Detection system for water level data and geological data of weakly cemented coal seam

Abstract

The invention discloses a detection system for water level data and geological data of a weakly cemented coal seam, and belongs to the technical field of mine safety monitoring. Comprising a hardware layer and an algorithm layer, the hardware layer constructs a water surface-water body-water bottom three-in-one dynamic monitoring system, and the algorithm layer fuses multi-source data through a physical information neural network to realize global water bottom DEM inversion and dynamic early warning. And the physical information neural network inverts the global underwater DEM in a data driving item and physical driving item joint optimization mode by performing space-time alignment on the sonar point cloud, the unmanned aerial vehicle water level data and the optical fiber strain gradient and then performing weighted fusion. Sonar, optical fiber and water level data are fused through a PINN model, the crack recognition capacity is improved, and the side-scan sonar can detect cracks with the width larger than or equal to 2 cm; second-level data acquisition is achieved through the optical fiber sensor, the water inrush early warning response time is shortened, the response speed is increased, and the sonar sounding error is still controlled within 5 cm. The system robustness is increased, and the double-ring optical fiber network reduces the single-point fault influence range.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological engineering monitoring, more specifically, to a weakly cemented coal seam water level data and geological data detection system. BACKGROUND

[0002] Weakly cemented coal seam refers to a coal seam with low cementation strength between coal body and surrounding rock, loose structure and easy breakage. Due to low content of cementing materials (such as clay minerals and carbonates) or weak cementation, the coal seam has low compressive strength, and problems such as roof collapse and coal wall spalling are prone to occur during mining, especially when mining under aquifer, which can easily cause water inrush, ground subsidence and other disasters. Water bottom topographic deformation monitoring faces various challenges.

[0003] The prior art mainly includes: 1. Optical fiber sensor array monitoring technology: by arranging optical fiber curvature sensors into a strip array and fixing them in a vulcanized rubber base, the water bottom topographic deformation is analyzed by using the difference in optical fiber signal attenuation, which is suitable for deformation monitoring of underwater topography, tunnel surrounding rock and the like, and the precision can reach millimeter level.

[0004] 2. Satellite positioning technology (such as Beidou GNSS): by using high-precision satellite positioning (differential RTK / PPP technology) to obtain three-dimensional coordinate changes of the monitoring point, it is commonly used for horizontal displacement and settlement monitoring of structures such as dams and ports, and the static precision can reach millimeter level.

[0005] 3. Synthetic aperture radar interferometry (InSAR): satellite radar images are used to monitor ground deformation, and differential technology (D-InSAR) is used to improve precision.

[0006] 4. Pore pressure monitoring technology: by measuring the pore pressure changes of seabed sediments through a probe, the stability of rock-soil mass is indirectly reflected.

[0007] However, the existing technology has obvious problems and deficiencies: Contradiction between precision and coverage range: high-precision technologies (such as optical fiber and acoustic nodes) usually require dense deployment of sensors, which is costly and has small coverage range; large-scale technologies (such as InSAR and GNSS) are limited by underwater signal attenuation and are difficult to penetrate water to directly measure seabed topography.

[0008] Insufficient environmental adaptability: satellite positioning technology (GNSS) has serious signal attenuation in deep water, and traditional sonar has a significant decrease in depth measurement accuracy in high turbidity or noisy environments.

[0009] Poor real-time performance and dynamic response capability: traditional monitoring methods have long data update cycles, and disaster warning such as water inrush lags behind, which makes it difficult to meet the real-time monitoring needs of weakly cemented coal seam safety mining.

[0010] In view of this, the present application is proposed to solve the above technical problems. SUMMARY

[0011] The present application aims to provide a weakly cemented coal seam water level data and geological data detection system to solve the technical problems.

[0012] To achieve the above object, the present application provides the following technical solutions: A river embankment coal mining monitoring system based on unmanned ship and GNSS fusion, comprising: Hardware layer and algorithm layer, the hardware layer constructs a "water surface-water body-water bottom" three-in-one dynamic monitoring system, the algorithm layer realizes global water bottom DEM inversion and dynamic early warning by fusing multi-source data through physical information neural network, and the physical information neural network inverses global water bottom DEM in a data driving item and physical driving item joint optimization manner after weighting fusion of sonar point cloud, unmanned aerial vehicle water level data and optical fiber strain gradient through space-time alignment; The hardware layer comprises: An aerial unit, which is equipped with a millimeter wave radar and a multispectral camera on an unmanned aerial vehicle, is used for monitoring water surface elevation and water body leakage area; An underwater unit, which comprises a multi-beam sonar array, a pressure-resistant optical fiber sensor network and a floating water quality monitoring device, is used for water bottom topography scanning, rock stratum strain monitoring and water quality parameter acquisition; A communication relay unit realizes data aggregation, encryption and transmission through a water surface buoy and a Beidou satellite; The algorithm layer comprises: A data preprocessing module, which performs motion compensation and splicing registration on sonar point cloud and temperature drift correction on optical fiber strain data; A physical information neural network model, which models sonar point cloud, unmanned aerial vehicle water level data and optical fiber strain gradient uniformly, and inverses global water bottom DEM by constraining rock stratum continuity rules; A dynamic early warning module, which realizes water inrush risk early warning based on threshold judgment and GIS visualization.

[0013] Further, the millimeter wave radar of the aerial unit transmits high-frequency electromagnetic waves to calculate water level elevation, and the precision reaches centimeter level; the multispectral camera identifies water body leakage area through thermal infrared and visible light feature analysis.

[0014] Further, the multi-beam sonar array of the underwater unit adopts adaptive hopping technology for emission frequency, avoids 50-200Hz mining noise frequency band in real time, improves depth measurement precision by combining wavelet packet decomposition and motion compensation algorithm, and the point cloud density reaches one million points per second; the sonar array removes high-frequency noise by using wavelet packet decomposition first, and then compensates sonar motion track by EKF algorithm, so that the depth measurement precision is improved to ±5cm.

[0015] Further, the pressure-resistant optical fiber sensor is coated with a nanocolloid coating of silica nanoparticles, and the tensile strength is increased to 200 MPa. The double-loop network topology is used, and when a single point breaks, the data is transmitted through the backup path.

[0016] Further, in the communication relay unit, the water surface buoy parses the underwater sonar point cloud, optical fiber data and water quality information through the underwater acoustic communication protocol, and transmits the compressed and encrypted data to the ground control center in batches through the Beidou short message.

[0017] Further, in the data preprocessing module, the sonar motion trajectory is reconstructed based on IMU data through extended Kalman filtering, and multi-frame point cloud is registered by using iterative closest point algorithm; the temperature-strain transfer function model is trained through LSTM network, and the optical fiber strain gradient is corrected in real time.

[0018] Further, the loss function of the PINN model includes a data-driven term and a physics-driven term, and the weight is dynamically adjusted by an adaptive balancing algorithm. The course learning strategy is used to fit the static terrain first and then introduce dynamic data.

[0019] Further, the dynamic warning module sets the strain gradient exceeding the historical mean value 3σ as the warning threshold, and combines the GIS platform to visualize the DEM change and the rock stability evaluation result, so that the warning response time is ≤8 minutes.

[0020] Further, it also includes a turbidity compensation model, which inverses the water body suspended matter concentration (NTU) through sonar echo intensity, establishes a "turbidity-sound wave attenuation" relationship to correct the depth data, and the depth error is ≤5cm in the environment with turbidity >150NTU.

[0021] Further, it also includes a dynamic deployment and AUV cooperation strategy: when the optical fiber sensor detects that the strain rate >0.1% / min, the AUV carrying the side-scan sonar is automatically dispatched to the abnormal area, the scanning priority is dynamically adjusted according to the strain gradient, and the optimal scanning path is generated by combining the flow velocity and the sonar coverage.

[0022] Compared with the prior art, the beneficial effects of the present application are: (1) The monitoring accuracy is significantly improved, the underwater terrain resolution is improved, the multi-beam sonar (MBES) and the AUV side-scan sonar (SSS) work cooperatively, the underwater DEM resolution is improved from 1-5m of the traditional sonar to 0.1m, and the local key area (such as the fracture zone) can reach 1cm; the settlement inversion error is reduced, the PINN model is used to fuse the sonar, optical fiber and water level data, and the settlement inversion error is ≤5% (the traditional indirect model error is 10-20%); the fracture identification capability is improved, the side-scan sonar can detect the fracture with a width ≥2cm (the traditional sonar can only identify the fracture with a width ≥10cm).

[0023] (2) Real-time and response efficiency breakthrough, data update frequency upgrade, optical fiber sensor realizes second-level data acquisition, sonar dynamic scanning period is shortened to 10 minutes / time (traditional sonar ship needs several hours); water inrush early warning response time is reduced, the average time from anomaly detection to early warning publication is ≤8 minutes (traditional manual inspection needs ≥1 hour), the response speed is increased by 80%.

[0024] (3) Anti-interference and reliability enhancement, complex environment adaptability enhancement, in the environment of turbidity >150 NTU or noise >90 dB, the sonar depth error is still controlled within 5 cm; system robustness is increased, single point failure influence range is reduced by 70% by double ring optical fiber network, data interruption rate <1%; long-term stability is enhanced, nano-coated optical fiber works continuously for ≥6 months without performance attenuation in humid and high-stress environment.

[0025] (4) Comprehensive cost and energy efficiency optimization, equipment power consumption is reduced, lightweight sonar node (power consumption <20W) and edge computing optimization make the overall system energy consumption reduced by 30%; operation and maintenance cost is reduced, AUV autonomous inspection reduces manual intervention, annual operation and maintenance cost is reduced by 50%. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments.

[0027] A weakly cemented coal seam water level data and geological data detection system, comprising: A hardware layer and an algorithm layer, the hardware layer constructs a "water surface-water body-water bottom" three-in-one dynamic monitoring system, the algorithm layer realizes global water bottom DEM inversion and dynamic early warning by fusing multi-source data through a physical information neural network, and the physical information neural network inverses the global water bottom DEM in a manner of joint optimization of data driving items and physical driving items after time and space alignment and weighted fusion of sonar point cloud, unmanned aerial vehicle water level data and optical fiber strain gradient; The hardware layer comprises: An aerial unit, which is equipped with a millimeter wave radar and a multispectral camera by an unmanned aerial vehicle, is used for monitoring water surface elevation and water body leakage area; the millimeter wave radar emits high-frequency electromagnetic waves in a frequency band of 24-77 GHz, and the water surface elevation is calculated by a formula d = c wherein c is the speed of light, is the frequency difference, and B is the sweep bandwidth, For the sweep frequency period, the height error is corrected by combining the unmanned aerial vehicle GPS positioning data (accuracy ± 10 cm), and the accuracy reaches centimeter level; the multispectral camera adopts a U-Net deep learning model, based on thermal infrared (8-14 μm) and visible light (400-700 nm) band data, by identifying water temperature anomaly areas and spectral reflectance mutation points, analyzing and identifying water leakage areas, and the accuracy is ≥92%.

[0028] The underwater unit includes a multi-beam sonar array, a pressure-resistant optical fiber sensor network, and a floating water quality monitoring device, which is used for seabed topography scanning, rock layer strain monitoring, and water quality parameter collection; the multi-beam sonar array uses an adaptive hopping technology controlled by a pseudo-random sequence for emission frequency, scans the environmental noise spectrum every 5 seconds, avoids the 50-200 Hz mining noise frequency band in real time, automatically switches to the 300-400 Hz "clean frequency band" if the energy of the frequency band exceeds the threshold (-80 dBm), and monitors the interference of the new frequency band in real time through fast Fourier transform (FFT), ensuring that the signal-to-noise ratio is > 20 dB, and the point cloud density is up to one million points per second; the surface of the pressure-resistant optical fiber sensor is coated with a nanometer colloidal coating of silica nanoparticles (particle size 50-100 nm) through a chemical vapor deposition (CVD) process, the silica nanoparticles are mixed with silane coupling agents at 300°C and then deposited on the surface of the optical fiber, the coating thickness is controlled at 1-2 μm, the tensile strength is improved to 200 MPa, and there is no corrosion after 72 hours of salt spray testing (ISO9227 standard), a double-ring network topology is adopted, each optical fiber node is equipped with 2 optical couplers, forming a main ring and a standby ring, when a single point fails, the optical signal automatically switches to the standby path, the switching time is < 50 ms, and the data transmission delay increases by < 10%.

[0029] The communication relay unit realizes data aggregation, encryption, and transmission through a water surface buoy and a Beidou satellite; The algorithm layer includes: The data preprocessing module performs motion compensation and splicing registration on the sonar point cloud, and temperature drift correction on the optical fiber strain data; based on the IMU data, the sonar motion trajectory is reconstructed by the extended Kalman filter (EKF) of X(t)= T (position, velocity, attitude angle, and angular velocity), the multi-frame point cloud is registered by the iterative closest point (ICP) algorithm with the point-to-plane distance measurement, the iteration termination condition is that the mean square error (MSE) is < 0.01 m² or the iteration number is ≥ 20 times, and the initial transformation matrix is calculated by singular value decomposition (SVD); the temperature-strain transfer function model is trained by a LSTM network containing 2 layers of 128 neurons, an activation function ReLU, a loss function Huber loss, an optimizer Adam (learning rate 0.001), and 50 rounds of training based on 100,000 groups of temperature (-10℃-50℃) and strain (0-1000με) samples, and the optical fiber strain gradient is corrected in real time.

[0030] The Physical Information Neural Network (PINN) model unifies the modeling of sonar point clouds, UAV water level data, and fiber optic strain gradients to constrain the continuity of rock strata and invert the global seabed DEM. The loss function includes a data-driven term (MSE between measured elevation and predicted value) and the elasticity equilibrium equation. ( For stress tensor, (for physical forces) and continuity equations =0 ( The physical driving term (partial differential equation residual) for the velocity field is weighted by a dynamic weight update formula. (k=0.1 is the adjustment coefficient, L) data L physics The adaptive balancing algorithm is dynamically adjusted for both data and physical loss. A phased learning strategy is adopted. In the first phase, only 1,000 sets of static terrain sonar point cloud data are used for training for 10 rounds. In the second phase, dynamic data (including water level and strain) are added for another 40 rounds of training. The learning rate is reduced from 0.001 to 0.0001 to retrieve the full-domain underwater DEM.

[0031] The dynamic early warning module uses threshold judgment and GIS visualization to provide early warning of water inrush risk. It sets the warning threshold as a strain gradient exceeding the historical average by 3σ. WebGL technology is used to render DEM changes in real time, with displacement represented by color gradients (red > 10cm). Topographic data is loaded via ArcGIS API, with a refresh rate of 1 time per minute. Based on the Mohr-Coulomb criterion, a safety factor F is calculated. s = (c represents cohesion,) For normal stress, It is the internal friction angle. To assess rock stratum stability (for shear stress), when F s An orange alert is triggered when the value is less than 1.2. The alert response time is ≤8 minutes, based on the DEM changes visualized on the GIS platform.

[0032] 1. Anti-interference sonar design and turbidity compensation algorithm: Adaptive frequency hopping technology avoids mining noise frequencies in real time, automatically switching to the 300-400Hz "clean band" during sudden noise events. The turbidity compensation model uses the least squares method to fit the error between the sonar echo intensity I and the theoretical attenuation, iteratively solving for the NTU value (initial value set at 50 NTU, convergence condition...). The concentration of suspended solids in water is inverted using empirical formulas. ( The "turbidity-acoustic attenuation" relationship is established for the attenuation coefficient a=0.005 dB / (m·NTU) and b=0.1 dB / m to correct the sonar propagation time The depth error is ≤5 cm in a turbidity>150 NTU environment.

[0033] 2. Dynamic deployment and AUV cooperation strategy: When the optical fiber sensor detects a strain rate>0.1% / min, the AUV carrying a side-scan sonar is automatically dispatched to the abnormal area. The improved Dijkstra algorithm is used, taking the flow velocity (unit: m / s) as the edge weight W= , and the target point is the area with the maximum strain gradient. The path search time is <10 seconds. According to the strain gradient, the priority is calculated by the formula P= · ( is the strain rate, is the distance from the AUV to the abnormal point) and the scanning order is arranged in descending order of P value. The optimal scanning path is generated by combining the flow velocity and the sonar coverage.

[0034] The millimeter wave radar of the air unit transmits high-frequency electromagnetic waves to calculate the water level elevation with an accuracy of centimeters. The multispectral camera identifies water leakage areas through thermal infrared and visible light feature analysis.

[0035] The multi-beam sonar array of the underwater unit uses adaptive hopping technology for transmission frequency, which can avoid the 50-200Hz mining noise frequency band in real time. The wavelet packet decomposition and motion compensation algorithm are used to improve the depth measurement accuracy, with a point cloud density of one million points per second. The sonar array first removes high-frequency noise using wavelet packet decomposition, and then compensates the sonar motion trajectory using EKF algorithm to improve the depth measurement accuracy to ±5 cm.

[0036] The pressure-resistant optical fiber sensor is coated with a nanometer colloidal coating of silica nanoparticles with a particle size of 50-100 nm, and the tensile strength is improved to 200 MPa. The double-ring network topology is used, and when a single point breaks, the data is transmitted through the backup path.

[0037] In the communication relay unit, the surface buoy parses the underwater sonar point cloud, optical fiber data, and water quality information through the underwater acoustic communication protocol, and transmits them to the ground control center in batches after compression and encryption through the Beidou short message.

[0038] The communication protocol uses CMUP (underwater acoustic communication unified protocol), and the data frame format includes the header (device ID, frame number), data segment (sonar point cloud / strain data), and check bit (CRC-32). The compression algorithm uses the Zstandard compression algorithm, with a compression ratio of 5:1, which meets the length limit of the Beidou short message (≤120 bytes). Encryption algorithm: SM4 national encryption algorithm is used, and the key is pre-assigned by the ground control center through the quantum key distribution system; Batch strategy: large amount of sonar point cloud is split by 1024 points / batch, transmitted to the ground control center through TCP / IP over Short Message protocol, and the retransmission mechanism adopts ARQ automatic request for retransmission (timeout time 5s).

[0039] In the data preprocessing module, the sonar motion trajectory is reconstructed based on IMU data through extended Kalman filter (EKF), and multi-frame point cloud is registered by using iterative closest point (ICP) algorithm; the temperature-strain transfer function model is trained through LSTM network, and the optical fiber strain gradient is corrected in real time.

[0040] Among them, the input and output: the input is the temperature value (unit: ℃) of 10 continuous time steps, and the output is the strain correction value (unit: με) corresponding to the time step; Normalization: Min-Max normalization is used to map the temperature range [-10, 50] to [0, 1]; Evaluation index: the root mean square error (RMSE) and the determination coefficient (R²) are used to evaluate the model performance, and the RMSE of the test set is required to be less than 0.5 με and the R² is required to be greater than 0.95.

[0041] The loss function of the PINN model includes a data-driven term (MSE of measured elevation and predicted value) and a physics-driven term (partial differential equation residual), and the weight is dynamically adjusted through an adaptive balancing algorithm. The course learning strategy is used to fit the static terrain first and then introduce dynamic data.

[0042] The dynamic warning module sets the strain gradient exceeding the historical mean value 3σ as the warning threshold, and combines the GIS platform to visualize the DEM change and the rock stability evaluation result, and the warning response time is less than or equal to 8 minutes.

[0043] It should be noted that the interface protocol: the OGC Web Map Service (WMS) standard is used, and the warning data is pushed to the GIS platform in real time through the RESTful API interface; Visualization: the stability evaluation result is superimposed on the DEM in the form of a heat map, the safety factor Fs is mapped to a color gradient (green ≥ 1.5, yellow 1.2-1.5, red < 1.2), and the risk area coordinates and level are marked.

[0044] The detection system provided by the embodiment of the application further includes a turbidity compensation model, which inverses the water body suspended matter concentration (NTU) through the sonar echo intensity, establishes a "turbidity-sound wave attenuation" relationship formula to correct the depth measurement data, and the depth measurement error is less than or equal to 5 cm in the turbidity> 150 NTU environment.

[0045] The detection system provided by the embodiment of the application further comprises a dynamic deployment and AUV cooperation strategy: when the optical fiber sensor detects that the strain rate is greater than 0.1% / min, the AUV is automatically dispatched to carry a side scan sonar (SSS) to the abnormal area, the scanning priority is dynamically adjusted according to the strain gradient, and the optimal scanning path is generated by combining the flow velocity and the sonar coverage range.

[0046] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A system for detecting water level data and geological data in weakly cemented coal seams, characterized in that, include: The system consists of a hardware layer and an algorithm layer. The hardware layer constructs a three-in-one dynamic monitoring system for "water surface-water body-bottom". The algorithm layer uses a physical information neural network to fuse multi-source data to achieve full-domain bottom DEM inversion and dynamic early warning. The physical information neural network inverts the full-domain bottom DEM by performing spatiotemporal alignment and weighted fusion of sonar point cloud, UAV water level data and fiber strain gradient, and by jointly optimizing data-driven and physical-driven terms. The hardware layer includes: The aerial unit, consisting of a drone equipped with millimeter-wave radar and a multispectral camera, is used to monitor water surface elevation and areas of water leakage. The underwater unit includes a multibeam sonar array, a pressure-resistant fiber optic sensor network, and a floating water quality monitoring device, used for underwater topographic scanning, rock strain monitoring, and water quality parameter acquisition. The communication relay unit uses surface buoys to achieve data aggregation, encryption, and transmission with BeiDou satellites; The algorithm layer includes: The data preprocessing module performs motion compensation and stitching registration on the sonar point cloud and temperature drift correction on the fiber optic strain data. The physical information neural network model integrates sonar point cloud, UAV water level data and fiber strain gradient to constrain the continuity of rock strata and invert the whole-domain underwater DEM. The dynamic early warning module enables early warning of water inrush risk based on threshold determination and GIS visualization.

2. The detection system for water level data and geological data of weakly cemented coal seams according to claim 1, characterized in that, The airborne unit's millimeter-wave radar emits high-frequency electromagnetic waves to calculate water level elevation with centimeter-level accuracy; the multispectral camera identifies water seepage areas through thermal infrared and visible light feature analysis.

3. The detection system for water level data and geological data of weakly cemented coal seams according to claim 2, characterized in that, The underwater unit's multi-beam sonar array uses adaptive frequency switching technology to avoid the 50-200Hz mining noise band in real time. Combined with wavelet packet decomposition and motion compensation algorithms, it improves depth measurement accuracy, achieving a point cloud density of millions of points per second. The sonar array first removes high-frequency noise using wavelet packet decomposition and then compensates for the sonar's motion trajectory using the EKF algorithm, improving depth measurement accuracy to ±5cm.

4. The detection system for water level data and geological data of weakly cemented coal seams according to claim 3, characterized in that, The surface of the pressure-resistant fiber optic sensor is coated with a nanocolloid coating of silica nanoparticles, which increases the tensile strength to 200MPa. It adopts a double-ring mesh topology layout, and the data is transmitted through a backup path in the event of a single point breakage.

5. The detection system for water level data and geological data of weakly cemented coal seams according to claim 4, characterized in that, In the communication relay unit, the surface buoy parses underwater sonar point clouds, fiber optic data and water quality information through underwater acoustic communication protocols, and transmits them to the ground control center in batches via Beidou short messages after compression and encryption.

6. The detection system for water level data and geological data of weakly cemented coal seams according to claim 5, characterized in that, In the data preprocessing module, the sonar trajectory is reconstructed based on IMU data using extended Kalman filtering, and multi-frame point clouds are registered using the iterative nearest point algorithm; the temperature-strain transfer function model is trained using an LSTM network to correct the fiber strain gradient in real time.

7. The detection system for water level data and geological data of weakly cemented coal seams according to claim 6, characterized in that, The loss function of the PINN model includes data-driven and physics-driven terms, and the weights are dynamically adjusted through an adaptive balancing algorithm. A course learning strategy is adopted to first fit static terrain and then introduce dynamic data.

8. The detection system for water level data and geological data of weakly cemented coal seams according to claim 7, characterized in that, The dynamic early warning module sets the early warning threshold to a strain gradient exceeding the historical average of 3σ. Combined with the DEM changes visualized on the GIS platform and the rock strata stability assessment results, the early warning response time is ≤8 minutes.

9. The detection system for water level data and geological data of weakly cemented coal seams according to claim 8, characterized in that, It also includes a turbidity compensation model, which uses sonar echo intensity to invert the concentration of suspended solids (NTU) in the water, establishes a "turbidity-sound attenuation" relationship to correct the depth sounding data, and achieves a depth sounding error of ≤5cm in environments with turbidity >150NTU.

10. The detection system for water level data and geological data of weakly cemented coal seams according to claim 9, characterized in that, It also includes a dynamic deployment and AUV collaboration strategy: when the fiber optic sensor detects a strain rate > 0.1% / min, the AUV is automatically dispatched to carry the side-scan sonar to the abnormal area, the scanning priority is dynamically adjusted according to the strain gradient, and the optimal scanning path is generated by combining the water flow velocity and the sonar coverage area.