An artificial intelligence-based optical fiber embankment underwater piping seepage patrol system

Abstract

The application discloses a kind of based on artificial intelligence's optical fiber embankment underwater piping seepage patrol system, it is related to embankment patrol management technical field, including patrol management center, the patrol management center communication connection has following module, wherein: data acquisition module, for using optical fiber sensing network real-time monitoring the underwater piping seepage situation of embankment, and the data collected are preprocessed;Intelligent analysis and diagnosis module, based on artificial intelligence algorithm identifies the characteristic mode of piping seepage, judges whether there is danger.The application utilizes optical fiber sensing network real-time monitoring underwater piping seepage situation, can quickly capture the physical quantity data related to seepage, analyzes data by AI big model based on artificial intelligence algorithm, accurately identifies the characteristic mode of piping seepage, judges whether there is danger, compared with artificial patrol, greatly improve the efficiency and accuracy of danger discovery, can discover hidden danger in time in piping early stage, effectively reduce the risk of embankment breach.

Description

Technical Field

[0001] This invention relates to the field of dike patrol and management technology, specifically to an artificial intelligence-based fiber optic dike underwater piping and seepage inspection system. Background Technology

[0002] Dikes are important water conservancy facilities, and their safety is directly related to the flood control safety of the surrounding areas and the safety of people's lives and property. Piping often occurs in the foundation of the dike or underwater parts, and there are no obvious surface signs in the early stage, making it difficult to detect by traditional methods. If piping is not dealt with in time, it may lead to the collapse of the dike. Traditional dike inspection mainly relies on manual inspection, which is inefficient, labor-intensive, and difficult to detect hidden seepage and piping problems in time. Fiber optic monitoring technology can monitor without damaging the dike structure, reduce the workload of manual inspection, and improve management efficiency.

[0003] In existing technologies, during the inspection of piping and seepage, the output of large AI models is mostly a binary classification result of safety / danger, lacking quantitative analysis of the evolution path of the hazard. The decision-making scheme relies heavily on the experience of maintenance personnel, lacks scientific basis, and is prone to resource waste or inadequate handling. Therefore, the problem we want to solve is to render the state of the dike in real time through a digital twin engine to form a visual decision sandbox, and combine reinforcement learning and multi-objective optimization algorithms, inputting the current hazard state and resource constraints, and outputting the optimal decision scheme to form a closed-loop decision support process. To this end, we propose an artificial intelligence-based fiber optic dike underwater piping and seepage inspection system. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage, in order to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] An artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage includes a monitoring management center, which is communicatively connected to the following modules:

[0007] The data acquisition module is used to monitor underwater piping and seepage in the dike in real time using fiber optic sensor networks, and to preprocess the acquired data.

[0008] The intelligent analysis and diagnosis module uses artificial intelligence algorithms to identify the characteristic patterns of piping and seepage, determine whether there is a danger, and combine digital twin technology to build a digital twin model of the dike, analyze and predict the evolution path of the danger, and simulate the development trend of the danger under different conditions.

[0009] The visualization decision support module is used to render the digital twin model of the dike in real time, dynamically display the changes in the dike's status, build a visualization decision sandbox, and generate and display the optimal decision-making scheme.

[0010] The feedback optimization module is used to feed back the execution effect of the optimal decision-making plan to the system. Based on the actual emergency response results and feedback information, the system is evaluated and optimized to form a closed-loop decision support process.

[0011] A further improvement to the technical solution of the present invention is that the data acquisition module includes:

[0012] Fiber optic sensor networks are laid at key sections of the dike to detect underwater piping and seepage in real time and obtain physical data related to the seepage.

[0013] It receives data collected by fiber optic sensor networks and performs preprocessing operations, including data cleaning, noise reduction, and format conversion, to ensure data integrity and availability.

[0014] The preprocessed data is transmitted to the patrol management center via communication equipment and stored in the database to provide data support for subsequent analysis.

[0015] A further improvement of the technical solution of the present invention is that: the intelligent analysis and diagnosis module includes a hazard diagnosis unit and a hazard evolution prediction unit;

[0016] The hazard diagnosis unit analyzes the pre-processed data based on artificial intelligence algorithms to identify the characteristic patterns of piping and seepage and determine whether there is a hazard.

[0017] The aforementioned hazard evolution prediction unit is used to combine digital twin technology, utilize historical and real-time data to construct a digital twin model of the dike, analyze the evolution path of the hazard, and determine the risk trend of the hazard.

[0018] A further improvement to the technical solution of the present invention is that the hazard diagnosis unit includes:

[0019] The pre-processed data is input into a large AI model pre-built based on artificial intelligence algorithms to extract features related to piping and seepage.

[0020] The extracted features are compared with a pre-defined library of piping and seepage feature patterns to find feature pattern types with high matching degree.

[0021] Based on the pattern matching results, combined with the set matching threshold and judgment rules, determine whether there is a risk of piping and seepage in the current dike and output a diagnostic conclusion.

[0022] A further improvement to the technical solution of the present invention is that the hazard evolution prediction unit includes:

[0023] By combining digital twin technology and utilizing historical and real-time data, a digital twin model of the dike is constructed to reflect the physical characteristics and operational status of the dike in real time.

[0024] Based on the digital twin model, the evolution path of the hazard is quantitatively analyzed, and the changing trend of the hazard at different time scales is analyzed, including the changes in seepage flow, the expansion of seepage area and the stability changes of the dike structure. Then, the risk trend coefficient is calculated in combination with the preset risk threshold to analyze whether the risk of the hazard exceeds the risk threshold, and thus determine the risk trend of the hazard. The prediction results are output to the visualization decision support module to provide support for decision-making.

[0025] A further improvement to the technical solution of this invention lies in the following: the risk tendency analysis process for the aforementioned dangerous situation is as follows:

[0026] Real-time data is acquired from the digital twin model, including the physical characteristics and operational status of the dike, to ensure that the digital twin model can reflect the current status of the physical dike in real time. Using historical and real-time data from the digital twin model, the trend of seepage flow over time is analyzed, the change of seepage flow is quantitatively analyzed to obtain the rate of change of seepage flow, the expansion trend of seepage area is analyzed, including the change of area and range of seepage area, the expansion of seepage area is quantitatively analyzed to obtain the speed and direction of seepage area expansion, and the stability changes of dike structure are analyzed, including stress distribution, deformation, etc., the stability of dike is quantitatively analyzed to obtain the trend of stability change.

[0027] Based on the results of the quantitative analysis, risk-related factors are extracted, including seepage flow, seepage area, and levee structural stability trend value. Weights are assigned to each risk factor based on historical data. The quantitative results of each risk factor are combined using weighted summation to calculate the risk trend coefficient. Based on the magnitude of the risk trend coefficient, the risk trend of the hazard is determined, and it is judged whether the hazard tends to stabilize or deteriorate.

[0028] The results of the quantitative analysis and the risk tendency coefficients are organized into a structured data format, including the seepage flow change curve, the expansion map of the seepage area, the stability change map of the dike structure, and the specific values ​​of the risk tendency coefficients. The organized data is then transmitted to the visualization decision support module through the internal communication interface of the system to ensure that the data transmission is accurate and that the format meets the requirements of the visualization module.

[0029] A further improvement of the technical solution of the present invention is that: the visualization decision support module includes a visualization rendering unit and a decision scheme generation unit;

[0030] The visualization rendering unit uses a digital twin engine to render the real-time status of the dike, the danger information, and the prediction results in real time, forming a visualized decision-making sand table, including three-dimensional models, dynamic charts, early warning information, and other forms.

[0031] The decision-making scheme generation unit is used to generate the optimal decision scheme based on factors such as the state of danger and resource constraints, combined with reinforcement learning and multi-objective optimization algorithms, and by considering multiple objectives of risk, cost and benefit.

[0032] A further improvement to the technical solution of the present invention is that the visualization rendering unit includes:

[0033] The system receives real-time status of dikes, information on potential dangers, and forecast results, performs format conversion and integration to adapt the data to the visualization engine, and provides a reliable data foundation for rendering.

[0034] Using a digital twin engine, a three-dimensional model of the dike is constructed based on the received data. The hazard information and prediction results are then integrated into the three-dimensional model of the dike to create a visual decision-making sand table. Different colors and symbols are used to mark the location of piping and seepage, and dynamic charts are used to show the changing trend of seepage flow.

[0035] Presenting a visual decision-making sandbox to decision-makers, the rendered visualization content is displayed in an intuitive interface, providing a variety of interactive functions such as zooming, rotating, and querying, to help relevant personnel view and analyze the status of dikes and potential dangers from multiple angles.

[0036] A further improvement to the technical solution of the present invention is that the decision scheme generation unit includes:

[0037] Collect information on the current emergency situation, including the location of seepage, flow rate, and stability of the dike, while clarifying resource constraints such as manpower, materials, and time, and analyzing and determining the objectives and limitations of decision-making.

[0038] By using reinforcement learning algorithms, the feedback of different decision actions in dangerous environments is simulated, and the decision-making strategy is adjusted to adapt to dynamic dangerous situations. Combined with multi-objective optimization algorithms, under the premise of satisfying resource constraints, the decision-making schemes involved are searched and evaluated to generate multiple feasible decision-making schemes. Then, the decision-making scheme that performs best under the comprehensive objectives of risk, cost and benefit is selected.

[0039] The optimal decision-making plan generated is comprehensively evaluated to verify its effectiveness and feasibility in reducing risks, controlling costs, and improving efficiency. After confirming that the plan meets the actual handling conditions and requirements, the decision-making plan is output in a clear and specific form to provide clear operational guidance for the emergency response personnel.

[0040] A further improvement to the technical solution of the present invention is that the feedback optimization module includes:

[0041] Collect key data during the emergency response process based on the optimal decision-making scheme, including changes in seepage flow, improvement in dike stability, and resource utilization, and then feed back the implementation effect of the optimal decision-making scheme to the system in real time as feedback information;

[0042] Based on the collected feedback on the implementation effect and the actual handling results of the emergency, a comprehensive evaluation of the system is conducted. The deviation between the actual implementation results of the decision-making plan and the expected goals is analyzed, and the shortcomings of the system in prediction, decision-making or resource allocation are identified. Based on the evaluation results, the system algorithm and parameters are optimized to improve the accuracy and efficiency of decision-making.

[0043] Based on feedback information, the digital twin model of the dike is updated and maintained synchronously, and parameters such as physical characteristics and operating status that do not match the actual situation are corrected. This ensures that the digital twin model accurately reflects the real-time status of the dike, providing support for subsequent risk prediction and decision-making, and forming a complete closed loop.

[0044] Due to the adoption of the above technical solution, the technical progress achieved by this invention compared to the prior art is as follows:

[0045] 1. This invention provides an artificial intelligence-based fiber optic dike underwater piping and seepage inspection system. It utilizes a fiber optic sensor network to monitor underwater piping and seepage in real time, and can quickly capture seepage-related physical data. Through AI large-scale model based on artificial intelligence algorithms, the system analyzes the data, accurately identifies the characteristic patterns of piping and seepage, and determines whether there is a danger. Compared with manual inspection, it greatly improves the efficiency and accuracy of danger detection, and can detect hidden dangers in the early stage of piping, effectively reducing the risk of dike collapse.

[0046] 2. This invention provides an artificial intelligence-based fiber optic dike underwater piping and seepage inspection system. Combining digital twin technology, it uses historical and real-time data to construct a digital twin model of the dike, quantitatively analyzes and predicts the evolution path of the danger, and calculates the risk tendency coefficient by simulating the development trend of the danger under different conditions. This intuitively reflects the risk trend of the danger, which helps decision-makers to understand the development trend of the danger in advance, formulate targeted response strategies, and improve the scientificity and effectiveness of the danger handling. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0048] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0049] Figure 2 This is a flowchart illustrating the risk trend analysis of the hazards of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1, as Figure 1 , Figure 2 As shown, this invention provides an artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage, including a monitoring management center. The monitoring management center is communication-connected to the following modules, wherein:

[0052] The data acquisition module is used to monitor underwater piping and seepage in the dike in real time using a fiber optic sensor network. It also preprocesses the acquired data. A fiber optic sensor network is laid in key parts of the dike to sense underwater piping and seepage in real time, acquire seepage-related physical quantity data, receive data collected by the fiber optic sensor network, and perform preprocessing operations including data cleaning, noise reduction, and format conversion to ensure data integrity and usability. The preprocessed data is then transmitted to the inspection and management center through communication equipment and stored in the database to provide data support for subsequent analysis.

[0053] The specific tasks of the data acquisition module are as follows:

[0054] A fiber optic sensor network is deployed at key sections of the dike to ensure sensor coverage of all potential piping and seepage risk areas. The fiber optic sensors monitor underwater piping and seepage in real time, acquiring data on related physical quantities such as seepage location, flow rate, water pressure changes, and temperature changes. The raw data collected by the fiber optic sensor network undergoes preprocessing, including data cleaning, noise reduction, and format conversion. Data cleaning removes invalid, duplicate, and abnormal data to ensure accuracy; filtering algorithms reduce noise impact and improve data quality; and format conversion transforms the collected data into a unified format. The preprocessed data is then transmitted to the patrol management center via communication equipment (wireless communication modules or fiber optic communication links). At the patrol management center, the received data is stored in a database to support subsequent analysis and decision-making. The communication equipment must have high reliability and low latency to ensure timely and accurate data transmission, and the database must have efficient data storage and retrieval capabilities to support large-scale data storage and rapid querying.

[0055] The intelligent analysis and diagnosis module uses artificial intelligence algorithms to identify the characteristic patterns of piping and seepage, determine whether there is a danger, and combine digital twin technology to build a digital twin model of the dike to analyze and predict the evolution path of the danger and simulate the development trend of the danger under different conditions. The intelligent analysis and diagnosis module includes a danger diagnosis unit and a danger evolution prediction unit.

[0056] Among them, the hazard diagnosis unit analyzes the pre-processed data based on artificial intelligence algorithms, identifies the characteristic patterns of piping and seepage, and determines whether there is a hazard. The pre-processed data is input into a pre-built AI model based on artificial intelligence algorithms to extract features related to piping and seepage. The extracted features are compared with a pre-set piping and seepage feature pattern library to find feature pattern types with high matching degree. Based on the pattern matching results, combined with the set matching threshold and judgment rules, it is determined whether there is a piping and seepage hazard in the current dike and outputs a diagnostic conclusion.

[0057] The specific tasks of the hazard diagnosis unit are as follows:

[0058] The preprocessed data is input into a pre-built AI model based on artificial intelligence algorithms. The AI ​​model analyzes the input data and extracts features related to piping and seepage, including seepage location, seepage flow rate, water pressure changes, temperature changes, and seepage rate. The construction process of the AI ​​model involves: collecting a large amount of levee monitoring data, including underwater seepage-related physical quantity data collected by fiber optic sensors, such as seepage location, seepage flow rate, water pressure changes, and temperature changes; collecting historical hazard data, including known piping and seepage events and their characteristic patterns; collecting data under normal conditions for the model to learn the characteristics of normal operation; and labeling the collected data to distinguish between normal conditions and different types of piping and seepage hazards. The labeling work combines expert experience and historical data. The collected data undergoes cleaning, noise reduction, and normalization to ensure data quality and consistency. Finally, the data is divided into training, validation, and test sets. For model training, validation, and testing, features related to piping and leakage are extracted based on the characteristics of the physical quantity data. A support vector machine (SVM) machine learning algorithm is selected, and the model's input, hidden, and output layers are designed. The input layer receives preprocessed feature data, and the output layer outputs the diagnostic results of piping and leakage. Specific network structures, such as the number of layers, neurons, and activation functions, are designed according to the selected algorithm. The model is trained using training set data, and its parameters are adjusted through algorithm optimization to enable the model to learn patterns in the data. During training, the model is validated using validation set data, and hyperparameters are adjusted to optimize model performance. A loss function is selected to measure the difference between the model's predictions and the true labels. The trained model is evaluated using test set data, with evaluation metrics including accuracy, recall, F1 score, and ROC curve. The model's performance is analyzed, its shortcomings are identified, and the model is optimized based on the evaluation results. Optimization methods include adjusting the model structure, adjusting hyperparameters, using regularization techniques, and data augmentation to obtain a trained AI model. The extracted features are compared with a pre-defined piping and seepage feature pattern library, which stores typical feature patterns for various known piping and seepage situations. These typical feature patterns are constructed using historical data and expert experience. The similarity between the extracted features and each pattern in the feature pattern library is calculated to find feature pattern types with high matching degrees. It should be noted that the feature pattern library needs to be updated and maintained regularly to ensure it contains the latest piping and seepage feature patterns. Based on the feature comparison results, combined with the set matching threshold and judgment rules, it is determined whether there is a piping and seepage hazard in the current dike. If the matching degree is higher than the set matching threshold, a piping and seepage hazard is judged to exist; if it is lower than the matching threshold, the current state is judged to be safe. Based on the judgment results, a diagnostic conclusion is output, including whether there is a hazard and possible seepage locations.

[0059] The hazard evolution prediction unit is used to combine digital twin technology with historical and real-time data to construct a digital twin model of the dike, analyze the evolution path of the hazard, determine the risk trend of the hazard, and combine digital twin technology with historical and real-time data to construct a digital twin model of the dike to reflect the physical characteristics and operational status of the dike in real time. Based on the digital twin model, the evolution path of the hazard is quantitatively analyzed, and the changing trend of the hazard at different time scales is analyzed, including changes in seepage flow, expansion of seepage area, and changes in the stability of the dike structure. Then, a risk trend coefficient is calculated and combined with a preset risk threshold to analyze whether the risk of the hazard exceeds the risk threshold, thereby determining the risk trend of the hazard. The prediction results are output to the visualization decision support module to provide support for decision-making.

[0060] The specific tasks of the hazard evolution prediction unit are as follows:

[0061] Comprehensive historical data on the dikes is collected, encompassing past monitoring records, maintenance information, and disaster reports. Simultaneously, real-time operational data collected through the data acquisition module is acquired. Combining this with digital twin technology, the integrated historical and real-time data is used to construct a digital twin model reflecting the physical characteristics and operational status of the dikes. This achieves a high degree of mapping between the physical dike and the virtual model. Based on the constructed digital twin model, the evolution path of potential hazards is quantitatively analyzed. From a temporal perspective, the changing trends of hazards at different time scales are analyzed, including the dynamic changes in seepage flow, observing whether it gradually increases, decreases, or remains stable, and the expansion of the seepage area, thus clarifying... The speed and direction of seepage spread over time, as well as the changes in the stability of the dike structure, are used to assess the bearing capacity and deformation of the dike at different stages of the risk development. By comprehensively considering multiple factors such as seepage flow, seepage area expansion, and changes in the stability of the dike structure, a risk tendency coefficient is calculated, which intuitively reflects the risk tendency of the risk. The larger the value, the higher the probability of the risk developing in an unfavorable direction and the greater the risk level. The coefficient is compared with the preset risk threshold to analyze whether the risk of the risk exceeds the risk threshold. Then, the prediction results of the risk evolution trend obtained from the quantitative analysis and the calculated risk tendency coefficient are systematically organized to form a data report and output to the visualization decision support module.

[0062] Furthermore, the analysis process for the risk trend of the dangerous situation is as follows:

[0063] Real-time data is acquired from the digital twin model, including the physical characteristics and operational status of the dike, ensuring that the digital twin model can reflect the current state of the physical dike in real time. Using historical and real-time data from the digital twin model, the trend of seepage flow over time is analyzed, and the change in seepage flow is quantitatively analyzed to obtain the rate of change. The expansion trend of the seepage area is analyzed, including changes in the area and extent of the seepage area, and the expansion of the seepage area is quantitatively analyzed to obtain the speed and direction of the expansion. The stability changes of the dike structure are analyzed, including stress distribution and deformation, and the stability of the dike is quantitatively analyzed to obtain the trend of stability changes. Based on the results of the quantitative analysis, risk-related factors are extracted. The data includes seepage flow, seepage area, and levee structural stability trend value. Based on historical data, each risk factor is assigned a weight, and the quantitative results of each risk factor are combined using weighted summation to calculate the risk trend coefficient. Based on the magnitude of the risk trend coefficient, the risk trend of the hazard is determined, and it is judged whether the hazard tends to stabilize or worsen. The results of the quantitative analysis and the risk trend coefficient are organized into a structured data format, including the seepage flow change curve, the seepage area expansion map, the levee structural stability change map, and the specific value of the risk trend coefficient. Then, the organized data is transmitted to the visualization decision support module through the internal communication interface of the system to ensure that the data transmission is accurate and the format meets the requirements of the visualization module.

[0064] The formula for the change in leakage flow rate is as follows:

[0065] ;

[0066] In the formula, In time Leakage flow rate at that time The initial leakage flow rate, i.e. Traffic flow at that time The leakage flow rate change rate reflects the rate at which the leakage flow rate changes over time. It is a time variable;

[0067] The formula for the expansion of the leakage area is as follows:

[0068] ;

[0069] In the formula, In time Area of ​​the leakage zone at that time The area of ​​the initial leakage zone, i.e. Area at time, The average radius of the leaking area reflects the extent of the leak's expansion. The rate of expansion of the leakage area reflects how quickly the leakage area expands over time.

[0070] The formula for the stability change of the dike structure is as follows:

[0071] ;

[0072] In the formula, In time The stability trend value of the dike structure at a given time reflects the degree of stability of the dike structure. This represents the initial tendency value for the stability of the dike structure. The rate of change of dike structural stability reflects the speed at which the stability of the dike structure changes over time.

[0073] The formula for the risk tendency coefficient is as follows:

[0074] ;

[0075] In the formula, In time Risk tendency coefficient at that time , where are weighting coefficients, representing the relative importance of seepage flow, seepage area, and levee structural stability tendency value to risk, respectively. , This represents the maximum permissible leakage flow rate. This represents the maximum permissible area of ​​the leakage zone. This represents the minimum allowable value for the structural stability of the dike, over time. If the leakage flow increases, the leakage area expands, or the stability of the dike structure decreases, then... An increase in the value indicates an increase in risk; conversely, a decrease in the value indicates a decrease in risk.

[0076] The visualization decision support module is used to render the digital twin model of the dike in real time, dynamically display the changes in the dike's status, build a visualization decision sandbox, and generate and display the optimal decision-making scheme.

[0077] The feedback optimization module is used to feed back the execution effect of the optimal decision-making plan to the system. Based on the actual emergency response results and feedback information, the system is evaluated and optimized to form a closed-loop decision support process. At the same time, the digital twin model of the dike is updated and maintained according to the actual situation.

[0078] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: preferably, the visualization decision support module includes a visualization rendering unit and a decision scheme generation unit;

[0079] The visualization rendering unit uses a digital twin engine to render the real-time status, hazard information, and prediction results of the dike, forming a visualized decision-making sandbox. This sandbox includes various formats such as 3D models, dynamic charts, and early warning information, allowing decision-makers to intuitively understand the current status and hazard situation of the dike. It receives real-time data on the dike's status, hazard information, and prediction results from the system, performs format conversion and integration to adapt the data to the visualization engine, providing a reliable data foundation for rendering. Using the digital twin engine, it constructs a 3D model of the dike based on the received data and integrates the hazard information and prediction results into the 3D model, creating a visualized decision-making sandbox. Different colors and markers are used to indicate the location of piping and seepage, and dynamic charts display the changing trend of seepage flow. The visualized decision-making sandbox is presented to decision-makers through an intuitive interface, providing various interactive functions such as zooming, rotating, and querying, assisting relevant personnel in viewing and analyzing the dike's status and hazard situation from multiple angles.

[0080] The specific tasks of the visualization rendering unit are as follows:

[0081] The system receives real-time status data of the dikes from within, including information reflecting the current physical condition of the dikes such as water level, water pressure, and soil moisture. It also acquires hazard information, including the location and severity of diagnosed seepage and piping, as well as hazard evolution prediction data, such as seepage flow trends and risk directional coefficients. The system converts the different types of data received into a format that the visualization engine can recognize and process. It integrates various data types, removes redundant information, and ensures the correlation and consistency between data. Using a digital twin engine, a 3D model of the dike is constructed based on the processed data. This 3D model recreates the dike's geometry, structural features, and surrounding environment, including its slope, height, materials, and information on nearby rivers and topography, allowing decision-makers to intuitively understand the situation. This system aims to understand the overall condition of the dikes and then cleverly integrate the information on potential hazards and forecasts into a 3D model of the dikes to create a visual decision-making sandbox. Different colors and markers are used to indicate the location of piping and seepage, and dynamic charts are used to display the changing trends of seepage flow. At the same time, it presents the risk trends of different areas, enabling decision-makers to grasp the levee's hazard status. The rendered visualization content is presented to decision-makers in a clear and intuitive interface, providing them with a variety of interactive functions. For example, the zoom function allows decision-makers to view local details or the overall overview of the dike as needed, the rotation function allows decision-makers to observe the dike and hazard situation from different angles, and the query function allows decision-makers to quickly obtain detailed information on specific locations or specific hazards. Through the provided interactive functions, it assists relevant personnel in viewing and analyzing the dike status and hazard situation from multiple angles and in all aspects.

[0082] The decision-making scheme generation unit is used to generate optimal decision schemes based on factors such as the state of the emergency and resource constraints, combined with reinforcement learning and multi-objective optimization algorithms, to comprehensively consider multiple objectives of risk, cost, and benefit. This provides specific guidance for emergency response, reduces resource waste, and improves the effectiveness and efficiency of emergency response. It collects current emergency status information, including leakage location, flow rate, and dike stability, while clarifying resource constraints such as manpower, materials, and time. It analyzes and determines the objectives and limitations of the decision-making process, uses reinforcement learning algorithms to simulate the feedback of different decision actions in the emergency environment, adjusts decision strategies to adapt to dynamic emergencies, and, combined with multi-objective optimization algorithms, searches and evaluates relevant decision schemes under the premise of satisfying resource constraints, generating multiple feasible decision schemes. It then selects the optimal decision scheme under the comprehensive objectives of risk, cost, and benefit, conducts a comprehensive evaluation of the generated optimal decision scheme, and verifies its effectiveness and feasibility in reducing risk, controlling costs, and improving benefits. After confirming that the scheme meets the actual response conditions and requirements, it outputs the decision scheme in a clear and specific form, providing clear operational guidance for emergency response personnel.

[0083] The specific tasks of the decision-making scheme generation unit are as follows:

[0084] This process comprehensively collects detailed information on the current state of the dike, including the location of seepage, accurate to specific dike sections and coordinates; seepage flow rate, its magnitude and trend obtained through real-time monitoring data; dike stability, assessing the dike's bearing capacity and deformation based on structural monitoring data; and defining available resource constraints, including: manpower (number and skill level of available emergency personnel); material resources (reserves and supply capacity of sand, gravel, timber, and emergency equipment); and time constraints (allowed time for emergency response and key time points). Based on the collected information and resource constraints, the process analyzes and determines decision-making objectives, including effectively controlling seepage within a specified timeframe and ensuring the structural safety of the dike. The model is designed to be identical to the real-world environment, while clearly defining the constraints in the decision-making process, namely, compliance with relevant safety regulations. Using reinforcement learning algorithms, a simulation model of the hazardous environment is constructed. Different decision actions are input into the model as the agent's behavior, simulating the feedback effects of these actions in the hazardous environment. This includes changes in seepage flow and improvements in dike stability after reinforcement measures are taken. The decision-making strategy is continuously adjusted based on the feedback results, enabling the agent to adapt to the dynamically changing hazardous environment and improving the accuracy and effectiveness of decision-making. Specifically, real-time monitoring data of the dike, including key information such as seepage location, seepage flow, and dike stability, is integrated. Through digital twin technology, this real-time monitoring data is mapped onto a virtual 3D model, constructing a model that mirrors the real-world environment. A virtual environment consistent with the actual dikes is constructed. Based on this, Geographic Information System (GIS) data is introduced to provide the model with background information such as terrain and landforms, making the simulation environment closer to real-world scenarios. Simultaneously, historical hazard data and expert experience are combined to parameterize the hazard development patterns in the model. A reinforcement learning algorithm is integrated, using different decision actions as inputs to the agent. By simulating the implementation of these decisions in the virtual environment, the impact on key indicators such as seepage flow and dike stability is evaluated. Corresponding feedback is provided based on a pre-set reward function. The agent continuously adjusts its decision-making strategy based on the feedback results, gradually learning to make optimal decisions in dynamically changing hazard environments. Through extensive simulation training, the agent masters… To determine which actions can minimize risk and maximize benefits under different hazardous conditions, after initial construction, the hazardous environment simulation model undergoes rigorous verification and optimization. By comparing and analyzing with actual hazardous cases, the model's predictions are verified to ensure consistency with reality, identify biases and shortcomings, and adjust and optimize parameters and algorithms to further improve accuracy and reliability. Simultaneously, a multi-objective optimization algorithm is introduced to comprehensively evaluate decision-making schemes within the model, ensuring an optimal balance between risk, cost, and benefit while meeting resource constraints. Through repeated verification and optimization, the required hazardous environment simulation model is finally constructed.Combining multi-objective optimization algorithms, this study comprehensively searches and evaluates various decision-making schemes under resource constraints. Considering multiple objectives such as risk, cost, and benefit, it aims to reduce leakage risk while minimizing material and manpower costs, thereby improving the overall efficiency of emergency response. The scores of different schemes on each objective are calculated, generating multiple feasible decision-making schemes. These schemes are then comprehensively compared and analyzed to select the optimal scheme that best performs under the combined risk, cost, and benefit objectives. This optimal scheme maximizes the achievement of decision-making objectives under existing resource conditions, minimizing the impact of the emergency on the dike and surrounding environment. A comprehensive evaluation of the optimal scheme is then conducted, analyzing, from a risk reduction perspective, whether the leakage flow is effectively controlled and the dike stability is improved after implementation. Significant improvements are made; from the perspective of cost control, the efficiency of material and human resource utilization is assessed, and whether the total cost of the plan is within the budget; from the perspective of improving efficiency, the role of the plan in reducing economic losses is considered, and the feasibility of the optimal decision plan under actual handling conditions and requirements is verified, including checking whether the required materials can be supplied in a timely manner, whether personnel have the corresponding operational skills, and whether the implementation of the plan complies with relevant safety regulations, to ensure that the plan can be smoothly implemented in reality. Then, the optimal decision plan that has been evaluated and verified is output in a clear and specific form to provide clear operational guidance for the emergency response personnel. The output content includes detailed information such as the steps of decision-making action, time arrangement, allocation of required resources, and responsible persons, so that the response personnel can accurately understand and implement the plan, thereby improving the efficiency and effectiveness of emergency response.

[0085] The feedback optimization module includes:

[0086] Key data collected during the emergency response process based on the optimal decision-making scheme include changes in seepage flow, improvement in dike stability, and resource utilization. The implementation effect of the optimal decision-making scheme is then fed back to the system in real time as feedback information. Based on the collected feedback on implementation effect and actual emergency response results, the system is comprehensively evaluated. The deviation between the actual implementation results of the decision-making scheme and the expected goals is analyzed, and the shortcomings of the system in prediction, decision-making, or resource allocation are identified. Based on the evaluation results, the system algorithm and parameters are optimized to improve the accuracy and efficiency of decision-making. Combined with the feedback information, the digital twin model of the dike is updated and maintained synchronously, and parameters such as physical characteristics and operating status that do not conform to the actual situation are corrected. This ensures that the digital twin model accurately reflects the real-time status of the dike, providing support for subsequent emergency prediction and decision-making, forming a complete closed loop.

[0087] The specific tasks of the feedback optimization module are as follows:

[0088] In the process of emergency response based on the optimal decision-making scheme, key data is collected comprehensively, including: seepage flow variation data, which is recorded in real time at different times and locations using high-precision flow monitoring equipment to record the values ​​and fluctuations of seepage flow; data on the improvement of dike stability, which is obtained by using structural monitoring sensors to acquire parameters such as displacement, stress, and strain of the dike to assess the dynamic changes in dike stability; and resource usage data, which records in detail the time of manpower input, the use of personnel skills, the types and quantities of materials consumed, and the time nodes of their use. The collected key data is then used as feedback information on the implementation effect of the optimal decision-making scheme, and is fed back in real time. The system is designed to ensure timely acquisition of the actual situation at the emergency response site. Based on the collected feedback on execution effectiveness and the actual emergency response results, a comprehensive evaluation of the system is conducted. The actual execution results of the decision-making plan are compared with the expected goals, analyzed from multiple dimensions including risk control, cost-effectiveness, and time efficiency. Deficiencies in the system's prediction, decision-making, or resource allocation stages are identified. The accuracy of the prediction of the emergency's development trend is examined, and any significant prediction deviations and their causes are analyzed. The rationality and adaptability of the decision-making plan are assessed, and logical flaws in the decision-making process are identified. Regarding resource allocation, the timeliness and rationality of resource allocation are reviewed. There are instances of resource waste or shortage. Based on the system evaluation results, targeted optimizations are made to the algorithms and parameters in the system. For example, the reward function and exploration strategy in the reinforcement learning algorithm are adjusted to improve the agent's decision-making ability in hazardous environments. The weight allocation of the multi-objective optimization algorithm is optimized to achieve a better balance between risk, cost, and benefit objectives in the decision-making scheme. By optimizing the algorithm parameters, the accuracy and efficiency of the system's decision-making are improved, enabling the system to more accurately predict the development trend of hazardous situations, generate more effective decision-making schemes, shorten decision-making time, and improve the response speed to hazardous situations. Combined with feedback information, the digital twin model of the dike is further optimized. The system is continuously updated and maintained. Based on actual monitored data such as changes in seepage flow and improvements in dike stability, physical property parameters in the model that do not match the actual situation are corrected, such as the permeability coefficient and mechanical performance parameters of dike materials. The operating status parameters in the model are also updated, such as water level, water pressure, and soil moisture. This ensures that the digital twin model can accurately reflect the real-time status of the dike. By continuously updating and maintaining the digital twin model, it is ensured that it always remains consistent with the actual situation of the dike. This allows the system to make accurate predictions and decisions based on the latest model data, forming a complete closed-loop feedback optimization mechanism, and continuously improving the overall performance of the system and its ability to respond to dike emergencies.

[0089] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage, comprising a monitoring and management center, characterized in that, The patrol management center has the following communication connection modules, including: The data acquisition module is used to monitor underwater piping and seepage in the dike in real time using fiber optic sensor networks, and to preprocess the acquired data. The intelligent analysis and diagnosis module uses artificial intelligence algorithms to identify the characteristic patterns of piping and seepage, determine whether there is a danger, and combine digital twin technology to build a digital twin model of the dike to analyze and predict the evolution path of the danger. The visualization decision support module is used to render the digital twin model of the dike in real time, dynamically display the changes in the dike's status, build a visualization decision sandbox, and generate and display the optimal decision-making scheme. The feedback optimization module is used to feed back the execution effect of the optimal decision-making plan to the system, and to evaluate and optimize the system based on the actual emergency response results and feedback information.

2. The artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage as described in claim 1, characterized in that: The data acquisition module includes: Fiber optic sensor networks are laid at key sections of the dike to detect underwater piping and seepage in real time and obtain physical data related to the seepage. It receives data collected by fiber optic sensor networks and performs preprocessing operations including data cleaning, noise reduction, and format conversion. The pre-processed data is transmitted to the patrol management center via communication equipment and stored in the database.

3. The artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage as described in claim 1, characterized in that: The intelligent analysis and diagnosis module includes a hazard diagnosis unit and a hazard evolution prediction unit; The hazard diagnosis unit analyzes the pre-processed data based on artificial intelligence algorithms to identify the characteristic patterns of piping and seepage and determine whether there is a hazard. The aforementioned hazard evolution prediction unit is used to combine digital twin technology, utilize historical and real-time data to construct a digital twin model of the dike, analyze the evolution path of the hazard, and determine the risk trend of the hazard.

4. The artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage as described in claim 3, characterized in that: The hazard diagnosis unit includes: The pre-processed data is input into a large AI model pre-built based on artificial intelligence algorithms to extract features related to piping and seepage. The extracted features are compared with a pre-defined library of piping and seepage feature patterns to find feature pattern types with high matching degree. Based on the pattern matching results, combined with the set matching threshold and judgment rules, determine whether there is a risk of piping and seepage in the current dike and output a diagnostic conclusion.

5. The artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage as described in claim 3, characterized in that: The hazard evolution prediction unit includes: By combining digital twin technology and utilizing historical and real-time data, a digital twin model of the dike is constructed to reflect the physical characteristics and operational status of the dike in real time. Based on the digital twin model, the evolution path of the hazard is quantitatively analyzed, and the changing trend of the hazard at different time scales is analyzed, including the changes in seepage flow, the expansion of seepage area and the stability changes of the dike structure. Then, the risk trend coefficient is calculated in combination with the preset risk threshold to analyze whether the risk of the hazard exceeds the risk threshold, and thus determine the risk trend of the hazard. The prediction results are output to the visualization decision support module.

6. The artificial intelligence-based fiber optic dike waterproofing piping and seepage inspection system according to claim 5, characterized in that: The risk trend analysis process for the aforementioned dangerous situation is as follows: Real-time data is obtained from the digital twin model, including the physical characteristics and operational status of the dike. Using historical and real-time data from the digital twin model, the trend of seepage flow over time is analyzed, the change of seepage flow is quantitatively analyzed to obtain the rate of change of seepage flow, the expansion trend of seepage area is analyzed, including the change of area and range of seepage area, the expansion of seepage area is quantitatively analyzed to obtain the speed and direction of seepage area expansion, and the stability change of dike structure is analyzed, the stability of dike is quantitatively analyzed to obtain the trend of stability change. Based on the results of the quantitative analysis, risk-related factors are extracted, including seepage flow, seepage area, and levee structural stability trend value. Weights are assigned to each risk factor based on historical data. The quantitative results of each risk factor are combined using weighted summation to calculate the risk trend coefficient. Based on the magnitude of the risk trend coefficient, the risk trend of the hazard is determined, and it is judged whether the hazard tends to stabilize or deteriorate. The results of the quantitative analysis and the risk tendency coefficients are organized into a structured data format, including the change curve of seepage flow, the expansion map of the seepage area, the change map of the stability of the dike structure, and the specific values ​​of the risk tendency coefficients. The organized data is then transmitted to the visualization decision support module through the internal communication interface of the system.

7. The artificial intelligence-based fiber optic dike waterproofing piping and seepage inspection system according to claim 3, characterized in that: The visualization decision support module includes a visualization rendering unit and a decision scheme generation unit; The visualization rendering unit uses a digital twin engine to render the real-time status of the dike, the danger information, and the prediction results in real time, forming a visualized decision-making sand table. The decision-making scheme generation unit is used to generate the optimal decision scheme based on the situation status, resource constraints, reinforcement learning and multi-objective optimization algorithms, and by comprehensively considering multiple objectives such as risk, cost and benefit.

8. The artificial intelligence-based fiber optic dike waterproofing piping and seepage inspection system according to claim 7, characterized in that: The visualization rendering unit includes: The system receives real-time status of dikes, information on potential dangers, and forecast results, performs format conversion and integration, and adapts the data to the visualization engine. Using a digital twin engine, a three-dimensional model of the dike is constructed based on the received data. The hazard information and prediction results are then integrated into the three-dimensional model of the dike to create a visual decision-making sand table. Different colors and symbols are used to mark the location of piping and seepage, and dynamic charts are used to show the changing trend of seepage flow. Presenting a visual decision-making sandbox to decision-makers, the rendered visual content is displayed in an intuitive interface, providing a variety of interactive functions to help relevant personnel view and analyze the status of dikes and potential dangers from multiple perspectives.

9. The artificial intelligence-based fiber optic dike waterproofing piping and seepage inspection system according to claim 7, characterized in that: The decision-making scheme generation unit includes: Collect information on the current state of danger, clarify resource constraints, and analyze and determine the objectives and limitations of decision-making; By using reinforcement learning algorithms, the feedback of different decision actions in dangerous environments is simulated, and the decision-making strategy is adjusted to adapt to dynamic dangerous situations. Combined with multi-objective optimization algorithms, under the premise of satisfying resource constraints, the decision-making schemes involved are searched and evaluated to generate multiple feasible decision-making schemes. Then, the decision-making scheme that performs best under the comprehensive objectives of risk, cost and benefit is selected. The optimal decision-making plan generated is comprehensively evaluated to verify its effectiveness and feasibility in reducing risks, controlling costs, and improving efficiency. After confirming that the plan meets the actual handling conditions and requirements, the decision-making plan is output in a clear and specific form to provide clear operational guidance for the emergency response personnel.

10. The artificial intelligence-based fiber optic dike waterproofing system for detecting piping and seepage as described in claim 9, characterized in that: The feedback optimization module includes: Collect key data during the emergency response process based on the optimal decision-making plan, and then feed back the execution effect of the optimal decision-making plan to the system in real time as feedback information; Based on the collected feedback on the execution effect and the actual results of emergency response, a comprehensive evaluation of the system is conducted. The deviation between the actual execution results of the decision-making plan and the expected goals is analyzed, and the shortcomings of the system in prediction, decision-making or resource allocation are identified. Based on the evaluation results, the system algorithm and parameters are optimized. Based on feedback information, the digital twin model of the dike is updated and maintained synchronously, correcting physical characteristics and operational parameters in the model that do not match the actual situation, so that the digital twin model reflects the real-time status of the dike, providing support for subsequent risk prediction and decision-making, and forming a complete closed loop.

Images