A method and system for real-time monitoring of mining subsidence

Abstract

This invention discloses a method and system for real-time monitoring of mining-induced subsidence, comprising: an in-situ acoustic and electrical sensing module, an edge intelligent analysis module, a multi-field coupling inference module, and a dynamic early warning and response module. The in-situ acoustic and electrical sensing module is used to non-invasively collect acoustic emission signals and apparent resistivity time-series data of micro-fractures within the overburden in underground working faces, surrounding rock of roadways, and surface subsidence-affected areas. This method and system improves the positioning accuracy of deep overburden micro-fractures to 0.3m by dynamically adjusting electrode spacing, a temperature compensation module, and a hybrid redundant transmission link. It solves the industry problem of incomplete sensing dimensions and missing deep information in traditional single monitoring methods. Through lightweight CNN noise removal and deep learning rapid inversion technology, the accuracy of effective acoustic emission event identification is increased to over 97%, significantly reducing cloud data transmission volume and computational load, and enabling local real-time processing of monitoring data.

Description

Technical Field

[0001] This invention relates to the field of mine safety monitoring technology, and in particular to a method and system for real-time monitoring of mining subsidence. Background Technology

[0002] Coal, as a core pillar of my country's energy system, has long occupied a dominant position in the primary energy consumption structure, and its safe and efficient mining is directly related to the stable operation of the national economy. As coal mining in my country gradually extends to deeper levels, the evolution of overlying strata fracturing, water-conducting fracture development, and surface subsidence under the influence of mining exhibits complex characteristics of strong nonlinearity, strong concealment, and multi-field coupling. If monitoring is not timely, it can easily induce major disasters such as water inrush, face collapse, and damage to surface structures, posing a serious threat to coal mine safety and the surrounding ecological environment. Currently, traditional monitoring methods include combining data on subsidence that has already occurred in the mining area to be monitored, setting monitoring areas within the mining area, and achieving online monitoring of these areas through neural networks and preset monitoring frequencies.

[0003] Existing subsidence monitoring methods based on matching and screening historical mining area databases still rely on empirical judgments based on existing historical data. They are completely unable to adapt to the real-time dynamic changes in mining conditions at the current working face and cannot achieve dynamic perception of subsidence evolution within the overburden. Other surface subsidence monitoring schemes that use InSAR satellite remote sensing can only obtain subsidence rates at the millimeter level on the surface and cannot detect the early signs of fracture deep within the overburden. The early warning is extremely delayed, and risks are often only identified after several centimeters of obvious subsidence has already appeared on the surface, completely missing the best window of opportunity for disaster prevention and control. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a method and system for real-time monitoring of mining subsidence. The purpose is to solve problems such as the inability to detect early-stage fracture signals deep within the overlying strata, the extremely high lag in early warning, and the fact that risks are often only identified after significant subsidence of several centimeters has already occurred on the surface, thus completely missing the optimal window for disaster prevention and control.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by the present invention is as follows: a method and system for real-time monitoring of mining subsidence, comprising: an acoustic and electrical in-situ sensing module, an edge intelligent analysis module, a multi-field coupling deduction module, and a dynamic early warning and handling module;

[0006] The acoustic-electric in-situ sensing module is used to simultaneously deploy acoustic emission sensor arrays and high-density electrical resistivity monitoring electrodes in the overburden borehole, roadway surrounding rock, and surface subsidence-affected area of ​​the downhole working face. It non-invasively collects acoustic emission signals and apparent resistivity time-series data of micro-fractures inside the overburden, realizing in-situ continuous sensing of overburden fracture and water-conducting fracture development. The electrode deployment spacing of the acoustic-electric in-situ sensing module is dynamically adjusted according to the overburden lithology. In the weak rock layer area, the deployment spacing is increased to 2m to ensure the identification accuracy of low-resistivity fracture anomaly areas.

[0007] The edge intelligent analysis module is used to aggregate acoustic emission signals and apparent resistivity data through the well edge gateway, and complete the local acoustic emission event location and apparent resistivity anomaly area identification locally, while only uploading the subsidence-related abnormal feature data to the cloud.

[0008] The multi-field coupling simulation module is used to construct a numerical model of the three fields of mining stress, resistivity and subsidence displacement. It uses the real-time feature data after edge analysis as dynamic constraints and iteratively solves to output a full-space subsidence evolution cloud map from deep overburden to the surface.

[0009] The dynamic early warning and response module is used to spatially correlate the full-space subsidence evolution cloud map with the mining engineering map and the surface protection object layer to generate a graded subsidence early warning report, which is simultaneously pushed to the ground dispatch center and the terminals of underground workers to achieve early detection and early response to subsidence disasters.

[0010] Furthermore, the acoustic-electric in-situ sensing module includes an acoustic emission acquisition subunit and a high-density electrical resistivity subunit. The acoustic emission acquisition subunit adopts a 16-channel synchronous acquisition architecture with a sampling rate of 1MHz. Through a time difference positioning algorithm, the three-dimensional positioning accuracy of overburden micro-fracture events is controlled within 0.3m, accurately capturing minute fracture signals inside the overburden. The high-density electrical resistivity subunit uses a Wenner quadrupole device to complete the apparent resistivity data acquisition, performing a full-domain scan every 5 minutes. Through a two-dimensional inversion algorithm, it outputs the apparent resistivity profile of the overburden, accurately identifying the dynamic development boundary of the water-conducting fracture zone.

[0011] Furthermore, the edge intelligent analysis module includes an acoustic emission positioning subunit and a resistivity inversion subunit. The acoustic emission positioning subunit deploys a lightweight CNN model within the edge gateway to automatically eliminate strong noise interference generated by mining machinery, thereby improving the recognition accuracy of effective acoustic emission events to over 97%. The resistivity inversion subunit uses a pre-trained deep learning inversion model to replace the traditional iterative inversion algorithm, reducing the inversion time of a single frame of apparent resistivity profile from 10 minutes to 2 seconds, thus achieving real-time identification of overburden electrical anomalies.

[0012] Furthermore, the multi-field coupled simulation module includes a model building subunit and a dynamic solution subunit. The model building subunit pre-imports the mechanical parameters, electrical parameters and historical mining data of the overlying rock in the mining area to complete the offline calibration of the three-field coupled numerical model. The initial simulation results have a fitting degree of no less than 94% with the historical measured subsidence data. The dynamic solution subunit adopts a GPU parallel acceleration architecture, takes the acoustic emission positioning results obtained by edge analysis and the apparent resistivity anomaly area as the dynamic boundary condition input, and outputs a full-space three-dimensional subsidence displacement cloud map every 3 seconds.

[0013] Furthermore, the dynamic early warning and response module includes a risk grading subunit, which constructs a three-dimensional coupled early warning index system of "acoustic emission event rate - apparent resistivity change rate - subsidence displacement gradient", classifies the early warning level into three levels: attention, warning, and danger, and matches the corresponding response plan for targeted push.

[0014] Furthermore, the acoustic-electric in-situ sensing module also integrates a micro temperature compensation module, which automatically eliminates the interference of downhole environmental temperature changes on apparent resistivity monitoring data and controls the systematic error of long-term monitoring data to within 2%.

[0015] Furthermore, the acoustic and electrical in-situ sensing module and the edge intelligent analysis module transmit hybrid data through mining flame-retardant shielded cables and fiber optic redundant links. The acoustic emission signal is transmitted using cable analog transmission, while the apparent resistivity digital signal is transmitted using fiber optics. A signal synchronization calibrator is set at the edge gateway entrance to ensure that the spatiotemporal synchronization accuracy of the two types of data is not less than 0.5ms, providing a unified time reference for subsequent multi-field coupling simulations.

[0016] Furthermore, the acoustic-electric in-situ sensing module, the edge intelligent analysis module, and the dynamic early warning and handling module are all equipped with a working condition self-diagnosis subunit. The working condition self-diagnosis subunit monitors the power supply status, channel connectivity, electrode grounding resistance, and sensor sensitivity of each module in real time. When any parameter exceeds the preset threshold, it automatically generates equipment abnormality alarm information and pushes it to the ground dispatch center. At the same time, it marks the credibility coefficient of the monitoring data for the corresponding time period, which is used by the multi-field coupled inference module for weighted correction during iterative solution.

[0017] The beneficial effects of this invention are as follows:

[0018] This invention discloses a real-time monitoring method and system for mining subsidence. By dynamically adjusting the electrode spacing, temperature compensation module, and hybrid redundant transmission link, the system improves the positioning accuracy of deep overburden micro-fractures to 0.3m, and controls the long-term monitoring system error to within 2%. It solves the industry problem of incomplete perception dimensions and lack of deep information in traditional single monitoring methods. Through lightweight CNN noise removal and deep learning fast inversion technology, the accuracy of effective acoustic emission event identification is improved to over 97%, and the single-frame apparent resistivity profile inversion time is compressed from 10 minutes to 2 seconds. This significantly reduces the amount of cloud data transmission and computing load, and enables local real-time processing of monitoring data.

[0019] This invention discloses a real-time monitoring method and system for mining-induced subsidence. It constructs a three-field coupled numerical model of mining stress, resistivity, and subsidence displacement, and combines it with GPU parallel acceleration technology to output a full-space three-dimensional subsidence evolution cloud map every 3 seconds. The initial simulation results have a fitting degree of no less than 94% with historical measured data, achieving full-space dynamic simulation from deep overburden fracture to surface subsidence. A three-dimensional coupled hierarchical early warning system is established, and early warning information is pushed to the ground dispatch center and underground workers' terminals. Combined with a full-process working condition self-diagnosis and data credibility weighted correction mechanism, the timeliness and accuracy of subsidence disaster early warning are significantly improved, effectively supporting intelligent and safe mining. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method and system for real-time monitoring of mining subsidence according to the present invention;

[0021] Figure 2 This is a dynamic early warning framework diagram of a real-time monitoring method and system for mining subsidence according to the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. Identical components are indicated by the same reference numerals.

[0023] It should be noted that the terms “front,” “back,” “left,” “right,” “up,” and “down” used in the following description refer to the directions shown in the attached diagram, while the terms “inside” and “outside” refer to the directions toward or away from the geometric center of a specific component, respectively.

[0024] To make the content of this invention easier to understand, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings.

[0025] like Figures 1 to 2 As shown, a method and system for real-time monitoring of mining subsidence includes: an acoustic and electrical in-situ sensing module, an edge intelligent analysis module, a multi-field coupling inference module, and a dynamic early warning and response module;

[0026] The acoustic-electric in-situ sensing module is used to simultaneously deploy acoustic emission sensor arrays and high-density electrical resistivity monitoring electrodes in the overburden borehole, roadway surrounding rock, and surface subsidence-affected area of ​​the downhole working face. It non-invasively collects acoustic emission signals and apparent resistivity time-series data of micro-fractures inside the overburden, realizing in-situ continuous sensing of overburden fracture and water-conducting fracture development. The electrode deployment spacing of the acoustic-electric in-situ sensing module is dynamically adjusted according to the overburden lithology. In the weak rock layer area, the deployment spacing is increased to 2m to ensure the identification accuracy of low-resistivity fracture anomaly areas.

[0027] The edge intelligent analysis module is used to aggregate acoustic emission signals and apparent resistivity data through the well edge gateway, and complete the local acoustic emission event location and apparent resistivity anomaly area identification locally, while only uploading the subsidence-related abnormal feature data to the cloud.

[0028] The multi-field coupling simulation module is used to construct a numerical model of the three fields of mining stress, resistivity and subsidence displacement. It uses the real-time feature data after edge analysis as dynamic constraints and iteratively solves to output a full-space subsidence evolution cloud map from deep overburden to the surface.

[0029] The dynamic early warning and response module is used to spatially correlate the full-space subsidence evolution cloud map with the mining engineering map and the surface protection object layer to generate a graded subsidence early warning report, which is simultaneously pushed to the ground dispatch center and the terminals of underground workers to achieve early detection and early response to subsidence disasters.

[0030] The acoustic-electric in-situ sensing module includes an acoustic emission acquisition subunit and a high-density electrical resistivity subunit. The acoustic emission acquisition subunit adopts a 16-channel synchronous acquisition architecture with a sampling rate of 1MHz. Through a time difference positioning algorithm, the three-dimensional positioning accuracy of overburden micro-fracture events is controlled within 0.3m, accurately capturing minute fracture signals inside the overburden. The high-density electrical resistivity subunit uses a Wenner quadrupole device to complete the apparent resistivity data acquisition, performing a full-domain scan every 5 minutes. Through a two-dimensional inversion algorithm, it outputs the apparent resistivity profile of the overburden, accurately identifying the dynamic development boundary of the water-conducting fracture zone.

[0031] The edge intelligent analysis module includes an acoustic emission localization subunit and a resistivity inversion subunit. The acoustic emission localization subunit deploys a lightweight CNN model within the edge gateway to automatically eliminate strong noise interference generated by mining machinery, improving the recognition accuracy of effective acoustic emission events to over 97%. The resistivity inversion subunit uses a pre-trained deep learning inversion model to replace the traditional iterative inversion algorithm, reducing the inversion time of a single frame of apparent resistivity profile from 10 minutes to 2 seconds, and realizing real-time identification of overburden electrical anomalies.

[0032] The multi-field coupled simulation module includes a model building subunit and a dynamic solution subunit. The model building subunit pre-imports the mechanical and electrical parameters of the overlying rock in the mining area and historical mining data to complete the offline calibration of the three-field coupled numerical model. The initial simulation results have a fitting degree of no less than 94% with the historical measured subsidence data. The dynamic solution subunit adopts a GPU parallel acceleration architecture and uses the acoustic emission positioning results obtained from edge analysis and the apparent resistivity anomaly area as the input of dynamic boundary conditions. It outputs a full-space three-dimensional subsidence displacement cloud map every 3 seconds.

[0033] The dynamic early warning and response module includes a risk classification subunit. The risk classification subunit constructs a three-dimensional coupled early warning indicator system of "acoustic emission event rate - apparent resistivity change rate - subsidence displacement gradient", which divides the early warning level into three levels: attention, warning, and danger, and matches the corresponding response plan for targeted push.

[0034] The acoustic and electrical in-situ sensing module also integrates a micro temperature compensation module, which automatically eliminates the interference of downhole environmental temperature changes on apparent resistivity monitoring data and controls the systematic error of long-term monitoring data to within 2%.

[0035] The acoustic and electrical in-situ sensing module and the edge intelligent analysis module transmit hybrid data through mining flame-retardant shielded cables and fiber optic redundant links. The acoustic emission signal is transmitted using cable analog transmission, while the apparent resistivity digital signal is transmitted using fiber optics. A signal synchronization calibrator is set at the edge gateway entrance to ensure that the spatiotemporal synchronization accuracy of the two types of data is not less than 0.5ms, providing a unified time reference for subsequent multi-field coupling simulations.

[0036] The acoustic and electrical in-situ sensing module, the edge intelligent analysis module, and the dynamic early warning and handling module are all equipped with a working condition self-diagnosis subunit. The working condition self-diagnosis subunit monitors the power supply status, channel connectivity, electrode grounding resistance, and sensor sensitivity of each module in real time. When any parameter exceeds the preset threshold, it automatically generates equipment abnormality alarm information and pushes it to the ground dispatch center. At the same time, it marks the credibility coefficient of the monitoring data for the corresponding time period, which is used by the multi-field coupled inference module for weighted correction during iterative solution.

[0037] Embodiments of this invention patent:

[0038] The application scenario of this embodiment is a typical deep fully mechanized mining face in a mining area. The working face is buried at a depth of 480m, with a strike length of 1920m and a dip length of 245m. The overburden consists of four layers of weak mudstone with thicknesses of 7.2m, 11.5m, 6.8m, and 8.3m respectively. The surface is covered by a cluster of villages and buildings and the protection zone of a water conservancy project's water conveyance branch. This is a high-risk confined water mining face as clearly defined in the "Coal Mine Safety Regulations". Traditional single-point subsidence monitoring methods have repeatedly shown problems with delayed early warnings, posing a dual safety hazard of overburden water inrush and surface building cracking.

[0039] Step 1: On-site deployment and debugging of the acoustic and electrical in-situ sensing module

[0040] Measurement point deployment: In the three overburden observation boreholes completed in advance above the working face, 16 acoustic emission sensors were deployed at intervals along the borehole axis to form a 16-channel synchronous acquisition array; a total of 142 high-density electrical resistivity tomography (EPR) monitoring electrodes were deployed in the surrounding rock of the two sides of the roadway and the surface subsidence area of ​​the working face. For the corresponding depth range of the four layers of weak mudstone, the electrode deployment spacing was strictly increased to 2m, and the electrode deployment spacing in the other stable medium-hard sandstone areas was set to 5m. This ensured the identification accuracy of low-resistivity fracture anomaly areas and avoided redundant measurement points and wasted resources.

[0041] Hardware parameter calibration: The hardware sampling rate of the acoustic emission acquisition subunit was fixed at 1MHz. The time difference joint positioning algorithm was used to complete the field calibration. By manually tapping at different depths of the borehole to simulate the micro-fracture signal of the overburden, the positioning algorithm parameters were repeatedly calibrated. Finally, the three-dimensional positioning accuracy of the micro-fracture event of the overburden was measured to be controlled at 0.27m, which can accurately capture the micro-fracture signal with a particle size of more than 0.1mm inside the overburden. The high-density electrical resistivity subunit was connected and debugged using a Wenner quadrupole device. It was set to automatically trigger a full-domain scan every 5 minutes. The apparent resistivity profile of the overburden was output through the basic two-dimensional inversion algorithm. Field measurements showed that it can accurately identify the dynamic development boundary of the water-conducting fracture zone with a thickness greater than 1.2m.

[0042] Temperature compensation function verification: An NTC micro temperature compensation module is integrated inside each electrical resistivity acquisition electrode. It automatically acquires the downhole ambient temperature every 10 seconds and automatically corrects the apparent resistivity monitoring value through a pre-calibrated "temperature-resistivity" correction formula. The results of continuous 72-hour downhole simulated environment test show that the module can completely eliminate the monitoring data drift caused by downhole ambient temperature fluctuations. The actual measured system error of continuous long-term monitoring for 6 months is controlled at 1.6%, which is far below the preset 2% threshold.

[0043] Step 2: Establishment of Hybrid Redundant Transmission Links and Spatiotemporal Synchronization Calibration

[0044] Link deployment: The acoustic emission analog signal is transmitted using MHYV mining flame-retardant shielded cable, which fully preserves the details of the original 1MHz high-frequency signal and avoids the loss of micro-fragmentation features due to high-frequency signal attenuation; the apparent resistivity digital signal is transmitted using GYTA single-mode optical fiber, which completely resists the strong electromagnetic interference from underground mining machinery and eliminates the problem of digital signal packet loss.

[0045] Synchronization accuracy calibration: A nanosecond-level signal synchronization calibrator was installed at the signal inlet of the well edge gateway. A hardware timing chip was used to uniformly mark the timestamps of the acoustic emission signal and the apparent resistivity signal. Through multiple rounds of on-site synchronization tests, the spatiotemporal synchronization accuracy of the two types of data was finally controlled at 0.3ms, which is far higher than the preset requirement of 0.5ms. This provides a completely unified time reference for subsequent multi-field coupling simulations and completely avoids the distortion of simulation results caused by time misalignment of multi-source data.

[0046] Step 3: On-site deployment and model optimization of the edge intelligent parsing module

[0047] Hardware deployment: The edge intelligent parsing module is deployed in the intrinsically safe edge gateway for mining in the well dispatch room. The gateway is equipped with an 8-core CPU and 16GB of RAM, which fully meets the real-time computing power requirements of the lightweight model.

[0048] On-site optimization of the noise removal model: The pre-trained lightweight CNN model was imported into the edge gateway, and on-site noise samples of 12 types of mining machinery, such as coal mining machines, scraper conveyors, and tunneling machines, were imported into the mine. The model was then fine-tuned and optimized for the second time. On-site test results for 7 consecutive days showed that the model can automatically remove 93% of the strong noise interference from mining machinery, and the accuracy rate of effective acoustic emission event recognition reached 98.5%, which is far higher than the preset threshold of 97%.

[0049] Deployment of deep learning inversion model: The pre-trained lightweight U-Net deep learning inversion model is deployed to the edge gateway to replace the traditional finite difference iterative inversion algorithm. The inversion time of a single frame of 142 channels of apparent resistivity profile is only 1.7 seconds, which is much lower than the 10 minutes of the traditional algorithm, realizing millisecond-level real-time identification of overburden electrical anomalies.

[0050] Feature data upload configuration: Set data filtering rules in the edge gateway to upload only settlement-related abnormal feature data such as the three-dimensional positioning coordinates of acoustic emission events, the range of apparent resistivity anomalies, and the rate of change of resistivity in the anomalies to the cloud. The measured cloud data upload volume is reduced by 93% compared to the original full data, and there is no problem of cloud data congestion or delay.

[0051] Step 4: Model calibration and accelerated debugging of multi-field coupled inference module

[0052] Offline model calibration: A three-field coupled numerical model of mining-induced stress-resistivity-subsidence displacement was constructed in the cloud server. The mechanical and electrical parameters of the overlying rock in the mining area and the measured data of mining subsidence over the past 6 years were pre-imported to complete the offline calibration of the model. The initial simulation results and the measured goodness of fit with the historical subsidence data reached 96.2%, which is much higher than the preset threshold of 94%.

[0053] GPU parallel acceleration debugging: The parallel acceleration architecture is built using NVIDIA A10 GPU. The real-time acoustic emission positioning results obtained from edge analysis and the apparent resistivity anomaly area are used as dynamic boundary conditions input to the model to replace the traditional fixed boundary inference mode. On-site tests show that a full-space three-dimensional subsidence displacement cloud map covering an area of ​​240m×120m×500m can be output every 3 seconds, clearly showing the complete evolution process from deep overburden fracture to surface subsidence.

[0054] Step 5: Configure hierarchical rules and conduct linkage tests for the dynamic early warning and response module.

[0055] Early warning indicator system configuration: Construct a three-dimensional coupled early warning indicator system of "acoustic emission event rate - apparent resistivity change rate - subsidence displacement gradient", and set three levels of early warning thresholds:

[0056] Level 1 warning: Acoustic emission event rate exceeds 20 times / hour, apparent resistivity change rate exceeds 5%, subsidence displacement gradient exceeds 1mm / m. Once triggered, the message is pushed to the ground control center, reminding them to increase the monitoring scan frequency from every 5 minutes to every 2 minutes.

[0057] Warning level alert: When the acoustic emission event rate exceeds 90 times / hour, the apparent resistivity change rate exceeds 16%, and the subsidence displacement gradient exceeds 3.5mm / m, the personnel in the corresponding underground work area will be notified to suspend mining operations and check the stability of the overburden.

[0058] Danger level warning: When the acoustic emission event rate exceeds 210 times / hour, the apparent resistivity change rate exceeds 32%, and the subsidence displacement gradient exceeds 8.5 mm / m, the warning will be immediately pushed to the explosion-proof terminals of all personnel working underground and the emergency evacuation plan will be activated.

[0059] Spatiotemporal correlation configuration: The spatiotemporal coordinates of the full-space subsidence evolution cloud map with the mining engineering plan map of the mining area at a precision of 1:1000, as well as the surface village and water conveyance branch protection layers, are completed to ensure that the early warning information can be accurately located to the specific mining area and surface protection objects.

[0060] Linked push test: 12 simulated early warning push tests were completed on site. The average time from the generation of the early warning report to the push to the terminal of the downhole operator was only 1.2 seconds, which fully meets the timeliness requirements of disaster early warning.

[0061] Step 6: Verification of the full-process operating condition self-diagnosis mechanism

[0062] The system incorporates a self-diagnostic subunit for operating conditions in the acoustic-electric in-situ sensing module, edge intelligent analysis module, and dynamic early warning and handling module. This subunit monitors the power supply status, channel connectivity, electrode grounding resistance, and sensor sensitivity of each module in real time. The system simulates three fault scenarios of excessive electrode grounding resistance and two scenarios of abnormal channel connectivity. In each scenario, the system automatically generates equipment abnormality alarm information within one second and pushes it to the ground dispatch center. At the same time, it automatically marks the credibility coefficient of the monitoring data for the corresponding time period and feeds it back to the multi-field coupled inference module to complete the data weighting correction, thus completely avoiding the interference of abnormal monitoring data with the inference results.

[0063] During the continuous operation of this embodiment for 8 months, it successfully provided an early warning of a local subsidence risk caused by overburden fracture 1.8 hours in advance and an early warning of an abnormal development risk of water-conducting fracture zone 2.4 hours in advance. No subsidence or water inrush disasters occurred. Compared with traditional monitoring methods, the early warning time was increased by more than 9 times, which fully meets the real-time monitoring and safety control needs of subsidence in high-risk "three-under" mining faces.

[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method and system for real-time monitoring of mining-induced subsidence, characterized in that, include: The module includes an in-situ acoustic and electrical sensing module, an edge intelligent analysis module, a multi-field coupling deduction module, and a dynamic early warning and response module. The acoustic-electric in-situ sensing module is used to simultaneously deploy acoustic emission sensor arrays and high-density electrical resistivity monitoring electrodes in the overburden borehole, roadway surrounding rock, and surface subsidence-affected area of ​​the downhole working face. It non-invasively collects acoustic emission signals and apparent resistivity time-series data of micro-fractures inside the overburden, realizing in-situ continuous sensing of overburden fracture and water-conducting fracture development. The electrode deployment spacing of the acoustic-electric in-situ sensing module is dynamically adjusted according to the overburden lithology. In the weak rock layer area, the deployment spacing is increased to 2m to ensure the identification accuracy of low-resistivity fracture anomaly areas. The edge intelligent analysis module is used to aggregate acoustic emission signals and apparent resistivity data through the well edge gateway, and complete the local acoustic emission event location and apparent resistivity anomaly area identification locally, while only uploading the subsidence-related abnormal feature data to the cloud. The multi-field coupling simulation module is used to construct a numerical model of the three fields of mining stress, resistivity and subsidence displacement. It uses the real-time feature data after edge analysis as dynamic constraints and iteratively solves to output a full-space subsidence evolution cloud map from deep overburden to the surface. The dynamic early warning and response module is used to spatially correlate the full-space subsidence evolution cloud map with the mining engineering map and the surface protection object layer to generate a graded subsidence early warning report, which is simultaneously pushed to the ground dispatch center and the terminals of underground workers to achieve early detection and early response to subsidence disasters.

2. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The acoustic-electric in-situ sensing module includes an acoustic emission acquisition subunit and a high-density electrical resistivity subunit. The acoustic emission acquisition subunit adopts a 16-channel synchronous acquisition architecture with a sampling rate of 1MHz. It uses a time-difference positioning algorithm to control the three-dimensional positioning accuracy of overburden micro-fracture events within 0.3m, accurately capturing minute fracture signals inside the overburden. The high-density electrical resistivity subunit uses a Wenner quadrupole device to complete the apparent resistivity data acquisition, performing a full-domain scan every 5 minutes. It outputs the apparent resistivity profile of the overburden through a two-dimensional inversion algorithm, accurately identifying the dynamic development boundary of the water-conducting fracture zone.

3. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The edge intelligent analysis module includes an acoustic emission positioning subunit and a resistivity inversion subunit. The acoustic emission positioning subunit deploys a lightweight CNN model within the edge gateway to automatically eliminate strong noise interference generated by mining machinery, improving the recognition accuracy of effective acoustic emission events to over 97%. The resistivity inversion subunit uses a pre-trained deep learning inversion model to replace the traditional iterative inversion algorithm, reducing the inversion time of a single frame of apparent resistivity profile from 10 minutes to 2 seconds, and realizing real-time identification of overburden electrical anomalies.

4. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The multi-field coupled simulation module includes a model building subunit and a dynamic solution subunit. The model building subunit pre-imports the mechanical and electrical parameters of the overlying rock in the mining area and historical mining data to complete the offline calibration of the three-field coupled numerical model. The initial simulation results have a fitting degree of no less than 94% with the historical measured subsidence data. The dynamic solution subunit adopts a GPU parallel acceleration architecture and uses the acoustic emission positioning results obtained from edge analysis and the apparent resistivity anomaly area as dynamic boundary condition inputs. It outputs a full-space three-dimensional subsidence displacement cloud map every 3 seconds.

5. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The dynamic early warning and response module includes a risk grading subunit. The risk grading subunit constructs a three-dimensional coupled early warning index system of "acoustic emission event rate - apparent resistivity change rate - subsidence displacement gradient", which divides the early warning level into three levels: attention, warning, and danger, and matches the corresponding response plan for targeted push.

6. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The acoustic and electrical in-situ sensing module also integrates a micro temperature compensation module, which automatically eliminates the interference of downhole environmental temperature changes on apparent resistivity monitoring data and controls the systematic error of long-term monitoring data to within 2%.

7. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The acoustic and electrical in-situ sensing module and the edge intelligent analysis module transmit hybrid data through mining flame-retardant shielded cables and fiber optic redundant links. The acoustic emission signal is transmitted using cable analog transmission, while the apparent resistivity digital signal is transmitted using fiber optics. A signal synchronization calibrator is set at the edge gateway entrance to ensure that the spatiotemporal synchronization accuracy of the two types of data is not less than 0.5ms, providing a unified time reference for subsequent multi-field coupling simulations.

8. The method and system for real-time monitoring of mining subsidence according to claim 1, characterized in that: The acoustic-electric in-situ sensing module, edge intelligent analysis module, and dynamic early warning and handling module are all equipped with a working condition self-diagnosis subunit. The working condition self-diagnosis subunit monitors the power supply status, channel connectivity, electrode grounding resistance, and sensor sensitivity of each module in real time. When any parameter exceeds the preset threshold, it automatically generates equipment abnormality alarm information and pushes it to the ground dispatch center. At the same time, it marks the credibility coefficient of the monitoring data for the corresponding time period, which is used by the multi-field coupled inference module for weighted correction during iterative solution.

Images