A real-time monitoring method for surface subsidence in coal mining areas integrating multiple sensors

Abstract

The present invention discloses a method for real-time monitoring of surface subsidence in coal mining areas that integrates multiple sensors, and relates to the technical field of subsidence monitoring. The method comprises: constructing a heterogeneous sensor network and collecting raw monitoring data in real time, transmitting the data to a monitoring module via a wireless sensor network for preprocessing and unified processing of time and space benchmarks to obtain multi-source heterogeneous data; using a data fusion algorithm to deeply fuse the multi-source heterogeneous data to generate optimal estimated fusion information; calculating subsidence characteristic parameters based on the fusion information, and inputting the subsidence characteristic parameters into a subsidence trend prediction model to determine subsidence trend prediction data; comparing the multi-source heterogeneous data and the subsidence trend prediction data with multiple preset thresholds: when any value in the multi-source heterogeneous data exceeds a corresponding threshold interval, or when the subsidence trend prediction data exceeds a corresponding threshold interval, triggering an early warning message. The present invention can improve the accuracy of real-time monitoring of surface subsidence in coal mining areas.

Description

Technical Field

[0001] The present invention relates to the technical field of subsidence monitoring, and in particular to a multi-sensor integrated real-time monitoring method for surface subsidence in coal mining areas. Background Art

[0002] During coal mining, surface subsidence is one of the common and serious geological disasters. Accurate and real-time monitoring of surface subsidence in coal mining areas is crucial to ensuring safe production in mining areas, the safety of life and property of surrounding residents, and the stability of the ecological environment.

[0003] Traditional methods for monitoring surface subsidence in coal mining areas have numerous limitations. Single-sensor monitoring methods, such as those relying solely on GNSS receivers, can obtain the absolute three-dimensional coordinates and change data of monitoring points, but their limited spatial resolution makes it difficult to fully reflect the overall surface deformation. Ground-based InSAR equipment can obtain high-resolution surface deformation field information, but factors such as atmospheric errors and temporal decoherence affect absolute displacement accuracy. Inclinometers and crack meters can respectively monitor changes in surface tilt angle and crack width, but lack collaborative analysis with other data, making it difficult to fully understand surface subsidence.

[0004] Furthermore, existing monitoring systems lack effective fusion methods for data processing. Each sensor data is analyzed independently, failing to fully leverage the complementary advantages of multi-source data, resulting in inaccurate monitoring results. Therefore, a new real-time monitoring method for surface subsidence in coal mining areas is urgently needed. Summary of the Invention

[0005] The purpose of the present invention is to provide a real-time monitoring method for surface subsidence in coal mining areas integrating multiple sensors, aiming to solve or improve at least one of the above-mentioned technical problems.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A multi-sensor integrated real-time monitoring method for surface subsidence in coal mining areas, comprising:

[0008] Constructing a heterogeneous sensor network; the heterogeneous sensor network is composed of multiple types of sensor nodes deployed in the target coal mining area; the sensor nodes include GNSS receiver nodes, ground-based InSAR equipment, tilt sensor nodes, and crack meter nodes;

[0009] The raw monitoring data is collected in real time based on the heterogeneous sensor network and transmitted to the monitoring module via the wireless sensor network; the monitoring data includes the absolute three-dimensional coordinates of the monitoring points and coordinate change data, surface deformation field information, monitoring point tilt angle change data, and crack width change data;

[0010] In the monitoring module, the received raw monitoring data is pre-processed and subjected to unified time and space benchmark processing to obtain multi-source heterogeneous data;

[0011] A data fusion algorithm is used to deeply fuse the multi-source heterogeneous data to generate optimal estimated fusion information; the fusion information includes a three-dimensional deformation field, a tilt field, and a fracture development state map of the coal mining area surface; the fusion algorithm includes calibrating the ground-based InSAR deformation results using GNSS absolute displacement and spatially constraining or interpolating GNSS points using the ground-based InSAR high-spatial-resolution deformation field;

[0012] Calculating settlement characteristic parameters based on the fusion information, and inputting the settlement characteristic parameters into a settlement trend prediction model to determine settlement trend prediction data; the settlement trend prediction model is constructed based on an LSTM network and a self-attention mechanism;

[0013] The multi-source heterogeneous data and the settlement trend prediction data are compared with multi-level preset thresholds: when any value in the multi-source heterogeneous data exceeds the corresponding threshold interval, or when the settlement trend prediction data exceeds the corresponding threshold interval, an early warning message is triggered.

[0014] Optionally, the GNSS receiver nodes include a base station deployed in a stable area and a monitoring station deployed in the mining impact area; the ground-based InSAR equipment is deployed at a commanding height or a fixed platform in the mining area, and the scanning range covers the entire working face and the main area of ​​concern.

[0015] Optionally, the monitoring module performs preprocessing and time-space reference unified processing on the received raw monitoring data to obtain multi-source heterogeneous data, specifically including:

[0016] Preprocessing is first performed in the monitoring module: the absolute three-dimensional coordinates and coordinate change data of the monitoring points are solved to obtain a high-precision three-dimensional coordinate sequence; the surface deformation field information is subjected to interference processing, phase unwrapping, atmospheric correction, and geocoding to generate a surface deformation map; the tilt angle change data of the monitoring points and the crack width change data are filtered, denoised, and outliers are eliminated;

[0017] The preprocessed data are unified into the same spatiotemporal benchmark to obtain multi-source heterogeneous data.

[0018] Optionally, the adopting of a data fusion algorithm to perform deep fusion on the multi-source heterogeneous data to generate optimal estimated fusion information specifically includes:

[0019] Performing spatial correlation and gridding processing on the multi-source heterogeneous data to obtain spatial matching data;

[0020] Based on the spatial matching data, the deformation results of the ground-based InSAR are calibrated using the GNSS absolute displacement, and the GNSS points are sparsely spatially constrained or interpolated using the high spatial resolution deformation field of the ground-based InSAR to obtain the solved tilt value;

[0021] The calculated tilt amount is verified using the tilt sensor data, and the crack meter data is correlated with the calculated tilt amount. The correlated data are spatially and temporally fused using a Kalman filter or an extended Kalman filter to generate optimal estimated fusion information.

[0022] Optionally, the settlement characteristic parameters include settlement amount, settlement rate, settlement range, settlement center position, inclination and curvature.

[0023] Optionally, the training process of the settlement trend prediction model includes:

[0024] Obtaining training data; the training data includes historical multi-source heterogeneous data and corresponding prediction labels;

[0025] Build a pre-trained network based on LSTM network and self-attention mechanism;

[0026] The training data is input into the pre-trained network, and the goal is to minimize the loss between the network output and the predicted label. The gradient descent strategy is used for iterative training, and the trained network is determined as the sedimentation trend prediction model; wherein the training loss adopts the cross entropy loss function.

[0027] Optionally, the triggering method of the warning information includes at least one of the following: popping up a window on the monitoring interface of the manager and displaying the corresponding warning level and area, and sending SMS and / or email alarm information to designated personnel; automatically retrieving and displaying video surveillance images near the warning point; generating a warning report containing the warning location, time, indicator value, status and suggestions.

[0028] Optionally, the monitoring module adopts an edge computing gateway or a cloud monitoring platform.

[0029] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0030] The present invention discloses a method for real-time monitoring of surface subsidence in coal mining areas using integrated multiple sensors. The method comprises constructing a heterogeneous sensor network and collecting raw monitoring data in real time. The data is transmitted to a monitoring module via a wireless sensor network for preprocessing and unified processing based on a time-space benchmark to obtain multi-source heterogeneous data. A data fusion algorithm is used to deeply fuse the multi-source heterogeneous data to generate optimally estimated fusion information. Subsidence characteristic parameters are calculated based on the fusion information, and the subsidence characteristic parameters are input into a subsidence trend prediction model to determine subsidence trend prediction data. The multi-source heterogeneous data and the subsidence trend prediction data are compared with multiple preset thresholds. When any value in the multi-source heterogeneous data exceeds a corresponding threshold interval, or when the subsidence trend prediction data exceeds a corresponding threshold interval, an early warning message is triggered. The present invention can improve the accuracy of real-time monitoring of surface subsidence in coal mining areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 This is a flow chart of the real-time monitoring method for surface subsidence in coal mining areas integrating multiple sensors according to the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] The purpose of the present invention is to provide a real-time monitoring method for surface subsidence in coal mining areas integrating multiple sensors, aiming to solve or improve at least one of the above-mentioned technical problems.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] like Figure 1 As shown, the present invention provides a method for real-time monitoring of surface subsidence in coal mining areas integrating multiple sensors, comprising:

[0037] Step 100: Construct a heterogeneous sensor network; the heterogeneous sensor network is composed of multiple types of sensor nodes deployed in the target coal mining area; the sensor nodes include GNSS receiver nodes, ground-based InSAR equipment, tilt sensor nodes and crack meter nodes.

[0038] Step 200: Collecting raw monitoring data in real time based on the heterogeneous sensor network and transmitting it to the monitoring module through the wireless sensor network; the monitoring data includes the absolute three-dimensional coordinates and coordinate change data of the monitoring point, surface deformation field information, monitoring point tilt angle change data and crack width change data.

[0039] Step 300: In the monitoring module, the received raw monitoring data is pre-processed and subjected to unified time and space benchmark processing to obtain multi-source heterogeneous data.

[0040] Step 400: Deeply fuse the multi-source heterogeneous data using a data fusion algorithm to generate optimal estimated fusion information; the fusion information includes a three-dimensional deformation field, a tilt field, and a fracture development state map of the coal mining area surface; the fusion algorithm includes calibrating the ground-based InSAR deformation results using GNSS absolute displacement, and spatially constraining or interpolating GNSS points using the ground-based InSAR high-spatial-resolution deformation field.

[0041] Step 500: Calculate settlement characteristic parameters based on the fusion information, and input the settlement characteristic parameters into a settlement trend prediction model to determine settlement trend prediction data; the settlement trend prediction model is constructed based on an LSTM network and a self-attention mechanism.

[0042] Step 600: Compare the multi-source heterogeneous data and the settlement trend prediction data with multi-level preset thresholds: when any value in the multi-source heterogeneous data exceeds the corresponding threshold interval, or when the settlement trend prediction data exceeds the corresponding threshold interval, trigger an early warning message.

[0043] As a specific implementation example, a fully mechanized coal face at a coal mine in North China is used as an example. The mining depth is 350 meters, the strike length is 2.1 kilometers, and the dip width is 280 meters. The surface is covered with villages, high-voltage towers, and provincial roads, necessitating real-time subsidence monitoring to prevent disasters.

[0044] First, a multi-sensor network was deployed. GNSS receiver nodes included two base stations and 18 monitoring stations. Each base station was located on stable bedrock outside the mining area. Monitoring stations were located along the main section of the working face, around villages, and along roads. They used a BeiDou-3 / GPS dual-frequency system with a sampling rate of 1 Hz and a positioning accuracy of ±2 mm (RTK mode). A ground-based InSAR device was installed at the top of the waste rock mass (+85 m above sea level), covering the entire working face and village area. It used the Ku-band, had a 30-minute scanning cycle, and a resolution of 1 m x 1 m. Tilt sensor nodes were installed on 12 GNSS monitoring piers and 8 village house foundations. They used dual-axis sensors with a range of ±10°, an accuracy of 0.001°, and a sampling rate of 1 time / minute. Crack gauge nodes were deployed on 10 cracks in village house walls and 5 expansion joints on roads. They had a range of 50 mm, an accuracy of 0.1 mm, and a sampling rate of 1 time / minute. An edge computing gateway or cloud-based monitoring platform served as the monitoring module.

[0045] Second, GNSS monitoring stations transmit raw observations (carrier phase and pseudorange) in real time to the base station. Network RTK calculations generate a 3D coordinate change sequence. Ground-based InSAR scans occur every 30 minutes, with raw echo data transmitted directly to the monitoring module via the 4G network. Inclinometer / crackmeter data is aggregated via a LoRa gateway and uploaded to the monitoring module every minute.

[0046] High-precision GNSS receiver nodes are used to obtain the absolute three-dimensional coordinates (longitude, latitude, and elevation) of monitoring points and their changes, providing a high-precision, real-time absolute displacement benchmark. Ground-based InSAR equipment conducts periodic (minute- to hourly) radar scans to acquire large-scale, high-spatial-resolution surface deformation (line-of-sight displacement) information. Tilt sensor nodes measure the tilt angle and rate of change of monitoring points. Crack gauge nodes measure changes in the opening and closing width of cracks. All nodes are equipped with wireless communication modules (such as 4G / 5G, LoRa, and NB-IoT) and positioning modules (for node calibration).

[0047] Collected data is transmitted in real time or near real time via a wireless sensor network, utilizing star, mesh, or hybrid topologies, to an edge computing gateway deployed at the mine site or directly to a cloud-based monitoring platform. Priority is given to ensuring real-time performance (in seconds or minutes) of key variables such as GNSS, inclination, and crack meters.

[0048] The collected data is then fused. Preprocessing of the received raw data occurs on the edge gateway or cloud platform. This includes solving the GNSS data to obtain precise point positioning (PPP) or differential positioning (RTK / PPK), further acquiring a high-precision 3D coordinate sequence, and performing quality checks (multipath and cycle slip repair). Ground-based InSAR data undergoes interferometric processing, phase unwrapping, atmospheric correction, and geocoding to generate surface deformation maps (displacement contour maps). Dip and crack data undergo filtering, denoising, unit conversion, and outlier removal. Other relevant data undergoes format conversion and validity checks. The preprocessed data is then unified to the same spatiotemporal reference (CGCS2000 coordinate system or UTC).

[0049] Furthermore, we conduct deep fusion of multi-source heterogeneous data. This mainly includes the following steps:

[0050] Spatial matching: Spatial correlation and gridding of monitoring results from different sensors (point GNSS / inclinometer / crackmeter, surface InSAR).

[0051] Data complementation and enhancement: GNSS absolute displacement is used to calibrate the deformation results of ground-based InSAR (resolving phase unwrapping ambiguities and atmospheric residual errors), significantly improving the absolute accuracy and reliability of InSAR.

[0052] The high spatial resolution deformation field of ground-based InSAR is used to spatially interpolate or constrain sparse GNSS points to generate a more refined and continuous subsidence basin model (including vertical subsidence and horizontal displacement components).

[0053] Use tilt sensor data to calculate the horizontal displacement gradient of a local area (combined with baseline information provided by GNSS / InSAR), or verify the tilt calculated by GNSS / InSAR.

[0054] The crack meter data is used to directly reflect the degree of local damage and is correlated with the settlement / tilt data to determine the structural stability.

[0055] Fusion Algorithm: Using a Kalman filter (KF) or extended Kalman filter (EKF) fusion algorithm, we perform spatiotemporal fusion of heterogeneous data from multiple sources to generate an optimal estimate of the mining area's three-dimensional surface deformation field (vertical settlement, east-west horizontal displacement, and north-south horizontal displacement), tilt field, and fracture development status map. The fusion process weights the accuracy, reliability, and spatiotemporal characteristics (sampling rate, resolution) of different sensors.

[0056] Finally, conduct settlement analysis and early warning:

[0057] On the cloud-based monitoring platform, the fused 3D deformation field, tilt field, and crack information are displayed in real-time via map overlays, charts, and 3D models. Parameters such as settlement, rate, extent, center of settlement, tilt, and curvature are calculated in real time for key areas. Using machine learning models, based on historical fused data and current mining progress, short-term predictions of settlement trends at key points or areas are made. Multiple warning thresholds (such as those for settlement rate, cumulative settlement, tilt angle, and crack width change) can be set. When real-time monitoring data or predicted results exceed thresholds, or when abnormal accelerated deformation is detected, the system automatically triggers multiple warning levels (prompts, warnings, and alarms). Warning information is delivered to relevant management personnel and responsible individuals via platform pop-ups, text messages, emails, and audio and visual alarms. Video surveillance footage near warning points is automatically retrieved to assist in manual verification of on-site conditions. Warning reports are generated, including the warning location, time, trigger indicator values, screenshots of the current status, and recommended measures.

[0058] As a more specific early warning process:

[0059] The platform displays the fused surface settlement contour map, horizontal displacement vector map, and key point time series curves (settlement amount, rate, inclination angle, and crack width) in real time.

[0060] For key areas such as above the center of the working face, village areas, roads, etc., the maximum settlement and average settlement rate (in the past 1 hour, 6 hours, and 24 hours) are calculated in real time.

[0061] Setting warning thresholds (example):

[0062] Warning level (yellow): Sedimentation rate > 5 mm / day or cumulative sedimentation > 50 mm or tilt angle change > 0.1°.

[0063] Warning level (orange): Settlement rate > 10 mm / day or cumulative settlement > 100 mm or inclination angle change > 0.2° or crack widening > 2 mm.

[0064] Alert level (red): Settlement rate > 20 mm / day or accelerated deformation occurs (such as the rate doubles) or the inclination angle changes by > 0.5° or the crack widens sharply (> 5 mm) or it is predicted that the settlement in the next 24 hours will exceed the structural safety threshold.

[0065] When the monitoring value in a certain area triggers the threshold, the system automatically:

[0066] The warning area is highlighted on the platform map, and a warning window pops up. Warning text messages and emails are sent to the mine dispatch room, the head of the geodetic department, and the mobile phones of the relevant team leaders. The video surveillance screen near the warning point is automatically switched to the main screen for viewing, and a warning report is generated for archiving.

[0067] All the above raw data, pre-processed data, fusion results, and warning records are stored in the cloud database, supporting historical data query, retrospective analysis, and report generation.

[0068] Therefore, the present invention has the following beneficial effects:

[0069] Multi-source complementarity and comprehensive information: By integrating data from multiple sensors such as GNSS (high-precision absolute displacement), ground-based InSAR (large-scale high-resolution deformation field), inclination (local tilt), and crackmeter (local damage), we can obtain multi-dimensional, full-factor information such as three-dimensional displacement (vertical and horizontal), tilt, and cracking of surface subsidence, and more comprehensively depict the subsidence process and damage morphology.

[0070] High precision and strong reliability: InSAR is calibrated using the GNSS absolute reference to significantly improve the absolute accuracy of the deformation field; multi-source data are mutually verified to reduce the risk of single sensor error or failure, and improve the reliability and robustness of monitoring results.

[0071] Strong real-time performance: Through high-frequency data collection (especially GNSS, inclinometers, and crack meters) and wireless real-time transmission, combined with edge computing for preliminary processing, key indicators (such as point displacement and tilt) can be monitored from seconds to minutes. Ground-based InSAR can also be updated at the hourly level, meeting the real-time monitoring needs of dynamic settlement in coal mining.

[0072] High degree of automation: From data collection, transmission, processing, fusion, analysis to early warning, the entire process is automated, greatly reducing manual intervention, lowering operation and maintenance costs, and improving efficiency.

[0073] Wide spatial coverage and high resolution: Ground-based InSAR provides wide-area coverage, while GNSS and point sensors provide high-precision control of key points. After fusion, they can generate a detailed sedimentation field with a spatial resolution far higher than that of a single GNSS network.

[0074] Timely and effective early warning: Based on multi-indicator fusion analysis and trend prediction, intelligent and multi-level early warning is realized. Early warning information is pushed quickly and through various channels, and linked to video confirmation, which significantly improves disaster response speed and disposal efficiency.

[0075] Provide support for scientific decision-making: Provide real-time, comprehensive, and high-precision settlement data to serve the optimization of mining plans, subsidence area management planning, building protection, disaster risk assessment and emergency plan formulation.

[0076] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0077] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for real-time monitoring of surface subsidence in coal mining areas using integrated multi-sensor technology, characterized in that: include: Building heterogeneous sensor networks; The heterogeneous sensor network is composed of multiple types of sensor nodes deployed in the target coal mining area; the sensor nodes include GNSS receiver nodes, ground-based InSAR equipment, tilt sensor nodes and crack meter nodes; The raw monitoring data is collected in real time based on the heterogeneous sensor network and transmitted to the monitoring module via the wireless sensor network; the monitoring data includes the absolute three-dimensional coordinates of the monitoring points and coordinate change data, surface deformation field information, monitoring point tilt angle change data, and crack width change data; In the monitoring module, the received raw monitoring data is pre-processed and subjected to unified time and space benchmark processing to obtain multi-source heterogeneous data; Using a data fusion algorithm to deeply fuse the multi-source heterogeneous data to generate optimal estimated fusion information; The fused information includes the three-dimensional deformation field, tilt field, and fracture development state map of the coal mining area surface; the fusion algorithm includes using GNSS absolute displacement to calibrate the ground-based InSAR deformation results, and using the ground-based InSAR high-spatial-resolution deformation field to spatially constrain or interpolate GNSS points; Calculating settlement characteristic parameters based on the fusion information, and inputting the settlement characteristic parameters into a settlement trend prediction model to determine settlement trend prediction data; the settlement trend prediction model is constructed based on an LSTM network and a self-attention mechanism; Comparing the multi-source heterogeneous data and the settlement trend prediction data with multiple preset thresholds: triggering an early warning message when any value in the multi-source heterogeneous data exceeds a corresponding threshold interval, or when the settlement trend prediction data exceeds a corresponding threshold interval; The monitoring module performs preprocessing and time-space benchmarking on the received raw monitoring data to obtain multi-source heterogeneous data, specifically including: Preprocessing is first performed in the monitoring module: the absolute three-dimensional coordinates and coordinate change data of the monitoring points are solved to obtain a high-precision three-dimensional coordinate sequence; the surface deformation field information is subjected to interference processing, phase unwrapping, atmospheric correction and geocoding to generate a surface deformation map; the tilt angle change data of the monitoring points and the crack width change data are filtered, denoised and outliers are eliminated; Unify the preprocessed data to the same spatiotemporal benchmark to obtain multi-source heterogeneous data; The method of using a data fusion algorithm to deeply fuse the multi-source heterogeneous data to generate optimal estimated fusion information specifically includes: Performing spatial correlation and gridding processing on the multi-source heterogeneous data to obtain spatial matching data; Based on the spatial matching data, the deformation results of the ground-based InSAR are calibrated using the GNSS absolute displacement, and the GNSS points are sparsely spatially constrained or interpolated using the high spatial resolution deformation field of the ground-based InSAR to obtain the solved tilt value; The calculated tilt amount is verified using the tilt sensor data, and the crack meter data is correlated with the calculated tilt amount. The correlated data are spatially and temporally fused using a Kalman filter or an extended Kalman filter to generate optimal estimated fusion information.

2. The method for real-time monitoring of surface subsidence in coal mining areas with integrated multi-sensor technology according to claim 1, characterized in that: The GNSS receiver nodes include base stations located in stable areas and monitoring stations located in mining-affected areas. The ground-based InSAR equipment is deployed at commanding heights or fixed platforms in the mining area, with the scanning range covering the entire working face and key areas of concern.

3. The method for real-time monitoring of surface subsidence in coal mining areas with integrated multi-sensor technology according to claim 1, characterized in that: The settlement characteristic parameters include settlement amount, settlement rate, settlement range, settlement center position, inclination and curvature.

4. The method for real-time monitoring of surface subsidence in coal mining areas with integrated multi-sensor technology according to claim 1, characterized in that: The training process of the settlement trend prediction model includes: Acquire training data; the training data includes historical multi-source heterogeneous data and corresponding prediction labels; Build a pre-trained network based on LSTM network and self-attention mechanism; The training data is input into the pre-trained network, and the goal is to minimize the loss between the network output and the predicted label. The gradient descent strategy is used for iterative training, and the trained network is determined as the sedimentation trend prediction model; wherein the training loss adopts the cross entropy loss function.

5. The method for real-time monitoring of surface subsidence in coal mining areas with integrated multi-sensor technology according to claim 1, characterized in that: The triggering method of the warning information includes at least one of the following: popping up a window on the manager's monitoring interface and displaying the corresponding warning level and area, and sending SMS and / or email alarm information to designated personnel; automatically retrieving and displaying video surveillance images near the warning point; generating a warning report containing the warning location, time, indicator value, status and suggestions.

6. The method for real-time monitoring of surface subsidence in coal mining areas with integrated multi-sensor technology according to claim 1, characterized in that: The monitoring module adopts an edge computing gateway or a cloud monitoring platform.

Images