A method and system for monitoring and predicting a crack flow

Abstract

The application discloses a method and system for monitoring and predicting a crack flow, and the method comprises the following steps: obtaining historical monitoring data and real-time monitoring data obtained by current monitoring; simulating an impending crack flow by using an X-BEACH model according to the historical monitoring data and the real-time monitoring data to obtain simulation crack flow data; correcting the simulation crack flow data by using a machine learning model to obtain predicted crack flow data; calculating a crack flow prediction score according to the predicted crack flow data, and determining a crack flow early warning of a corresponding grade according to the crack flow prediction score. The application can accurately and comprehensively monitor and predict the crack flow, and can be widely applied to the technical field of crack flow monitoring.

Description

Technical Field

[0001] This application relates to the field of crack flow monitoring technology, and in particular to a method and system for monitoring and predicting crack flow. Background Technology

[0002] The existing crack flow monitoring technology has the following problems: (1) The current crack flow monitoring system is not comprehensive enough and cannot obtain multi-source monitoring data. The accuracy of the monitoring data needs to be verified. (2) When the flow velocity is large, the monitoring data is not accurate enough. Remote sensing images and UAV shooting have limited resolution. (3) The prediction results obtained by the current empirical prediction and numerical simulation prediction methods are not accurate enough and the prediction time is long. The reliability of empirical prediction needs to be improved; in terms of numerical model prediction, the time-averaged wave model cannot fully consider some characteristics of nearshore wave-generated currents, such as the nonlinear influence of waves and wave-current interaction. Due to the limitation of wave scale, the wave time-domain model has a large amount of computation, low efficiency and long time consumption, so it can only be used in a very small area. (4) There are many models that can simulate crack flows, but most of them are in the laboratory stage. So far, no model can predict crack flows on a large scale. The reason is that the model has high computational efficiency but poor accuracy or good simulation effect but too much computation. (5) The complete observation content of the rift system needs to cover the flow velocity, topography, tide level and incident wave elements in the open sea of ​​the rift channel and sandbar. Underwater topographic surveying is costly in terms of manpower, financial resources and time. Summary of the Invention

[0003] In view of this, this application provides a method and system for monitoring and predicting crack flow, so as to accurately and comprehensively monitor and predict crack flow.

[0004] One aspect of this application provides a method for monitoring and predicting fracture flow, comprising:

[0005] Acquire historical monitoring data and real-time monitoring data obtained from current monitoring;

[0006] The X-BEACH model is used to simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data to obtain simulated crack flow data;

[0007] The simulated fracture flow data is corrected using a machine learning model to obtain predicted fracture flow data;

[0008] The predicted fracture flow data is used to calculate the fracture flow prediction score, and the corresponding level of fracture flow warning is determined based on the fracture flow prediction score.

[0009] Optionally, acquiring the currently monitored real-time fracture flow data includes:

[0010] Real-time wind, wave and current field data of the monitored area are acquired using X-band marine radar.

[0011] The crack flow data of the crack flow area in the monitoring region is obtained by using a drone;

[0012] The underwater flow and topography data of the monitored area are obtained by unmanned surface vessels.

[0013] Optionally, the step of using the X-BEACH model to simulate the impending fracture flow based on the historical monitoring data and the real-time monitoring data to obtain simulated fracture flow data includes:

[0014] The historical monitoring data and the real-time monitoring data are merged into a dataset, and outliers are removed, the format is standardized, the data units are standardized, and the data is smoothed using filtering techniques to remove noise or high-frequency oscillations. Then, principal component analysis is used to perform dimensionality reduction on the dataset to obtain the processed dataset.

[0015] Set the time step, spatial grid, initial conditions, and model boundary conditions of the X-Beach2D model, and then configure the model configuration file, historical data file, and bottom friction parameter file to obtain the configured X-Beach2D model.

[0016] The processed dataset is input into the configured X-Beach2D model to obtain the simulated fracture flow data.

[0017] Optionally, the step of using a machine learning model to correct the simulated fracture flow data to obtain predicted fracture flow data includes:

[0018] The simulated fracture flow data is fused with the residuals of the corresponding historical periods using the machine learning model to obtain the predicted fracture flow data.

[0019] Optionally, the step of calculating a fracture flow prediction score based on the predicted fracture flow data and determining a corresponding level of fracture flow warning based on the fracture flow prediction score includes:

[0020] Weights are assigned to each data item in the predicted fracture flow data, and the sum of the weights is 1.

[0021] The standardized score of each data point is multiplied by its corresponding weight to obtain the actual score for each data point.

[0022] Summing the actual scores of each term yields the fracture flow prediction score;

[0023] The corresponding level of crack flow warning is determined based on the risk range in which the crack flow prediction score is located.

[0024] Optionally, the method further includes:

[0025] Periodically acquire monitoring data within a set time period as detection data;

[0026] If the detection data is discontinuous or contains outliers, the status of the monitoring equipment will be detected and automatically repaired. If automatic repair is not possible, an alarm will be sent to maintenance personnel.

[0027] Optionally, the method further includes:

[0028] The predicted fracture flow data is periodically compared with the actual fracture flow data, and multiple indicators are used to quantify the difference between the predicted fracture flow data and the actual fracture flow data; the indicators include RMSE, correlation coefficient, and bias; each indicator corresponds to a threshold range;

[0029] When a difference in one of the indicators is found to correspond to the threshold range, an automatic calibration measurement is triggered and the parameters of the machine learning model are calibrated.

[0030] Another aspect of this application provides a system for monitoring and predicting cracked flows, comprising:

[0031] The first module is used to acquire historical monitoring data and real-time monitoring data obtained from current monitoring.

[0032] The second module is used to simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data using the X-BEACH model, and to obtain simulated crack flow data.

[0033] The third module is used to correct the simulated crack flow data using a machine learning model to obtain predicted crack flow data.

[0034] The fourth module is used to calculate the crack flow prediction score based on the predicted crack flow data, and to determine the corresponding level of crack flow warning based on the crack flow prediction score.

[0035] Another aspect of this application provides an electronic device, including a processor and a memory;

[0036] The memory is used to store programs;

[0037] The processor executes the program to implement the aforementioned method.

[0038] Another aspect of this application provides a computer-readable storage medium storing a program that is executed by a processor to implement the aforementioned method.

[0039] This application also discloses a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of an electronic device can read the computer instructions from the computer-readable storage medium and execute the computer instructions to cause the electronic device to perform the aforementioned method.

[0040] This application includes at least the following beneficial effects:

[0041] This application acquires historical monitoring data and current real-time monitoring data; it uses the X-BEACH model to simulate impending crack flows based on the historical and real-time monitoring data, obtaining simulated crack flow data; combining historical and real-time monitoring data enables a more accurate simulation of impending crack flows, thus obtaining accurate simulated crack flow data; then, a machine learning model is used to correct the simulated crack flow data, obtaining predicted crack flow data; the corrected predicted crack flow data is closer to the actual crack flows and more consistent with reality; a crack flow prediction score is calculated based on the predicted crack flow data, and the corresponding crack flow warning level is determined based on the crack flow prediction score. Based on the accurate crack flow prediction score, the corresponding crack flow warning level can be determined, reducing false warnings. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating a method for monitoring and predicting fracture flow provided in an embodiment of this application;

[0044] Figure 2 A flowchart illustrating the workflow of a crack flow integrated monitoring and prediction system provided in this application embodiment;

[0045] Figure 3 A specific schematic diagram of a crack flow provided in an embodiment of this application;

[0046] Figure 4 This is a schematic diagram of a process for acquiring monitoring data provided in an embodiment of this application;

[0047] Figure 5 This application provides a schematic diagram of a data preprocessing process.

[0048] Figure 6 A schematic diagram illustrating a process for configuring an XBeach model, provided as an embodiment of this application;

[0049] Figure 7 A network architecture diagram of a machine learning model provided in an embodiment of this application;

[0050] Figure 8 An example diagram of weight allocation provided for an embodiment of this application;

[0051] Figure 9 A risk warning level map provided for embodiments of this application;

[0052] Figure 10 This is a structural block diagram of a crack flow monitoring and prediction system provided in an embodiment of this application. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart.

[0055] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0057] To facilitate understanding of the embodiments of this application, the keywords that may be involved in the embodiments of this application are explained as follows:

[0058] The relevant crack flow monitoring and prediction methods and systems mainly include two aspects: monitoring methods and prediction methods.

[0059] 1. Monitoring methods include on-site observation and remote sensing imagery.

[0060] (1) On-site observation methods: fixed-point measurement and float tracing:

[0061] Fixed-point measurements collect data from fixed points in the flow field using a fixed velocity meter, but this method cannot cover the entire flow field and is relatively inadequate for measuring the planar flow structure. Doppler instruments and Doppler sonar can be used to measure the velocity and pulsation mechanism of split flows, and a few stations can also be used to measure velocity and pressure.

[0062] Float tracing compensates for the limitations of fixed-point measurements. By analyzing the float's trajectory, the velocity distribution and detailed planar flow structure of the entire flow field can be obtained. The kinematic characteristics of the fractured flow can be analyzed by using visual observation or methods employing dyes and float tracing, supplemented by camera recording of the float's motion. GPS positioning technology can also be used to record the float's trajectory, and Lagrange multiplication methods can be applied to better track the detailed flow field of the fractured flow, thereby analyzing its flow field characteristics.

[0063] (2) Remote sensing images:

[0064] By utilizing satellite remote sensing imagery, information such as image time, geomorphological development, and wave breaking can be extracted. Waves break into whitecaps in shallow waters or sandbars, appearing as bright areas in the image. The greater the intensity of wave breaking, the greater the corresponding brightness. Appropriate image interpretation techniques can be used for observational research on nearshore rift systems and geomorphology, including wave elements such as incident wave height, period, direction, and runoff, as well as surface flow fields and sandbar topographic evolution.

[0065] A complete observation of the rift system requires encompassing flow velocity, topography, tidal levels, and incident wave elements from the open sea at the rift trough and sandbar locations. Combining coastline detection technology with video imagery to invert underwater beach topography offers a rapid and highly accurate method for acquiring beach topography. Simultaneously, radar technology is also being explored for inverting nearshore rift flow fields and underwater topography.

[0066] 2. Prediction methods include empirical forecasting and numerical simulation.

[0067] (1) Empirical forecasting is based on statistical analysis to establish the relationship between crack flow and certain factors, and then infers the crack flow situation from these factors.

[0068] The LURCS (LUshine Rip Current Scale) method was developed by statistically analyzing rift drowning and rescue data from southeastern Florida beaches in the United States, along with wind direction, wind speed, wave height, and tide time. This method reduces or even eliminates the weight of wind, while increasing the consideration of swell, wave direction, and tide level, significantly improving the objectivity of forecasts. For the empirical forecasting method CAP-LURCS (Coastal Andhra Pradesh LURCS) for the Indian coast, this method can automatically forecast rifts hourly with higher accuracy, but it cannot predict the location of rift occurrences, and its reliability still needs improvement.

[0069] (2) Numerical Simulation: There are two main types of models suitable for simulating fractured flow: time-averaged wave models and time-domain wave models. Time-averaged wave models average the fluid motion equations over the wave period, directly obtaining the time-averaged flow by solving the equations, while the effect of waves is considered through radiation stress. Time-domain wave models calculate the flow field directly in the time domain, obtaining the time-averaged flow field by averaging the velocity of fluid particles in the wave over time. Due to the limitation of wave scale, this type of model has lower computational efficiency, but it can compensate for some of the shortcomings of the time-averaged wave models.

[0070] ①FUNWAVE model:

[0071] The FUNWAVE model is a time-domain model that incorporates wave-induced momentum flux and horizontal turbulent mixing effects, enabling it to address phase interactions and simulate nearshore circulation. The core equations of the FUNWAVE model employ the fully nonlinear dispersion equations proposed by Wei et al., derived from the depth-averaged three-dimensional Euler equations under irrotational and shallow-water assumptions. FUNWAVE is one of the most advanced models for simulating fracture flows, capable of simulating rapidly changing fracture flows. However, FUNWAVE is only suitable for laboratory simulations of fracture flows and cannot be used for operational forecasting because its computational demands are too high; simulating even a small segment of fracture flow often requires a very long time.

[0072] ②XBeach model:

[0073] The XBeach model is a time-domain model that is processed in parallel by several modules. Its main modules are the hydrodynamic module and the geomorphological dynamic module. The hydrodynamic module contains two sub-modules: a shortwave module and a current module, while the geomorphological dynamic module also contains two sub-modules: a morphology module and a sediment transport module. This highly modular and parallel processing approach not only significantly improves computational efficiency, but also fully considers various influencing factors in the coupled computation between modules, thereby improving simulation accuracy.

[0074] Reference Figure 1This application provides a method for monitoring and predicting fracture flow, including steps S100 to S130, as follows:

[0075] S100: Acquire historical monitoring data and real-time monitoring data obtained from current monitoring.

[0076] Specifically, this embodiment can obtain historical or actual monitoring data of the monitoring area through various monitoring devices. Historical monitoring data can also be obtained from other historical materials, such as research reports or news reports on the monitoring area.

[0077] Furthermore, S100 may include:

[0078] Real-time wind, wave and current field data of the monitored area are acquired using X-band marine radar.

[0079] The crack flow data of the crack flow area in the monitoring region is obtained by using a drone;

[0080] The underwater flow and topography data of the monitored area are obtained by unmanned surface vessels.

[0081] Specifically, such as Figure 2 As shown in the diagram, this embodiment provides a flowchart of a comprehensive crack flow monitoring and prediction system. The monitoring component consists of an X-band radar, a float tracer, a drone, and an unmanned surface vessel (USV) equipped with a GNSS receiver, ADCP, and sonar system. The prediction component is implemented using a numerical model combined with machine learning, including steps 1.1 to 1.3:

[0082] 1.1 X-band radar acquires wind, wave and current data in the monitored area.

[0083] X-band radar can provide high-resolution sea surface reflectivity data for detecting the direction and intensity of ocean currents. The data acquisition process is as follows:

[0084] First, the monitoring area is delineated, and then an X-band marine radar system is selected for monitoring. The radar deployment location, such as a land station, is chosen based on environmental conditions. Next, the radar equipment is installed and configured, ensuring its stability and performance. Before formal monitoring, the radar system is calibrated and tested to ensure accuracy. Data collection begins, acquiring information about the ocean surface by sending and receiving signals through the radar system. The data is then processed, decoded, and analyzed to generate useful information such as wind speed, wave height, wave direction, and ocean current speed. Finally, the data is visualized, presented in the form of images or maps.

[0085] Real-time acquisition of sea surface wind, wave, and current field data using X-band ocean radar provides reliable data for subsequent numerical models combined with machine learning to predict crack currents. The resulting images, visualized from this data, can be used for real-time monitoring of crack currents.

[0086] 1.2 Fracture flow identification and fracture flow data collection.

[0087] The water flow velocity within the region is divided into two components, u and v. The u component represents the velocity parallel to the shore, and the v component represents the velocity perpendicular to the shore. A positive v component indicates that the flow direction is offshore to seaward, which is also the velocity direction of the main part of the rift current. If X-band radar detects a velocity profile perpendicular to the shore where the v component is entirely positive, then the boundary is defined by extending this profile area along both sides of the shore until this profile can no longer be detected. The boundary is further defined based on the furthest extension of the positive offshore v component. Figure 3 As shown in the image, a rift flow was detected in the area. A red rift flow warning was issued to the relevant authorities to alert tourists and to take emergency measures.

[0088] like Figure 4 As shown, in order to further capture valuable rift flow data, an automatic release device was set up to release colored float tracers in the rift flow risk area. At the same time, a drone equipped with a CCD camera lens was deployed to take pictures of the risk area, and the resulting image set was further analyzed to obtain surface flow characteristics.

[0089] Each drone is equipped with two 1024×1280 resolution CCD cameras, capturing images at 10 frames per second, with a camera range of 11m×15m. The number of drones deployed is selected based on the size of the risk area; under normal circumstances, three drones traveling 100m-200m offshore can completely cover the risk area. The floats are made of environmentally friendly, biodegradable materials, green, circular, thin blocks, 10cm in diameter and 3cm thick. Fluorescent components are added to the surface of the floats to ensure high visibility even at night. The average density of the floats is slightly lighter than water, ensuring stable buoyancy and movement with the current. The number of floats released each time depends on the size of the rift zone (average 10 floats / m²). 2 ).

[0090] By analyzing the buoy trajectory, the planar flow structure of the entire rift can be further obtained. The specific analysis steps are as follows: (1) Use image processing tools such as MATLAB to find and record the positions of all buoys in the image. (2) Find the correspondence between the same buoys in adjacent frames and obtain their motion trajectories. (3) Convert the image pixels to two-dimensional space. (4) Use the buoy velocity to deduce the velocity of multiple points in the flow field. Calculating the velocity of a fixed point in space using the buoy velocity is a conversion from the Lagrangian method to the Eulerian method. The Lagrangian velocity of the buoy is obtained from the trajectory, which is equal to the instantaneous flow field velocity at the location of the buoy.

[0091] Finally, by comparing the data acquired by X-band radar and performing error analysis, a more accurate flow structure of the crack flow in this region was obtained, and the flow data was used for subsequent prediction work.

[0092] 1.3 Dynamic acquisition of regional vertical flow structure and seabed topography by unmanned surface vessels.

[0093] For acquiring the vertical flow structure of the entire monitoring area, an unmanned surface vessel (USV) equipped with an Advanced Digital Current Profiler (ADCP) was used. The ADCP provides information such as underwater flow velocity and direction, enabling a more accurate description of the spatial distribution of offshore currents. For acquiring seabed topography, an USV equipped with a multibeam sonar system and a side-scan sonar system was used. All observation information was transmitted to the land-based control center, where it was processed to obtain digital elevation models of the underwater flow structure and seabed topography. (For beaches with significant topographic variations, accurate and high-resolution topographic data is crucial for successful numerical simulations, especially observational data on surface flow fields and underwater topography in nearshore areas under storm surge conditions.) The specific implementation process is as follows:

[0094] (1) Select an unmanned surface vessel (USV) suitable for the mission. Initial selection should be based on historical data of the monitoring area's depth and current velocity, while also considering mission duration and equipment carrying capacity. Ensure the USV has a stable platform to guarantee the accuracy of ADCP and sonar equipment. A solar power system is required to support long-term, uninterrupted monitoring missions.

[0095] (2) Install the ADCP and sonar equipment on the unmanned surface vessel (USV), ensuring the equipment is fixed and stable. Connect the equipment to the USV's power supply and data acquisition system. Install a GNSS receiver to track the USV's position in real time. Calibrate the positioning system to ensure the accuracy of the position information.

[0096] (3) Use mission planning software to plan the unmanned surface vessel's (USV) route to ensure coverage of the entire area. Pre-program or remotely control the USV's startup, navigation, docking, and data acquisition. Start the USV, allowing it to navigate according to the predetermined plan and operate the ADCP and sonar systems. Regularly record and store vertical current data and seabed topography data acquired from these devices.

[0097] (4) Transfer the collected data to a computer. Process and analyze the data using specialized software, including converting ADCP data into vertical current profiles and sonar data into seabed topographic maps. Generate information on vertical flow structure and seabed topography through graphical visualization or numerical analysis. Regularly check data quality and make real-time decisions during mission execution. Regularly inspect and maintain ADCP, sonar, and positioning equipment to ensure their stable performance.

[0098] S110: Using the X-BEACH model, simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data to obtain simulated crack flow data.

[0099] Specifically, in this embodiment, the simulated crack flow may vary in size, or it may not occur at all.

[0100] Furthermore, S110 may include:

[0101] The historical monitoring data and the real-time monitoring data are merged into a dataset, and outliers are removed, the format is standardized, the data units are standardized, and the data is smoothed using filtering techniques to remove noise or high-frequency oscillations. Then, principal component analysis is used to perform dimensionality reduction on the dataset to obtain the processed dataset.

[0102] Set the time step, spatial grid, initial conditions, and model boundary conditions of the X-Beach2D model, and then configure the model configuration file, historical data file, and bottom friction parameter file to obtain the configured X-Beach2D model.

[0103] The processed dataset is input into the configured X-Beach2D model to obtain the simulated fracture flow data.

[0104] Specifically, since the purpose of the hydrodynamic simulation process is to predict the occurrence of crack flows, in this embodiment, considering both the rational use of computational resources and the restoration of crack flow details, the X-Beach2D model is used to perform hydrodynamic simulation of the nearshore area, including steps 2.1 to 2.3:

[0105] 2.1 Dataset preparation and preprocessing.

[0106] Through long-term monitoring and review of local data, multi-year, multi-source datasets and hydrodynamic parameters for the entire region were obtained, including topographic data (coastline and seabed topography), wave data (wave height, wave period, wave direction), tidal data (tidal height, tidal period), hydrodynamic parameters (bottom friction coefficient, seawater density, seawater viscosity), and wind data (wind speed, wind direction). The time-series data underwent initial screening, selecting data with high spatiotemporal resolution, good continuity, and long coverage periods. To further ensure data quality, accuracy, and usability, preprocessing was performed, such as... Figure 5 As shown.

[0107] (1) Examine the dataset to identify and handle missing data, outliers, or erroneous data points; handle duplicate records or duplicate data points;

[0108] (2) Since the datasets obtained in the early stage were acquired through multiple channels, a unified data format conversion was performed; data units were standardized to ensure that the data are comparable in the analysis;

[0109] (3) Use filtering techniques to smooth the data to remove noise or high-frequency oscillations in order to better identify trends and patterns; for time series data (waves, wind, tides), process the timestamps to ensure data consistency;

[0110] (4) Merge data from multiple data sources into a consistent dataset for input into the model; after data merging, use dimensionality reduction techniques (principal component analysis) to reduce data complexity while retaining key information;

[0111] (5) Use visualization tools and techniques to explore the characteristics, distribution and trends of the data to assist in subsequent analysis and interpretation.

[0112] 2.2 Model settings, please refer to the following for details. Figure 6 .

[0113] (1) Time step setting:

[0114] Taking into account the resolution of the model validation data and the model's simulation accuracy, an appropriate time step should be selected. For subsequent crack flow identification, to capture as many details as possible, a time step of approximately 1 second can be initially chosen. The time step can be adjusted during subsequent validation.

[0115] (2) Spatial grid settings:

[0116] Define the spatial grid resolution of the model, including both horizontal and vertical directions. The choice of resolution should be determined based on the scale of the simulation area and computational resources. For subsequent crack flow identification, a horizontal grid resolution of approximately 10m and a vertical grid resolution of approximately 0.5m can be selected. Furthermore, the vertical grid resolution setting should consider the bottom boundary layer to ensure that bottom interactions can be described. The grid settings should be adjusted based on subsequent testing.

[0117] (3) Setting initial conditions:

[0118] Based on the collected data, the starting conditions of the model are determined, including water depth (obtained through DEM), water surface height (using local tidal forecast data), current velocity (obtained through monitoring equipment), bottom friction (obtained through literature review or field measurement), waves (including wave height, wavelength, wave direction, and wave breaking parameters, obtained through monitoring equipment), and model start time.

[0119] (4) Model boundary conditions:

[0120] The boundary conditions of the model are determined, including those for open boundaries and land boundaries. Since radar monitoring equipment can acquire real-time data 24 hours a day, the real-time data will be integrated into the model boundary conditions after subsequent model optimization to achieve real-time simulation. This will then be combined with machine learning to predict transient phenomena such as rift flows in real time.

[0121] (5) Prepare the input file:

[0122] Based on the acquired data and determined parameters, create the input files required for the X-BEACH model. These files include the model configuration file, historical data file, and bottom friction parameter file. Before running the model, perform checks and debugging to ensure that the model settings are correct.

[0123] 2.3 X-BEACH Model Execution and Validation.

[0124] Run the X-BEACH model using its executable file. Input the data files, model configuration, and relevant parameter files. The model will simulate hydrodynamic processes according to the time step (1 s) and grid resolution specified in the configuration file. After the model runs, X-BEACH will generate a series of output files: xbeach_output.nc: contains hydrodynamic parameters such as water depth, flow velocity, wave height, and bottom friction; xbeach_moments.nc: contains instantaneous wave parameters; other output files include bottom sediment parameters and files for visualization. During model execution, monitor the model's progress and check for any error messages. Ensure the model is running correctly and keep records. Once the model is complete, use data visualization tools (such as Paraview software for Python) to process the simulation results and the corresponding measured data for the simulation period. Adjust the model configuration, parameters, or input data through comparison and verification, and then rerun the simulation. After multiple simulations and optimizations, achieve the best results from the numerical model simulation.

[0125] After model tuning is complete, historical data is input into the numerical model to simulate future data, and the simulated data is used as input to the machine learning model.

[0126] S120: The simulated crack flow data is corrected using a machine learning model to obtain the predicted crack flow data.

[0127] Specifically, due to the existence of various complex physical processes in the ocean, such as eddies and turbulence, coupled with the uncertainties in ocean observation data and numerical models, numerical simulations cannot completely and accurately reflect the actual distribution of fracture flows, and thus contain certain errors and biases. To further improve prediction accuracy, machine learning is used to further optimize the prediction results, that is, machine learning models are used to correct the errors in the simulated fracture flow data, thereby obtaining the predicted fracture flow data.

[0128] Furthermore, S120 may include:

[0129] The simulated fracture flow data is fused with the residuals of the corresponding historical periods using the machine learning model to obtain the predicted fracture flow data.

[0130] Specifically, the network architecture of the machine learning model in this embodiment consists of a convolutional neural network (CNN), a long short-term memory network (LSTM), and residual connections, as detailed below. Figure 7 As shown.

[0131] exist Figure 7 In the network architecture shown, both simulated and measured data from the numerical model enter from the input layer. Low-level generalized spatial features are extracted through convolutional layer 1, followed by batch normalization by a Batch Normalization (BN) layer. Parallel convolutional layers then further extract detailed and abstract spatial features, which are passed through another BN layer before being fed to an LSTM layer for time-series processing. The data then enters a concat layer, where residual connections are passed to the concat layer for concatenation. This process is repeated in the next convolutional layer to fully extract and process the information features. Finally, all processed information is fed to a fully connected layer to integrate the features extracted by the previous layers and map these features to the sample label space. The introduction of residual connections allows the network to be stacked to deeper layers, avoiding the vanishing gradient problem and significantly improving network performance. Furthermore, as a key part of the network architecture, residual connections enable the network to learn the residuals between simulated and measured data. The initial parameter settings are as follows:

[0132] (1) Convolutional layers 1, 4, and 7 use wider convolutional kernels (1*32) and a larger number of convolutional kernels (128), convolutional layers 2, 5, and 8 use slightly narrower convolutional kernels (1*16) and a slightly smaller number of convolutional kernels (64), and convolutional layers 3, 6, and 9 use even narrower convolutional kernels (1*8) and a smaller number of convolutional kernels (32); the activation function used is the ReLU (Rectified Linear Unit) activation function, which accelerates the training of the network and improves the accuracy of prediction.

[0133] (2) The number of hidden units in the LSTM layer is set to 64.

[0134] (3) The initial learning rate can be set to 0.001 and adjusted according to the performance during the training process; the mean square error is used as the loss function to reflect the difference between the prediction result and the true value, and it is easy to calculate and optimize; Adam is used as the optimization algorithm.

[0135] Training process: The numerical model simulation results and corresponding historical measured data are input into the machine learning network. After long-term training, the network learns the residual patterns between the simulation results and the measured data. If the amount of data allows, the model can be trained by simulating data from multiple years, and the average of the residuals obtained each year can be used as the final result, allowing the model to fully learn the residual patterns across different time periods. The residual obtained during training is calculated as: Measured value - Numerical model simulation result.

[0136] Prediction Process: A pre-trained machine learning network, having learned the residual patterns, is used. The numerical model simulates future data, which is then input into the network. The network identifies and analyzes the data, fusing it with historical residuals. The output is a corrected result, including water depth, flow velocity, wave height, and friction parameters. The prediction result equals the numerical model simulation result plus the residuals.

[0137] Automatic prediction: During the model setup process, real-time monitoring data has been integrated into the model boundary conditions. Using an automated program, the data for the next three days is simulated daily using updated historical data, and the results are input into the machine learning model for further optimization of the output.

[0138] S130: Calculate the crack flow prediction score based on the predicted crack flow data, and determine the corresponding level of crack flow warning based on the crack flow prediction score.

[0139] Specifically, the crack flow prediction score can reflect the hazard of the predicted crack flow. The higher the crack flow prediction score, the more dangerous the impending crack flow is. Therefore, the embodiments of this application determine the corresponding level of crack flow warning based on the hazard of the crack flow.

[0140] Furthermore, S130 may include:

[0141] Weights are assigned to each data item in the predicted fracture flow data, and the sum of the weights is 1.

[0142] The standardized score of each data point is multiplied by its corresponding weight to obtain the actual score for each data point.

[0143] Summing the actual scores of each term yields the fracture flow prediction score;

[0144] The corresponding level of crack flow warning is determined based on the risk range in which the crack flow prediction score is located.

[0145] Specifically, to quantify the impact of each data point on the occurrence of fracture flow, this embodiment proposes a method using weights to represent the relative importance of each data point, thereby assessing the magnitude of fracture flow risk. Optionally, the specific values ​​of these weights are determined based on research experience, experimental data, and recommendations from domain experts. The specific weight allocation strategy is as follows:

[0146] Water depth: Based on previous research and observations, water depth has been identified as one of the main influencing factors in the occurrence of rift flows. Shallower waters are more prone to rift flows and are therefore given a higher weight.

[0147] Flow velocity: Flow velocity is another key factor, and higher flow velocities are generally associated with cracking phenomena. Previous studies have shown that flow velocity has a significant impact on crack formation, and therefore it is given a high weight.

[0148] Wave height: Variations in wave height can influence the occurrence of fracture flows. Higher waves may lead to more intense hydrodynamic conditions, which contribute to fracture flow formation, and therefore should be given appropriate weight.

[0149] Flow direction: Abrupt changes in flow direction may increase the likelihood of rifting. Although its impact is relatively small, it is still considered and assigned a moderate weight.

[0150] Bottom friction: Bottom friction can affect water flow velocity and direction, but it is generally less important than water depth and flow velocity. Therefore, it is given a relatively low weight.

[0151] Other factors: Considering the possibility of other factors, but the specific impact of these factors is difficult to determine, a small portion of the weight is reserved to cover the impact of other potential factors.

[0152] The weights are summed to 1. Based on the data output from the machine learning, each factor is standardized to a value between 0 and 1, according to historical maximum values ​​and empirical judgment. Finally, a weighted sum is calculated, and the probability of a crack flow is estimated based on the overall score. The higher the overall score, the greater the probability of a crack flow.

[0153] Here is a specific example: a shallow sea area near a coastline, with the following conditions:

[0154] Water depth: Average depth is 3 meters. Current velocity: Average current velocity is 1.5 meters per second. Wave height: Waves generated by a large storm have an average height of 2.5 meters. Current direction: Current direction changes rapidly over short periods due to the influence of storms. Bottom friction: Bottom friction is low; the seabed is sandy.

[0155] In this case, the weighting of examples is given, and the probability of crack flow occurring is estimated based on the weights.

[0156] Figure 8 An example of weight allocation is shown.

[0157] Water depth score: The water depth is 3 meters, which can be standardized to 0.6 (3 meters / maximum water depth 5 meters).

[0158] Flow velocity score: The flow velocity is 1.5 m / s, which can be normalized to 0.5 (1.5 m / s / maximum flow velocity 3 m / s).

[0159] Wave height score: The wave height is 2.5 meters, which can be normalized to 0.5 (2.5 meters / maximum wave height of 5 meters).

[0160] Flow direction score: Abrupt changes in flow direction, standardized to 0.2 (determined by comparing with historical rift occurrence statistics).

[0161] Bottom friction score: Bottom friction is relatively low, standardized to 0.8 (judged by comparing with historical rift occurrence statistics).

[0162] Based on the weighting, the overall score is calculated as follows: Overall score = (0.278*0.6) + (0.278*0.5) + (0.167*0.5) + (0.111*0.2) + (0.111*0.8) = 0.5003.

[0163] In this example case, the overall score is 0.5003, which means the probability of a crack flow is 50.03%.

[0164] The maximum water depth, maximum flow velocity, and maximum wave height refer to the historically observed maximum water depth and flow velocity in the predicted area. These values ​​are determined based on specific environmental conditions and historical data, and are used to standardize or normalize the predicted values ​​to convert them into relative scores between 0 and 1. The purpose of standardization or normalization is to transform data from different ranges into relative, uniform measures so that they can be compared in the analysis. Bottom friction and flow direction changes account for a relatively small proportion, requiring empirically standardized scores based on comparisons with historical fracture flow occurrences.

[0165] Furthermore, after determining the crack flow prediction score, the corresponding level of crack flow warning is determined based on the crack flow prediction score.

[0166] The raft flow risk assessment results will be used to issue tiered early warnings. Warning information will be released through the meteorological department. Areas with a raft flow risk below 50% will not be subject to action. Areas with a raft flow risk of 50%–70% will receive a blue raft flow warning, alerting residents. Areas with a risk of 70%–85% will receive a yellow raft flow warning, indicating a high risk requiring preventative measures. Areas with a risk of 85%–100% will receive a red raft flow warning, indicating an extremely high risk requiring immediate action to prevent people from approaching the area. Specific details are as follows... Figure 9 As shown.

[0167] In one alternative implementation, if the monitoring equipment detects that a crack flow is occurring, a red alert is immediately issued, the relevant area is sealed off, personnel are notified to evacuate, and emergency measures are taken simultaneously.

[0168] Since the changes in the data acquired by the monitoring equipment are uncertain, and the performance of the X-BEACH model may degrade over time, this embodiment may also include automated monitoring and calibration steps to maintain the reliability of crack flow prediction.

[0169] Furthermore, this embodiment may include:

[0170] Periodically acquire monitoring data within a set time period as detection data;

[0171] If the detection data is discontinuous or contains outliers, the status of the monitoring equipment will be detected and automatically repaired. If automatic repair is not possible, an alarm will be sent to maintenance personnel.

[0172] Specifically, this embodiment monitors in real time for discontinuities or anomalies in data acquisition. If a problem occurs, automated measures are taken: first, the device status is checked and repairs are attempted; if repairs are unsuccessful, an alarm is sent to a professional for manual handling. Once the problem is resolved and data acquisition returns to normal, the problematic data range is identified through automated steps such as interpolation, data correction, or supplementation using historical data.

[0173] Furthermore, this embodiment may include:

[0174] The predicted fracture flow data is periodically compared with the actual fracture flow data, and multiple indicators are used to quantify the difference between the predicted fracture flow data and the actual fracture flow data; the indicators include RMSE, correlation coefficient, and bias; each indicator corresponds to a threshold range;

[0175] When a difference in one of the indicators is found to correspond to the threshold range, an automatic calibration measurement is triggered and the parameters of the machine learning model are calibrated.

[0176] Specifically, the accuracy of the prediction results is checked periodically. RMSE, correlation coefficient, and bias are used to quantify the difference between the model output and the measured data. A threshold range is set for each indicator; when the value exceeds the threshold range, automatic calibration is triggered. The machine learning model calibration strategy is as follows:

[0177] The parameter tuning method employs Automated Machine Learning (AutoML) tools, which automatically select and adjust the parameters of the machine learning model, typically combining multiple optimization algorithms. After calibration begins, an initial solution is randomly generated. An adaptive strategy is used, adjusting the step size or range based on performance improvements during the calibration process. Calibration ends when performance metrics no longer improve. Parallel computation is used to improve efficiency during the calibration process. An anomaly detection mechanism is implemented; if calibration fails to converge or performance degrades, anomaly handling is triggered: attempting recalibration from a different starting point or notifying relevant personnel. The results of the automated calibration process are recorded, including changes in calibration parameters, performance evaluation results, and calibration success or failure information. A history is established to track the model calibration history and improvements. An integrated notification and reporting mechanism ensures relevant personnel are informed of the process status and results, including emails, SMS messages, and alerts on monitoring dashboards.

[0178] Based on calibration results and performance evaluations, the automated calibration process is continuously iterated and improved to enhance its efficiency and effectiveness. Ultimately, a robust automated calibration process is established to ensure that numerical models combined with machine learning predictions consistently maintain high accuracy and adapt to changing environments and new data.

[0179] This embodiment combines X-band radar, buoy tracer, UAV, and unmanned surface vessel (USV) (equipped with GNSS receiver, ADCP, and sonar system) for monitoring. The X-band radar acquires wind, wave, and current field data for the monitored area. When a rift occurs, tracer material is released, and combined with UAV imagery, data on the rift zone is obtained. For vertical flow structure and underwater topography data acquisition, an USV equipped with a GNSS receiver, ADCP, and sonar system is used. GNSS provides location information, ADCP provides flow pattern information, and the sonar system provides topographic information. For forecasting, multi-source monitoring data is combined, numerical models simulate future data, machine learning models are used for correction, and finally, the rift risk is assessed and forecasted.

[0180] This embodiment has the following beneficial effects:

[0181] (1) The integrated monitoring system of X-band radar-buoy tracer-UAV-unmanned surface vessel (equipped with GNSS receiver, ADCP and sonar system) obtains multi-source monitoring data and improves the accuracy of monitoring data.

[0182] (2) The unmanned surface vessel is equipped with ADCP and sonar system to solve the problem of underwater flow and terrain information in one stop. The unmanned surface vessel can dynamically change its position according to the plan to obtain accurate data of different areas. It is less affected by environmental interference, has a high degree of automation, and greatly improves efficiency.

[0183] (3) The method of combining numerical models with machine learning models can improve prediction accuracy and efficiency.

[0184] (4) Automated monitoring and calibration ensure the reliability of the entire monitoring and prediction system.

[0185] Reference Figure 10 This invention provides a system for monitoring and predicting cracked flows, comprising:

[0186] The first module is used to acquire historical monitoring data and real-time monitoring data obtained from current monitoring.

[0187] The second module is used to simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data using the X-BEACH model, and to obtain simulated crack flow data.

[0188] The third module is used to correct the simulated crack flow data using a machine learning model to obtain predicted crack flow data.

[0189] The fourth module is used to calculate the crack flow prediction score based on the predicted crack flow data, and to determine the corresponding level of crack flow warning based on the crack flow prediction score.

[0190] The specific implementation of this monitoring and prediction system is basically the same as the specific implementation of the monitoring and prediction method described above, and will not be repeated here.

[0191] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described monitoring and prediction method.

[0192] Specifically, electronic devices can be user terminals or servers.

[0193] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described monitoring and prediction methods.

[0194] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0195] This application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of an electronic device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the electronic device to perform... Figure 1 The method shown.

[0196] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0197] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0198] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0199] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0200] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0201] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0202] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0203] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0204] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for monitoring and predicting fracture flow, characterized in that, include: Acquire historical monitoring data and real-time monitoring data obtained from current monitoring; The X-BEACH model is used to simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data to obtain simulated crack flow data; The simulated fracture flow data is corrected using a machine learning model to obtain predicted fracture flow data; A crack flow prediction score is calculated based on the predicted crack flow data, and a corresponding level of crack flow warning is determined based on the crack flow prediction score. Obtain the real-time fracture flow data obtained from current monitoring, including: Real-time wind, wave and current field data of the monitored area are acquired using X-band marine radar. The crack flow data of the crack flow area in the monitoring region is obtained by using a drone; The underwater flow and topographic data of the monitored area were acquired using an unmanned surface vessel. The method further includes: The predicted fracture flow data is periodically compared with the actual fracture flow data, and multiple indicators are used to quantify the difference between the predicted fracture flow data and the actual fracture flow data; the indicators include RMSE, correlation coefficient, and bias; each indicator corresponds to a threshold range; When the difference of one of the indicators exceeds the corresponding threshold range, an automatic calibration measurement is triggered and the parameters of the machine learning model are calibrated. The water flow velocity within the region is divided into two components, u and v. The u component represents the water flow velocity parallel to the shore, and the v component represents the velocity component perpendicular to the shore. A positive v component indicates that the flow velocity direction is from the shore to the sea. If the X-band radar detects that the v component is positive on a certain velocity profile perpendicular to the shore, then the boundary is defined by extending the region along the shore to both sides until this profile can no longer be detected. The boundary is also defined based on the furthest extension of the positive v component from the shore, and it is determined that a rift flow has occurred in this region.

2. The method for monitoring and predicting fracture flow according to claim 1, characterized in that, The method of using the X-BEACH model to simulate the impending fracture flow based on the historical monitoring data and the real-time monitoring data to obtain simulated fracture flow data includes: The historical monitoring data and the real-time monitoring data are merged into a dataset, and outliers are removed, the format is standardized, the data units are standardized, and the data is smoothed using filtering techniques to remove noise or high-frequency oscillations. Then, principal component analysis is used to perform dimensionality reduction on the dataset to obtain the processed dataset. Set the time step, spatial grid, initial conditions, and model boundary conditions of the X-Beach2D model, and then configure the model configuration file, historical data file, and bottom friction parameter file to obtain the configured X-Beach2D model. The processed dataset is input into the configured X-Beach2D model to obtain the simulated fracture flow data.

3. The method for monitoring and predicting fracture flow according to claim 1, characterized in that, The step of using a machine learning model to correct the simulated fracture flow data to obtain predicted fracture flow data includes: The simulated fracture flow data is fused with the residuals of the corresponding historical periods using the machine learning model to obtain the predicted fracture flow data.

4. The method for monitoring and predicting fracture flow according to claim 1, characterized in that, The step of calculating a fracture flow prediction score based on the predicted fracture flow data and determining a corresponding level of fracture flow warning based on the fracture flow prediction score includes: Weights are assigned to each data item in the predicted fracture flow data, and the sum of the weights is 1. The standardized score of each data point is multiplied by its corresponding weight to obtain the actual score for each data point. Summing the actual scores of each term yields the fracture flow prediction score; The corresponding level of crack flow warning is determined based on the risk range in which the crack flow prediction score is located.

5. The method for monitoring and predicting fracture flow according to claim 1, characterized in that, The method further includes: Periodically acquire monitoring data within a set time period as detection data; If the detection data is discontinuous or contains outliers, the status of the monitoring equipment will be detected and automatically repaired. If automatic repair is not possible, an alarm will be sent to maintenance personnel.

6. A system for monitoring and predicting fracture flow, characterized in that, include: The first module is used to acquire historical monitoring data and real-time monitoring data obtained from current monitoring. The second module is used to simulate the impending crack flow based on the historical monitoring data and the real-time monitoring data using the X-BEACH model, and to obtain simulated crack flow data. The third module is used to correct the simulated crack flow data using a machine learning model to obtain predicted crack flow data. The fourth module is used to calculate the crack flow prediction score based on the predicted crack flow data, and to determine the corresponding level of crack flow warning based on the crack flow prediction score. This includes acquiring the real-time fracture flow data obtained from current monitoring, including: Real-time wind, wave and current field data of the monitored area are acquired using X-band marine radar. The crack flow data of the crack flow area in the monitoring region is obtained by using a drone; The underwater flow and topographic data of the monitored area were acquired using an unmanned surface vessel. The predicted fracture flow data is periodically compared with the actual fracture flow data, and multiple indicators are used to quantify the difference between the predicted fracture flow data and the actual fracture flow data; the indicators include RMSE, correlation coefficient, and bias; each indicator corresponds to a threshold range; When the difference of one of the indicators exceeds the corresponding threshold range, an automatic calibration measurement is triggered and the parameters of the machine learning model are calibrated. The water flow velocity within the region is divided into two components, u and v. The u component represents the water flow velocity parallel to the shore, and the v component represents the velocity component perpendicular to the shore. A positive v component indicates that the flow velocity direction is from the shore to the sea. If the X-band radar detects that the v component is positive on a certain velocity profile perpendicular to the shore, then the boundary is defined by extending the region along the shore to both sides until this profile can no longer be detected. The boundary is also defined based on the furthest extension of the positive v component from the shore, and it is determined that a rift flow has occurred in this region.

7. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the method as described in any one of claims 1 to 5.

Images