Reservoir project safety management system and method

Abstract

The invention relates to the technical field of hydraulic engineering, in particular to a reservoir engineering safety management system and method, and the method comprises the steps: obtaining a fracture tomographic image of a drainage gallery wall through a through-wall ground penetrating radar, and synchronously collecting the infiltration linear displacement dynamic data of a distributed optical fiber sensor; performing space-time alignment on the two types of data, constructing a fracture space topological relation and seepage conduction path model, and generating a defect-seepage coupling map; a coupling atlas analysis module and a space-time convolutional neural network module are used for identifying the sudden change trend of the displacement rate of the infiltration line, and a gallery structure failure early warning signal is triggered; after early warning, a reinforcement work order is automatically generated, the unmanned aerial vehicle is called to carry out a high-definition image rechecking task, a rechecking result is fed back to correct the seepage conduction path model, and dynamic self-correction is achieved. The method has the capabilities of accurate modeling, intelligent early warning and real-time closed-loop regulation and control, and can significantly improve the structural safety management level of reservoir engineering.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy projects, and in particular to a reservoir project safety management system and method. Background Art

[0002] As a major infrastructure crucial to the national economy and people's livelihood, the structural safety of reservoir projects directly impacts flood control and storage, water resource scheduling, and the safety and stability of downstream areas. Drainage corridors, key structures within reservoir gravity or earth-rock dams, are crucial channels for guiding seepage, mitigating seepage pressure, and preventing piping and structural degradation at the dam foundation. During long-term operation, microcracks may develop in the corridor walls due to factors such as concrete aging, seepage erosion, and thermal stress. These microcracks, under the influence of high water pressure, can form potential seepage channels, further inducing deformation, abnormal displacement, and even overall dam instability.

[0003] Existing reservoir corridor safety monitoring methods primarily rely on single-point infiltration line monitoring or discrete crack observation, making it difficult to comprehensively perceive the spatial distribution of wall cracks and seepage behavior. While some methods incorporate fiber optic sensors, they lack the ability to structurally model the data and reveal the dynamic coupling between crack connectivity and seepage conduction paths. Furthermore, most systems lack predictive mechanisms and real-time response capabilities for sudden changes. Early warnings rely on manual intervention, resulting in delayed responses and difficulty in dynamically correcting the models. This leads to problems such as misjudgments and missed detections, limiting their adaptability and practicality in high-risk hydraulic structure scenarios. Summary of the Invention

[0004] The present invention provides a reservoir engineering safety management system and method, which realizes the coordinated monitoring, intelligent identification and early warning disposal of drainage corridor structural hidden dangers and seepage status.

[0005] The reservoir project safety management method includes the following steps: S1: Use a through-wall ground-penetrating radar to obtain tomographic images of cracks inside the drainage corridor wall, and simultaneously use distributed fiber optic sensors deployed on the back surface of the retaining wall to collect dynamic data on the displacement of the infiltration line; S2: performing spatiotemporal alignment of the fracture tomography image and the dynamic displacement data of the infiltration line, constructing a seepage conduction path model based on the spatial topological relationship of the fracture, and outputting a defect-seepage coupling map; S3: Inputting the defect-seepage coupling map into the seepage line mutation prediction model. If the seepage line mutation prediction model identifies that the displacement rate of the seepage line exceeds the dynamic threshold within two consecutive monitoring cycles and spreads along the seepage conduction path, a corridor structure failure warning signal is triggered; S4: A reinforcement work order including positioning coordinates is automatically generated based on the corridor structure failure warning signal, and a drone is simultaneously started to conduct high-definition image review of the warning area, and the review results are fed back to the seepage conduction path model for self-correction.

[0006] Optionally, the S1 includes: S11: a scanning track grid of a through-the-wall ground penetrating radar is laid out on the surface of the drainage gallery wall, and the through-the-wall ground penetrating radar is controlled to move and scan along the track grid at a constant speed of 0.2 m / s to obtain an initial crack tomography image inside the wall; S12: Based on the crack distribution density in the initial crack tomography image, the coordinates of the infiltration line section on the back surface of the retaining wall that needs to be monitored are determined, and the distributed optical fiber sensor in the corresponding section is activated to enter the high-frequency sampling mode; S13: Activate distributed fiber optic sensors to collect dynamic displacement data of the infiltration line in key monitoring areas, and simultaneously control the through-wall ground penetrating radar to perform a secondary fine scan of the high-density fracture area to obtain high-resolution fracture tomography images; S14: Fusion of high-frequency acquired dynamic data of infiltration line displacement and high-resolution fracture tomography images to generate a time-stamped synchronous monitoring dataset.

[0007] Optionally, the S2 includes: S21: performing a spatiotemporal alignment operation on the fracture tomography image and the dynamic displacement data of the infiltration line, unifying the spatial coordinates of the two types of data into the engineering absolute coordinate system through the Beidou spatiotemporal positioning chip, synchronizing the timestamps to the UTC standard time, and outputting the fracture tomography image with the unified coordinates and the dynamic displacement data of the infiltration line with the synchronized time stamp; S22: Extract the three-dimensional coordinates and opening parameters of the fracture endpoints from the fracture tomography image based on the unified coordinates, and construct a fracture space topological relationship network diagram; S23: Mapping the dynamic displacement data of the infiltration line at the synchronous time scale to the topological relationship network diagram of the fracture space, calculating the seepage conduction probability of each connected edge under the action of the infiltration line pressure gradient, and generating a seepage conduction path model with weight parameters; S24: Based on the generated seepage conduction path model, a defect-seepage coupling map is output with the conduction direction, risk weight, and corresponding infiltration line displacement vector marked.

[0008] Optionally, the invasion line mutation prediction model includes a coupled graph analysis module, a spatiotemporal convolutional neural network module and a mutation probability calculation module.

[0009] Optionally, the S3 includes: S31: extracting the seepage conduction path weight, the infiltration line displacement vector, and the conduction direction parameter in the defect-seepage coupling map through a coupling map analysis module to generate structured prediction input data; S32: Input the structured prediction input data into the spatiotemporal convolutional neural network module, fuse the historical working condition data to construct a three-dimensional feature tensor, and output the spatiotemporal change feature matrix of the infiltration line displacement rate.

[0010] Optionally, the S3 further includes: S33: performing residual analysis on the spatiotemporal variation characteristic matrix of the seepage line displacement rate and the real-time collected dynamic data of the seepage line displacement through a mutation probability calculation module to generate a predicted value of the seepage line displacement rate and a mutation probability index for each seepage conduction path; S34: When the predicted displacement rate of the infiltration line of the seepage conduction path exceeds the dynamic threshold within two consecutive monitoring cycles, and its mutation probability index is greater than the preset diffusion threshold, the corridor structure failure warning signal is triggered.

[0011] Optionally, the S4 includes: S41: Analyze the risk path location coordinates based on the corridor structure failure warning signal, call the emergency response plan library to match the historical reinforcement plan for the location coordinates, and generate a reinforcement work order including the location coordinates, risk level, and recommended reinforcement measures; S42: Send positioning coordinates and image acquisition instructions to the UAV control terminal, start the UAV to perform high-definition image review tasks in the warning area, and obtain millimeter-level precision review images including crack width and expansion direction.

[0012] Optionally, the S4 further includes: S43: Bind the HD image review results to the reinforcement work order to generate a comprehensive disposal report, push it to the responsible person's mobile terminal, and start the disposal countdown; S44: Extract the crack extension characteristic parameters from the high-definition image review results and feed them back to the seepage conduction path model to correct the weight coefficients of the crack space topology relationship, completing the model self-correction.

[0013] The reservoir engineering safety management system is used to implement the above-mentioned reservoir engineering safety management method and includes the following modules: Crack image acquisition module: used to control the movement of the through-wall ground penetrating radar along the scanning track grid to obtain tomographic images of cracks inside the drainage corridor wall, and synchronously collect the spatial position and timestamp information of the equipment during operation; The infiltration line displacement monitoring module is used to activate the distributed optical fiber sensors deployed in the key monitoring section of the back water surface of the retaining wall to collect the dynamic data of the infiltration line displacement; Defect-seepage coupling map generation module: used to perform spatiotemporal alignment on the fracture tomography image and the infiltration line displacement dynamic data, construct a seepage conduction path model, and output a defect-seepage coupling map; The infiltration line mutation prediction model module is used to input the defect-seepage coupling map into the infiltration line mutation prediction model, integrate historical operating condition data to generate a spatiotemporal variation characteristic matrix of the infiltration line displacement rate, and output the predicted value of the infiltration line displacement rate and mutation probability index of each seepage conduction path, and trigger the corridor structure failure warning signal based on the dynamic threshold; Reinforcement work order generation module: used to analyze the risk path location coordinates based on the corridor structure failure warning signal, call the emergency response plan library to match the historical reinforcement plan for the location coordinates, and generate a reinforcement work order including the location coordinates, risk level and recommended reinforcement measures; UAV image review module: used to control the UAV to perform high-definition image review tasks according to the positioning coordinates in the reinforcement work order, collect millimeter-level review images including crack width and expansion direction, and extract crack expansion characteristic parameters; Seepage conduction path model self-correction module: used to feed back the crack extension characteristic parameters to the seepage conduction path model, correct the weight coefficient in the crack space topological relationship, and realize model self-correction.

[0014] Beneficial effects of the present invention: The present invention uses a through-wall ground-penetrating radar and distributed fiber optic sensors to collaboratively acquire tomographic images of cracks inside drainage corridor walls and dynamic data on infiltration line displacements. It also integrates high-resolution radar scanning and high-frequency fiber optic sampling to form a synchronous monitoring data set with a timestamp. This effectively enhances the ability to jointly perceive microscopic structural defects and seepage conditions within reservoir projects, and enables high-precision capture of potential instability hazards.

[0015] The present invention constructs a network diagram of the spatial topological relationship of cracks based on spatiotemporal alignment, and superimposes the infiltration line displacement data to establish a seepage conduction path model with weight parameters. It further outputs a defect-seepage coupling map with the conduction direction, risk weight and infiltration line displacement vector marked. Combining the coupling map analysis module with the spatiotemporal convolutional neural network model can accurately predict the mutation trend and timely identify the risk of structural failure, effectively solving the problems of fuzzy coupling relationship between cracks and seepage and delayed early warning in traditional monitoring methods.

[0016] The present invention automatically generates a reinforcement work order after a sudden warning and triggers a drone to conduct high-definition image review of the warning area. It corrects the weight coefficient of the fracture space topology relationship in the seepage conduction path model based on the characteristic parameters of the crack expansion, realizes the dynamic self-correction and feedback evolution of the model, and constructs a closed-loop mechanism for structure identification, warning judgment and model updating, which significantly enhances the intelligence, responsiveness and robustness of reservoir project safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1Schematic diagram of a method flow in an embodiment of the present invention; Figure 2 Schematic diagram of the system flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0020] It should be noted that references in the specification to "one embodiment," "an embodiment," "exemplary embodiments," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment will include such specific features, structures, or characteristics. Furthermore, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0021] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0022] like Figure 1 As shown in FIG, the reservoir project safety management method includes the following steps: S1: A through-the-wall ground-penetrating radar is used to obtain tomographic images of cracks inside the drainage corridor wall. Distributed fiber optic sensors deployed on the back surface of the retaining wall are used to collect dynamic displacement data of the infiltration line. Specifically: S11, Ground-penetrating Radar Scanning Track Grid Layout and Initial Crack Image Acquisition: A scanning track grid was installed on the inner wall surface of the drainage corridor. The track grid is a rectangular grid with 0.3-meter horizontal and vertical spacing. The total track length is customized according to the corridor wall dimensions and uses a magnetic mounting structure to adapt to the concrete surface. The ground-penetrating radar is installed on a track-mounted trolley and is equipped with a high-frequency antenna array with a center frequency of 1.2 GHz and a spacing of 5 cm. It uses linear polarization transmission, has a penetration depth of 3 meters or more, and can identify cracks with a resolution of 2 mm.

[0023] The ground-penetrating radar scans the entire wall at a constant speed of 0.2 m / s along the track grid. Using an ultra-wideband pulse signal transmission and echo reception mechanism, it images sensitive areas within the wall's dielectric constant, generating tomographic images of the wall's initial cracks. The data acquisition process is cached in real time by the FPGA control module and transmitted to an on-site industrial computer for time-domain inversion and reconstruction. The output is a grayscale tomographic image grid matrix with a resolution of 256×256.

[0024] S12, Identification of Key Infiltration Sections and High-Frequency Sensor Activation: After acquiring the initial fracture tomography image, a density clustering algorithm is used to cluster and identify areas with high dielectric gradients within the image, extracting the coordinates of sections with dense fracture distribution. These extracted coordinates are mapped to the backwater surface of the retaining wall in the corridor structure diagram to identify key infiltration line sections for monitoring. Distributed fiber optic sensors (using the BOTDA principle) are pre-embedded in these sections. The system sends activation commands via the serial port controller, activating them in high-frequency sampling mode, increasing the sampling frequency from 1Hz in basic mode to 10Hz, improving data timeliness and sensitivity to seepage changes.

[0025] S13, Synchronous High-Frequency Monitoring and Secondary Fine Radar Scanning: In key monitoring areas, distributed fiber-optic sensors collect real-time dynamic displacement data from the infiltration line. The Brillouin frequency shift at each sampling point on the sensing fiber is recorded and converted into microstrain data along the fiber axis. The system calculates the first-order derivative of the continuous sampling results as a displacement rate reference.

[0026] At the same time, the through-the-wall ground-penetrating radar was repositioned to an area with dense fractures in the initial image for a second scan. This round of scanning employed a sampling strategy with a lower rate (0.05 m / s) and a higher pulse overlap ratio (≥80%), further improving the signal-to-noise ratio and image resolution, producing a high-resolution fracture tomography image with a resolution of 512 × 512 pixels.

[0027] S14, Synchronous Monitoring Dataset Fusion and Time Stamping: The acquired dynamic displacement data of the infiltration line are aligned with the high-resolution fracture tomography images according to the timestamps and organized using a unified data structure. Each data set includes: timestamp (accurate to the second), the number and position coordinates of the infiltration line displacement sensor point, the microstrain value and displacement rate at the corresponding time point, the synchronously acquired fracture image sub-image (sliced ​​by coordinate region), and structural parameters such as the maximum fracture length and density index in the fracture image.

[0028] Finally, a synchronous monitoring data set with time stamps is formed, which serves as the basic input data for subsequent seepage path modeling and failure early warning.

[0029] S2: Temporally and spatially align the fracture tomography image with the dynamic displacement data of the infiltration line, construct a seepage conduction path model based on the spatial topological relationship of the fracture, and output a defect-seepage coupling map, specifically: S21, Spatiotemporal Alignment: Performs spatiotemporal alignment of the fracture tomography images and the dynamic displacement data of the infiltration line output from S1. The system integrates the Beidou spatiotemporal positioning chip (model BD980), which offers centimeter-level 3D spatial positioning accuracy and nanosecond-level time synchronization. The specific alignment process is as follows:

[0030] Unified spatial coordinates: The two-dimensional pixel coordinates in the fracture tomography image are mapped to the engineering absolute coordinate system (using the ENU projection format) through a calibration transformation matrix. At the same time, the positions of each sampling point collected by the distributed fiber optic sensor are also converted to the same coordinate system to ensure a one-to-one correspondence between the two types of data in the spatial dimension. Timestamp synchronization: Extract the original timestamps of the two types of data and synchronize them to the UTC standard time output by the Beidou system to achieve high-precision time alignment.

[0031] The output results are: fracture tomography images with unified coordinates and dynamic displacement data of the infiltration line with synchronized time scale.

[0032] S22, construction of the fracture space topology relationship network diagram: After completing the spatiotemporal alignment, an image analysis method based on structural edge extraction and topological skeleton extraction is used to analyze the fracture tomography image with unified coordinates: Extract the three-dimensional coordinates of the first and last endpoints of each identifiable crack; The opening parameters of each crack (unit: mm) are calculated using the image grayscale gradient inversion algorithm; Whether there is a connectivity relationship between adjacent endpoints is determined based on spatial proximity and geometric continuity.

[0033] Thus, a topological relationship network diagram of the crack space is constructed, where: Nodes represent the endpoints of cracks; Edges represent paths with potential percolation connectivity between adjacent endpoints.

[0034] This network diagram provides basic topological structure support for subsequent path modeling and risk propagation analysis.

[0035] S23, Seepage conduction path model construction: Seepage conduction path model generation The dynamic displacement data of the infiltration line at the synchronous time scale is mapped to the above-mentioned fracture space topology network diagram to construct the seepage conduction path model. The specific process is as follows:

[0036] The spatial position of each infiltration line sampling point is mapped to the relevant fracture node in the topological network diagram; Calculate the displacement difference and spatial distance between the sensing points to form an approximate infiltration line pressure gradient; Combining the fracture aperture and connection distance, a seepage conduction probability model of each connected edge under the action of the infiltration line pressure gradient is established.

[0037] The model is represented as a weighted directed graph and is defined as follows: ; in, is the set of crack endpoints, is the set of connected edges between adjacent endpoints, is the weight set of edges, representing the percolation conduction probability, is the set of conduction directions of the edge, is the set of displacement vectors of the edge's infiltration line. Calculated as:

[0038] ; in, is the experience adjustment coefficient, is the rock mass permeability coefficient, Fracture segment The average opening of is the Euclidean distance between the two endpoints, is the displacement difference between the sensing points on the wetting line, is the sampling time interval.

[0039] Defined as: ; in is the three-dimensional coordinate of the crack endpoint, and sgn indicates the positive or negative direction.

[0040] It is calculated by weighting the microstrain direction and amplitude of the two connected sensing points, indicating the dominant deformation direction and magnitude induced by seepage on the current path.

[0041] The final generated percolation conduction path model is used for subsequent coupling map construction and mutation identification.

[0042] S24: Defect-percolation coupling map generation: Based on the percolation conduction path model, the final defect-percolation coupling map is output. The map includes the following elements:

[0043] Conduction direction: determined by the percolation conduction path model express; Risk weight: by edge weight Calculated by comprehensive evaluation of crack opening; Wetting line displacement vector: derived from the seepage conduction path model It indicates the dynamic trend of local displacement; The graph is organized and stored in the form of a graph database, with timestamp identification and spatial positioning capabilities. It serves as the input basis for subsequent mutation identification in S3, and also supports self-correction operations in S4 based on the review results.

[0044] S3: Input the defect-seepage coupling map into the seepage line mutation prediction model. If the seepage line mutation prediction model identifies that the displacement rate of the seepage line exceeds the dynamic threshold within two consecutive monitoring cycles and spreads along the seepage conduction path, a corridor structure failure warning signal is triggered. Specifically: The invasion line mutation prediction model includes a coupled graph analysis module, a spatiotemporal convolutional neural network module, and a mutation probability calculation module.

[0045] S31, structured prediction input data generation: The coupled graph analysis module analyzes the defect-percolation coupled graph output from S2 edge by edge to extract the key elements that constitute the prediction model input, including: Percolation path weight (from the edge weight in the percolation path model) ; The displacement vector of the wetting line (indicates the actual displacement direction and amplitude along the conduction path, marked as ); Conduction direction parameter (i.e. the directed edge direction vector defined in the graph ).

[0046] To facilitate neural network processing, the system encodes the above information into a structured data set in the following format: Each percolation conduction path As a sample unit The characteristic fields include: 、 (split into 3-dimensional components), (split into 3-dimensional components), Current timestamp, The coordinates of the path's start and end points.

[0047] The system updates the structured prediction input data for each monitoring period (such as 10-minute intervals) and provides it to subsequent models as an input sample sequence.

[0048] S32, Spatiotemporal Feature Tensor Construction and Feature Extraction: The structured prediction input data generated in S31 is fed into a spatiotemporal convolutional neural network module, which uses a three-dimensional convolutional neural network architecture (3D-CNN) that integrates spatial structural features with time series dynamics. The network consists of the following three layers:

[0049] 1. Input layer: accepts The input tensor of is the number of percolation conduction path samples, is the channel dimension of each sample (including path weights, vector components, etc.), is the length of the continuous time window (e.g. 12 monitoring cycles); 2. Spatiotemporal convolution layer: using 3 The convolution kernel is used to perform joint feature extraction and extract joint patterns between paths and time series; 3. Output layer: Output a tensor-shaped spatiotemporal variation feature matrix of the displacement rate of the infiltration line, with a dimension of , where each element represents the The percolation conduction path is In addition, the module integrates historical operating data (such as water level and rainfall in the same period) into additional channels for convolution processing to improve prediction accuracy.

[0050] S33, calculation of mutation probability index and generation of rate prediction value: Through the mutation probability calculation module, the outputted spatiotemporal variation characteristic matrix of the infiltration line displacement rate is compared and analyzed path by path with the real-time collected dynamic data of the infiltration line displacement.

[0051] 1. Residual analysis: For each seepage conduction path, calculate the predicted value in the latest two monitoring cycles With real-time observations The residuals are: ; 2. Calculation of mutation probability index: Based on the residual value sequence and historical volatility level, the exponential sliding statistical model (such as Z-score, normalized relative difference) is used to calculate the mutation probability index of the path. , its expression is as follows: ; in, For path The sliding average of the residual series, For path Sliding standard deviation of the residual series.

[0052] The final output is: the predicted displacement rate of the infiltration line of each seepage conduction path , and its corresponding mutation probability index .

[0053] S34, early warning determination and signal triggering: When the following two conditions are met, the corridor structure failure early warning signal can be triggered: During two consecutive monitoring cycles, the predicted displacement rate of the infiltration line of one or more seepage conduction paths exceeds the dynamic threshold (e.g., the historical 95th percentile) automatically updated by the system based on historical statistics; The mutation probability index corresponding to the same path exceeds the system preset diffusion threshold (for example, the diffusion threshold is set to 2.5).

[0054] If the above conditions are met, the system generates a corridor structure failure warning signal through the warning interface and pushes it to the data fusion control platform as the triggering basis for S4.

[0055] S4: Automatically generate reinforcement work orders including positioning coordinates based on corridor structure failure warning signals, simultaneously launch drones to conduct high-definition image review of the warning area, and feed the review results back to the seepage conduction path model for self-correction. Specifically: S41, Reinforcement Work Order Generation: When the corridor structure failure warning signal from step S3 is triggered, the system automatically analyzes the risk path location coordinates marked in the warning signal and extracts the corresponding 3D engineering absolute coordinates (ENU format). The system then initiates the reinforcement response process, which includes the following steps:

[0056] Plan matching: The system calls a pre-loaded emergency response plan library, which is indexed by spatial regions and stores historical reinforcement measures, construction records, and response time information. Each plan entry is associated with a central coordinate point, applicable risk level range, applicable lithologic environment label, and recommended treatment process.

[0057] Matching mechanism: Using a fast spatial matching algorithm based on KD-Tree, the current risk path location coordinates are used as the query point to retrieve the historical plan area in the emergency response plan database that is closest in spatial distance and has the best match between risk level and lithologic conditions. The matching process uses the following objective function:

[0058] ; in, Indicates the best matching historical plan area, Locate the coordinates for the current risk path, The first The center coordinates of the plan, is the Euclidean distance, which indicates the spatial proximity between the current risk path location coordinates and the historical plan coordinates. Indicates the The warning level difference penalty item corresponding to the historical area. If the historical warning level of the area is lower than the current level, the value will increase. Indicates the The degree of difference in environmental factors of different regions, such as whether the lithology and hydrological conditions match, is usually obtained by a discrete scoring function (e.g., perfect match is recorded as 0 and mismatch is recorded as 1). and is a weighting coefficient that controls the balance between spatial distance and warning level / environmental factors.

[0059] Reinforcement Work Order Generation: Outputs a reinforcement work order with the following fields: Risk path positioning coordinates; Risk level (e.g., Level I / II / III, determined based on the mutation probability index and pathway weight); Recommend reinforcement measures (such as grouting sealing, carbon fiber reinforcement, double-layer composite lining); response time limits; Responsible person assignment information.

[0060] S42, HD image verification task execution: To verify the actual damage status of the warning area, the system sends the following verification task parameters to the UAV control terminal: Risk path positioning coordinates; Execution instruction type: image acquisition; Shooting requirements: millimeter-level resolution and adaptability to low-light conditions; Route planning mode: surround scanning + vertical fixed-point overlooking shooting.

[0061] The system controls the drone, equipped with a high-resolution visible light and near-infrared composite camera, to initiate a high-definition image review task for the designated coordinate area. The captured images are transmitted in real time to the on-site edge server for processing. The processing flow includes:

[0062] Crack width extraction: using multi-scale edge enhancement and sub-pixel fitting methods, the resolution reaches 0.3mm; Extension direction calculation: Based on the main crack axis fitting, output direction vector and angle value; Interference filtering: Eliminate abnormal samples such as occlusion, reflection, and water marks.

[0063] The final output is the review image in the form of images and structured data for subsequent analysis and processing.

[0064] S43, Comprehensive Disposal Report Generation and Delivery: The system combines the reinforcement work order generated in S41 with the high-definition image review results obtained in S42 to create a structured comprehensive disposal report, including: Risk path positioning coordinates; Risk level; Crack width and propagation direction (with image annotation); Recommend reinforcement measures; Response time countdown module (such as setting a 24-hour countdown); Information on the responsible unit and person; Summary of historical records of disposal of similar risk areas.

[0065] The comprehensive disposal report is pushed to the responsible person's mobile terminal through the DMS (Decision Management System) interface, such as mobile phone APP, tablet-specific client, etc.; at the same time, the system starts the corresponding disposal countdown process, supporting supervision and scheduling, response tracking and automatic reminder mechanism.

[0066] S44, model self-correction: To improve the model's subsequent prediction accuracy and dynamic adaptability, the system feeds back the crack propagation characteristic parameters from the high-definition image verification results obtained in S42 to the previously constructed seepage conduction path model and performs model self-correction operations, specifically including the following: 1. Mapping and positioning: Based on the coordinate information extracted from the fracture image, the extended fracture is matched with the edges in the seepage conduction path model; 2. Weight coefficient correction: the weight coefficient of the topological relationship of the crack space corresponding to the path For dynamic correction, the following weighted update formula is used: ; in, represents the weight in the original seepage conduction path model, The actual transmission strength index derived from the image crack width and expansion rate, Indicates the correction of the fusion factor to control the proportion of new and old information.

[0067] 3. Synchronous update of direction vector: If the deviation between the expansion direction shown in the image and the conduction direction in the model exceeds a set threshold (such as 15°), the conduction direction parameter vector of the path is synchronously updated.

[0068] 4. Topology fine-tuning: If new cracks or fractures are detected in the review image, the system automatically adds nodes and edges to expand the original topological relationship map of the crack space. This self-correction mechanism provides a more reliable data foundation for the next round of infiltration line mutation prediction, achieving a data-driven closed-loop model evolution.

[0069] like Figure 2 As shown, the reservoir engineering safety management system is used to implement the above-mentioned reservoir engineering safety management method, including the following modules: Crack image acquisition module: used to control the movement of the through-wall ground penetrating radar along the scanning track grid to obtain tomographic images of cracks inside the drainage corridor wall, and synchronously collect the spatial position and timestamp information of the equipment during operation; Wetting line displacement monitoring module: used to activate distributed fiber optic sensors deployed in the key monitoring section on the back surface of the retaining wall to collect dynamic data on the displacement of the wetting line; Defect-seepage coupling map generation module: used to perform spatiotemporal alignment of fracture tomography images and dynamic displacement data of infiltration lines, construct a seepage conduction path model, and output defect-seepage coupling maps; The seepage line mutation prediction model module is used to input the defect-seepage coupling map into the seepage line mutation prediction model, integrate historical operating condition data to generate the spatiotemporal variation characteristic matrix of the seepage line displacement rate, and output the predicted value of the seepage line displacement rate and mutation probability index of each seepage conduction path. It triggers the corridor structure failure warning signal based on the dynamic threshold; Reinforcement work order generation module: used to analyze the risk path location coordinates based on the corridor structure failure warning signal, call the emergency response plan library to match the historical reinforcement plan for the location coordinates, and generate a reinforcement work order including the location coordinates, risk level and recommended reinforcement measures; UAV image review module: used to control the drone to perform high-definition image review tasks based on the positioning coordinates in the reinforcement work order, collect millimeter-level review images including crack width and propagation direction, and extract crack propagation characteristic parameters; Seepage conduction path model self-correction module: used to feed back the crack extension characteristic parameters to the seepage conduction path model, correct the weight coefficients in the crack space topology relationship, and realize model self-correction.

[0070] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0071] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A reservoir engineering safety management method, characterized in that: The following steps are involved: S1: Use a through-wall ground-penetrating radar to obtain tomographic images of cracks inside the drainage corridor wall, and simultaneously use distributed fiber optic sensors deployed on the back surface of the retaining wall to collect dynamic data on the displacement of the infiltration line; S2: performing spatiotemporal alignment of the fracture tomography image and the dynamic displacement data of the infiltration line, constructing a seepage conduction path model based on the spatial topological relationship of the fracture, and outputting a defect-seepage coupling map; S3: Inputting the defect-seepage coupling map into the seepage line mutation prediction model. If the seepage line mutation prediction model identifies that the displacement rate of the seepage line exceeds the dynamic threshold within two consecutive monitoring cycles and spreads along the seepage conduction path, a corridor structure failure warning signal is triggered; S4: A reinforcement work order including positioning coordinates is automatically generated based on the corridor structure failure warning signal, and a drone is simultaneously started to conduct high-definition image review of the warning area, and the review results are fed back to the seepage conduction path model for self-correction.

2. The reservoir engineering safety management method according to claim 1, characterized in that: Said S1 comprises: S11: a scanning track grid of a through-the-wall ground penetrating radar is laid out on the surface of the drainage gallery wall, and the through-the-wall ground penetrating radar is controlled to move and scan along the track grid at a constant speed of 0.2 m / s to obtain an initial crack tomography image inside the wall; S12: Based on the crack distribution density in the initial crack tomography image, the coordinates of the infiltration line section on the back surface of the retaining wall that needs to be monitored are determined, and the distributed optical fiber sensor in the corresponding section is activated to enter the high-frequency sampling mode; S13: The distributed fiber optic sensors are activated to collect dynamic displacement data of the infiltration line in the key monitoring section. At the same time, the through-wall ground penetrating radar is controlled to perform a secondary fine scan of the fracture area to obtain fracture tomography images. S14: Fusion of high-frequency acquired infiltration line displacement dynamic data and fracture tomography images to generate a time-stamped synchronous monitoring dataset.

3. The reservoir engineering safety management method according to claim 2, characterized in that: The S2 includes: S21: performing a spatiotemporal alignment operation on the fracture tomography image and the dynamic displacement data of the infiltration line, unifying the spatial coordinates of the two types of data into the engineering absolute coordinate system through the Beidou spatiotemporal positioning chip, synchronizing the timestamps to the UTC standard time, and outputting the fracture tomography image with the unified coordinates and the dynamic displacement data of the infiltration line with the synchronized time stamp; S22: Extract the three-dimensional coordinates and opening parameters of the fracture endpoints from the fracture tomography image based on the unified coordinates, and construct a fracture space topological relationship network diagram; S23: Mapping the dynamic displacement data of the infiltration line at the synchronous time scale to the topological relationship network diagram of the fracture space, calculating the seepage conduction probability of each connected edge under the action of the infiltration line pressure gradient, and generating a seepage conduction path model with weight parameters; S24: Based on the generated seepage conduction path model, a defect-seepage coupling map is output with the conduction direction, risk weight, and corresponding infiltration line displacement vector marked.

4. The reservoir engineering safety management method according to claim 3, characterized in that: The invasion line mutation prediction model includes a coupled graph analysis module, a spatiotemporal convolutional neural network module and a mutation probability calculation module.

5. The reservoir engineering safety management method according to claim 4, characterized in that: The S3 includes: S31: extracting the seepage conduction path weight, the infiltration line displacement vector, and the conduction direction parameter in the defect-seepage coupling map through a coupling map analysis module to generate structured prediction input data; S32: Input the structured prediction input data into the spatiotemporal convolutional neural network module, fuse the historical working condition data to construct a three-dimensional feature tensor, and output the spatiotemporal change feature matrix of the infiltration line displacement rate.

6. The reservoir engineering safety management method according to claim 5, characterized in that: Said S3 further comprises: S33: performing residual analysis on the spatiotemporal variation characteristic matrix of the seepage line displacement rate and the real-time collected dynamic data of the seepage line displacement through a mutation probability calculation module to generate a predicted value of the seepage line displacement rate and a mutation probability index for each seepage conduction path; S34: When the predicted displacement rate of the infiltration line of the seepage conduction path exceeds the dynamic threshold within two consecutive monitoring cycles, and its mutation probability index is greater than the preset diffusion threshold, the corridor structure failure warning signal is triggered.

7. The reservoir engineering safety management method according to claim 6, characterized in that: The S4 includes: S41: Analyze the risk path location coordinates based on the corridor structure failure warning signal, call the emergency response plan library to match the historical reinforcement plan for the location coordinates, and generate a reinforcement work order including the location coordinates, risk level, and recommended reinforcement measures; S42: Send positioning coordinates and image acquisition instructions to the UAV control terminal, start the UAV to perform high-definition image review tasks in the warning area, and obtain millimeter-level precision review images including crack width and expansion direction.

8. The reservoir engineering safety management method according to claim 7, characterized in that: Said S4 further comprises: S43: Bind the HD image review results to the reinforcement work order to generate a comprehensive disposal report, push it to the responsible person's mobile terminal, and start the disposal countdown; S44: Extract the crack extension characteristic parameters from the high-definition image review results and feed them back to the seepage conduction path model to correct the weight coefficients of the crack space topology relationship, completing the model self-correction.

9. A reservoir engineering safety management system, used to implement the reservoir engineering safety management method according to any one of claims 1 to 8, characterized in that: Includes the following modules: Crack image acquisition module: used to control the movement of the through-wall ground penetrating radar along the scanning track grid to obtain tomographic images of cracks inside the drainage corridor wall, and synchronously collect the spatial position and timestamp information of the equipment during operation; The infiltration line displacement monitoring module is used to activate the distributed optical fiber sensors deployed in the key monitoring section of the back water surface of the retaining wall to collect the dynamic data of the infiltration line displacement; Defect-seepage coupling map generation module: used to perform spatiotemporal alignment on the fracture tomography image and the infiltration line displacement dynamic data, construct a seepage conduction path model, and output a defect-seepage coupling map; The infiltration line mutation prediction model module is used to input the defect-seepage coupling map into the infiltration line mutation prediction model, integrate historical operating condition data to generate a spatiotemporal variation characteristic matrix of the infiltration line displacement rate, and output the predicted value of the infiltration line displacement rate and mutation probability index of each seepage conduction path, and trigger the corridor structure failure warning signal based on the dynamic threshold; Reinforcement work order generation module: used to analyze the risk path location coordinates based on the corridor structure failure warning signal, call the emergency response plan library to match the historical reinforcement plan for the location coordinates, and generate a reinforcement work order including the location coordinates, risk level and recommended reinforcement measures; UAV image review module: used to control the UAV to perform high-definition image review tasks according to the positioning coordinates in the reinforcement work order, collect millimeter-level review images including crack width and expansion direction, and extract crack expansion characteristic parameters; Seepage conduction path model self-correction module: used to feed back the crack extension characteristic parameters to the seepage conduction path model, correct the weight coefficient in the crack space topological relationship, and realize model self-correction.

Images