Mine microearthquake positioning control system and method

Abstract

The invention relates to the technical field of information processing, in particular to a mine microseism positioning control system and method.The mine microseism positioning control method comprises the steps that microseism signals are synchronously collected through a geophone, an accelerometer, an acoustic emission sensor and an optical fiber sensor, and data timestamps are aligned based on a GPS / Beidou clock; blind source separation is carried out on the signals after wavelet filtering, independent seismic source components are extracted, multi-sensor features are fused through a graph neural network, and a space-time joint feature matrix is constructed; classifying micro-seismic event types by using a ResNet-LSTM hybrid model, generating a seismic source coordinate initial solution based on deep reinforcement learning, and iteratively converging to an optimal solution through a particle swarm optimization algorithm; and mapping a positioning result to a three-dimensional geological grid model, simulating an energy diffusion path, triggering a three-level alarm according to the event energy density, and automatically pushing an emergency instruction to a mine safety management system. Therefore, the problem of low precision of the existing mine microseism positioning control system is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of information processing, and particularly to a mine microseismic positioning control system and method. Background Art

[0002] The mine microseismic positioning technology is of great significance for preventing geological disasters and monitoring illegal mining. Most of the existing technologies rely on the Geiger classical algorithm. Its linear positioning method is easily restricted by the distribution of stations, and the velocity field model often adopts a homogenization hypothesis, making it difficult to adapt to complex geological structures.

[0003] Existing mine microseismic positioning systems have problems such as the positioning accuracy being limited by the velocity field hypothesis, high hardware deployment costs, and insufficient real-time performance. Especially under complex geological conditions, the non-uniformity of the signal propagation path makes it difficult to calibrate the velocity model, and existing systems are difficult to quickly adapt to the dynamically changing mine environment.

[0004] In addition, traditional methods are sensitive to noise, have low accuracy, lack an intelligent early warning function, and cannot meet the real-time and flexible requirements of modern mine safety monitoring. Therefore, there is an urgent need for an innovative solution that integrates new sensor technologies, intelligent algorithms, and dynamic calibration mechanisms. Summary of the Invention

[0005] The purpose of the present invention is to provide a mine microseismic positioning control system and method, aiming to solve the problem of low accuracy of the existing mine microseismic positioning control system.

[0006] To achieve the above purpose, in the first aspect, the present invention provides a mine microseismic positioning control system, including a multi-sensor array module, a data acquisition and preprocessing module, a machine learning model module, a positioning module, and a visualization and alarm module. The multi-sensor array module, the data acquisition and preprocessing module, the machine learning model module, the positioning module, and the visualization and alarm module are connected in sequence;

[0007] The multi-sensor array module is used to collect microseismic signals inside the mine;

[0008] The data acquisition and preprocessing module is used to synchronously collect, denoise, and extract features from the signals;

[0009] The machine learning model module receives the preprocessed data through a PCIe interface and performs event recognition and position prediction;

[0010] The positioning module receives the output result of the machine learning model and calculates the source coordinates through an optimization algorithm;

[0011] The visualization and alarm module communicates with the positioning module through Ethernet, displays the positioning result in real time, and triggers a hierarchical alarm.

[0012] Among them, the multi-sensor array module includes a geophone unit, an accelerometer unit, a transmitting sensor unit, and an optical fiber sensor unit;

[0013] The geophone unit is used to capture low-frequency micro-seismic waves and record the first arrival times and waveform characteristics of P-waves and S-waves;

[0014] The accelerometer unit monitors the instantaneous impact energy generated by rock fracture;

[0015] The transmitting sensor unit is used to capture high-frequency acoustic emission signals during rock fracture for analyzing the process of micro-crack propagation;

[0016] The optical fiber sensor unit monitors the rock mass strain and temperature changes in real time.

[0017] Among them, the data acquisition and preprocessing module includes a signal synchronous acquisition unit, a data cleaning and noise reduction unit, and a feature extraction and annotation unit, and the signal synchronous acquisition unit, the data cleaning and noise reduction unit, and the feature extraction and annotation unit are connected in sequence;

[0018] The signal synchronous acquisition unit ensures the time alignment of multi-sensor data based on the high-precision clock synchronization of GPS / Beidou.

[0019] The data cleaning and noise reduction unit is used to separate the effective signal from the background noise, and combines blind source separation and independent component analysis to eliminate irrelevant seismic source signals;

[0020] The feature extraction and annotation unit is used to extract the time-domain features, frequency-domain features, and time-frequency joint features of the signal, and based on semi-supervised learning, label the unlabeled data to construct a training data set.

[0021] Among them, the machine learning model module includes a feature fusion unit, an event recognition unit, and a location prediction unit, and the feature fusion unit, the event recognition unit, and the location prediction unit are connected in sequence;

[0022] The feature fusion unit models the spatial topological relationship between sensors through a graph neural network and fuses heterogeneous sensor features;

[0023] The event recognition unit extracts the local spatial features of the signal through a residual network (ResNet), identifies the types of micro-seismic events, captures the time dependence of the signal, and distinguishes real events from noise interference.

[0024] The location prediction unit constructs a positioning model based on deep reinforcement learning. The input features include sensor coordinates, signal propagation time delay, and attenuation coefficient, and the output is the three-dimensional coordinates (x, y, z) of the seismic source and the confidence score.

[0025] Among them, the positioning module includes an initial positioning unit and an optimized positioning unit, and the initial positioning unit is connected to the optimized positioning unit;

[0026] The initial positioning unit calculates the source candidate area through the arrival time difference of P-waves, reducing the search space of the optimization algorithm;

[0027] The optimized positioning unit takes the mean square error as the objective function, iteratively updates the particle swarm parameters, and realizes the global optimal solution of the source coordinates.

[0028] Among them, the visualization and alarm module includes a 3D geological modeling unit, a dynamic rendering unit, an alarm triggering unit, and a data interface unit, and the 3D geological modeling unit, the dynamic rendering unit, the alarm triggering unit, and the data interface unit are connected in sequence;

[0029] The 3D geological modeling unit integrates the mine geological database and constructs a 3D grid model;

[0030] The dynamic rendering unit: realizes real-time interaction on the browser side through WebGL, and supports the playback of the source trajectory and the simulation of energy diffusion;

[0031] The alarm triggering unit triggers yellow / orange / red three-level alarms according to the event energy and the positioning density threshold, pushes alarm information to the mine safety management system and the mobile terminal, and synchronously starts the emergency shutdown protocol.

[0032] The data interface unit supports data interconnection with third-party platforms and realizes the integrated analysis of the overall mine safety situation.

[0033] In the second aspect, a mine microseismic positioning control method is used for the mine microseismic positioning control system described in the first aspect, and includes the following steps:

[0034] Synchronously collect microseismic signals through geophones, accelerometers, acoustic emission sensors, and fiber optic sensors, and align the data timestamps based on GPS / Beidou clocks;

[0035] Perform blind source separation on the wavelet-filtered signals, extract independent source components, and fuse multi-sensor features through a graph neural network to construct a spatio-temporal joint feature matrix;

[0036] Use a ResNet-LSTM hybrid model to classify microseismic event types, generate an initial solution of the source coordinates based on deep reinforcement learning, and iteratively converge to the optimal solution through a particle swarm optimization algorithm;

[0037] Map the positioning result to a three-dimensional geological grid model, simulate the energy diffusion path, trigger a three-level alarm according to the event energy density, and automatically push an emergency instruction to the mine safety management system.

[0038] A mine microseismic positioning control system of the present invention includes a multi-sensor array module, a data acquisition and preprocessing module, a machine learning model module, a positioning module, and a visualization and alarm module. The present invention synchronously collects microseismic signals through geophones, accelerometers, acoustic emission sensors, and fiber optic sensors, and aligns the data timestamps based on the GPS / Beidou clock; performs blind source separation on the wavelet-filtered signals, extracts independent source components, and fuses multi-sensor features through a graph neural network to construct a spatio-temporal joint feature matrix; uses a ResNet-LSTM hybrid model to classify microseismic event types, generates an initial solution of the source coordinates based on deep reinforcement learning, and iteratively converges to the optimal solution through a particle swarm optimization algorithm; maps the positioning result to a three-dimensional geological grid model, simulates the energy diffusion path, triggers a three-level alarm according to the event energy density, and automatically pushes an emergency instruction to the mine safety management system. Thus, the problem of low accuracy of the existing mine microseismic positioning control system is solved. Brief Description of the Drawings

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0040] Figure 1 It is a schematic diagram of a mine microseismic positioning control system provided by the present invention.

[0041] Figure 2 It is a schematic diagram of the multi-sensor array module.

[0042] Figure 3 It is a schematic diagram of the data acquisition and preprocessing module.

[0043] Figure 4 It is a schematic diagram of the machine learning model module.

[0044] Figure 5 It is a schematic diagram of the positioning module.

[0045] Figure 6 It is a schematic diagram of the visualization and alarm module.

[0046] Figure 7 It is a flowchart of a mine microseismic positioning control method provided by the present invention.

[0047] In the figure: 1 - Multi - sensor array module, 2 - Data acquisition and pre - processing module, 3 - Machine learning model module, 4 - Positioning module, 5 - Visualization and alarm module, 11 - Seismic detector unit, 12 - Accelerometer unit, 13 - Transmitting sensor unit, 14 - Fiber optic sensor unit, 21 - Signal synchronous acquisition unit, 22 - Data cleaning and noise reduction unit, 23 - Feature extraction and annotation unit, 31 - Feature fusion unit, 32 - Event recognition unit, 33 - Position prediction unit, 41 - Initial positioning unit, 42 - Optimization positioning unit, 51 - 3D geological modeling unit, 52 - Dynamic rendering unit, 53 - Alarm trigger unit, 54 - Data interface unit. Specific implementation manners

[0048] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention, and should not be construed as a limitation of the present invention.

[0049] Please refer to Figures 1 to 6 , in the first aspect, the present invention provides a mine micro - seismic positioning control system, including a multi - sensor array module 1, a data acquisition and pre - processing module 2, a machine learning model module 3, a positioning module 4, and a visualization and alarm module 5. The multi - sensor array module 1, the data acquisition and pre - processing module 2, the machine learning model module 3, the positioning module 4, and the visualization and alarm module 5 are connected in sequence;

[0050] The multi - sensor array module 1 is used to collect micro - seismic signals inside the mine;

[0051] The data acquisition and pre - processing module 2 is used to synchronously collect, denoise, and extract features from the signals;

[0052] The machine learning model module 3 receives the pre - processed data through a PCIe interface and performs event recognition and position prediction;

[0053] The positioning module 4 receives the output result of the machine learning model and calculates the source coordinates through an optimization algorithm;

[0054] The visualization and alarm module 5 communicates with the positioning module 4 through Ethernet, and displays the positioning result in real time and triggers a hierarchical alarm.

[0055] In this embodiment, the present invention synchronously collects microseismic signals through geophones, accelerometers, acoustic emission sensors and fiber optic sensors, and aligns the data timestamps based on the GPS / Beidou clock; performs blind source separation on the wavelet-filtered signals, extracts independent source components, and fuses multi-sensor features through a graph neural network to construct a spatio-temporal joint feature matrix; uses a ResNet-LSTM hybrid model to classify microseismic event types, generates an initial solution of the source coordinates based on deep reinforcement learning, and iteratively converges to the optimal solution through a particle swarm optimization algorithm; maps the positioning results to a three-dimensional geological grid model, simulates the energy diffusion path, triggers a three-level alarm according to the event energy density, and automatically pushes emergency instructions to the mine safety management system. Thus, the problem of low accuracy of the existing mine microseismic positioning control system is solved.

[0056] Further, the multi-sensor array module 1 includes a geophone unit 11, an accelerometer unit 12, a transmitting sensor unit 13 and a fiber optic sensor unit 14;

[0057] The geophone unit 11 is used to capture low-frequency microseismic waves and record the arrival times and waveform characteristics of P-waves and S-waves;

[0058] The accelerometer unit 12 monitors the instantaneous impact energy generated by rock fracture;

[0059] The transmitting sensor unit 13 is used to capture high-frequency acoustic emission signals during rock fracture for analyzing the process of microcrack propagation;

[0060] The fiber optic sensor unit 14 monitors the rock mass strain and temperature changes in real time.

[0061] In this embodiment, the geophone unit 11 adopts a broadband velocity geophone (0.1 - 50 Hz), with a sensitivity ≥ 20 V / m / s and a dynamic range > 120 dB.

[0062] P / S wave identification algorithm: Based on the STA / LTA (Short-Term Average / Long-Term Average) ratio method:

[0063]

[0064] where N STA = 0.1 s, N LTA = 5 s, and the threshold λ th = 3.0.

[0065] The fiber optic sensor unit 14 adopts a strain demodulation algorithm: Based on Brillouin optical time domain reflectometry (BOTDR) technology, with a strain accuracy of ±5 με.

[0066] Δ VB = C ∈·Δ ∈ +C T ·ΔT

[0067] where C ∈ = 0.048 MHz / με, C T = 1.0 MHz / °C, and strain and temperature effects are separated through temperature compensation.

[0068] Furthermore, the data acquisition and preprocessing module 2 includes a signal synchronous acquisition unit 21, a data cleaning and noise reduction unit 22, and a feature extraction and annotation unit 23, and the signal synchronous acquisition unit 21, the data cleaning and noise reduction unit 22, and the feature extraction and annotation unit 23 are connected in sequence;

[0069] The signal synchronous acquisition unit 21 ensures time alignment of multi-sensor data based on high-precision clock synchronization of GPS / Beidou.

[0070] The data cleaning and noise reduction unit 22 is used to separate valid signals from background noise, and combines blind source separation and independent component analysis to eliminate irrelevant seismic source signals;

[0071] The feature extraction and annotation unit 23 is used to extract time-domain features, frequency-domain features, and time-frequency joint features of signals, and tag unlabeled data based on semi-supervised learning to construct a training data set.

[0072] In this embodiment, the signal synchronous acquisition unit 21 adopts a clock synchronization protocol: the IEEE 1588v2 Precision Time Protocol (PTP) is adopted, the synchronization error <1 μs, and multi-node cascading is supported. Anti-aliasing design: According to the Nyquist theorem, the sampling rate f s ≥ 2.56×f max , and the sampling rate (1 - 100 kHz) is dynamically configured.

[0073] The wavelet threshold denoising of the data cleaning and noise reduction unit 22 selects the Db4 wavelet basis, the decomposition level J = 8J = 8, and the threshold function is the improved SUREShrink:

[0074]

[0075] where σ j : the standard deviation of the detail coefficients of the j-th layer, N j : the number of coefficients.

[0076] Blind source separation (BSS): Based on the FastICA algorithm, the objective function is to maximize negative entropy:

[0077] J(y) = [E{G(y)} - E{G(v)}] 2

[0078] Among them, G(u) = tanh(u), and the unmixing matrix W is solved by the Newton iteration method.

[0079] For the feature extraction and annotation unit 23, the wavelet packet energy entropy:

[0080]

[0081] Mel-frequency cepstral coefficients (MFCC): 24-dimensional features, covering the human ear sensitive frequency band of 20 Hz - 4 kHz.

[0082] Furthermore, the machine learning model module 3 includes a feature fusion unit 31, an event recognition unit 32, and a location prediction unit 33, and the feature fusion unit 31, the event recognition unit 32, and the location prediction unit 33 are connected in sequence;

[0083] The feature fusion unit 31 models the spatial topological relationship between sensors through a graph neural network and fuses heterogeneous sensor features;

[0084] The event recognition unit 32 extracts the local spatial features of the signal through a residual network (ResNet), identifies the types of microseismic events, captures the time dependence of the signal, and distinguishes real events from noise interference.

[0085] The location prediction unit 33 constructs a positioning model based on deep reinforcement learning. The input features include sensor coordinates, signal propagation delay, and attenuation coefficient, and the output is the three-dimensional coordinates (x, y, z) of the seismic source and the confidence score.

[0086] In this embodiment, the design of the graph neural network (GNN) of the feature fusion unit 31:

[0087] Node features: Sensor coordinates, signal-to-noise ratio, geological parameters (such as rock layer hardness).

[0088] Edge weight: The reciprocal of the sensor spacing w ij = 1 / d ij , and the propagation attenuation coefficient α ij = e -kdij .

[0089] Aggregation function:

[0090]

[0091] The residual block structure of the event recognition unit 32: Each residual block contains 2 convolutional layers (kernel = 3×1, stride = 1) + batch normalization (BN) + ReLU.

[0092] The skip connection compensates for the vanishing gradient: y = F(x) + x.

[0093] LSTM Time Series Modeling:

[0094] The input gate, forget gate, and output gate share parameters, the hidden layer dimension = 128, and the time step = 100.

[0095] The state space of the position prediction unit 33: st = (Δt ij , A peak , f dominant , vp)

[0096] Action space: Three-dimensional coordinate increment Δ x , Δ y , Δ z ∈[-1,1]m;

[0097] Reward function:

[0098] Furthermore, the positioning module 4 includes an initial positioning unit 41 and an optimized positioning unit 42, and the initial positioning unit 41 is connected to the optimized positioning unit 42;

[0099] The initial positioning unit 41 calculates the source candidate area through the P-wave arrival time difference, reducing the search space of the optimization algorithm;

[0100] The optimized positioning unit 42 takes the mean square error as the objective function, iteratively updates the particle swarm parameters, and realizes the global optimal solution of the source coordinates.

[0101] In this embodiment, the initial positioning unit 41 adopts the particle swarm optimization (PSO) algorithm, and the parameter settings are as follows:

[0102] The number of particles = 200, the number of iterations = 500, the learning factors c1 = c2 = 2.0, and the inertia weight ω linearly decreases from 0.9 to 0.4.

[0103] Objective function:

[0104]

[0105] Furthermore, the visualization and alarm module 5 includes a three-dimensional geological modeling unit 51, a dynamic rendering unit 52, an alarm trigger unit 53, and a data interface unit 54, and the three-dimensional geological modeling unit 51, the dynamic rendering unit 52, the alarm trigger unit 53, and the data interface unit 54 are connected in sequence;

[0106] The three-dimensional geological modeling unit 51 integrates the mine geological database and constructs a three-dimensional grid model;

[0107] The dynamic rendering unit 52: realizes real-time interaction on the browser side through WebGL, and supports the playback of the source trajectory and the simulation of energy diffusion;

[0108] The alarm trigger unit 53 triggers yellow / orange / red three-level alarms according to the event energy and the positioning density threshold, pushes alarm information to the mine safety management system and the mobile terminal, and synchronously starts the emergency shutdown protocol.

[0109] The data interface unit 54 supports data intercommunication with third-party platforms to achieve integrated analysis of the overall mine safety situation.

[0110] In this embodiment, the 3D geological modeling unit 51 adopts a grid division algorithm: Delaunay triangulation, with a grid size ≤ 1 m, integrating the data of the strike and dip of rock strata.

[0111] Energy diffusion simulation: Based on the finite difference method (FDM) of wave equation:

[0112]

[0113] Three-level threshold:

[0114]

[0115]

[0116] Please refer to Figure 7 , in the second aspect, a mine microseismic positioning control method for the mine microseismic positioning control system described in the first aspect includes the following steps:

[0117] S1 Synchronously collects microseismic signals through geophones, accelerometers, acoustic emission sensors and fiber optic sensors, and aligns the data timestamps based on the GPS / Beidou clock;

[0118] Specifically, the geophone detects the propagation of seismic waves and can capture the P-waves and S-waves generated by microseismic events. The accelerometer measures the vibration acceleration inside the mine and can reflect the intensity and dynamic changes of microseismic events. The acoustic emission sensor detects the acoustic emission signals generated when the rock inside the mine fractures and is suitable for monitoring events such as rock bursts. The fiber optic sensor uses the optical time domain reflectometry (OTDR) technology of optical fiber to real-time monitor the strain and temperature changes inside the mine and can detect the tiny deformations caused by microseismic events. Using the GPS / Beidou clock synchronization technology, it ensures that the data collected by all sensors has a unified time reference. Through the high-precision clock module, the sampling timestamps of different sensors are accurately aligned, and the error is controlled within milliseconds. This step is the basis for subsequent data processing and fusion, ensuring the temporal consistency of multi-source data. The collected original signals are preliminarily processed, including filtering, denoising and normalization. S2 performs blind source separation on the wavelet-filtered signals, extracts independent source components, and fuses multi-sensor features through a graph neural network to construct a spatio-temporal joint feature matrix;

[0119] Specifically, the independent component analysis (ICA) algorithm is used to perform blind source separation on the wavelet-filtered signal. ICA can decompose the mixed signal into multiple independent source components, thereby separating the characteristic signals of different microseismic events. These characteristics can reflect the physical properties of microseismic events. A graph neural network (GNN) model is constructed, and the feature data of multiple sensors are used as node inputs. GNN can learn the spatial relationship and temporal dependence between sensors, and extract spatio-temporal joint features through graph convolution operations.

[0120] S3 uses a ResNet-LSTM hybrid model to classify microseismic event types, generates an initial solution for the source coordinates based on deep reinforcement learning, and iteratively converges to the optimal solution through the particle swarm optimization algorithm;

[0121] Specifically, a hybrid model of ResNet (residual network) and LSTM (long short-term memory network) is used to classify microseismic events. ResNet can extract the deep features of the signal, and LSTM can capture the dynamic changes of time series data. Through the hybrid model, different types of microseismic events (such as rock bursts, rock bumps, etc.) can be accurately distinguished. The deep reinforcement learning (DRL) algorithm is used to generate an initial solution for the source coordinates. DRL can find the optimal source location through trial-and-error learning. PSO finds the optimal solution by simulating the foraging behavior of bird flocks. In microseismic positioning, PSO can gradually adjust the source coordinates according to the initial solution and finally converge to the optimal solution.

[0122] S4 maps the positioning results to a three-dimensional geological grid model, simulates the energy diffusion path, and triggers a three-level alarm according to the event energy density, and automatically pushes emergency instructions to the mine safety management system.

[0123] Specifically, the positioning results are mapped into the three-dimensional geological grid model of the mine. This model can intuitively display the spatial distribution and geological background of microseismic events. Numerical simulation techniques (such as the finite element method) are used to simulate the energy diffusion path of microseismic events. Through simulation, the possible affected range of microseismic events can be predicted, providing a basis for emergency response. A three-level alarm mechanism is triggered according to the energy density of microseismic events. Level 1 alarm: The energy density is low, prompting mine workers to pay attention. Level 2 alarm: The energy density is medium, requiring preventive measures to be taken. Level 3 alarm: The energy density is high, triggering an emergency evacuation instruction. The alarm information and emergency instructions are automatically pushed to the mine safety management system. The system can notify mine managers and on-site workers in real time to ensure that measures are taken in a timely manner to ensure mine safety.

[0124] The above-disclosed is only a preferred embodiment of a mine microseismic positioning control system and method of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.

Claims

1. A mine microseismic positioning control system, characterized in that it includes a multi-sensor array module, a data acquisition and preprocessing module, a machine learning model module, a positioning module, and a visualization and alarm module, and the multi-sensor array module, the data acquisition and preprocessing module, the machine learning model module, the positioning module, and the visualization and alarm module are connected in sequence; The multi-sensor array module is used to collect microseismic signals inside the mine; The data acquisition and preprocessing module is used to synchronously collect, denoise, and extract features from the signals; The machine learning model module receives the preprocessed data through the PCIe interface and performs event recognition and position prediction; The positioning module receives the output result of the machine learning model and calculates the source coordinates through an optimization algorithm; The visualization and alarm module communicates with the positioning module through Ethernet, displays the positioning result in real time, and triggers hierarchical alarms.

2. The mine microseismic positioning control system according to claim 1, characterized in that the multi-sensor array module includes a geophone unit, an accelerometer unit, a transmitting sensor unit, and an optical fiber sensor unit; The geophone unit is used to capture low-frequency microseismic waves and record the arrival times and waveform characteristics of P-waves and S-waves; The accelerometer unit monitors the instantaneous impact energy generated by rock fracture; The transmitting sensor unit is used to capture high-frequency acoustic emission signals during rock fracture and is used to analyze the process of microcrack propagation; The optical fiber sensor unit monitors the rock mass strain and temperature changes in real time.

3. The mine microseismic positioning control system according to claim 2, characterized in that the data acquisition and preprocessing module includes a signal synchronous acquisition unit, a data cleaning and denoising unit, and a feature extraction and annotation unit, and the signal synchronous acquisition unit, the data cleaning and denoising unit, and the feature extraction and annotation unit are connected in sequence; The signal synchronous acquisition unit ensures the time alignment of multi-sensor data based on the high-precision clock synchronization of GPS / Beidou. The data cleaning and denoising unit is used to separate effective signals from background noise, and combines blind source separation and independent component analysis to eliminate irrelevant source signals; The feature extraction and annotation unit is used to extract signal time-domain features, frequency-domain features), and time-frequency joint features, and label unlabeled data based on semi-supervised learning to construct a training data set.

4. The mine microseismic positioning control system according to claim 3, characterized in that the machine learning model module includes a feature fusion unit, an event recognition unit, and a position prediction unit, and the feature fusion unit, the event recognition unit, and the position prediction unit are connected in sequence; The feature fusion unit models the spatial topological relationship between sensors through a graph neural network and fuses heterogeneous sensor features; The event recognition unit extracts the local spatial features of the signal through a residual network (ResNet), identifies the types of microseismic events, captures the time dependence of the signal, and distinguishes real events from noise interference. The position prediction unit constructs a positioning model based on deep reinforcement learning. The input features include sensor coordinates, signal propagation delay, and attenuation coefficient, and the output is the three-dimensional coordinates (x, y, z) of the seismic source and a confidence score.

5. The mine microseismic positioning control system according to claim 4, wherein the positioning module includes an initial positioning unit and an optimized positioning unit, and the initial positioning unit is connected to the optimized positioning unit; the initial positioning unit calculates the candidate area of the seismic source through the P-wave arrival time difference to narrow the search space of the optimization algorithm; the optimized positioning unit takes the mean square error as the objective function, iteratively updates the particle swarm parameters, and realizes the global optimal solution of the seismic source coordinates.

6. The mine microseismic positioning control system according to claim 5, wherein the visualization and alarm module includes a three-dimensional geological modeling unit, a dynamic rendering unit, an alarm trigger unit, and a data interface unit, and the three-dimensional geological modeling unit, the dynamic rendering unit, the alarm trigger unit, and the data interface unit are connected in sequence; the three-dimensional geological modeling unit integrates the mine geological database and constructs a three-dimensional grid model; the dynamic rendering unit: realizes real-time interaction on the browser side through WebGL, and supports the playback of the seismic source trajectory and the simulation of energy diffusion; the alarm trigger unit triggers three-level alarms of yellow / orange / red according to the event energy and the positioning density threshold, pushes the alarm information to the mine safety management system and the mobile terminal, and synchronously starts the emergency shutdown protocol. the data interface unit supports data interconnection with a third-party platform to realize the integrated analysis of the overall safety situation of the whole mine.

7. A mine microseismic positioning control method for the mine microseismic positioning control system according to any one of claims 1-6, characterized in that, including the following steps: Synchronously collect microseismic signals through geophones, accelerometers, acoustic emission sensors, and fiber optic sensors, and align the data timestamps based on the GPS / Beidou clock; Perform blind source separation on the wavelet-filtered signals, extract independent seismic source components, and fuse multi-sensor features through a graph neural network to construct a spatio-temporal joint feature matrix; Use a ResNet-LSTM hybrid model to classify the types of microseismic events, generate an initial solution of the seismic source coordinates based on deep reinforcement learning, and iteratively converge to the optimal solution through a particle swarm optimization algorithm; Map the positioning results to the three-dimensional geological grid model, simulate the energy diffusion path, trigger three-level alarms according to the event energy density, and automatically push emergency instructions to the mine safety management system.

Images