Underriver levee coal mining monitoring system and method based on integration of unmanned ship and GNSS (Global Navigation Satellite System)

Abstract

The invention discloses an under-river-levee coal mining monitoring system and method based on integration of an unmanned ship and a GNSS (Global Navigation Satellite System), and belongs to the technical field of mine safety monitoring. The monitoring system comprises a multi-source data collaborative acquisition module, a dynamic error compensation module, a multi-path effect suppression unit and an inertial navigation compensation unit, the multi-path effect suppression unit is used for constructing a water surface reflection signal attenuation model and suppressing GNSS multi-path effect interference in combination with a multi-band adaptive switching mechanism, and an adaptive data fusion engine is used for performing data fusion on the multi-source data collaborative acquisition module and the dynamic error compensation module. The data processing module is used for carrying out space-time reference unification, dynamic weight distribution and deformation settlement on multi-source data, and the prediction decision module is used for setting a multi-stage early warning mechanism according to the curvature radius and the deformation rate and executing a corresponding control instruction. According to the invention, through the GNSS receiving system, dynamic error compensation and other technologies, errors of positioning, water depth measurement and the like are significantly reduced; the fusion precision and the data integrity are improved; according to the system-level energy efficiency optimization, power consumption and computing resource occupation are reduced through multi-band switching and a lightweight algorithm.

Description

Technical Field

[0001] This invention relates to the field of mine safety monitoring technology, and more specifically, to a monitoring system and method for coal mining under river embankments based on the integration of unmanned vessels and GNSS. Background Technology

[0002] Due to the constraints of coal resource occurrence conditions, some mining areas need to mine coal resources below riverbeds due to geological limitations. However, the movement of overlying strata during mining can easily cause cracks, subsidence, or seepage channels in the riverbed. When the ratio of mining depth to riverbed bedrock thickness is lower than a critical value, it may lead to major safety accidents such as collapse and embankment breach. To address these risks, the "Detailed Rules for Water Prevention and Control in Coal Mines" clearly requires the construction of a real-time deformation monitoring system to ensure safe mining. However, actual monitoring faces dual technical challenges: at the data acquisition level, traditional single sensors have significant limitations—although GNSS can achieve positioning through satellite signals, its accuracy is easily affected by water reflection and multipath effects caused by bridge obstruction. While real-time dynamic positioning (RTK) technology can theoretically achieve centimeter-level accuracy through base station correction, signal attenuation and phase shift problems still exist in complex terrain. At the data fusion level, multi-source monitoring data suffers from inconsistent spatiotemporal benchmarks and static weight allocation mechanisms that cannot adapt to dynamic environments, leading to amplified fusion errors. Although simple fusion of multiple sensors in existing technical solutions can improve data reliability, problems such as insufficient time synchronization and dynamic response lag still restrict the effectiveness of the monitoring system.

[0003] Existing technologies mainly include single GNSS positioning schemes, unmanned surface vessel (USV) periodic patrol survey schemes, satellite remote sensing monitoring schemes, and simple multi-sensor fusion schemes. Single GNSS positioning schemes primarily rely on RTK or Precise Point Positioning (PPP) technology to obtain the USV's position. This method suffers from multipath effects and signal attenuation. Water surface reflection causes signal phase shift, resulting in a positioning error of 15-30 cm. Furthermore, the GNSS signal strength decreases by 2-3 dB for every meter increase in water depth, leading to a data loss rate greater than 15%. The unmanned surface vessel (USV) periodic survey scheme uses a USV equipped with a multibeam echo sounder to scan underwater terrain. This method cannot achieve real-time capture, and the data loss rate exceeds 40% when the water flow velocity is greater than 1.5 m / s. The satellite remote sensing monitoring scheme uses InSAR technology to acquire large-scale surface deformation data. This method suffers from low spatiotemporal resolution, a revisit period of ≥3 days, inability to capture transient deformations during mining operations, and poor water penetration, resulting in radar signals being unable to effectively detect underwater terrain changes. The multi-sensor simple fusion scheme is a fundamental method in data fusion technology, mainly achieving preliminary integration of multi-source data through low-complexity algorithms. However, this method suffers from inconsistent spatiotemporal references; when the sensor timestamp deviation is greater than 10 ms, the fusion error is amplified by more than 30%, and the fixed weighting coefficients cannot adapt to dynamic environments. Traditional riverbank monitoring relies on manual inspections and fixed sensors, which suffers from low data acquisition efficiency, limited coverage, and poor real-time performance. Existing unmanned surface vessel (USV) technology is mostly used for hydrological surveys but lacks deep integration with GNSS high-precision positioning and the ability to dynamically model riverbank deformation caused by coal mining. Therefore, we propose a coal mining monitoring system and method for riverbanks based on the fusion of USV and GNSS. Summary of the Invention

[0004] The purpose of this invention is to provide a monitoring system and method for coal mining under river embankments based on the fusion of unmanned vessels and GNSS, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A coal mining monitoring system under a river embankment based on the fusion of unmanned surface vessel and GNSS includes a multi-source data collaborative acquisition module, which is used to collect water parameters by carrying multiple sensors on an unmanned surface vessel, deploying ground reference stations and unmanned surface vessel-borne mobile stations to form a local differential network, and using the PTP precision clock protocol to achieve spatiotemporal synchronization; The dynamic error compensation module includes a multipath effect suppression unit and an inertial navigation compensation unit. The multipath effect suppression unit is used to construct a water surface reflection signal attenuation model and combine a multi-band adaptive switching mechanism to suppress GNSS multipath effect interference. The inertial navigation compensation unit is used to suppress inertial navigation error and compensate for GNSS multipath effect through an INS / GNSS tightly coupled model. An adaptive data fusion engine is used to unify the spatiotemporal benchmark, dynamically assign weights, and calculate deformation for multi-source data. The dynamic weight assignment is based on the sliding window standard deviation and sensor state thresholds, and the deformation calculation adopts the Timoshenko beam mechanics model combined with dynamic boundary conditions of mining. The prediction and decision-making module is used to set up a multi-level early warning mechanism and execute corresponding control commands based on the radius of curvature and deformation rate.

[0006] Preferably, in the multi-source data collaborative acquisition module, the ground reference station uses three dual-frequency receivers, and the ionospheric error is eliminated by joint calculation of L1 / L5 dual-frequency signals. The unmanned surface vessel continuously scans along the preset route at a speed of 0.5m / s. The multipath effect suppression unit uses a CNN phase residual recognition algorithm to extract GNSS observation sequence features and identify phase residual patterns to trigger the reconfiguration of reflection model parameters.

[0007] Preferably, the adaptive data fusion engine includes a data preprocessing layer, a dynamic weight allocator, and a deformation settlement model. The data preprocessing layer is used for spatiotemporal synchronization and anomaly detection of the data. The dynamic weight allocator is used to achieve dynamic weight optimization of multiple sensors based on the sliding window standard deviation weight allocation mechanism combined with GNSS multipath factor and INS angular velocity variance threshold. The deformation settlement model is used to simplify the river embankment into a Timoshenko beam mechanics model and to update the boundary conditions in real time with the mining face advance speed.

[0008] Preferably, the formula for the water surface reflection signal attenuation model of the multipath effect suppression unit is: ; Where h is the antenna height. For the satellite elevation angle, The attenuation coefficient is... The wavelength of the GNSS signal. For significant wave height; The formula for the INS / GNSS tightly coupled model of the inertial navigation compensation unit is: , ; in, The state vector is 15-dimensional, including three components: position error, velocity error, attitude angle error, gyroscope bias, and accelerometer bias. , These are process noise and observation noise, respectively. Here is the error state transition matrix. This is the noise-driven matrix.

[0009] Preferably, the weight calculation formula for the dynamic weight allocator is: ; in, For the first The sensor at the first The standard deviation of each sliding window is automatically reduced to below 0.3 when the GNSS multipath factor MP ≥ 0.5; if the INS angular velocity variance > 0.1 (° / s)², its weight is reduced to below 0.2. The differential equation of the Timoshenko beam mechanics model for the deformation calculation model is: ; in, The equivalent load distribution caused by coal mining.

[0010] Preferably, the multi-level early warning mechanism of the prediction and decision-making module includes: Yellow alert: Radius of curvature R < 500m and deformation rate > 3mm / h; Orange alert: Curvature radius R < 300m and deformation rate ≥ 5mm / h for 10 minutes; Red alert: Curvature radius R < 200m or deformation rate ≥ 10mm / h for 5 minutes.

[0011] Preferably, the multi-source data collaborative acquisition module also includes a multi-band adaptive GNSS receiving system and an underwater terrain dynamic compensation mechanism; The multi-band adaptive GNSS receiving system introduces an L2C band signal processing module on the basis of a dual-band receiver, constructs a joint settlement model of L1 / L5 / L2C three bands, and uses an adaptive band selection algorithm based on signal-to-noise ratio to prioritize the use of the L5+L2C band combination to suppress ionospheric delay error during ionospheric disturbance periods. The underwater terrain dynamic compensation mechanism adds a pressure-type wave height meter to the bottom of the unmanned vessel to measure the changes in the hull's draft in real time. By establishing a hull attitude-wave height compensation model that includes pitch and roll angle Kalman filter estimation, the water depth measurement error is reduced from ±5cm to ±1cm. Preferably, the multipath effect suppression unit of the dynamic error compensation module constructs a training set containing typical water surface reflection patterns for feature extraction by the CNN algorithm.

[0012] Preferably, it also includes a distributed edge computing architecture, in which edge computing units are mounted on the unmanned vessel, 80% of the data processing tasks are offloaded to local nodes, and 5G slicing technology is used to prioritize the uploading of critical data and offload computing tasks.

[0013] A method for monitoring coal mining under river embankments based on unmanned surface vessel and GNSS fusion, characterized by the following steps: S1. Construct a multi-source data collaborative acquisition system, collect water parameters by carrying multiple sensors on an unmanned vessel, deploy a local differential network with ground reference stations and unmanned vessel-borne mobile stations, and use the PTP precision clock protocol to achieve spatiotemporal synchronization. S2. Construct a dynamic error compensation algorithm, suppress GNSS multipath effect through water surface reflection modeling and multi-band adaptive switching, and suppress inertial navigation drift error by using INS / GNSS tightly coupled model; S3. Construct an adaptive data fusion engine to unify the spatiotemporal benchmarks of multi-source data, dynamically allocate weights based on the sliding window standard deviation, and calculate deformation based on the Timoshenko beam model. S4. Set up a predictive decision system, set multi-level early warning thresholds based on the radius of curvature and deformation rate, and trigger corresponding operations.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) In terms of high-precision data acquisition and error suppression system, this invention reduces the positioning error caused by multipath effect from 15-30cm in the traditional scheme to within 0.1m through dual-frequency / tri-frequency GNSS receiving system and dynamic error compensation algorithm. When SWH≥0.5m, water surface reflection modeling makes the multipath error attenuation rate of L5 band reach 80%. The inertial navigation tight coupling model suppresses the drift error of pure INS. The attitude angle error can still be controlled within 0.05° within 30 seconds of GNSS signal loss, which is 3 times more accurate than the traditional loose coupling model. The underwater terrain dynamic compensation mechanism optimizes the water depth measurement error of multibeam detector from ±5cm to ±1cm through hull attitude-wave height correction model, effectively eliminating the influence of pitch / roll angle on underwater terrain data.

[0015] (2) In terms of the dynamic adaptive multi-source data fusion performance, the present invention improves the fusion accuracy by using a dynamic weight allocation algorithm based on a sliding window. When the GNSS multipath factor MP ≥ 0.5, the system automatically reduces the weight of GNSS data and increases the INS weight to above 0.7, ensuring that millimeter-level deformation perception can still be maintained in complex environments. The spatiotemporal synchronization module compresses the timestamp deviation of multiple sensors to within 1 ms through the PTP protocol. Combined with cubic spline interpolation technology, the spatiotemporal reference unification error is reduced to 0.2 mm / m. The anomaly detection module, based on the 3σ principle and sensor health index, makes the outlier misjudgment rate < 0.5%, thus improving data integrity.

[0016] (3) In terms of deformation analysis capability driven by both physics and data, the Timoshenko beam model combined with the dynamic boundary conditions of mining achieves a deformation rate prediction error of ≤0.5mm / h, which is more accurate than the traditional static elastic model; the data-driven adaptive fusion engine can identify micro deformation signals at the level of 0.3mm with a response delay of <0.1 seconds, meeting the real-time monitoring requirements of transient deformation in mining.

[0017] (4) In terms of graded early warning and proactive safety control, the three-level early warning mechanism reduces the false alarm rate from 15% of the traditional single threshold scheme to less than 2% by using the combined criteria of radius of curvature and deformation rate. When a red warning is triggered, the system can complete the transmission of the coal mining machine shutdown command within 50 ms, which improves the response speed compared to manual intervention and effectively avoids dam failure accidents.

[0018] (5) In terms of system-level energy efficiency optimization, the multi-band adaptive switching technology reduces the power consumption of the GNSS module and can still maintain 24-hour continuous monitoring in the ionospheric disturbance scenario; the lightweight fusion algorithm controls the computing resource occupancy rate to below 15%, supports the deployment of edge computing devices, and reduces cloud dependence. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Example 1: A coal mining monitoring system for riverbanks based on the fusion of unmanned surface vessels (USVs) and GNSS includes a multi-source data collaborative acquisition module. This module uses multiple sensors mounted on the USV to collect water parameters such as flow velocity, water depth, and riverbank morphology. The multiple sensors include a multibeam echo sounder, a fiber optic inertial navigation system, and a millimeter-wave radar. A local differential network is formed by deploying a ground reference station and an USV-borne mobile station, and spatiotemporal synchronization is achieved using the PTP precision clock protocol. In the multi-source data collaborative acquisition module, the ground reference station uses three dual-frequency receivers. Ionospheric errors are eliminated through joint calculation of L1 / L5 dual-frequency signals. The USV continuously scans along a preset route at a speed of 0.5 m / s. By equipping the unmanned vessel with a multibeam detector, fiber optic inertial navigation system, and millimeter-wave radar, it continuously scans along a preset route at a speed of 0.5 m / s. To form a local differential network with the mobile unmanned vessel's mobile station, three ground reference stations are deployed using dual-frequency receivers, and the influence of ionospheric errors is eliminated through joint calculation of L1 / L5 dual-frequency signals. The BeiDou timestamp and laser scanning frame number are embedded in the data packet header, and the PTP precision clock protocol is used to achieve time and space synchronization.

[0021] The dynamic error compensation module includes a multipath effect suppression unit and an inertial navigation compensation unit. The multipath effect suppression unit is used to construct a water surface reflection signal attenuation model and combine a multi-band adaptive switching mechanism to suppress GNSS multipath effect interference. The inertial navigation compensation unit is used to suppress inertial navigation error and compensate for GNSS multipath effects through an INS / GNSS tightly coupled model. The multipath effect suppression unit uses a CNN phase residual recognition algorithm to extract GNSS observation sequence features and identify phase residual patterns to trigger the reconfiguration of reflection model parameters.

[0022] Specifically, multipath effects directly threaten positioning accuracy and reliability. On calm water, multipath errors fluctuate periodically, leading to low-frequency noise in deformation monitoring. During unmanned surface vessel movement, water surface ripples can also randomize multipath errors, causing jumps in positioning results or data loss. To address these issues, this solution uses water surface reflection modeling to suppress multipath errors. The formula for the water surface reflection signal attenuation model of the multipath effect suppression unit is: Where h is the antenna height. For the satellite elevation angle, The attenuation coefficient is... The wavelength of the GNSS signal. This represents the significant wave height. As the significant wave height increases, the exponential term causes the reflected signal to attenuate.

[0023] In this application, to effectively address the coordination issues of pure inertial navigation error accumulation and GNSS multipath interference, the proposed solution utilizes an INS / GNSS tightly coupled model to suppress inertial navigation drift errors, compensate for GNSS multipath effects, and improve fusion accuracy in dynamic environments. The formula for the INS / GNSS tightly coupled model of the inertial navigation compensation unit is as follows: , ;in, The state vector is 15-dimensional, including three components: position error, velocity error, attitude angle error, gyroscope bias, and accelerometer bias. , These are process noise and observation noise, respectively. Here is the error state transition matrix. This is the noise-driven matrix.

[0024] An adaptive data fusion engine is used to unify spatiotemporal references, dynamically allocate weights, and calculate deformation for multi-source data. Dynamic weight allocation is based on the sliding window standard deviation and sensor state thresholds, while deformation calculation employs the Timoshenko beam mechanics model combined with dynamic mining boundary conditions. This engine is responsible for dynamically fusing multi-source sensor data from unmanned surface vessels with GNSS data, addressing challenges such as inconsistent spatiotemporal references, significant differences in noise characteristics, and dynamic environmental changes. The desired objectives of this solution are: to dynamically adjust fusion weights based on real-time sensor confidence levels, achieving dynamic weight allocation; to eliminate timestamp and spatial coordinate errors between sensors, achieving spatiotemporal reference unification; and to maintain fusion stability and improve anti-interference capabilities under abnormal conditions such as multipath effects and signal obstruction.

[0025] The adaptive data fusion engine includes a data preprocessing layer, a dynamic weight allocator, and a deformation settlement model. The data preprocessing layer is used for spatiotemporal synchronization and anomaly detection of the data. The dynamic weight allocator is used to achieve dynamic weight optimization of multiple sensors based on the sliding window standard deviation weight allocation mechanism combined with GNSS multipath factor and INS angular velocity variance threshold. The deformation settlement model is used to simplify the river embankment into a Timoshenko beam mechanics model and to update the boundary conditions in real time with the mining face advance speed.

[0026] A spatiotemporal synchronization module and an anomaly detection module are introduced in the data preprocessing layer. The spatiotemporal synchronization module uses the PTP protocol to unify the clocks of all sensors and converts GNSS, INS, and radar data into a unified local coordinate system for the riverbank. It also performs cubic spline interpolation on asynchronous data to ensure consistent data stream frequency. The anomaly detection module is based on 3D... In principle, when sensor data deviates from the mean by more than 3 times the standard deviation, it is judged as an outlier and removed. The health status of the sensor is quantified by indicators such as GNSS carrier-to-noise ratio and INS angular velocity variance.

[0027] Specifically, to achieve dynamic weight allocation, this scheme uses the Extended Kalman Filter (EKF) to calculate the confidence scores of each sensor in real time, and employs a sliding window method to calculate the confidence weights of each sensor in real time. The specific weight calculation formula for the dynamic weight allocator is as follows: ;in, For the first The sensor at the first The standard deviation of the sliding window is automatically reduced to below 0.3 when the GNSS multipath factor MP ≥ 0.5. If the INS angular velocity variance > 0.1 (° / s)², its weight is reduced to below 0.2.

[0028] Specifically, to fuse heterogeneous data from multiple sensors and transform it into millimeter-level deformation parameters of the river embankment, and to achieve high-precision deformation sensing and disaster early warning in dynamic environments through synergistic optimization driven by both physical and data, this scheme simplifies the river embankment into a Timoshenko beam model and establishes deformation differential equations. The Timoshenko beam mechanical model differential equations for the deformation calculation model are as follows: ;in, The boundary conditions are updated in real time to reflect the equivalent load distribution caused by coal mining, taking into account the advance speed of the mining face.

[0029] The prediction and decision-making module is used to set up a multi-level early warning mechanism and execute corresponding control commands based on the radius of curvature and deformation rate.

[0030] Specifically, the multi-level early warning mechanism of the predictive decision-making module includes: Yellow alert: Radius of curvature R < 500m and deformation rate > 3mm / h; Orange alert: Curvature radius R < 300m and deformation rate ≥ 5mm / h for 10 minutes; Red alert: Curvature radius R < 200m or deformation rate ≥ 10mm / h for 5 minutes. When a red alert is triggered, a STOP command is automatically sent to the coal mining machine to forcibly stop it.

[0031] A method for monitoring coal mining under river embankments based on unmanned surface vessel and GNSS fusion includes the following steps: S1. Construct a multi-source data collaborative acquisition system, collect water parameters by carrying multiple sensors on an unmanned vessel, deploy a local differential network with ground reference stations and unmanned vessel-borne mobile stations, and use the PTP precision clock protocol to achieve spatiotemporal synchronization. S2. Construct a dynamic error compensation algorithm, suppress GNSS multipath effect through water surface reflection modeling and multi-band adaptive switching, and suppress inertial navigation drift error by using INS / GNSS tightly coupled model; S3. Construct an adaptive data fusion engine to unify the spatiotemporal benchmarks of multi-source data, dynamically allocate weights based on the sliding window standard deviation, and calculate deformation based on the Timoshenko beam model. S4. Set up a predictive decision system, set multi-level early warning thresholds based on the radius of curvature and deformation rate, and trigger corresponding operations.

[0032] This application also includes: introducing an L2C band signal processing module to construct a three-frequency joint settlement model; adding a pressure-type wave height meter to establish a hull attitude-wave height compensation model; and adopting a distributed edge computing architecture to reduce data processing latency. Based on a dual-frequency receiver, an L2C band signal processing module is introduced to construct the three-frequency joint settlement model. Through an adaptive frequency band selection algorithm based on signal-to-noise ratio, the L5+L2C band combination is preferentially used during ionospheric disturbance periods to suppress ionospheric delay errors. Furthermore, when water surface reflection causes a multipath error in the L1 band to >0.3m, the system automatically switches to the L5 band, reducing the multipath error to within 0.1m.

[0033] The training set is constructed using CNN to extract features from GNSS observation sequences, and phase residual patterns are identified to trigger reconfiguration of reflection model parameters, thereby improving the accuracy of the multimodal early warning mechanism.

[0034] Example 2: The difference from Embodiment 1 is that the multi-source data collaborative acquisition module also includes a multi-band adaptive GNSS receiving system and an underwater terrain dynamic compensation mechanism.

[0035] The multi-band adaptive GNSS receiving system introduces an L2C band signal processing module on the basis of a dual-band receiver, constructs a joint settlement model of L1 / L5 / L2C three bands, and uses an adaptive band selection algorithm based on signal-to-noise ratio to prioritize the use of the L5+L2C band combination to suppress ionospheric delay error during ionospheric disturbance periods. When water surface reflection causes the multipath error of the L1 band to be greater than 0.3m, it automatically switches to the L5 band, reducing the multipath error to within 0.1m.

[0036] The underwater terrain dynamic compensation mechanism adds a pressure-type wave height meter to the bottom of the unmanned vessel to measure the changes in the hull's draft in real time. By establishing a hull attitude-wave height compensation model that includes pitch and roll angle Kalman filter estimation, the water depth measurement error is reduced from ±5cm to ±1cm.

[0037] To effectively avoid errors caused by relative displacement between the onboard multibeam echo sounder and the water surface reference plane due to varying hull attitudes during unmanned surface vessel (USV) navigation, a pressure-type wave height meter will be installed on the bottom of the USV to measure changes in hull draft in real time. A hull attitude-wave height compensation model will be established, with the specific formula as follows: ;in, The pitch angle, This is the roll angle. By using Kalman filtering for dynamic estimation, the water depth measurement error was reduced from ±5cm to ±1cm.

[0038] In order to enhance the impact of multipath effects, the multipath effect suppression unit of the dynamic error compensation module constructs a training set containing typical water surface reflection patterns, uses CNN to extract GNSS observation sequence features, identifies the "fishhook" phase residual when the ship turns, and automatically reconfigures the reflection model parameters.

[0039] It also includes a distributed edge computing architecture, where edge computing units are mounted on the unmanned surface vessel (USV), offloading 80% of data processing tasks to local nodes. Combined with 5G slicing technology, this prioritizes the uploading of critical data and offloads computational tasks, reducing the time for fusion computing and early warning decision-making from 1.2 seconds to 0.3 seconds. Since riverbank deformation monitoring requires sub-meter level response, and traditional centralized cloud computing suffers from communication latency, edge computing offloads 80% of data processing tasks to the local nodes of the USV, reducing the time for fusion computing and early warning decision-making from 1.2 seconds to 0.3 seconds. Therefore, to reduce latency, this solution equips the USV with edge computing units, enabling 80% of data preprocessing and fusion computing to be completed locally, prioritizing the uploading of critical data through 5G slicing technology, and implementing a computational task offloading strategy. This reduces end-to-end latency and ensures accurate and efficient data transmission.

[0040] Multimodal early warning mechanism: Based on the original plan, the early warning mechanism is improved. When a yellow warning is issued, a high-density scanning mode is automatically activated; when an orange warning is issued, unmanned vessels are automatically dispatched to form a monitoring array; when a red warning is issued, an audible and visual alarm is triggered and an emergency evacuation route plan is automatically generated.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A monitoring system for coal mining under river embankments based on the fusion of unmanned surface vessels and GNSS, characterized in that, include: The multi-source data collaborative acquisition module is used to collect water parameters by carrying multiple sensors on an unmanned vessel, deploying ground reference stations and unmanned vessel-borne mobile stations to form a local differential network, and using the PTP precision clock protocol to achieve spatiotemporal synchronization. The dynamic error compensation module includes a multipath effect suppression unit and an inertial navigation compensation unit. The multipath effect suppression unit is used to construct a water surface reflection signal attenuation model and combine it with a multi-band adaptive switching mechanism to suppress GNSS multipath effect interference. The inertial navigation compensation unit is used to suppress inertial navigation error and compensate for GNSS multipath effect through an INS / GNSS tight coupling model. An adaptive data fusion engine is used to unify the spatiotemporal benchmark, dynamically assign weights, and calculate deformation for multi-source data. The dynamic weight assignment is based on the sliding window standard deviation and sensor state thresholds, and the deformation calculation adopts the Timoshenko beam mechanics model combined with dynamic boundary conditions of mining. The prediction and decision-making module is used to set up a multi-level early warning mechanism and execute corresponding control commands based on the radius of curvature and deformation rate.

2. The riverbank coal mining monitoring system based on unmanned vessel and GNSS fusion as described in claim 1, characterized in that: In the multi-source data collaborative acquisition module, the ground base station uses three dual-frequency receivers to eliminate ionospheric errors through joint calculation of L1 / L5 dual-frequency signals. The unmanned surface vessel continuously scans along the preset route at a speed of 0.5 m / s. The multipath effect suppression unit uses a CNN phase residual recognition algorithm to extract GNSS observation sequence features and identifies phase residual patterns to trigger the reconfiguration of reflection model parameters.

3. The riverbank coal mining monitoring system based on unmanned vessel and GNSS fusion as described in claim 1, characterized in that: The adaptive data fusion engine includes a data preprocessing layer, a dynamic weight allocator, and a deformation settlement model. The data preprocessing layer is used for spatiotemporal synchronization and anomaly detection of the data. The dynamic weight allocator is used to achieve dynamic weight optimization of multiple sensors based on a sliding window standard deviation weight allocation mechanism combined with GNSS multipath factor and INS angular velocity variance threshold. The deformation settlement model is used to simplify the river embankment into a Timoshenko beam mechanics model and integrate the boundary conditions updated in real time with the advance speed of the mining face.

4. The riverbank coal mining monitoring system based on unmanned vessel and GNSS fusion as described in claim 1, characterized in that: The formula for the water surface reflection signal attenuation model of the multipath effect suppression unit is: ； Where h is the antenna height. For the satellite elevation angle, The attenuation coefficient is... The wavelength of the GNSS signal. For significant wave height; The formula for the INS / GNSS tightly coupled model of the inertial navigation compensation unit is: ， ； in, The state vector is 15-dimensional, including three components: position error, velocity error, attitude angle error, gyroscope bias, and accelerometer bias. , These are process noise and observation noise, respectively. Here is the error state transition matrix. This is the noise-driven matrix.

5. A coal mining monitoring system for riverbanks based on the fusion of unmanned vessels and GNSS as described in claim 3, characterized in that: The weight calculation formula for the dynamic weight allocator is as follows: ； in, For the first The sensor at the first The standard deviation of each sliding window is automatically reduced to below 0.3 when the GNSS multipath factor MP ≥ 0.5; if the INS angular velocity variance > 0.1 (° / s)², its weight is reduced to below 0.

2. The differential equation of the Timoshenko beam mechanics model of the deformation calculation model is: ； in, The equivalent load distribution caused by coal mining.

6. The riverbank coal mining monitoring system based on unmanned vessel and GNSS fusion as described in claim 1, characterized in that: The multi-level early warning mechanism of the prediction and decision-making module includes: Yellow alert: Radius of curvature R < 500m and deformation rate > 3mm / h; Orange alert: Curvature radius R < 300m and deformation rate ≥ 5mm / h for 10 minutes; Red alert: Curvature radius R < 200m or deformation rate ≥ 10mm / h for 5 minutes.

7. A coal mining monitoring system for riverbanks based on unmanned surface vessels and GNSS fusion as described in claim 2, characterized in that: The multi-source data collaborative acquisition module also includes a multi-band adaptive GNSS receiving system and an underwater terrain dynamic compensation mechanism; The multi-band adaptive GNSS receiving system introduces an L2C band signal processing module on the basis of a dual-band receiver, constructs an L1 / L5 / L2C three-band joint settlement model, and uses an adaptive band selection algorithm based on signal-to-noise ratio to prioritize the use of the L5+L2C band combination to suppress ionospheric delay error during ionospheric disturbance periods. The underwater terrain dynamic compensation mechanism is achieved by installing a pressure-type wave height meter on the bottom of the unmanned vessel to measure the changes in the hull's draft in real time. By establishing a hull attitude-wave height compensation model that includes pitch and roll angle Kalman filter estimation, the water depth measurement error is reduced from ±5cm to ±1cm.

8. A coal mining monitoring system under river embankments based on unmanned surface vessel and GNSS fusion as described in claim 7, characterized in that: The multipath effect suppression unit of the dynamic error compensation module constructs a training set containing typical water surface reflection patterns for feature extraction by the CNN algorithm.

9. A coal mining monitoring system under river embankments based on the fusion of unmanned surface vessels and GNSS as described in claim 7, characterized in that: It also includes a distributed edge computing architecture, which equips unmanned ships with edge computing units to offload 80% of data processing tasks to local nodes, and combines 5G slicing technology to prioritize the uploading of critical data and offload computing tasks.

10. A method for monitoring coal mining under river embankments based on the fusion of unmanned surface vessels and GNSS, applied to the system described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Construct a multi-source data collaborative acquisition system, collect water parameters by carrying multiple sensors on an unmanned vessel, deploy a local differential network with ground reference stations and unmanned vessel-borne mobile stations, and use the PTP precision clock protocol to achieve spatiotemporal synchronization. S2. Construct a dynamic error compensation algorithm, suppress GNSS multipath effect through water surface reflection modeling and multi-band adaptive switching, and suppress inertial navigation drift error by using INS / GNSS tightly coupled model; S3. Construct an adaptive data fusion engine to unify the spatiotemporal benchmarks of multi-source data, dynamically allocate weights based on the sliding window standard deviation, and calculate deformation based on the Timoshenko beam model. S4. Set up a predictive decision system, set multi-level early warning thresholds based on the radius of curvature and deformation rate, and trigger corresponding operations.

Images