A levee underwater piping eruption detection and positioning system fusing distributed optical fiber acoustics

Abstract

The application discloses a kind of fusion distributed optical fiber acoustics embankment underwater pipe heaving detection positioning system, it is related to water conservancy engineering safety monitoring technical field, comprising: voiceprint signal perception preliminary screening module, for collecting original acoustic vibration signal in strong noise environment, and carry out preliminary purification and separation to it, obtain the intrinsic mode function of different frequency bands, separate out weak pipe heaving suspected signal submerged in background noise.The application maps the seepage network reasoned into high-precision fluid-structure coupling model by introducing digital twin technology, simulates the evolution process under extreme working conditions such as sudden change of water level in virtual space, so that the system can calculate the final breach form, size and position, and the consequences of pipe heaving disaster are pushed from the qualitative judgment to the quantitative prediction, realizing the dynamic deduction and quantitative prediction of pipe heaving disaster process based on physical mechanism.

Description

A Dike-based Underwater Piping Detection and Location System Integrating Distributed Fiber Acoustics Technical Field

[0001] This invention relates to the field of water conservancy project safety monitoring technology, specifically to a dike flooding underwater piping detection and location system that integrates distributed fiber optic acoustics. Background Technology

[0002] Underwater piping is one of the main causes of dike damage and poses a serious threat to the safety of water conservancy projects. Traditional underwater detection methods, such as manual inspection and sonar scanning, have certain limitations, especially in real-time monitoring and high-precision positioning. Distributed fiber optic acoustic technology has gradually become an important application technology in the detection of underwater piping in dikes because it can monitor structural changes over a large area in real time. By laying optical fibers inside or near the dike and using fiber optic acoustic sensors to capture changes in underwater sound waves, the location and intensity of piping can be accurately located.

[0003] In existing technologies, under strong environmental noise interference (such as continuous ship vibration in busy waterways), the system is easily submerged by continuous ship vibration noise, leading to the failure to detect the initial weak signals of piping. Furthermore, after identifying a single piping point, it is difficult to predict whether a connected seepage channel network has formed inside the dam foundation based on the fluctuation of its acoustic signal, making it impossible to effectively assess the systemic risk of underwater piping in dams. In addition, for the confirmed and located seepage channel network, it is difficult to dynamically predict the instability evolution process, and it cannot provide forward-looking guidance for emergency response decisions. Therefore, a dam underwater piping detection and location system integrating distributed fiber optic acoustics is proposed. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: a dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics, comprising the following modules:

[0005] The voiceprint signal sensing and screening module is used to collect raw acoustic vibration signals in a high-noise environment, and to perform preliminary purification and separation on them. It obtains the intrinsic mode functions of different frequency bands, separates the weak suspected piping signals submerged in the background noise, and provides purified data raw materials for accurate identification.

[0006] The knowledge graph construction module, based on the suspected signal frequency bands separated by the initial screening, transforms the intrinsic mode function components containing turbulence information into time spectrum graphs, and inputs them into a pre-trained convolutional neural network model to identify, extract, and locate the initial weak turbulent acoustic features of piping, constructing a knowledge graph of the dam foundation structure, transforming isolated monitoring data points into a structured knowledge network with semantic associations, and realizing the deep integration of monitoring information and prior knowledge;

[0007] The seepage network reasoning and judgment module, based on the knowledge graph of the dam foundation structure, simulates the expansion logic of seepage channels through graph algorithms. Starting from the identified isolated piping points, it actively reasones and determines whether a connected and hidden seepage channel network has been formed inside the dam, thus achieving the leap from point risk to surface risk.

[0008] The fluid-structure coupling modeling engine constructs a basic digital twin model that is consistent with the geometric and physical properties of the physical dam foundation based on digital twin technology. It synchronously maps the seepage channel network structure inferred from the dam foundation structure knowledge graph to the basic digital twin model, creating a fluid-structure coupling model of the dam in virtual space that evolves synchronously with the physical entity and reflects the actual internal seepage state.

[0009] The disaster process simulation module drives the fluid-structure coupling model of the dam to simulate the dynamic process of the scouring, expansion, and interconnection evolution of the existing seepage channel network under the scenario of sudden water level change, and calculates the final breach shape and location, realizing the quantitative prediction of the evolution trend of piping disaster, and providing forward-looking guidance for the formulation of accurate emergency response plans.

[0010] Preferably, the voiceprint signal sensing and screening module includes a distributed acoustic sensing unit and a modal decomposition and noise reduction unit;

[0011] The distributed acoustic sensing unit is used to continuously convert acoustic vibration signals in the environment, including ship vibration, water flow and potential piping turbulence, into a full-domain acoustic signal sequence by utilizing a distributed optical fiber sensing network deployed on the underwater foundation of the dam, thereby obtaining a raw acoustic data field containing rich information but with an extremely low signal-to-noise ratio.

[0012] The modal decomposition noise reduction unit uses an empirical modal decomposition algorithm to adaptively decompose the acoustic vibration signal of the original acoustic data field into intrinsic mode function components from high frequency to low frequency, and initially separate strong environmental noise and suspected signal frequency bands containing turbulence information.

[0013] Preferably, the execution steps of the distributed acoustic sensing unit include:

[0014] By deploying a distributed optical fiber sensor network on the underwater foundation of the dam, the system continuously collects and converts full-domain acoustic vibration signals, including ship vibration, water flow and potential piping turbulence, to form a raw acoustic data field, enabling all-weather, long-distance and seamless monitoring of the dam foundation status.

[0015] The empirical mode decomposition algorithm is used to adaptively decompose the original acoustic data field into a series of intrinsic mode function components arranged from high frequency to low frequency, effectively separating the signal features of different frequency bands and improving the targeting of signal processing;

[0016] Based on the frequency domain characteristics of the intrinsic mode function components, the strong environmental noise frequency band and the suspected signal frequency band containing turbulence information are initially separated, realizing the preliminary purification of the effective signal and significantly improving the accuracy and reliability of identification.

[0017] Preferably, the execution steps of the mode decomposition and noise reduction unit include:

[0018] Hilbert transform is performed on each eigenmode function component obtained from empirical mode decomposition to generate a time-frequency spectrum that characterizes the time-frequency energy of the signal, significantly improving the time-frequency resolution of the signal.

[0019] The time-spectrum image is input into a pre-trained convolutional neural network model, which automatically extracts deep acoustic features related to piping turbulence, achieving high-precision automatic feature mining.

[0020] Based on the output of the convolutional neural network model, the initial turbulence signal of the piping is accurately identified and spatially located, the piping point is located, and its main frequency, harmonic structure and pulse interval are extracted as the core feature entities of the acoustic print, which effectively improves the positioning accuracy and feature interpretability.

[0021] Preferably, the execution steps of the knowledge graph construction module include:

[0022] The identified piping points and their core acoustic signature entities are associated with the geological structure entities and historical potential hazard points of the dam to initially form a risk association network and improve the semantic level of the data.

[0023] Based on the spatial and physical attribute relationships between entities, a knowledge graph of dam foundation structure is constructed to achieve structured integration and visual expression of multi-source information;

[0024] By using a graph database to store and manage the knowledge graph of the dam foundation structure, isolated monitoring data points are transformed into a structured knowledge network with semantic relationships, supporting efficient graph querying and hidden risk mining.

[0025] Preferably, the percolation network reasoning and determination module includes a graph structure reasoning unit and a risk determination unit;

[0026] The graph structure reasoning unit is used to perform reasoning calculations in the knowledge graph of dam foundation structure based on the pattern of real-time acoustic vibration signals, running predefined seepage channel network formation rules and graph algorithms to discover potential seepage paths that are not directly monitored, and to determine whether there is a hydraulic connection between multiple piping points or hidden danger points.

[0027] The risk determination unit is used to integrate graph reasoning results with real-time acoustic vibration signals. If the current signal pattern is highly matched with the predefined seepage channel network in the dam foundation structure knowledge graph, it automatically determines that a connected seepage channel network has been formed inside the dam foundation and evaluates its connectivity and seepage efficiency.

[0028] Preferably, the execution steps of the graph structure reasoning unit include:

[0029] In the knowledge graph of the dam foundation structure, predefined rules for the formation of seepage channel networks based on seepage mechanics and geological conditions significantly improve the physical credibility of path reasoning;

[0030] Based on the real-time identification of the core acoustic features of piping points and their positions in the knowledge graph of dam foundation structure, the graph traversal and community discovery algorithm is used to efficiently uncover hidden seepage associations.

[0031] Based on the algorithm results and predefined rules, potential seepage paths that are not directly monitored are inferred, and the hydraulic connectivity between multiple points is determined, so as to realize the systematic identification from isolated points to network risks.

[0032] Preferably, the execution steps of the risk determination unit include:

[0033] The matching degree of the inferred potential seepage path is calculated with the real-time acoustic vibration signal, which significantly improves the accuracy of hidden channel identification. If the matching degree exceeds the preset matching threshold, it is determined that a connected seepage channel network has been formed inside the dam foundation. The matching degree calculation integrates the confidence of the path node signal, the signal synchronization strength, and the signal trend consistency.

[0034] Based on the topology of the seepage channel network, by quantifying the network density and global efficiency, its overall connectivity and permeability efficiency are evaluated, achieving a leap from qualitative to quantitative risk assessment and completing the qualitative identification of systemic risks.

[0035] Preferably, the execution steps of the fluid-structure coupling modeling engine include:

[0036] Based on digital twin technology, a high-precision basic digital twin model with the same geometric and physical properties as the physical dam foundation is constructed to achieve accurate mapping between the physical entity and the virtual model;

[0037] The seepage channel network structure determined by the knowledge graph of the dam foundation structure is synchronously mapped to the basic digital twin model, giving the model real seepage path information;

[0038] Initiate a multiphysics coupling solver in virtual space to establish a fluid-structure coupling model that reflects the actual internal seepage state, supporting the dynamic simulation and prediction of subsequent disaster processes.

[0039] Preferably, the execution steps of the disaster process simulation module include:

[0040] The fluid-structure coupling model is configured with various boundary condition scenarios, including sudden changes in water level and sustained high water level, to accurately simulate the dynamic impact of extreme hydrological conditions on the dam.

[0041] The driving fluid-structure coupling model simulates the dynamic process of scouring expansion, sediment transport and interconnection evolution of the existing seepage channel network under boundary conditions, intuitively presenting the nonlinear expansion and interconnection mechanism of seepage channels;

[0042] Based on the dynamic simulation results, the final breach shape, size and spatial location after the piping channel becomes unstable are calculated, disaster prediction data are generated, providing guidance for the formulation of emergency response plans, and the breach parameters are accurately quantified to provide key decision-making basis for emergency response plans.

[0043] This invention provides a dike-based underwater piping detection and location system integrating distributed fiber optic acoustics. It offers the following advantages:

[0044] (I) This underwater piping detection and location system for dikes, which integrates distributed fiber optic acoustics, effectively solves the problem of strong environmental noise such as busy waterways drowning out weak initial piping signals by using a hybrid algorithm of empirical mode decomposition and convolutional neural network. Based on signal purification and artificial intelligence criteria, it significantly improves the detection rate of initial piping signals in extremely complex acoustic environments and minimizes the risk of false alarms, thus achieving accurate identification and low false alarm warning of weak piping signals under strong noise background.

[0045] (II) This underwater piping detection and location system for dikes, which integrates distributed fiber optic acoustics, constructs a knowledge graph of the dike foundation structure by associating identified piping points with multi-source information such as geological structure and historical hazards. Based on the principle of seepage mechanics and predefined rules, it uses graph algorithms to infer potential seepage channel networks. This overcomes the limitation of traditional methods that can only identify single piping points. It can actively discover and determine whether a connected and hidden seepage channel network has been formed inside the dike. This makes risk assessment no longer limited to local hazards, but can qualitatively and quantitatively reveal the systemic instability risk faced by the entire dike foundation, achieving a leapfrog risk assessment from isolated point risk to systemic network risk.

[0046] (III) This kind of underwater piping detection and location system for dikes integrating distributed fiber optic acoustics, by introducing digital twin technology, maps the inferred seepage network onto a high-precision fluid-structure coupling model, and simulates the evolution process under extreme conditions such as sudden changes in water level in virtual space, so that the system can calculate the final breach shape, size and location, and advance the consequences of piping disasters from vague qualitative judgment to accurate quantitative prediction, realizing dynamic deduction and quantitative prediction of piping disaster process based on physical mechanism. Attached Figure Description

[0047] Figure 1 is a schematic diagram of the working process of a dike flooding detection and positioning system integrating distributed fiber optic acoustics according to the present invention.

[0048] Figure 2 is a data flow diagram of an underwater piping detection and location system for dikes that integrates distributed fiber optic acoustics according to the present invention. Detailed Implementation

[0049] 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, and 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.

[0050] Example 1, please refer to Figures 1 and 2. This invention provides a technical solution: a dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics, comprising the following modules:

[0051] The voiceprint signal sensing and screening module is used to collect raw acoustic vibration signals in a high-noise environment, and to perform preliminary purification and separation on them. It obtains the intrinsic mode functions of different frequency bands, separates the weak suspected piping signals submerged in the background noise, and provides purified data raw materials for accurate identification. The voiceprint signal sensing and screening module includes a distributed acoustic sensing unit and a mode decomposition and noise reduction unit.

[0052] The distributed acoustic sensing unit utilizes a distributed optical fiber sensor network deployed on the underwater foundation of the dam to continuously convert acoustic vibration signals from the environment, including ship vibration, water flow, and potential piping turbulence, into a global acoustic signal sequence. This enables all-weather, long-distance, and seamless monitoring of the dam foundation's condition, acquiring a raw acoustic data field containing rich information but with an extremely low signal-to-noise ratio. The distributed optical fiber sensor network deployed on the underwater foundation of the dam continuously collects and converts global acoustic vibration signals, including ship vibration, water flow, and potential piping turbulence, to form a raw acoustic data field. This enables all-weather, long-distance, and seamless monitoring of the dam foundation's condition. Employing an empirical mode decomposition algorithm, the raw acoustic data field is adaptively decomposed into a series of intrinsic mode function components arranged from high to low frequencies, effectively separating signal characteristics of different frequency bands and improving the targeting of signal processing. Based on the frequency domain characteristics of the intrinsic mode function components, strong environmental noise frequency bands and suspected signal frequency bands containing turbulence information are initially separated, achieving preliminary purification of effective signals and significantly improving the accuracy and reliability of identification.

[0053] The specific work involves deploying a distributed optical fiber sensor network on the underwater foundation of a dam in a dam safety monitoring scenario. This network continuously collects global acoustic vibration signals, possessing high sensitivity and a wide detection range. The global acoustic vibration signals encompass various components, including vibrations generated by ship navigation, water flow, and potential piping turbulence. These complex signals undergo preliminary transformation processing by the distributed optical fiber sensor network to form a raw acoustic data field. An empirical mode decomposition algorithm is then used to adaptively decompose the raw acoustic data field into a series of intrinsic mode function (EMF) components. These EMF components are arranged sequentially from high to low frequency, and each component contains characteristic information of the original acoustic vibration signal at different frequency bands. During the decomposition process, multiple screening processes are used to ensure that each EMF component meets specific conditions, including that the number of extrema and zero-crossings are equal or differ by at most one, and that they are locally symmetric. The upper and lower envelopes average to zero. After decomposition, N intrinsic mode function components are obtained, namely IMF1, IMF2, ..., IMFn, where IMF1 is the highest frequency component and IMFn is the lowest frequency component. Each component has a different frequency range and energy distribution. For each intrinsic mode function component obtained by decomposition, its frequency domain characteristics are analyzed. The frequency spectrum distribution of each intrinsic mode function component is obtained by transforming it from the time domain to the frequency domain through Fourier transform. Based on the typical differences in the frequency domain characteristics of different types of signals, the strong environmental noise frequency band and the suspected signal frequency band containing turbulence information are initially separated. The strong environmental noise frequency band has the characteristics of wide frequency range, relatively dispersed and stable energy distribution. The suspected signal frequency band containing turbulence information has a specific frequency concentration area and a unique energy distribution pattern. By setting frequency thresholds and energy thresholds, the strong environmental noise frequency band is removed from the original acoustic vibration signal, and the suspected signal frequency band containing turbulence information is retained, realizing the initial purification of the effective signal and improving the signal quality and usability.

[0054] The modal decomposition and denoising unit employs an empirical modal decomposition algorithm to adaptively decompose the acoustic vibration signal of the original acoustic data field into intrinsic mode function components from high frequency to low frequency. This initially separates strong environmental noise from suspected signal frequency bands containing turbulence information, achieving preliminary separation of noise from effective signals. Hilbert transform is applied to each intrinsic mode function component obtained from the empirical modal decomposition to generate a time-frequency spectrum representing the signal's time-frequency energy, significantly improving the signal's time-frequency resolution. The time-frequency spectrum is input into a pre-trained convolutional neural network model, which automatically extracts deep acoustic signature features related to piping turbulence, achieving high-precision automatic feature mining. Based on the output of the convolutional neural network model, the initial turbulence signal of the piping is accurately identified and spatially located, pinpointing the piping point and extracting its dominant frequency, harmonic structure, and pulse interval as core acoustic signature features, effectively improving positioning accuracy and feature interpretability.

[0055] The specific work involves performing Hilbert transform on a series of intrinsic mode function (IMF) components obtained after empirical mode decomposition (EMD). During the transform, the instantaneous frequency and amplitude of each IMF component at various times are calculated. Based on the calculation results, a time-frequency spectrum is constructed, with time as the horizontal axis, frequency as the vertical axis, and instantaneous amplitude represented by color or grayscale values. This time-frequency spectrum clearly presents the frequency distribution of the signal at different times and intuitively reflects the changes in the energy strength of each frequency component. The generated time-frequency spectrum is then input into a convolutional neural network model that has been trained with a large amount of data. This convolutional neural network model has a multi-layer structure, including convolutional layers, pooling layers, and fully connected layers. In the convolutional layers, multiple layers of different sizes are used to transform the signal. The convolution kernel performs convolution operations with the time-spectrum map, automatically learning and extracting local features from the time-spectrum map, including edge and texture feature patterns. The pooling layer downsamples the feature map output by the convolution layer, reducing the amount of data while retaining the main features, thus enhancing the robustness of the model. After multiple convolution and pooling operations, the fully connected layer integrates and classifies the extracted features, automatically extracting deep acoustic signature features related to piping turbulence. Finally, through the fully connected layer and the Softmax classifier, a confidence score representing the probability of the existence of the piping signal is output. When the confidence score exceeds the preset judgment threshold (0.95), the signal is confirmed to originate from a piping turbulence event. The deep acoustic signature features cover frequency dimension features, energy dimension features, and time dimension features. The frequency dimension features include frequency shift characteristics, frequency bandwidth characteristics, and harmonic component characteristics. Frequency shift characteristics refer to the fact that during the formation of piping turbulence, the water flow state changes, and the dominant frequency of the resulting acoustic signal differs from that of normal water flow or environmental noise. Fully connected layers can capture this frequency shift. Normal water flow's dominant frequency is concentrated in a lower frequency band, while piping turbulence shifts the dominant frequency to higher or lower frequency bands. Frequency bandwidth characteristics refer to the difference in frequency distribution range between the acoustic signal caused by piping turbulence and ordinary signals. Fully connected layers can analyze the frequency coverage range of the signal, i.e., the frequency bandwidth. Piping turbulence signals have wider or narrower frequency bandwidths, reflecting the complexity of its internal turbulent motion and the breadth of its energy distribution. Harmonic components... In addition to the dominant frequency component, the characteristics of the piping turbulence signal also include harmonic components. The fully connected layer can identify the frequency multiples, relative energy magnitudes, and number of harmonics. Some piping turbulence signals have obvious second and third harmonics, and the energy proportions of each harmonic show specific patterns. The harmonic component characteristics help to accurately identify piping turbulence. The energy dimension characteristics include energy distribution characteristics and energy variation characteristics. The energy distribution characteristics refer to the energy distribution in different frequency bands, which can reflect the degree of energy concentration and distribution pattern of the piping turbulence signal. The fully connected layer can analyze the energy proportion of the signal in each frequency sub-band. For example, the energy of the piping turbulence signal is highly concentrated in certain frequency bands, while the energy is weaker in other frequency bands, forming a unique energy distribution curve.Energy variation characteristics refer to the fact that the development of piping turbulence is a dynamic process, and the energy of its acoustic signal changes over time. The fully connected layer can capture the trend of energy change, such as sudden increases or decreases in energy, and periodic changes in energy. For example, in the early stages of piping, energy may gradually accumulate and increase, while after the piping development stabilizes, the energy exhibits relatively stable periodic fluctuations. Temporal dimensions include pulse interval characteristics and signal duration characteristics. Pulse interval characteristics refer to the fact that the acoustic signal generated by piping turbulence may appear in pulse form. The fully connected layer can measure the time interval between pulses. The pulse intervals under different piping conditions have different patterns; for example, the pulse intervals of some piping turbulence signals are relatively uniform, while others show random or periodic changes. Signal duration characteristics refer to the duration of the piping turbulence signal from start to finish, which is also an important feature. The fully connected layer can identify the start and end times of the signal and calculate its duration. Different types of piping or piping at different development stages may have different signal durations; a longer duration may indicate a more severe piping or a slower development. Spatial dimensions include signal propagation characteristics and spatial energy distribution characteristics. Signal propagation characteristics refer to the signal propagation characteristics of the acoustic signal generated by piping turbulence in the underwater foundation of a dam. When deploying a distributed fiber optic sensor network, the acoustic signals generated by the turbulent flow of piping are collected by sensors at different locations. The fully connected layer can analyze information such as the time difference and propagation path of the signal between different sensors, thereby inferring the location of the piping point and the direction of signal propagation. For example, by comparing the order in which multiple sensors receive the signals and combining this with the known sensor locations, the approximate location of the piping point relative to the sensors can be determined. Combining the spatial energy distribution characteristics with the signal energy information collected by each sensor in the sensor network, the fully connected layer can construct a spatial energy distribution map of the signal, showing the energy distribution near the piping point. The signals collected by the sensors are usually strong, while the signals collected by the sensors far from the piping point are weak. After the convolutional neural network model identifies the piping event, based on the time difference positioning principle of the distributed optical fiber sensor network, the piping point is accurately located by analyzing the time difference of the signal arriving at different sensing units and combining the propagation speed of light in the optical fiber. At the same time, the core acoustic features of the signal are extracted from the intrinsic mode function components corresponding to the identified signal, including: the dominant frequency value of the dominant oscillation frequency, the integer harmonic structure generated by nonlinear interaction, and the pulse interval reflecting the periodicity of turbulent bursts.

[0056] The expression for the confidence score is as follows: ;

[0057] In the formula: The confidence score represents the probability that the convolutional neural network model believes the input spectrogram originates from a piping turbulence event; For the target category index, in a binary classification problem, Specifically refers to a category of piping turbulence, set Piping, It is not piping; The target category score is the raw score output by the fully connected layer, corresponding to the piping turbulence category. The total number of categories; the total number of categories the model can recognize. That is: piping turbulence / non-piping turbulence; For the first The category score is the output of the fully connected layer, corresponding to the first category. The original scores for each category;

[0058] The knowledge graph construction module, based on the suspected signal frequency bands identified through initial screening, transforms the intrinsic mode function components containing turbulence information into time-spectrum maps. These maps are then input into a pre-trained convolutional neural network model to identify, extract, and locate the initial weak turbulent acoustic signature features of piping points. This effectively overcomes strong noise interference, achieving low false alarm and high detection rate identification of individual piping points. Simultaneously, the acoustic signature features of the first identified piping point are used as the core entity and associated with the geological structure entity and historical hazard point entity of the dam, constructing a knowledge graph of the dam's foundation structure. This transforms isolated monitoring data points into structured knowledge graphs with semantic relationships. The knowledge network enables deep integration of monitoring information and prior knowledge. It associates identified piping points and their core acoustic signatures with geological structures and historical hazard points of the dam, initially forming a risk association network and improving the semantic level of data. Based on the spatial and physical attribute relationships between entities, it constructs a knowledge graph of the dam's basic structure, realizing the structured integration and visual representation of multi-source information. The knowledge graph of the dam's basic structure is stored and managed using a graph database, transforming isolated monitoring data points into a structured knowledge network with semantic relationships, supporting efficient graph querying and hidden risk mining.

[0059] The specific work involves: after accurately identifying and locating the piping point and extracting its core acoustic signature features, deeply correlating it with the geological structure of the dam and historical potential hazard points. The geological structure includes parameters such as the lithology of the dam's strata, the strike and dip of geological faults, and the depth and variation of groundwater levels, reflecting the physical environmental characteristics of the dam foundation. Historical potential hazard points record the locations of cracks, the distribution of seepage points, information on weak soil layers, and the evolution of potential hazard points under different working conditions discovered during past dam inspections. By analyzing the relative spatial positions of the piping point and the geological structure, including whether the piping point is near a geological fault, etc., the work is further refined. Located above a soft soil layer, and combining the physical characteristics exhibited by the core acoustic signature features and historical hazard points, a multi-dimensional association is established between piping points, geological structures, and historical hazard points. Based on the spatial and physical attribute relationships between piping points, geological structures, and historical hazard points, a knowledge graph of dam foundation structures is constructed. In this knowledge graph, piping points, geological structures, and historical hazard points are defined as different types of nodes. Each node contains detailed parameter information, including the coordinates of the piping point, the core acoustic signature parameter value, the stratigraphic number and fault attitude parameters of the geological structure, and the discovery time and hazard type of the historical hazard point. Edges are defined based on the relationships between entities. Proximity edges exist between piping points and geological faults, while feature-based relationships, established based on similar acoustic signatures, connect piping points and historical seepage points. Through the definition and connection of nodes and edges, the various elements of the dam foundation structure are organized graphically, forming a complete knowledge graph encompassing spatial distribution, physical characteristics, and historical evolution. This intuitively displays the complex relationships and internal logic of the dam foundation structure. A graph database is used to store and manage the constructed dam foundation structure knowledge graph. With its unique graph storage structure, the graph database efficiently stores node and edge information and supports complex graph query and traversal operations. After importing the node and edge data from the dam foundation structure knowledge graph into the graph database, isolated monitoring data points (such as individual piping point monitoring data and geological structure measurement data) are transformed into a semantically related structured knowledge network through the graph database's association query function. In the structured knowledge network, geological structure information and historical hidden dangers related to a certain piping point can be queried, or clusters of piping points with similar voiceprint characteristics and their geological environments can be found. This facilitates long-term data preservation and rapid retrieval, providing strong knowledge support for applications such as dam safety assessment and risk prediction, and helping decision-makers to more comprehensively and deeply understand the operational status and potential risks of the dam foundation structure.

[0060] The seepage network reasoning and judgment module, based on the knowledge graph of the dam foundation structure, simulates the expansion logic of seepage channels through graph algorithms. Starting from the identified isolated piping points, it actively reasones and determines whether a connected and hidden seepage channel network has been formed inside the dam, thus achieving the leap from point risk to surface risk.

[0061] The fluid-structure coupling modeling engine constructs a basic digital twin model that is consistent with the geometric and physical properties of the physical dam foundation based on digital twin technology. It synchronously maps the seepage channel network structure inferred from the dam foundation structure knowledge graph to the basic digital twin model, creating a fluid-structure coupling model of the dam in virtual space that evolves synchronously with the physical entity and reflects the actual internal seepage state.

[0062] The disaster process simulation module drives the fluid-structure coupling model of the dam to simulate the dynamic process of the scouring, expansion, and interconnection evolution of the existing seepage channel network under the scenario of sudden water level change, and calculates the final breach shape and location, realizing the quantitative prediction of the evolution trend of piping disaster, and providing forward-looking guidance for the formulation of accurate emergency response plans.

[0063] Example 2, as shown in Figures 1 and 2, based on Example 1, the present invention provides a technical solution: the percolation network reasoning and judgment module includes a graph structure reasoning unit and a risk judgment unit;

[0064] The graph structure reasoning unit is used to perform reasoning calculations in the dam foundation structure knowledge graph based on real-time acoustic vibration signal patterns, running predefined seepage channel network formation rules and graph algorithms to discover potential seepage paths that are not directly monitored, and to determine whether there is a hydraulic connection between multiple piping points or hidden danger points. In the dam foundation structure knowledge graph, predefined seepage channel network formation rules based on seepage mechanics and geological conditions are used to significantly improve the physical credibility of path reasoning. Based on the acoustic core feature entities of the piping points identified in real time and their positions in the dam foundation structure knowledge graph, graph traversal and community discovery algorithms are run to efficiently discover hidden seepage associations. According to the algorithm results and predefined rules, potential seepage paths that are not directly monitored are inferred, and the hydraulic connectivity between multiple points is determined, realizing the systematic identification from isolated points to networked risks.

[0065] The specific work involves: deeply integrating seepage mechanics principles and geological structural constraints into the knowledge graph of dam foundation structures; predefining rules for the formation of seepage channel networks; and formalizing these rules into logical assertions and weighted relationships within the graph structure. Specifically, this includes: based on Darcy's law and flow network theory, defining that when multiple piping nodes or historically problematic nodes are located at the same low-permeability stratum boundary or share similar hydraulic gradient vector directions, the hydraulic connections between them have higher priority; and defining development rules for dominant seepage channels based on the fault attitude (strike, dip) and stratum lithology permeability coefficient, such as when piping... When a piping point is located within the fault influence zone and its dominant frequency characteristics indicate high-speed flow scouring, the node is assigned higher centrality and identified as a potential network hub. Simultaneously, a dynamic triggering mechanism is introduced: when a sudden change in groundwater level depth or a significant increase in the synchronicity of the piping point pulse intervals is detected, the connectivity probability estimate of related paths is automatically increased. Based on the real-time identified piping points and their spatial coordinates and acoustic signature entity attributes, a graph theory algorithm, including graph traversal and community detection algorithms, is executed on the dam foundation structure knowledge graph. The graph traversal algorithm (shortest path) starts from the newly identified piping point and follows a predefined "neighborhood" path. The algorithm explores the edges of "feature association" to search for other nodes reachable within a specific permeability coefficient threshold and geometric distance tolerance. The community detection algorithm (Louvain method) performs cluster analysis on the entire dam foundation structure knowledge graph based on the tightness of connections between nodes (weighted by voiceprint feature similarity, spatial proximity, and hydraulic gradient consistency). This identifies node clusters with dense internal connections and sparse external connections, revealing potential hidden seepage clusters composed of multiple isolated points. The algorithm outputs from graph traversal and community detection are combined with predefined seepage channel network formation rules to further... The algorithm calculates the comprehensive correlation strength between node pairs, which is a multi-parameter fusion index consisting of path length, the weight of each edge on the path, and the consistency of community affiliation. This quantifies the possibility of hydraulic connection between any two points. Finally, it infers potential seepage paths that are not directly detected by the distributed fiber optic sensor network but are highly suspected to exist based on geomechanical principles and graph topology. Based on the connectivity of the path and the stability of the community structure, it makes a comprehensive qualitative (connected / disconnected) and quantitative (connectivity strength) judgment on the hydraulic connectivity between multiple points, thus completing the leap from identifying isolated signals to identifying systemic network risks.

[0066] The risk assessment unit integrates graph reasoning results with real-time acoustic vibration signals. If the current signal pattern highly matches the predefined seepage channel network in the dam foundation knowledge graph, it automatically determines that a connected seepage channel network has been formed inside the dam foundation and assesses its connectivity and permeability efficiency. This enables qualitative identification of hidden systemic risks, raising the warning level from local piping to overall instability risk. The unit calculates the matching degree between the inferred potential seepage paths and real-time acoustic vibration signals, significantly improving the accuracy of hidden channel identification. If the matching degree exceeds a preset matching threshold, it determines that a connected seepage channel network has been formed inside the dam foundation. The matching degree calculation integrates the confidence of path node signals, signal synchronization strength, and signal trend consistency. Based on the topology of the seepage channel network, it assesses its overall connectivity and permeability efficiency by quantifying network density and global efficiency, achieving a leap from qualitative to quantitative risk assessment and completing the qualitative identification of systemic risks.

[0067] The specific work involves: in the safety monitoring of dam foundation structures, after inferring potential seepage paths, calculating the matching degree between these paths and real-time acoustic vibration signals. This matching degree calculation integrates the confidence level of path node signals, the synchronicity strength of acoustic signals between seepage points, and the consistency of the overall trend of acoustic signal energy changes at seepage points along the inferred path with the signals of other nodes along the path over the past hour. When the calculated matching degree exceeds a preset matching threshold, the path is determined to no longer be an isolated potential risk, but rather constitutes a physically connected, active seepage channel network, achieving a qualitative change from potential existence to actual formation, thus directly identifying hidden connected risks. After confirming the existence of the seepage channel network, its topology is quantitatively analyzed to assess the level of systemic risk. This analysis relies on graph theory topology parameters, including network density, node average degree, and network diameter. The overall connectivity robustness and permeability efficiency of the network are evaluated using these graph theory topology parameters, deepening the risk assessment from a simple qualitative judgment of connectivity to a quantitative description of its impact range and destructive potential.

[0068] The expression for the matching degree is as follows:

[0069] ;

[0070] In the formula: The matching degree represents the overall degree of matching between the inference path and the real-time acoustic signal; For the confidence level of the path node signal, the first For each piping point located on the inference path, the confidence score of its voiceprint signal is identified by the CNN model. The higher the value, the more certain the piping event at that point is. The total number of nodes on the path, and the number of identified piping points included in the currently evaluated inference seepage path; As an indicator of signal synchronization, the first The synchronicity intensity of acoustic signals between piping points is obtained by calculating the cross-correlation coefficient. The higher the value, the closer the hydraulic connection between the two points. The number of node pairs on the path is the total number of all possible node pairs calculated based on the number of nodes on the path. As a signal trend consistency indicator, the first The acoustic signal energy of a piping point located on the inference path shows consistency with the overall trend of signal changes at other nodes on the path in the near future. The number of nodes participating in the trend calculation; , where are weighting coefficients, representing the relative importance of the three dimensions of confidence, synchronicity, and trend consistency, respectively. , , It emphasizes the determinism of individual points and the synchronicity between points; when the matching degree is very high ( All or most of the piping points along the inference path continuously emit high-confidence acoustic signals. (Very high), and the signal exhibits strong synchronicity in time ( (Very high), while the intensity of the activity is also increasing simultaneously ( High); when the match is low ( The signal at the piping point along the path is weak or has low confidence. Low), the signals between points are chaotic and unrelated ( Low), and the activity trends also vary ( Low);

[0071] The expression for connectivity robustness is as follows:

[0072] ;

[0073] In the formula: Network density is the ratio of the actual number of connected edges in a network to the theoretically maximum possible number of connected edges. The actual number of edges in the network represents the number of actual hydraulic connectivity paths determined through reasoning. The total number of nodes in the network represents the total number of piping points and historical potential hazards in the seepage network; when As the graph approaches 1, it is nearly a fully connected graph, with very dense connections between nodes. This indicates a complex network of seepage channels forming a intricate mesh structure, exhibiting extremely high connectivity robustness. Even if some channels in the network are blocked, the water flow can quickly bypass them through numerous other paths, making the system difficult to be locally disrupted, but with a high risk of overall instability. When... When it approaches 0, it is close to a sparse graph or a star graph, with very few connections. This indicates that the seepage channels have a simple chain or tree structure with extremely low connectivity robustness. Once a critical channel is blocked, the entire network may be divided and the water flow path may be interrupted. However, if it is a star structure, the failure of the central node will cause the entire network to collapse.

[0074] The expression for penetration efficiency is as follows:

[0075] ;

[0076] In the formula: Global efficiency is a metric that measures the average transmission efficiency between all pairs of nodes in a network. For nodes and The shortest path length between nodes and The minimum number of edges required to traverse the path, if two points are not connected. ; For node-pair efficiency, the shorter the path between two points ( The smaller the value, the higher the transmission efficiency between them; when As it approaches 1, the network is nearly a complete graph, and there is a direct edge connecting any two points. This indicates extremely high infiltration efficiency, where water and sediment can move rapidly and almost unimpeded along the shortest path within the network, exhibiting enormous potential for erosion and damage, and easily developing into concentrated leakage channels; when When the value approaches 0, the network contains a large number of disconnected node pairs, or nodes require very long paths to connect, indicating extremely low penetration efficiency. Water flows slowly through the network, with long paths and high energy loss; although dangerous, its development speed and concentrated flushing capacity are relatively weak.

[0077] The execution steps of the fluid-structure coupling modeling engine include: based on digital twin technology, constructing a high-precision basic digital twin model that is consistent with the geometric and physical properties of the physical dam foundation, realizing accurate mapping between the physical entity and the virtual model, synchronously mapping the seepage channel network structure determined by the dam foundation structure knowledge graph to the basic digital twin model, giving the model real seepage path information, starting the multiphysics coupling solver in the virtual space, establishing a fluid-structure coupling model that reflects the real internal seepage state, and supporting the dynamic deduction and prediction of subsequent disaster processes;

[0078] The specific work involves: Based on digital twin technology, integrating multi-source heterogeneous data of dam foundations to construct a high-precision digital twin model that maintains consistency with the geometric and physical properties of the actual dam foundation. Specifically, a precise three-dimensional geometric model is established using geological survey data (borehole coordinates, stratigraphic interface elevation) and topographic mapping data (ground control point cloud). The boundary conditions and computational domain of the model are defined, and physical properties are assigned to the three-dimensional geometric model. Geological structural parameters, including the permeability tensor, porosity, effective soil stress parameters (cohesion, internal friction angle), and compression modulus, are used as unit attributes for spatial interpolation, forming a digital twin model that combines geometric accuracy and physical realism. The seepage channel network structure, determined by the dam foundation structure knowledge graph, is dynamically and synchronously mapped to the constructed digital twin model. The topological relationships and node attributes in the dam foundation structure knowledge graph are transformed into elements identifiable by the numerical model. The inferred potential seepage paths are then concretized into… The model identifies high-permeability channels and assigns their equivalent permeability coefficients based on the connectivity strength quantification in the dam foundation structure knowledge graph. Simultaneously, piping nodes along the path are positioned as internal source-sink points in the model, and their acoustic signature core features are used as inversion parameters to calibrate the initial velocity field within the channels, ensuring a high degree of consistency between the seepage network structure in the virtual space and the physical reality inferred from monitoring data. In the virtual space where structure and attribute mapping is completed, a multiphysics coupling solver is activated to establish a fluid-structure coupling model reflecting the actual internal seepage state. By simultaneously solving Darcy's law describing groundwater movement and Biot's consolidation theory governing equations describing the mechanical response of the soil skeleton, the seepage field and stress field are calculated in a fully coupled manner. The multiphysics coupling solver uses the finite element method for spatial discretization and employs an implicit time integration scheme for transient calculations, dynamically simulating the interaction between the pore water pressure field, seepage velocity field, and the effective stress field and displacement field of the soil, thus reproducing the fluid-structure coupling behavior dominated by seepage within the dam foundation.

[0079] Full coupling means that the equations for the seepage field and the stress field are solved simultaneously, and a change in either field will immediately affect the other field, which is described by the governing equations of Biot's consolidation theory.

[0080] The equilibrium equation is: ;

[0081] The continuity equation for seepage is: ;

[0082] In the formula: and For differential operators, It is divergence. It is the gradient, which describes the change of a physical quantity in space; The displacement vector field of the soil skeleton describes the direction and magnitude of movement of each point when the soil deforms due to force and seepage. The soil stiffness tensor is composed of the elastic modulus and Poisson's ratio parameters. It describes the soil's ability to resist deformation. The larger the value, the "harder" the soil. The tensor of strain is derived from the displacement field and describes the change in the local shape of the soil. This refers to the pore water pressure field, which is the pressure generated by water flowing through the pores of the soil. The effective stress coefficient of Biot (close to 1) represents the degree of contribution of pore water pressure to the overall stress of the soil. The Biot modulus represents the increase in pore pressure caused by the injection of fluid while keeping the total volume constant. The smaller the value, the easier the soil is to compress. For time; The inherent permeability tensor of soil is an intrinsic property determined by the soil structure, reflecting the ease with which fluids can pass through, and is independent of the properties of the fluid. The dynamic viscosity coefficient of the fluid;

[0083] The transient calculation is performed using an implicit time integration scheme, and its expression is as follows:

[0084]

[0085] In the formula: The time step is the time interval between one time point and the next time point in the calculation. and In time and pore water pressure; and In time and The displacement; The stiffness matrix of the system is obtained by discretizing the equilibrium equations using the finite element method. This is the strain-displacement matrix, used to relate strain and displacement; This is the transpose of the matrix;

[0086] The execution steps of the disaster process simulation module include: setting up various boundary condition scenarios for the fluid-structure coupling model, including sudden changes in water level and continuous high water level, accurately simulating the dynamic impact of extreme hydrological conditions on the dam, driving the fluid-structure coupling model to simulate the dynamic process of scouring expansion, sediment transport and interconnection evolution of the existing seepage channel network under the boundary condition scenarios, intuitively presenting the nonlinear expansion and interconnection mechanism of seepage channels, and calculating the final breach morphology, size and spatial location after the instability of the piping channel based on the dynamic simulation results, generating disaster prediction data, providing guidance for the formulation of emergency response plans, accurately quantifying breach parameters, and providing key decision-making basis for emergency response plans;

[0087] The specific work involves: setting various boundary condition scenarios based on historical hydrological data and extreme working condition predictions for the fluid-structure coupling model; defining external loads and constraints driving the model evolution, including but not limited to: applying time-varying water level-time series functions to the upstream boundary and lateral seepage boundary of the model to simulate sudden rises, falls, and sustained high water levels; setting pore water pressure distribution or escape gradient at the downstream boundary of the model; and constraining displacement boundary conditions for the soil skeleton, including fixing the bottom and restricting lateral movement. Simultaneously, the equivalent permeability tensor, initial porosity, and soil shear strength parameters of the seepage channel network are used as internal distribution parameters. The model imports attribute fields to ensure that the virtual scene accurately reflects the stress and seepage environment of the physical dam under specific hydrological events. A fluid-structure coupling model is driven to perform transient solutions under set boundary conditions, dynamically simulating the scouring expansion, sediment transport, and interconnection evolution of an existing seepage channel network. The simulation is achieved by simultaneously solving a set of coupled equations for the Darcy velocity field, pore water pressure field, and effective stress and displacement fields of the soil. The scouring process is categorized by introducing critical hydraulic gradient and sediment initiation velocity. When the actual hydraulic gradient exceeds the critical hydraulic gradient or the velocity exceeds the sediment initiation velocity, an event based on the erosion rate formula is triggered. (in To reduce soil quality, For water flow shear force, The sediment transport calculation (for soil erosion coefficient) dynamically updates the porosity and permeability coefficient of the region, thus providing positive feedback to the velocity field and forming a nonlinear intensification process of scouring expansion until the channel is fully formed. Based on the dynamic simulation results, through post-processing and quantitative analysis, the final breach morphology, size, and spatial location after the piping channel becomes unstable are calculated. Specifically, potential sliding surfaces are identified by extracting concentrated areas of the soil plastic strain field and abrupt change zones of the displacement field. The spatial location and extent of the breach are determined by analyzing the vector concentration of the seepage velocity field and the significant reduction area of ​​the effective stress field. Finally, the geometric morphology parameters of the breach (top width, depth, and expansion angle) and its spatial coordinates are quantitatively output, generating a disaster prediction dataset containing the breach development time series, final stable morphology, and key impact range. This provides forward-looking and quantitative decision-making guidance for formulating targeted emergency response plans (expansion areas, reinforcement priorities).

[0088] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dike-resistant underwater piping detection and location system integrating distributed fiber optic acoustics, characterized in that, The system includes the following modules: a primary acoustic signal sensing and screening module, used to collect raw acoustic vibration signals in a high-noise environment, perform preliminary purification and separation, obtain intrinsic mode functions of different frequency bands, and separate weak suspected piping signals submerged in background noise; and a knowledge graph construction module, based on the suspected signal frequency bands separated by the primary screening, converting the intrinsic mode function components containing turbulence information into time-spectrum graphs, and inputting them into a pre-trained convolutional neural network model to identify, extract, and locate the initial weak turbulent acoustic features of piping, and construct a knowledge graph of the dam foundation structure. The seepage network reasoning and judgment module, based on the knowledge graph of the dam foundation structure, simulates the expansion logic of seepage channels through graph algorithms. Starting from the identified isolated piping points, it actively reasones and determines whether a connected and hidden seepage channel network has been formed inside the dam. The fluid-structure coupling modeling engine constructs a basic digital twin model that is consistent with the geometric and physical properties of the physical dam foundation based on digital twin technology, and synchronously maps the seepage channel network structure inferred from the dam foundation structure knowledge graph to the basic digital twin model, creating a fluid-structure coupling model of the dam in virtual space. The disaster process simulation module drives the fluid-structure coupling model of the dam to simulate the dynamic process of scouring, expansion, and interconnection of the existing seepage channel network under the scenario of sudden water level change, and calculates the final breach shape and location, providing guidance for the formulation of emergency response plans.

2. The underwater piping detection and location system for dikes integrating distributed fiber optic acoustics as described in claim 1, characterized in that: The acoustic signature signal sensing and initial screening module includes a distributed acoustic sensing unit and a mode decomposition and denoising unit. The distributed acoustic sensing unit is used to convert the acoustic vibration signal of piping turbulence in the environment into a global acoustic signal sequence using a distributed optical fiber sensing network deployed on the underwater foundation of the dam, thereby acquiring the original acoustic data field. The mode decomposition and denoising unit uses an empirical mode decomposition algorithm to adaptively decompose the acoustic vibration signal of the original acoustic data field into intrinsic mode function components, initially separating strong environmental noise and suspected signal frequency bands containing turbulence information.

3. The underwater piping detection and location system for dikes integrating distributed fiber optic acoustics as described in claim 2, characterized in that: The execution steps of the distributed acoustic sensing unit include: continuously acquiring and converting global acoustic vibration signals containing ship vibration, water flow, and potential piping turbulence through a distributed optical fiber sensing network deployed on the underwater foundation of the dam to form a raw acoustic data field; adaptively decomposing the raw acoustic data field into a series of intrinsic mode function components arranged from high frequency to low frequency using an empirical mode decomposition algorithm; and preliminarily separating the strong environmental noise frequency band and the suspected signal frequency band containing turbulence information based on the frequency domain characteristics of the intrinsic mode function components.

4. The underwater piping detection and location system for dikes integrating distributed fiber optic acoustics as described in claim 2, characterized in that: The execution steps of the mode decomposition and noise reduction unit include: performing Hilbert transform on each intrinsic mode function component obtained from empirical mode decomposition to generate a time-frequency spectrum representing the time-frequency energy of the signal; inputting the time-frequency spectrum into a pre-trained convolutional neural network model, which automatically extracts deep acoustic signature features related to piping turbulence; and based on the output of the convolutional neural network model, accurately identifying and spatially locating the initial turbulence signal of the piping, locating the piping point, and extracting its dominant frequency, harmonic structure, and pulse interval as the core acoustic signature features.

5. A dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics as described in claim 2, characterized in that: The execution steps of the knowledge graph construction module include: associating the identified piping points and their core acoustic signature entities with the geological structure entities and historical hazard points of the dam to initially form a risk association network; constructing a dam foundation structure knowledge graph based on the spatial and physical attribute relationships between entities; and storing and managing the dam foundation structure knowledge graph using a graph database to transform isolated monitoring data points into a structured knowledge network with semantic associations.

6. The underwater piping detection and location system for dikes integrating distributed fiber optic acoustics according to claim 5, characterized in that: The seepage network reasoning and judgment module includes a graph structure reasoning unit and a risk judgment unit. The graph structure reasoning unit, based on real-time acoustic vibration signal patterns, runs predefined seepage channel network formation rules and graph algorithms within the dam foundation structure knowledge graph to perform reasoning calculations, discovering potential, unmonitored seepage paths, and determining whether there is a hydraulic connection between multiple piping points or potential hazard points. The risk judgment unit, by integrating the graph reasoning results with real-time acoustic vibration signals, automatically determines that a connected seepage channel network has formed within the dam foundation if the current signal pattern matches the predefined seepage channel network formation rules in the dam foundation structure knowledge graph, and assesses its connectivity and seepage efficiency.

7. A dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics as described in claim 6, characterized in that: The execution steps of the graph structure reasoning unit include: predefining the formation rules of the seepage channel network based on seepage mechanics and geological conditions in the dam foundation structure knowledge graph; running the graph traversal and community discovery algorithm based on the acoustic core feature entities of the piping points identified in real time and their positions in the dam foundation structure knowledge graph; and inferring potential seepage paths that have not been directly monitored based on the algorithm results and the predefined rules, and determining the hydraulic connectivity between multiple points.

8. A dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics as described in claim 6, characterized in that: The execution steps of the risk assessment unit include: calculating the matching degree between the inferred potential seepage path and the real-time acoustic vibration signal; if the matching degree exceeds a preset matching threshold, it is determined that a connected seepage channel network has been formed inside the dam foundation. The matching degree calculation integrates the confidence level of the path node signal, the signal synchronization strength, and the signal trend consistency. Based on the topology of the seepage channel network, its overall connectivity and seepage efficiency are evaluated to complete the qualitative identification of systemic risks.

9. A dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics as described in claim 1, characterized in that: The execution steps of the fluid-structure coupling modeling engine include: constructing a high-precision basic digital twin model that is consistent with the geometric and physical properties of the physical dam foundation based on digital twin technology; synchronously mapping the seepage channel network structure determined by the dam foundation structure knowledge graph to the basic digital twin model; and starting a multiphysics coupling solver in the virtual space to establish a fluid-structure coupling model that reflects the actual internal seepage state.

10. A dike flooding underwater piping detection and location system integrating distributed fiber optic acoustics as described in claim 1, characterized in that: The execution steps of the disaster process simulation module include: setting various boundary condition scenarios for the fluid-structure coupling model, including sudden changes in water level and sustained high water level; driving the fluid-structure coupling model to simulate the dynamic process of scouring and expansion, sediment transport and interconnection evolution of the existing seepage channel network under the boundary condition scenarios; and calculating the final breach shape, size and spatial location after the piping channel becomes unstable based on the dynamic simulation results, generating disaster prediction data to provide guidance for formulating emergency response plans.

Images