling anomaly in the corresponding three-dimensional spatial grid and the water-rock coupling anomaly region is identified.

[0052] Step S107: Based on the spatial distribution coordinates of the water-rock coupling anomaly area, track the adjacent water-rock coupling anomaly areas to generate the dynamic evolution trajectory of the water inflow channel.

[0053] Step S108: Obtain the extension direction vector of the dynamic evolution trajectory of the water inflow channel and calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face.

[0054] Step S109: If the spatial distance is less than the preset distance threshold, it is determined that the area is in the risk accumulation stage and a water inrush warning command is triggered.

[0055] As one implementation method in this embodiment, the process of acquiring microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data includes:

[0056] The first multi-source dataset was obtained by extracting microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data through time windows.

[0057] Calculate the root mean square of the first multi-source data set to obtain the first root mean square sequence;

[0058] The isolated forest algorithm is used to identify mutation points in the first root mean square sequence and determine the first anomalous mutation sequence.

[0059] When the number of mutation points in the first abnormal mutation sequence is greater than a preset mutation threshold, the energy value of the first abnormal mutation sequence is calculated to obtain the first energy sequence.

[0060] The first energy sequence is classified using the support vector machine algorithm to obtain state values;

[0061] When the state value is greater than a preset state threshold, the sampling point corresponding to the state value is marked to determine the danger zone within the mining area.

[0062] Step S101: Obtain microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data of the mining area.

[0063] Microseismic waves, strain values, and water ion concentration values ​​of the mining area are acquired. These values ​​are then truncated using a time window to obtain a first multi-source dataset. Based on this dataset, the root mean square (RMS) is calculated to obtain a first RMS sequence. For this first RMS sequence, an isolated forest algorithm is used to identify mutation points, determining the first anomalous mutation sequence. If the number of mutation points in the first anomalous mutation sequence exceeds a preset mutation threshold (e.g., 5), the energy value of the first anomalous mutation sequence is calculated to obtain a first energy sequence. The first energy sequence is then classified using a support vector machine algorithm to determine its corresponding state value. Based on the state value, if it exceeds a preset state threshold, the sampling point corresponding to that state value is marked, identifying the hazardous area within the mining area.

[0064] Specifically, when mining areas are disturbed, rock strata fracturing generates microseismic waves and strain changes, while the opening of underground water channels leads to abnormal water ion concentrations. By deploying sensors to collect these three types of data in real time, a sliding time window technique is used to extract continuous data. With a time window length of 10 minutes and a step size of 5 minutes, the extracted microseismic wave amplitudes, strain values, and various groundwater chemical ion concentrations (such as calcium ions and sulfate ions) are combined to form the first multi-source data set. Calculating the root mean square (RMS) of the data within this set effectively smooths high-frequency noise and highlights the overall fluctuation energy of the data, outputting the first RMS sequence. This multi-source data fusion and RMS processing method effectively reduces the false alarm probability of a single sensor and improves the stability of data characteristics.

[0065] For example, an isolated forest algorithm is introduced for unsupervised anomaly detection for the first root mean square sequence. This algorithm continuously divides the data space by constructing multiple random trees, and mutation points, due to their sparsity, are often located at shallower levels of these random trees. The root mean square sequence is input into the isolated forest model to identify mutation points that deviate from the normal baseline, constituting the first anomalous mutation sequence. A threshold of 5 mutation points is set; when the number of mutation points in the sequence exceeds this threshold (e.g., reaching 8), it indicates significant damage to the rock mass structure. At this point, the original data corresponding to the mutation points is subjected to square integral calculation to extract the first energy sequence reflecting the rock mass fracture intensity. This process filters out occasional interference, ensuring that subsequent analysis focuses on real disaster precursors.

[0066] It should be noted that after obtaining the first energy sequence, it is input as a feature vector into a pre-trained support vector machine (SVM) classification model. The SVM maps energy features of different levels of damage to specific safety state values ​​by finding the optimal hyperplane. The model output state value is set to an integer range of 1 to 4. When the output state value is 3 (e.g., corresponding to the "high risk" level), which is greater than the preset state threshold of 2, it is determined that the current area is at high risk. Subsequently, the corresponding sampling points are marked according to the spatial coordinates of the sensors that triggered this state value, and the envelope range of the sampling points is delineated in the three-dimensional geological model, ultimately determining the danger zone within the mining area. This mechanism realizes a closed loop from data perception to disaster spatial location, improving the accuracy of safety early warning.

[0067] As one implementation method in this embodiment, the process of extracting the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment includes:

[0068] A dimension-reduced feature matrix is ​​generated based on the microseismic monitoring signal and the distributed optical fiber strain signal;

[0069] The dimensionality-reduced feature matrix is ​​trained using the random forest algorithm to obtain a rock mass state classification model;

[0070] The reduced-dimensional feature matrix is ​​input into the rock mass state classification model for prediction to obtain the initial rock mass state classification result.

[0071] The characteristics of microfracture propagation and rock strata deformation are extracted from the initial rock mass state classification results, and the water enrichment characteristics are extracted by combining the groundwater chemical ion concentration data.

[0072] Step S102: Based on the microseismic monitoring signal, the distributed optical fiber strain signal, and the groundwater chemical ion concentration data, the random forest algorithm is used to extract the characteristics of rock mass microfracture propagation, rock stratum deformation, and water enrichment.

[0073] Acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data; generate a dimensionality-reduced feature matrix based on the microseismic monitoring signals and the distributed optical fiber strain signals; train the dimensionality-reduced feature matrix using a random forest algorithm to obtain a rock mass state classification model; predict the dimensionality-reduced feature matrix based on the rock mass state classification model to obtain an initial rock mass state classification result; extract rock mass microfracture propagation characteristics, rock stratum deformation characteristics, and water enrichment characteristics based on the initial rock mass state classification result.

[0074] For example, a sensor network is deployed at the mining face to collect microseismic monitoring signals at a frequency of 1000 Hz and distributed fiber optic strain signals with a spatial resolution of 1 meter. Simultaneously, water chemistry sensors acquire data on the concentrations of calcium and sulfate ions in groundwater. The amplitude characteristics of the collected microseismic monitoring signals and the frequency shift characteristics of the distributed fiber optic strain signals are fused to construct an initial high-dimensional feature set. Principal component analysis (PCA) is then used to process this initial high-dimensional feature set, setting a cumulative contribution rate threshold of 90% to extract the top five principal components, thereby generating a dimensionality-reduced feature matrix.

[0075] In one possible implementation, the generated dimensionality-reduced feature matrix is ​​divided into a training set and a test set. Using the training set as input, a random forest algorithm is employed to construct a rock mass state classification model. During model training, the number of decision trees is set to 100, the maximum tree depth to 10 layers, and node splitting is performed using the Gini impurity index. After training, the dimensionality-reduced feature matrix of the region to be tested is input into the rock mass state classification model for prediction, outputting an initial rock mass state classification result encompassing three categories: stable, metastable, and unstable states.

[0076] Specifically, for the initial rock mass state classification results identified as unstable, multi-dimensional feature extraction is further performed by combining groundwater chemical ion concentration data. The frequency and magnitude spatial distribution of microseismic events are extracted from microseismic monitoring signals to form characteristics of rock mass microfracture propagation. The spatial distribution of strain gradients is extracted from distributed optical fiber strain signals to construct rock strata deformation characteristics. Simultaneously, the regional distribution of calcium ion concentrations exceeding 200 mg / L and abnormally elevated sulfate ion concentrations is extracted to generate water enrichment characteristics. Finally, a comprehensive feature set reflecting the physical, mechanical, and hydrochemical evolution processes within the rock mass of the mining area is output.

[0077] As one implementation method in this embodiment, the process of constructing a three-dimensional spatial grid of the mine geological body and obtaining the initial seepage state corresponding to each three-dimensional spatial grid includes:

[0078] The characteristics of microfracture propagation in the rock mass, the characteristics of rock layer deformation, and the characteristics of water enrichment are mapped onto the nodes of a three-dimensional spatial grid.

[0079] Determine the node connectivity of adjacent three-dimensional spatial grids. If the node connectivity is greater than a preset connectivity threshold, it is determined to be a water-conducting grid; otherwise, it is determined to be a water-blocking grid.

[0080] The permeability field is determined based on the distribution of the water-conducting grid and the water-blocking grid; the seepage equation is solved using the permeability field to obtain the initial seepage state of each three-dimensional spatial grid.

[0081] like Figure 2As shown, in step S103, a three-dimensional spatial grid of the mine geological body is constructed by using the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment to obtain the initial seepage state corresponding to each three-dimensional spatial grid.

[0082] The microfracture propagation characteristics, rock strata deformation characteristics, and water enrichment characteristics of the mine geological body are obtained, and a three-dimensional spatial grid is generated based on these characteristics. The node connectivity of the three-dimensional spatial grid is determined. If the node connectivity is greater than a preset connectivity threshold, the three-dimensional spatial grid is determined to be a water-conducting grid; otherwise, it is determined to be a water-blocking grid. The permeability field is determined based on the distribution of the water-conducting and water-blocking grids, and the initial seepage state corresponding to each three-dimensional spatial grid is obtained through the permeability field.

[0083] In one possible implementation, considering the spatial structure of the mine geological body, including the working face and the bottom aquifer, the target exploration area is divided into multiple cubic three-dimensional spatial grids with sides of 10 meters. Previously extracted characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment are obtained as input data. A Kriging interpolation algorithm is used to map these discrete feature data to the central node of each three-dimensional spatial grid, thereby assigning specific geological attributes to each grid.

[0084] Specifically, after assigning grid attributes, it is necessary to quantify the fluid exchange capacity between grids. The microfracture density and deformation tensor of adjacent grid nodes are extracted, and the depth-first search algorithm in graph theory is used to calculate the fracture connectivity probability between adjacent nodes, which is defined as node connectivity. This connectivity is expressed as a percentage; a higher value indicates better fracture connectivity. A preset connectivity threshold of 60% is set. When the calculated node connectivity between a 3D spatial grid and its surrounding grids is 75, since 75 is greater than the threshold of 60, the 3D spatial grid is determined to be a water-conducting grid. If the calculated node connectivity is 30, it is determined to be a water-blocking grid. By traversing all grid nodes, the spatial labeling of the water-conducting and water-blocking properties of the entire mine geological body is completed.

[0085] For example, a three-dimensional permeability matrix for the entire geological body is constructed based on the distribution of the marked water-conducting and water-blocking grids. High permeability baseline values ​​are assigned to the water-conducting grids, and low permeability baseline values ​​are assigned to the water-blocking grids, forming a heterogeneous permeability field. Using this permeability field and the set boundary water pressure conditions as input parameters, the Darcy flow equation is solved using a finite difference algorithm. The boundary water pressure is set to 5 MPa. After iterative calculation, the velocity vector and hydraulic head height within each three-dimensional spatial grid are output. Through the above calculation process, the initial seepage state corresponding to each three-dimensional spatial grid is obtained. For example, the initial velocity at a certain water-conducting grid node is quantified as 2 m / day, thus realizing a numerical expression of the groundwater seepage path in the mine.

[0086] As one implementation method in this embodiment, the process of obtaining a continuous seepage evolution sequence and determining the water-rock coupling anomaly region based on the seepage gradient change value includes:

[0087] The initial seepage state is mapped to an initial seepage matrix;

[0088] The initial percolation matrix is ​​input into a long short-term memory network for processing to obtain the updated memory cell state.

[0089] The hidden layer state is determined based on the updated memory unit state, and the hidden layer state is mapped to obtain the evolutionary feature vector;

[0090] The evolutionary feature vectors are concatenated to obtain a continuous seepage evolution sequence;

[0091] Obtain adjacent time node data in the continuous seepage evolution sequence, and calculate the seepage gradient change value between adjacent time node data;

[0092] When the change value of the seepage gradient is greater than the preset gradient threshold, it is determined that there is a water-rock coupling anomaly in the corresponding three-dimensional spatial grid, and the water-rock coupling anomaly region is identified.

[0093] Step S104: Input the initial seepage state into the long short-term memory network to extract temporal features and obtain a continuous seepage evolution sequence.

[0094] Initial seepage state data is acquired, and the initial seepage state data is mapped to obtain an initial seepage matrix. The initial seepage matrix is ​​input into a long short-term memory network (LSTM) to obtain the updated memory unit state of the network. The hidden layer state is determined based on the updated memory unit state, and the hidden layer state is mapped to obtain an evolutionary feature vector. The evolutionary feature vector is concatenated to obtain a continuous seepage evolution sequence obtained by extracting temporal features from the initial seepage state input into the LSTM network.

[0095] In one possible implementation, the acquired initial seepage state data is mapped to obtain an initial seepage matrix. The input data consists of the initial water pressure and initial flow velocity values ​​for each node in the three-dimensional spatial grid of the mine geological body. A linear normalization algorithm is used to process the discrete seepage state data, proportionally mapping the physical range of water pressure values ​​from 0 to 50 MPa to the numerical range of 0 to 1. After mapping, these values ​​are reorganized into an initial seepage matrix of dimension 64×64×64 according to the coordinate topology of the three-dimensional spatial grid. This matrix intuitively reflects the spatial distribution structure of the seepage field of the mine geological body at the initial moment.

[0096] For example, the constructed initial seepage matrix is ​​input into a Long Short-Term Memory (LSTM) network for temporal feature extraction. For example, a time step of 6 hours is set (adjustable according to actual monitoring needs), and the initial seepage matrices from multiple consecutively collected time points are sequentially input into the network. The forgetting gate algorithm within the LTM network calculates the weights for retaining or discarding historical seepage information based on the hidden layer state of the previous time step and the initial seepage matrix of the current time step. Subsequently, the input gate algorithm determines which new information in the current seepage matrix needs to be recorded. The two algorithms are combined to calculate and output the updated memory unit state of the current network. This state contains the dynamic evolution of the seepage field of the mine geological body over time.

[0097] It should be noted that, based on the updated memory unit states, the hidden layer states are further determined and mapped to obtain evolutionary feature vectors. Using the output gate algorithm of the Long Short-Term Memory network, combined with the updated memory unit states, key feature information that plays a crucial role in subsequent seepage prediction is selected as the hidden layer states. Subsequently, a fully connected layer algorithm is used to perform nonlinear mapping and dimensionality reduction on these hidden layer states, outputting a one-dimensional evolutionary feature vector of length 256. The evolutionary feature vectors corresponding to 20 consecutive time steps are concatenated in chronological order to obtain a continuous seepage evolution sequence of dimension 5120 (i.e., 256 × 20). This sequence fully quantifies the entire seepage evolution process of the mine geological body over 120 hours.

[0098] Step S105: Obtain data from adjacent time nodes in the continuous seepage evolution sequence and calculate the seepage gradient change value between adjacent time node data.

[0099] Multiple time nodes are obtained from a continuous seepage evolution sequence. Adjacent time nodes are extracted to obtain first node data and second node data. Spatial coordinates and water pressure values ​​are extracted based on the first node data and the second node data. Spatial distance difference is calculated using the spatial coordinates, and water pressure difference is obtained using the water pressure values. A first seepage gradient and a second seepage gradient are obtained by calculating the ratio between the spatial distance difference and the water pressure difference. The seepage gradient change value between adjacent time node data is determined by differential processing of the first seepage gradient and the second seepage gradient.

[0100] For example, a time window sliding algorithm is used to extract the first and second node data, with days 10 and 11 being adjacent, from the sequence. The three-dimensional spatial coordinates of the underground rock mass monitoring points and the corresponding water pressure values ​​are then extracted from these node data using data parsing rules. The input to this process is continuous time series data, and the output is a structured dataset containing spatial location and hydrodynamic state at a specific timestamp.

[0101] In one possible implementation, based on the extracted spatial coordinates, the Euclidean distance algorithm is used to calculate the spatial distance difference between two different monitoring points. Simultaneously, the water pressure difference is obtained by subtracting the water pressure values ​​at the two monitoring points at the same time node. Assuming the spatial distance difference between the two points is 5 meters and the water pressure difference is 20 kPa, the seepage gradient of the area can be obtained by dividing the water pressure difference by the spatial distance difference. This joint spatial and water pressure calculation is performed on the data from the first and second nodes respectively, thereby outputting the first seepage gradient corresponding to the first time node and the second seepage gradient corresponding to the second time node.

[0102] It should be noted that after obtaining the seepage gradients at adjacent time points, the first and second seepage gradients are differentially processed using a first-order forward differencing algorithm. If the calculated first seepage gradient value is 4 kPa / m and the second seepage gradient value is 6 kPa / m, then the seepage gradient change value output after the differencing operation is 2 kPa / m. This quantitative output of the change value reflects the drastic dynamic evolution of the seepage field within adjacent time periods, revealing the transient variation patterns of groundwater velocity and direction, and providing a basic quantitative indicator for subsequent rock mass seepage evolution analysis.

[0103] like Figure 3 As shown, in step S106, if the change value of the seepage gradient is greater than the preset gradient threshold, it is determined that there is a water-rock coupling anomaly in the corresponding three-dimensional spatial grid, and the water-rock coupling anomaly region is identified.

[0104] The process involves obtaining pore water pressure values, calculated from fluid pressure and rock stress in a three-dimensional spatial grid, and then determining an initial seepage gradient value based on these values. Next, the process involves obtaining seepage gradient variation values, obtained by subtracting the initial seepage gradient value from the real-time seepage gradient value. If the seepage gradient variation value exceeds a preset threshold, a water-rock coupling anomaly grid is obtained. Finally, the process involves clustering the coordinate data of the water-rock coupling anomaly grid to determine the water-rock coupling anomaly region.

[0105] In one possible implementation, fluid pressure and rock stress data at each grid node are input into the constructed three-dimensional spatial grid model. Taking deep rock engineering as an example, if the fluid pressure monitored in a certain grid cell is 15 MPa, this fluid pressure is the pore water pressure at that location. Based on the obtained pore water pressure value, combined with the spatial distance between adjacent grid cells (e.g., 5 meters), the initial seepage gradient value (e.g., 0.5 kPa / m) is calculated by dividing the water pressure difference by the distance. This process establishes the baseline hydraulic state of the rock mass before it is subjected to severe external disturbance, providing a reliable comparison benchmark for subsequent dynamic monitoring.

[0106] For example, as underground engineering excavation or grouting activities progress, the seepage field within the rock mass dynamically evolves. The real-time seepage gradient is continuously collected, and the difference between this real-time gradient and the initial seepage gradient value is calculated to obtain the seepage gradient change value. Assuming the initial seepage gradient value for a specific grid is 0.5 kPa / m, and at a certain monitoring moment its real-time seepage gradient rises to 1.2 kPa / m, the seepage gradient change value output through the difference operation is 0.7 kPa / m. This change value can quantify the fluctuation of the fluid flow state in the rock mass fracture network, revealing the abrupt changes in the local hydraulic gradient.

[0107] It should be noted that, in order to accurately identify potential seepage instability risks within the rock mass, a preset threshold is introduced for conditional judgment. The preset threshold for this engineering scenario is set to 0.5 kPa / m, and the calculated seepage gradient change value of 0.7 is compared with this preset threshold. Since 0.7 is greater than the preset threshold of 0.5, it indicates that the fluid driving force in this area has significantly exceeded the rock mass's own seepage resistance, which may have triggered the expansion or connection of microfractures. Therefore, this three-dimensional spatial mesh is extracted and marked as a water-rock coupling anomaly mesh.

[0108] In one possible implementation, after identifying multiple discrete water-rock coupling anomaly grids in a wide spatial area, the three-dimensional spatial coordinate data of these anomaly grids are extracted as the input set. A density clustering algorithm is then used to perform spatial aggregation analysis on this coordinate data, grouping anomaly grids that are spatially close and whose distribution density meets a set standard into the same cluster. The cluster set output by the algorithm represents the contiguous spatial extent, thereby determining the water-rock coupling anomaly region. Through this spatial clustering process, microscopic grid anomalies can be transformed into macroscopic regional risk distributions, intuitively revealing concentrated areas of strong water-rock coupling.

[0109] As one implementation method in this embodiment, the process of tracking the dynamic evolution trajectory of the water inrush channel generated by adjacent water-rock coupling anomaly areas and triggering a water inrush early warning command includes:

[0110] Obtain the first spatial distribution coordinates of the first water-rock coupling anomaly region;

[0111] Calculate the first spatial distance based on the first spatial distribution coordinates, and determine the first adjacent abnormal node when the first spatial distance is less than the preset spatial distance threshold.

[0112] Based on the first adjacent abnormal node, path optimization is performed to obtain the first optimal tracking path;

[0113] The first optimal tracking path is mapped to obtain the first water inflow channel node, and the first water inflow channel node is fitted to generate the dynamic evolution trajectory of the first water inflow channel.

[0114] The extension direction vector is determined based on the dynamic evolution trajectory of the first water inflow channel, and the spatial distance between the extension direction vector and the safety boundary of the mining face is calculated; when the spatial distance is less than a preset distance threshold, a water inrush warning command is triggered.

[0115] Step S107: Based on the spatial distribution coordinates of the water-rock coupling anomaly area, track the adjacent water-rock coupling anomaly areas to generate the dynamic evolution trajectory of the water inflow channel.

[0116] Obtain the first spatial distribution coordinates of the first water-rock coupling anomaly region; if the first spatial distance calculated based on the first spatial distribution coordinates is less than a threshold, then determine that the first water-rock coupling anomaly regions are first adjacent anomaly nodes; optimize the first optimal tracking path based on the first adjacent anomaly nodes; map the first optimal tracking path to obtain the first water inflow channel node, and fit the first water inflow channel node to generate the dynamic evolution trajectory of the first water inflow channel, thereby realizing the dynamic evolution trajectory of the water inflow channel generated by tracking adjacent water-rock coupling anomaly regions based on the spatial distribution coordinates of the water-rock coupling anomaly region.

[0117] One possible implementation for obtaining the first spatial distribution coordinates of the first water-rock coupling anomaly region involves extracting the geometric center positions of each independent anomaly region as representative coordinate data. A set of three-dimensional spatial meshes of multiple anomaly regions, determined in the previous step, is input, and a centroid calculation algorithm is used to output the corresponding three-dimensional coordinate integer values ​​for each region. Subsequently, the first spatial distance between these coordinates is calculated.

[0118] For example, the Euclidean distance metric is used to calculate the straight-line distance between the center coordinates of any two anomalous regions. A spatial distance threshold of 50 meters is set. If the calculated distance between two anomalous regions is 30 meters, since 30 is less than 50, these two first water-rock coupling anomalous regions are determined to be highly spatially correlated and are considered first adjacent anomalous nodes. By traversing the coordinates of all anomalous regions, an underground spatial topology network consisting of multiple first adjacent anomalous nodes can be constructed.

[0119] It should be noted that the process of obtaining the first optimal tracking path based on the first adjacent anomaly node relies on a path optimization algorithm in graph theory. The constructed topological network is used as input data, where nodes represent anomaly regions and lines represent adjacency relationships. An ant colony optimization algorithm is employed to perform a global search on this network, using the spatial distance between nodes and the characteristics of rock mass micro-fracture propagation (such as microseismic event density) as heuristic information for iterative optimization. After multiple iterations, the algorithm outputs the shortest line connecting multiple key adjacent anomaly nodes, which is also the most likely to result in fluid flow, and this is the first optimal tracking path. This path objectively reflects the preferred flow direction of groundwater within rock mass fractures.

[0120] The first optimal tracing path is mapped to obtain the first water inflow channel node, and a dynamic evolution trajectory of the first water inflow channel is generated by fitting the path. The discrete path segments obtained through optimization are mapped back to the actual 3D geological model, and key turning points on the path are extracted as the first water inflow channel nodes. The coordinates of these discrete channel nodes are input and smoothed using a cubic spline curve fitting algorithm, outputting a continuous 3D spatial curve. This curve, representing the dynamic evolution trajectory of the first water inflow channel, visually demonstrates the spatial evolution process of underground water flowing from its source through various water-rock coupling anomaly areas to finally form a complete water-conducting channel.

[0121] Step S108: Obtain the extension direction vector of the dynamic evolution trajectory of the water inflow channel and calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face.

[0122] An initial evolution trajectory is obtained by fitting a sequence of microseismic event coordinates; an extension direction vector is determined based on the initial evolution trajectory by calculating coordinate differences; a boundary intersection model containing the extension direction vector is obtained by fusing the extension direction vector with a safety boundary model; the length of the line segment is calculated based on the intersection nodes in the boundary intersection model to obtain the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face.

[0123] In one possible implementation, the process of obtaining the initial evolution trajectory mainly relies on spatial geometric fitting of the microseismic event coordinate sequence acquired by downhole microseismic monitoring equipment.

[0124] Specifically, the input data is a set of three-dimensional spatial coordinates of microseismic events occurring continuously over a period of time, including multiple spatial points such as an abscissa of 1200.5, a ordinate of 850.0, and a depth coordinate (with the surface as the origin and downwards as negative) of -600.5. The coordinate sequence is processed using a least-squares spatial linear fitting algorithm to output a spatial linear equation that characterizes the fracturing trend of the rock strata, which serves as the initial evolution trajectory. This trajectory reflects the macroscopic development pattern of microfractures in the underground rock mass in three-dimensional space.

[0125] For example, when determining the extension direction vector, key nodes need to be extracted based on the acquired initial evolution trajectory. The fitting point of the microseismic event with the latest time attributes on the initial evolution trajectory is selected as the starting point, and the fitting point with an earlier time attribute is selected as the reference point. By calculating the coordinate differences between these two spatial nodes along the three coordinate axes, a three-dimensional spatial vector can be constructed. This extension direction vector clearly indicates the physical direction in which the rock fracture zone or potential water-conducting channel approaches the mining face.

[0126] It should be noted that obtaining the boundary intersection model containing the extension direction vectors is a crucial step in evaluating spatial relationships. The input is a pre-constructed safety boundary model of the mining face using 3D geological modeling software; this model is typically composed of spatial triangular mesh surfaces. The previously calculated extension direction vectors are converted into spatial rays, and a ray-polygon intersection detection algorithm is used to spatially fuse these rays with the safety boundary model. The output is the coordinates of the intersection nodes generated when the rays penetrate the surface of the safety boundary model, thus generating a complete boundary intersection model.

[0127] In one possible implementation, the length of a line segment is calculated based on the intersection nodes in the boundary intersection model, aiming to quantify the true spatial distance between the potential hazard source and the working face. Starting from the current endpoint of the initial evolution trajectory and ending at the calculated intersection node, the length of the straight line segment between the two points is calculated using a three-dimensional Euclidean distance algorithm. For example, if the calculated length of this line segment is 45.5 meters, this value represents the spatial distance between the extending direction vector and the pre-established safety boundary of the mining face, thus completing the quantitative calculation of the approximation degree of the potential water-conducting channel.

[0128] Step S109: If the spatial distance is less than the preset distance threshold, it is determined that the area is in the risk accumulation stage and a water inrush warning command is triggered.

[0129] The initial coordinate data of the mining face and the bottom of the aquifer are obtained, and a three-dimensional spatial model is generated based on the initial coordinate data. Evolutionary features are extracted based on the three-dimensional spatial model, and real-time spatial distance is calculated based on the evolutionary features. A preset distance threshold is obtained, and the real-time spatial distance is compared with the preset distance threshold. If the real-time spatial distance is less than the preset distance threshold, it is determined that the risk accumulation stage is in place and a water inrush warning command is triggered.

[0130] In one possible implementation, obtaining the initial coordinate data of the mining face and the bottom plate of the aquifer and generating a three-dimensional spatial model is the basis for the entire water hazard risk assessment.

[0131] Specifically, the elevation data of the top and bottom plates of the working face and the spatial location information of the underlying limestone aquifer are collected using geological exploration equipment as input data. For example, the input data includes the boundary coordinates of the mining working face with a strike length of 1500 meters and a dip length of 200 meters, as well as the coordinates of discrete points on the bottom plate of the aquifer at an elevation of -500 meters. The discrete coordinates are spatially meshed using the Kriging interpolation algorithm, and then the Delaunay triangulation algorithm is used to construct a three-dimensional surface of the geological interface, outputting a three-dimensional spatial model that includes the relative spatial relationship between the working face and the aquifer.

[0132] Evolutionary features are extracted and real-time spatial distances are calculated based on the generated 3D spatial model, as an example. The extraction of evolutionary features primarily focuses on dynamically changing regions of stress concentration or microfracture development in the rock strata within the model. Using rock strata deformation data from the 3D spatial model as input, a spatial clustering algorithm is used to identify the region with the largest deformation gradient as the evolutionary feature region. Subsequently, the geometric center coordinates of this evolutionary feature region are extracted, and a spatial straight-line distance algorithm is used to calculate the vertical spatial distance from this geometric center to the lowest point of the mining face floor, outputting the real-time spatial distance. For example, the calculated current real-time spatial distance is 45 meters.

[0133] It should be noted that the logic for triggering the early warning is to obtain a preset distance threshold and compare the real-time spatial distance with the preset distance threshold. The preset distance threshold is set comprehensively based on historical water inrush cases in the mine and rock mechanics parameters. The calculated real-time spatial distance of 45 meters is compared with the preset distance threshold of 60 meters. Since 45 meters is less than the preset distance threshold of 60 meters, it is determined that the current thickness of the bottom aquifer is insufficient to resist the high water pressure of the aquifer below, and the mine is in a stage of water hazard risk accumulation. At this time, a water inrush early warning command is directly triggered, and an early warning signal containing the risk location and the current distance difference is output, completing the risk status determination process.

[0134] Based on this, this invention achieves continuous sensing of the seepage state in underground mine spaces by integrating multi-source data such as distributed fiber optic strain monitoring and microseismic monitoring, overcoming the shortcomings of traditional point-based monitoring methods that have monitoring blind spots. Based on the rock mass micro-fracture propagation characteristics, rock stratum deformation characteristics, and water enrichment characteristics extracted by the random forest algorithm, combined with three-dimensional spatial grid modeling and time-series analysis using long short-term memory networks, it can accurately capture the dynamic evolution law of water-rock coupling during the formation of water inrush channels, solving the problem that discrete data points cannot reflect the continuous change process of the seepage field. By tracking abnormal water-rock coupling areas and generating the dynamic evolution trajectory of water inrush channels, and calculating the spatial distance between the trajectory extension direction and the safety boundary of the mining face, early warning commands can be triggered at the stage of water inrush risk accumulation. This changes the passive situation of traditional methods that can only issue alarms after a sudden increase in water volume, providing reliable technical support for early warning and emergency response to mine water inrush disasters.

[0135] Example 2

[0136] In this embodiment, a computer terminal device is provided, including:

[0137] One or more processors;

[0138] A memory, coupled to the processor, for storing one or more programs;

[0139] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above-described mine water inrush early warning method based on distributed optical fiber monitoring.

[0140] In this embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-described mine water inrush early warning method based on distributed optical fiber monitoring.

[0141] In this embodiment, an electronic device is also provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps of the above-described mine water inrush early warning method based on distributed optical fiber monitoring.

[0142] In this embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the above-described mine water inrush early warning method based on distributed optical fiber monitoring.

[0143] The aforementioned program can run on a processor or be stored in memory (or a computer-readable medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0144] These computer programs may also be loaded onto a computer or other programmable data processing equipment, causing a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented by different modules for different steps.

[0145] This embodiment provides such a device or system. The system, referred to as a mine water inrush early warning system based on distributed fiber optic monitoring, includes:

[0146] The data acquisition module is used to acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area.

[0147] The feature extraction module is used to process the microseismic monitoring signal, the distributed optical fiber strain signal and the groundwater chemical ion concentration data using the random forest algorithm to extract the characteristics of rock mass microfracture propagation, rock strata deformation and water enrichment.

[0148] The spatial grid modeling module is used to construct a three-dimensional spatial grid of the mine geological body through the characteristics of the rock mass microfracture propagation, the characteristics of the rock strata deformation, and the characteristics of water enrichment, and to obtain the initial seepage state corresponding to each three-dimensional spatial grid.

[0149] The temporal evolution analysis module is used to input the initial seepage state into a long short-term memory network for temporal feature extraction, obtain a continuous seepage evolution sequence, and determine the three-dimensional spatial grid with water-rock coupling anomalies based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, thereby identifying the water-rock coupling anomaly region.

[0150] The water inflow channel tracking module is used to track adjacent water-rock coupling anomaly areas based on the spatial distribution coordinates of the water-rock coupling anomaly area, and generate a dynamic evolution trajectory of the water inflow channel.

[0151] The early warning decision module is used to obtain the extension direction vector of the dynamic evolution trajectory of the water inrush channel, calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face, and trigger a water inrush early warning command based on the comparison result of the spatial distance and the preset distance threshold.

[0152] As one implementation method in this embodiment, the data acquisition module includes:

[0153] The data extraction unit is used to extract microseismic monitoring signals, distributed optical fiber strain signals and groundwater chemical ion concentration data through a time window to obtain the first multi-source data set.

[0154] The root mean square calculation unit is used to calculate the root mean square of the first multi-source data set to obtain the first root mean square sequence.

[0155] The mutation identification unit is used to identify mutation points in the first root mean square sequence using the isolated forest algorithm, and to determine the first abnormal mutation sequence.

[0156] An energy calculation unit is used to calculate the energy value of the first abnormal mutation sequence when the number of mutation points in the first abnormal mutation sequence is greater than a preset mutation threshold, and to obtain a first energy sequence.

[0157] The danger zone marking unit is used to classify the first energy sequence using a support vector machine algorithm to obtain a state value, and to mark the sampling point corresponding to the state value when the state value is greater than a preset state threshold, thereby determining the danger zone within the mining area.

[0158] As one implementation method in this embodiment, the feature extraction module includes:

[0159] A dimension reduction matrix generation unit is used to generate a dimension reduction feature matrix based on the microseismic monitoring signal and the distributed optical fiber strain signal.

[0160] The classification model training unit is used to train the dimensionality-reduced feature matrix using the random forest algorithm to obtain a rock mass state classification model.

[0161] The state prediction unit is used to input the dimensionality reduction feature matrix into the rock mass state classification model for prediction and to obtain the initial rock mass state classification result.

[0162] The multi-source feature extraction unit is used to extract the microfracture propagation features and rock strata deformation features from the initial rock mass state classification results, and to extract water enrichment features by combining the groundwater chemical ion concentration data.

[0163] As one implementation method in this embodiment, the spatial grid modeling module includes:

[0164] The feature mapping unit is used to map the rock mass microfracture propagation features, the rock strata deformation features, and the water enrichment features to nodes of a three-dimensional spatial grid.

[0165] The water-conducting property determination unit is used to determine the node connectivity of adjacent three-dimensional spatial grids. When the node connectivity is greater than a preset connectivity threshold, it is determined to be a water-conducting grid; otherwise, it is determined to be a water-blocking grid.

[0166] A permeability field construction unit is used to determine the permeability field based on the distribution of the water-conducting grid and the water-blocking grid;

[0167] The seepage state solution unit is used to solve the seepage equation through the permeability field to obtain the initial seepage state of each three-dimensional spatial grid.

[0168] As one implementation method in this embodiment, the time-series evolution analysis module includes:

[0169] A matrix mapping unit is used to map the initial seepage state into an initial seepage matrix;

[0170] The network memory update unit is used to input the initial percolation matrix into the long short-term memory network for processing to obtain the updated memory unit state.

[0171] The feature vector generation unit is used to determine the hidden layer state based on the state of the updated memory unit, and to map the hidden layer state to obtain the evolutionary feature vector.

[0172] An evolution sequence splicing unit is used to splice the evolution feature vectors to obtain a continuous seepage evolution sequence;

[0173] The gradient change calculation unit is used to acquire adjacent time node data in the continuous seepage evolution sequence and calculate the seepage gradient change value between adjacent time node data.

[0174] An abnormal region determination unit is used to determine that there is a water-rock coupling anomaly in the corresponding three-dimensional spatial grid when the seepage gradient change value is greater than a preset gradient threshold, and to identify the water-rock coupling anomaly region.

[0175] As one implementation method in this embodiment, the water inrush channel tracking module and the early warning decision module include:

[0176] The coordinate extraction unit is used to obtain the first spatial distribution coordinates of the first water-rock coupling anomaly region.

[0177] The adjacent node determination unit is used to calculate the first spatial distance based on the first spatial distribution coordinates, and determine the first adjacent abnormal node when the first spatial distance is less than a preset spatial distance threshold.

[0178] The optimal path search unit is used to perform path optimization based on the first adjacent abnormal node to obtain the first optimal tracking path.

[0179] The trajectory fitting and generation unit is used to map the first optimal tracking path to obtain the first water inrush channel node, and to fit the first water inrush channel node to generate the dynamic evolution trajectory of the first water inrush channel.

[0180] The approximation distance calculation unit is used to determine the extension direction vector based on the dynamic evolution trajectory of the first water inflow channel, and to calculate the spatial distance between the extension direction vector and the safety boundary of the mining face.

[0181] The early warning command triggering unit is used to trigger a water inrush early warning command when the spatial distance is less than a preset distance threshold.

[0182] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0183] The above implementation method solves the problem of mine water inrush early warning based on distributed optical fiber monitoring in related technologies, thereby ensuring that the problems existing in the prior art are resolved.

[0184] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of mine water inrush based on distributed optical fiber monitoring, characterized in that, Includes the following steps: Acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area; The random forest algorithm was used to process the microseismic monitoring signal, the distributed optical fiber strain signal, and the groundwater chemical ion concentration data to extract the characteristics of rock mass microfracture propagation, rock strata deformation, and water enrichment. A three-dimensional spatial grid of the mine geological body is constructed by the characteristics of the microfracture propagation of the rock mass, the characteristics of the deformation of the rock strata, and the characteristics of water enrichment, and the initial seepage state corresponding to each three-dimensional spatial grid is obtained. The initial seepage state is input into a long short-term memory network for temporal feature extraction to obtain a continuous seepage evolution sequence; Based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, a three-dimensional spatial grid with water-rock coupling anomalies is determined, and the water-rock coupling anomaly region is identified. Based on the spatial distribution coordinates of the water-rock coupling anomaly region, adjacent water-rock coupling anomaly regions are tracked to generate the dynamic evolution trajectory of the water inflow channel; Obtain the extension direction vector of the dynamic evolution trajectory of the water inflow channel, and calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face; A water inrush warning command is triggered based on the comparison result between the spatial distance and the preset distance threshold.

2. The method according to claim 1, characterized in that, The process of acquiring microseismic monitoring signals, distributed fiber optic strain signals, and groundwater chemical ion concentration data includes: The first multi-source dataset was obtained by extracting microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data through time windows. Calculate the root mean square of the first multi-source data set to obtain the first root mean square sequence; The isolated forest algorithm is used to identify mutation points in the first root mean square sequence and determine the first anomalous mutation sequence. When the number of mutation points in the first abnormal mutation sequence is greater than a preset mutation threshold, the energy value of the first abnormal mutation sequence is calculated to obtain the first energy sequence. The first energy sequence is classified using the support vector machine algorithm to obtain state values; When the state value is greater than a preset state threshold, the sampling point corresponding to the state value is marked to determine the danger zone within the mining area.

3. The method according to claim 1, characterized in that, The process of extracting the characteristics of microfracture propagation, rock strata deformation, and water enrichment in rock masses includes: A dimension-reduced feature matrix is ​​generated based on the microseismic monitoring signal and the distributed optical fiber strain signal; The dimensionality-reduced feature matrix is ​​trained using the random forest algorithm to obtain a rock mass state classification model; The reduced-dimensional feature matrix is ​​input into the rock mass state classification model for prediction to obtain the initial rock mass state classification result. The characteristics of microfracture propagation and rock strata deformation are extracted from the initial rock mass state classification results, and the water enrichment characteristics are extracted by combining the groundwater chemical ion concentration data.

4. The method according to claim 1, characterized in that, The process of constructing a three-dimensional spatial grid of the mine geological body and obtaining the initial seepage state corresponding to each three-dimensional spatial grid includes: The characteristics of microfracture propagation in the rock mass, the characteristics of rock layer deformation, and the characteristics of water enrichment are mapped onto the nodes of a three-dimensional spatial grid. Determine the node connectivity of adjacent three-dimensional spatial grids. If the node connectivity is greater than a preset connectivity threshold, it is determined to be a water-conducting grid; otherwise, it is determined to be a water-blocking grid. The permeability field is determined based on the distribution of the water-conducting grid and the water-blocking grid; the seepage equation is solved using the permeability field to obtain the initial seepage state of each three-dimensional spatial grid.

5. The method according to claim 1, characterized in that, The process of obtaining a continuous seepage evolution sequence and determining the water-rock coupling anomaly region based on seepage gradient changes includes: The initial seepage state is mapped to an initial seepage matrix; The initial percolation matrix is ​​input into a long short-term memory network for processing to obtain the updated memory cell state. The hidden layer state is determined based on the updated memory unit state, and the hidden layer state is mapped to obtain the evolutionary feature vector; The evolutionary feature vectors are concatenated to obtain a continuous seepage evolution sequence; Obtain adjacent time node data in the continuous seepage evolution sequence, and calculate the seepage gradient change value between adjacent time node data; When the change value of the seepage gradient is greater than the preset gradient threshold, it is determined that there is a water-rock coupling anomaly in the corresponding three-dimensional spatial grid, and the water-rock coupling anomaly region is identified.

6. The method according to claim 1, characterized in that, The process of tracking the dynamic evolution trajectory of water inrush channels generated by adjacent water-rock coupling anomaly areas and triggering water inrush early warning commands includes: Obtain the first spatial distribution coordinates of the first water-rock coupling anomaly region; Calculate the first spatial distance based on the first spatial distribution coordinates, and determine the first adjacent abnormal node when the first spatial distance is less than the preset spatial distance threshold. Based on the first adjacent abnormal node, path optimization is performed to obtain the first optimal tracking path; The first optimal tracking path is mapped to obtain the first water inflow channel node, and the first water inflow channel node is fitted to generate the dynamic evolution trajectory of the first water inflow channel. The extension direction vector is determined based on the dynamic evolution trajectory of the first water inrush channel, and the spatial distance between the extension direction vector and the safety boundary of the mining face is calculated; when the spatial distance is less than a preset distance threshold, a water inrush warning command is triggered.

7. A mine water inrush early warning system based on distributed optical fiber monitoring, characterized in that, The system for implementing the method of any one of claims 1-6 comprises: The data acquisition module is used to acquire microseismic monitoring signals, distributed optical fiber strain signals, and groundwater chemical ion concentration data in the mining area. The feature extraction module is used to process the microseismic monitoring signal, the distributed optical fiber strain signal and the groundwater chemical ion concentration data using the random forest algorithm to extract the characteristics of rock mass microfracture propagation, rock strata deformation and water enrichment. The spatial grid modeling module is used to construct a three-dimensional spatial grid of the mine geological body through the characteristics of the rock mass microfracture propagation, the characteristics of the rock strata deformation, and the characteristics of water enrichment, and to obtain the initial seepage state corresponding to each three-dimensional spatial grid. The temporal evolution analysis module is used to input the initial seepage state into a long short-term memory network for temporal feature extraction, obtain a continuous seepage evolution sequence, and determine the three-dimensional spatial grid with water-rock coupling anomalies based on the seepage gradient change values ​​between adjacent time nodes in the continuous seepage evolution sequence, thereby identifying the water-rock coupling anomaly region. The water inflow channel tracking module is used to track adjacent water-rock coupling anomaly areas based on the spatial distribution coordinates of the water-rock coupling anomaly area, and generate a dynamic evolution trajectory of the water inflow channel. The early warning decision module is used to obtain the extension direction vector of the dynamic evolution trajectory of the water inrush channel, calculate the spatial distance between the extension direction vector and the pre-established safety boundary of the mining face, and trigger a water inrush early warning command based on the comparison result of the spatial distance and the preset distance threshold.

8. A computer terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images