A real-time evolution prediction and assessment method for sluice risk

Abstract

The present invention discloses a method for predicting and evaluating the real-time evolution of sluice risks, which relates to the technical field of water conservancy engineering, comprising: synchronously collecting real-time multi-physical field data of a target area inside a foundation through an in-situ multi-parameter collaborative sensing network; extracting a multi-dimensional spatiotemporal feature sequence from the real-time multi-physical field data, and matching the multi-dimensional spatiotemporal feature sequence with a seepage evolution pattern record pre-constructed in a multi-physical field data fusion and knowledge base to identify a collaborative abnormal pattern characterizing the current seepage state; predicting its risk evolution trend based on the identified collaborative abnormal pattern; comprehensively evaluating the collaborative abnormal pattern and the risk evolution trend, and integrating a preset physical instability criterion to conduct a real-time evaluation of the sluice risk, and generating early warning information when the evaluation result meets the preset conditions. The present invention can achieve real-time perspective of the three-dimensional structure of the seepage network inside the foundation and its dynamic evolution, and can issue a critical state warning before disasters such as pipe bursts occur.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy projects, and in particular to a method for predicting and evaluating the real-time evolution of sluice risks. Background Art

[0002] As a key water conservancy hub that controls water flow and ensures flood control, water supply, and navigation, the safety of its foundation is paramount. Among the many factors that can cause sluice gate instability, hidden seepage in deep foundations has long been recognized as a major safety hazard. Under high water heads, seepage within the foundation is not static. It can gradually evolve from a diffuse Darcy flow, which poses no threat, to a non-Darcy rapid flow dominated by dominant channels. This can eventually deteriorate rapidly within a short period of time, forming a continuous piping channel, leading to structural damage to the foundation soil and even catastrophic dam failures such as "tower sweeping."

[0003] However, how to effectively monitor and provide early warning for this dynamic evolution from quantitative change to qualitative change is a major, long-standing, and unresolved technical bottleneck in this field. Existing technical means have fundamental limitations, and their inability is mainly reflected in:

[0004] Traditional monitoring methods fail to reveal the full picture of the network. Currently used techniques, such as embedding a small number of pressure gauges within the sluice foundation, are essentially "point-based" monitoring. Data from these discrete measurement points can only reflect pore water pressure at a localized location, completely failing to capture the complex morphology, connectivity, and dynamic evolution of seepage channels in three-dimensional space. The vast areas between measurement points become vast monitoring "blind spots," where advantageous seepage channels can develop and breed, rendering existing monitoring systems ineffective.

[0005] Existing geophysical exploration techniques are unable to achieve dynamic and accurate internal perspectives. Although geophysical exploration techniques such as the high-density resistivity method (ERT) have been tried for seepage detection, these techniques have serious shortcomings: First, their extremely low resolution makes it difficult to identify the early development of micro-fractures or dominant channels; second, their static or low-frequency measurements cannot capture the instantaneous and rapid changes in seepage conditions during critical moments such as floods; more importantly, their measurements are a comprehensive reflection of multiple geological factors and cannot accurately and independently infer the dynamic evolution of hydraulic conductivity parameters (such as the permeability field) that characterize the core characteristics of seepage, thus failing to provide accurate information for decision-making.

[0006] Models are disconnected from reality and lack deep coupling with physical mechanisms. Existing analytical methods typically rely on simplified, static seepage models, completely ignoring the complex dynamic coupling between seepage fields, stress fields, rock and soil structural damage, and multi-physics responses. This severe disconnect between these models and reality renders them unable to explain observed anomalies and make reliable predictions about future evolutionary trends. Summary of the Invention

[0007] The present invention provides a real-time evolution prediction and assessment method for sluice gate risks to solve the technical problems in the existing technology of blind spots in monitoring the hidden seepage risks of sluice deep foundations, inability to reveal the dynamic evolution process of the seepage network, serious lag in early warning, and inability to achieve accurate prediction and assessment in advance.

[0008] In view of the above problems, the present invention provides a method for predicting and evaluating the real-time evolution of sluice risk, comprising:

[0009] Through the in-situ multi-parameter collaborative sensing network deployed inside the deep foundation of the sluice, real-time multi-physical field data of the target area inside the foundation are synchronously collected;

[0010] Extracting a multidimensional spatiotemporal feature sequence from the real-time multi-physics field data, and matching the multidimensional spatiotemporal feature sequence with a seepage evolution pattern record pre-built in a multi-physics field data fusion and knowledge base to identify a collaborative anomaly pattern representing a current seepage state;

[0011] Based on the identified collaborative anomaly pattern, predict its risk evolution trend;

[0012] By integrating the collaborative abnormal pattern and the risk evolution trend and integrating the preset physical instability criteria, the sluice risk is evaluated in real time, and early warning information is generated when the evaluation result meets the preset conditions.

[0013] Preferably, the method also includes a step of pre-constructing the seepage evolution pattern record, which includes at least one of the following: by establishing a multi-physical field coupling mathematical model that can reflect the specific geological structure of the target sluice project, and performing multi-condition dynamic simulation of a preset typical seepage evolution path, recording the full-process multi-physical field response data of the virtual sensor, and summarizing and extracting the seepage evolution pattern record therefrom; or, by implementing a physical model test or in-situ controlled test that can simulate the specific seepage destruction process of the target sluice, arranging a sensor array in the test model and synchronously collecting multi-physical field data, and correlating the response characteristics observed in the test with the actual seepage evolution state to calibrate and generate the seepage evolution pattern record.

[0014] Preferably, the method further includes: before or during the step of collecting the real-time multi-physical field data, actively emitting controllable elastic wave signals through active excitation and auxiliary observation, and receiving them by the in-situ multi-parameter collaborative sensing network to enhance the detection capability of dynamic changes in the foundation medium structure.

[0015] Preferably, the method further includes: based on the real-time multi-physics field data and the physical model constraints in the knowledge base, through multi-physics field joint inversion, real-time reconstruction of the three-dimensional effective hydraulic conductivity characteristic field and the dominant seepage network topology structure inside the foundation; and presenting the three-dimensional effective hydraulic conductivity characteristic field and the dominant seepage network topology structure through a three-dimensional dynamic visualization interface.

[0016] The technical solution provided by this application has at least the following technical effects or advantages:

[0017] This invention uses an in-situ multi-parameter collaborative sensing network to acquire comprehensive information. Through multi-physics field data fusion and knowledge base, followed by subsequent diagnosis and inversion, it achieves dynamic, three-dimensional, and quantitative characterization of the implicit seepage network within the deep foundation of the sluice and its evolutionary process. This enables a shift in safety risk assessment from discrete observation to a global perspective.

[0018] By establishing a collaborative mechanism for identifying abnormal patterns and predicting risk evolution trends, this invention can accurately capture the physical precursors of catastrophic instability, such as piping, before they occur. This elevates risk management from passive response to proactive foresight, providing project managers with a crucial window for emergency response.

[0019] The present invention ultimately outputs more than just a simple alarm signal; it includes early warning information on risk level, spatial location, evolution pattern, and development trend. This provides a scientific and accurate basis for decision-making in daily maintenance, risk mitigation and reinforcement, and emergency response for sluice gates. It significantly enhances the scientific, intelligent, and refined nature of safe operation and maintenance, avoids unnecessary engineering investment, and reduces lifecycle operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a method for predicting and evaluating the real-time evolution of sluice risks according to the present invention. DETAILED DESCRIPTION

[0021] The present invention relates to a real-time evolution prediction and assessment method for sluice risks, aiming to solve the technical problems in the existing technology of monitoring the hidden seepage risk of deep foundations of sluices, such as the existence of "blind spots", the inability to reveal the dynamic evolution process of the seepage network, the serious lag in early warning, and the inability to achieve accurate prediction and assessment in advance.

[0022] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the drawings.

[0023] like Figure 1 A flow chart of a method for predicting and evaluating the real-time evolution of sluice risk is shown in FIG. The method includes the following steps:

[0024] Through the in-situ multi-parameter collaborative sensing network deployed inside the deep foundation of the sluice, real-time multi-physical field data of the target area inside the foundation are synchronously collected;

[0025] Specifically, an underground network capable of collecting multi-dimensional physical information in a long-term, stable, synchronous, and high-frequency manner is constructed within key areas within the foundation. These target areas are determined based on the sluice's engineering geological survey report, historical seepage disease data, and preliminary numerical simulation results of the seepage field. These areas typically include key locations such as the underside of the sluice floor, bypass seepage areas on both banks, the interior and exterior of important anti-seepage structures, and geological defect zones.

[0026] In a specific embodiment, the step of collecting the real-time multi-physics field data includes:

[0027] Synchronous acquisition is performed using a plurality of in-situ multi-parameter collaborative sensing probes arranged in a three-dimensional array in the target area inside the foundation;

[0028] Among them, the in-situ multi-parameter collaborative sensing probes are all composite detection units, which integrate: pore water pressure sensor, three-axis broadband acceleration sensor, temperature sensor and resistivity measurement unit.

[0029] Specifically, the in-situ multi-parameter collaborative sensing probe is designed to synchronously collect multiple key physical parameters at the same physical location or within a very close distance to ensure strong correlation of data in time and space, laying the foundation for subsequent collaborative abnormal pattern recognition.

[0030] Furthermore, the pore water pressure sensor preferably utilizes a high-precision micro-electromechanical system (MEMS) piezoresistive or capacitive sensor to accurately monitor the absolute value and dynamic changes of foundation pore water pressure, which is the most direct indicator for determining the driving force of seepage. The triaxial broadband acceleration sensor preferably utilizes a highly sensitive MEMS accelerometer, with a frequency band covering the range from passive acoustic emission or microseismic signals (typically generated by soil particle friction and microcrack expansion) to actively excited elastic wave signals, enabling the capture of microstructural damage in the medium and detecting changes in its mechanical properties. The temperature sensor preferably utilizes a Class A precision platinum resistor to monitor the foundation temperature field and indirectly invert the seepage path and flow velocity information through the thermal convection effect caused by seepage. The resistivity measurement unit preferably utilizes a micro-four-electrode arrangement to measure the apparent resistivity of the rock and soil mass within a small area surrounding the probe. Changes in the apparent resistivity effectively reflect changes in the medium's water saturation, pore connectivity, and pore water properties.

[0031] In another specific embodiment, the method further comprises:

[0032] Before or during the step of collecting the real-time multi-physical field data, a controllable elastic wave signal is actively emitted through an active excitation and auxiliary observation system and received by the in-situ multi-parameter collaborative sensing network to enhance the detection capability of dynamic changes in the foundation medium structure.

[0033] Specifically, in order to further improve the detection accuracy and the reliability of inversion, this method adopts an active excitation step. By installing a well-controlled elastic wave excitation source in a special excitation hole outside the monitoring area or at a specific depth, such as a well-controlled electric spark source or a piezoelectric ceramic transducer array, elastic waves with controllable energy and known waveforms are regularly and on demand emitted to the foundation medium. The three-axis broadband acceleration sensor in the in-situ multi-parameter collaborative sensing probe is responsible for receiving the elastic wave signal after propagation through the medium. By analyzing the travel time, amplitude attenuation and waveform changes of the wave, tomography and other technologies can be used to more accurately invert the elastic modulus, density, Poisson's ratio and other mechanical parameters inside the foundation and their dynamic changes, providing stronger constraints for the assessment of soil structure damage.

[0034] Extracting a multidimensional spatiotemporal feature sequence from the real-time multi-physics field data, and matching the multidimensional spatiotemporal feature sequence with a seepage evolution pattern record pre-built in a multi-physics field data fusion and knowledge base to identify a collaborative abnormal pattern representing a current seepage state;

[0035] Specifically, first, online feature engineering is performed on the collected original data or preliminarily processed data of each physical field to extract the key characteristic parameters that can more directly and sensitively reflect the seepage evolution state, and construct a multidimensional feature vector sequence arranged in chronological order, that is, a multidimensional spatiotemporal feature sequence.

[0036] Furthermore, the multi-physics data fusion and knowledge base stores a large number of seepage evolution pattern records, which uniquely and quantitatively associate abstract, invisible seepage evolution stages with a set of measurable, specific multi-physics synergistic response characteristic spectra.

[0037] In a specific embodiment, the seepage evolution pattern record includes:

[0038] A list of associated key physical field feature parameters;

[0039] Quantitative reference intervals or statistical distribution models defined for the key characteristic parameters in the list;

[0040] A rule or association model is used to describe the coordinated change relationship between characteristic parameters of different physical fields under the seepage evolution mode.

[0041] Specifically, the seepage evolution pattern records are structured digital records, which include:

[0042] 1. Basic model information: This includes a unique model identifier and a clear textual description of the seepage evolution model, such as: "initial stable pore seepage", "initial formation of linear dominant seepage channels along specific geological structures", "development of piping channels in sand and gravel foundations accompanied by an accelerated period of fine particle initiation", etc.

[0043] 2. Multi-physics field collaborative response characteristic spectrum:

[0044] The associated key physical field characteristic parameter list: explicitly lists the key physical field characteristic parameters that are most indicative in this mode. Specifically, the list is selected and combined from the following sets:

[0045] Acoustic characteristic parameters: for example, the cumulative number of acoustic emission events, the average event energy, the logarithmic distribution slope of the event energy, the statistical distribution of the peak frequency of the event waveform, and the clustering density of the sound source events in three-dimensional space.

[0046] Electrical characteristic parameters: for example, average apparent resistivity value, rate of change of apparent resistivity, apparent resistivity spectrum characteristics under multi-frequency excitation, and spatial gradient of apparent resistivity between adjacent monitoring units.

[0047] Thermal characteristic parameters: for example, current temperature value, deviation from background temperature, temperature change rate, and temperature gradient along and perpendicular to the water flow direction.

[0048] Hydraulic characteristic parameters: for example, the excess pore water pressure value, the amplitude and dominant frequency of pore water pressure fluctuations in the short term, and the amplitude and phase lag time of the pore water pressure response to changes in external hydraulic head.

[0049] The quantitative reference interval or statistical distribution model: for each key characteristic parameter in the list, its typical numerical range or key parameters of the probability density function under this mode are given.

[0050] The rules or association models of the cooperative change relationship describe how the characteristic parameters of different physical fields change cooperatively under this mode. It can be expressed in the following forms:

[0051] Rule sets based on logical conditions: for example, “IF (resistivity change rate <-X% / day) AND (acoustic emission high energy event rate > Y events / hour) THEN (highly consistent with the synergistic characteristics of mode A)”.

[0052] Mathematical models based on statistical correlations: for example, describing the functional relationship between resistivity changes and saturation changes.

[0053] Synchronicity or hysteresis in the time series: for example, “A significant decrease in resistivity is usually observed first, followed by a peak in acoustic emission activity within XX hours.”

[0054] In another specific embodiment, the method further includes the step of pre-building the seepage evolution pattern record, the step including at least one of the following:

[0055] By establishing a multi-physics field coupling mathematical model that can reflect the specific geological structure of the target sluice project, and performing parameterized design and multi-condition dynamic simulation of the preset typical seepage evolution path, the full-process multi-physics field response data of the virtual sensor is recorded, and the seepage evolution pattern record is summarized and extracted from it;

[0056] Or by implementing a physical model test or in-situ controlled test that can simulate the specific seepage destruction process of the target sluice, a sensor array is deployed in the test model and multi-physical field data is collected synchronously, and the response characteristics observed in the test are associated with the actual seepage evolution state to calibrate and generate the seepage evolution pattern record.

[0057] Specifically, to ensure the scientific nature and reliability of the knowledge base, its initialization and construction process is extremely rigorous:

[0058] 1. Based on high-fidelity multi-physics field coupling numerical simulation:

[0059] First, based on the detailed geological survey report of the target sluice, a sophisticated hydraulic-geotechnical-acoustic-electrical-thermal multi-physics field coupling mathematical model is established.

[0060] Secondly, for several key implicit seepage evolution paths, a set of key control parameters and their reasonable values ​​that can characterize the entire process are set.

[0061] Then, a non-steady-state multi-physics coupling simulation is performed for each combination in the parameter space, and the time series data of all physical field responses obtained by the virtual sensor are recorded.

[0062] Finally, signal processing and pattern recognition algorithms were used to analyze massive simulation data, and the results with similar multi-physical field collaborative response evolution characteristics and clearly corresponding to specific seepage stages were summarized into a seepage evolution pattern record.

[0063] Calibration and validation based on large-scale or in-situ controlled physical model tests:

[0064] Design and implement large-scale indoor physical model tests or small-scale in-situ tests that can realistically simulate key seepage failure processes.

[0065] A sensor array similar to that in actual engineering is deployed in the test model, boundary conditions are precisely controlled, and multi-physics field data is synchronously collected at high frequency until the model undergoes expected penetration damage.

[0066] After the test, the actual development and damage mechanism of the internal seepage network were accurately determined through excavation dissection and dye tracing.

[0067] The multi-physics response characteristics observed in the experiment are accurately correlated with the actual evolution state, which is used to strictly calibrate and verify the seepage evolution pattern records obtained through numerical simulation.

[0068] Knowledge base maintenance and optimization mechanism:

[0069] After actual operation, the parameters and rules of relevant records in the knowledge base were fine-tuned using real case data verified by engineering surveys. At the same time, domain experts were organized to regularly review and inject knowledge into the knowledge base, and a strict version management system was established to ensure that the quality of the knowledge base kept pace with the times.

[0070] In a preferred embodiment, the step of matching to identify the collaborative anomaly pattern includes:

[0071] Converting the collaborative response rules into a set of executable logic judgment rules;

[0072] Calculating the degree of membership of the current feature value in the multidimensional spatiotemporal feature sequence to the logic judgment rule;

[0073] By adopting weighted rules and fuzzy logic reasoning, the activation strength of all logical judgment rules and their preset weights are integrated to calculate the matching confidence score between the current monitoring area and each of the seepage evolution pattern records to identify the collaborative anomaly pattern.

[0074] The specific steps are as follows: First, the "coordinated response rules" in the knowledge base are converted into a computer program. Then, for the real-time acquired feature values, their degree of conformity with the "normal range" defined in the rules is calculated, i.e., the "membership degree" (a value between 0 and 1). Finally, using methods such as fuzzy comprehensive evaluation, the degree of conformity of all rules is comprehensively considered, along with pre-set expert weights (for example, changes in water pressure may be given a higher weight than changes in temperature). Ultimately, a similarity score, i.e., the "match confidence score," is calculated between the current state and each "seepage evolution pattern" in the knowledge base (such as the "piping incubation period"). The pattern with the highest score is the most likely coordinated anomaly pattern at the moment.

[0075] In another preferred embodiment, the matching step to identify the collaborative anomaly pattern can also employ dynamic state sequence decoding based on a probabilistic graphical model (e.g., a hidden Markov model). Specifically, a hidden Markov model is pre-constructed, whose hidden state set corresponds to several key, ordered percolation evolution stages defined in a knowledge base. The model's emission and transition probabilities are trained using historical or simulated data. In real-time applications, the Viterbi algorithm is used to dynamically decode the most probable hidden state sequence, i.e., the most likely percolation evolution stage, experienced by the currently monitored area over the past period of time, based on the input multi-dimensional spatiotemporal feature sequence.

[0076] Based on the identified collaborative anomaly pattern, predict its risk evolution trend;

[0077] Specifically, after identifying the current collaborative anomaly pattern, this step uses advanced time series prediction models (e.g., long short-term memory networks) to learn the historical time series data of key risk indicators in the pattern (such as the matching confidence score with high-risk patterns, indirect estimates of local permeability, acoustic emission energy release rate, etc.), so as to predict their possible change trends in a shorter time window in the future (e.g., the next few hours to days).

[0078] Furthermore, after the step of matching to identify the collaborative anomaly pattern, the method further includes:

[0079] All monitoring units in the target area inside the foundation are regarded as nodes in graph theory, and a seepage activity network representing the spatial distribution and correlation status of the dominant seepage activity area is constructed and dynamically updated in real time according to the matching confidence score of each monitoring unit;

[0080] Graph theory analysis algorithms are used to quantitatively analyze the topological structure of the percolation activity network and its evolution characteristics over time.

[0081] Specifically, to assess risk from a global perspective, all monitoring units represented by in-situ multi-parameter collaborative sensing probes are treated as nodes in a graph. Based on the matching confidence score calculated in real time for each node, an edge representing the association of seepage activity is defined between multiple spatially adjacent nodes whose scores indicate a high risk of seepage activity. The weight of the edge is determined based on factors such as the similarity and distance between the node scores. This results in a weighted "seepage activity network" that is constructed and dynamically updated in real time. Graph theory analysis algorithms (such as connected component identification, shortest path algorithms, and key node identification algorithms) are then used to quantitatively analyze the network's topology (e.g., the presence of highly active, nearly interconnected paths or network clusters extending from upstream to downstream) and its temporal evolution. This "point-to-network" analysis is key to assessing global and structural risks and provides an important basis for predicting future risk trends.

[0082] By integrating the collaborative abnormal pattern and the risk evolution trend and integrating the preset physical instability criteria, the sluice risk is evaluated in real time, and early warning information is generated when the evaluation result meets the preset conditions.

[0083] Specifically, this step is the decision output of this method. The strategy of multi-criteria fusion is adopted to avoid the risk of misjudgment of a single indicator.

[0084] In a preferred embodiment, the step of performing real-time assessment of sluice risk includes:

[0085] Comprehensively evaluate the following criteria: whether the risk status of the local monitoring unit continues to be at or rapidly enters the preset danger precursor level, whether the global topology of the seepage activity network reaches or approaches the critical penetration threshold, and the degree of proximity between the calculated local hydraulic gradient or energy dissipation index and the physical instability criterion;

[0086] And determine the risk assessment results based on the comprehensive satisfaction of the multiple criteria.

[0087] Specifically, the final risk assessment and warning triggering integrate at least three levels of information: First, the risk status of the local monitoring unit, namely, whether the confidence score of the coordinated anomaly pattern matching of a single monitoring point remains high or rapidly increases. Second, the global topological criticality of the "dominant seepage activity area network," namely, whether the network's connectivity, permeability, or expansion rate has reached or is approaching its critical percolation threshold. Third, the degree of conformity with known physical instability criteria. This involves comparing indirectly inferred parameters such as hydraulic gradient with classic soil mechanics instability criteria (such as the Tayzaghi piping criterion) to assess their degree of similarity. Only when information from multiple levels collectively points to a hazard does the system dynamically assess and output a corresponding warning level (e.g., daily attention, general warning, severe warning, and serious alarm) through a pre-set, expert-adjustable decision logic system (e.g., a decision table based on weighted scoring or a simple neural network classifier). Detailed warning information is then generated, including the risk location, type, and evolutionary trend.

[0088] In a preferred embodiment, the method further comprises:

[0089] Based on the real-time multi-physics field data and the physical model constraints in the knowledge base, a three-dimensional effective hydraulic conductivity characteristic field and a dominant seepage network topology structure inside the foundation are reconstructed in real time through multi-physics field joint inversion;

[0090] The three-dimensional effective hydraulic conductivity characteristic field and the dominant seepage network topology are presented through a three-dimensional dynamic visualization interface.

[0091] Specifically, a multi-physics joint inversion engine is used to deeply integrate and calculate all collected data, directly inverting the distribution of key physical parameters such as the three-dimensional permeability field, saturation field, and porosity field within the foundation. Then, through a three-dimensional dynamic visualization platform, these invisible geological parameter fields and the identified dominant seepage network are intuitively presented to engineering technicians in the form of animations, profiles, and streamlines, allowing them to clearly "see" the location and form of seepage risks and their changes over time, providing the most powerful data support for final risk assessment and disposal decisions.

[0092] In a specific example, a 40-year-old concrete gravity dam had known minor seepage issues around the dam. A small amount of seepage had been recorded near the dam foundation corridor (observation corridor) on the right bank, particularly during high water levels during flood season. The operations team was concerned that with dam aging and increasing extreme weather events, this weak area could develop into a concentrated, dominant seepage channel, ultimately leading to piping or seepage damage.

[0093] Step One: Deployment of the Sensing Layer: Comprehensive coverage of the entire dam foundation is unnecessary. Based on historical data and geological analysis, eight "in-situ multi-parameter collaborative sensing probes" were deployed in a "pin" pattern, using drill holes, in the identified highest-risk area—around the right bank dam foundation corridor and downstream. This number is sufficient to form a localized, high-density monitoring network with basic three-dimensional resolution capabilities.

[0094] The "active excitation and auxiliary observation system" will not be deployed for the time being, and we will rely entirely on passive synchronous acquisition.

[0095] Step 2: Knowledge Base Construction: There is no need to build a massive knowledge base containing dozens of patterns. First, focus on the core issues to be verified and only build the three most critical "percolation evolution pattern records":

[0096] PEMR-01: Stable permeation mode (baseline state)

[0097] PEMR-02: Saturation uniform rise mode (normal response at high water level)

[0098] PEMR-03: Early pattern formation along the dominant microcrack channel (core warning target)

[0099] The knowledge base was constructed primarily using a high-fidelity multi-physics coupled numerical simulation approach. A refined local numerical model encompassing only the right-bank dam section and foundation was established. By setting different boundary conditions and microcrack parameters within the model and running several simulations, the "multi-physics synergistic response characteristic spectra" corresponding to the three aforementioned modes were extracted and calibrated.

[0100] Step 3: Implementation of diagnosis and warning engine Diagnosis algorithm: Adopt the relatively easy-to-implement "matching confidence scoring method based on weighted rules and fuzzy logic reasoning".

[0101] Early warning logic: Set up a simplified two-layer early warning logic:

[0102] Attention (blue warning): The matching confidence score of any probe detected status with "PEMR-03" (early stage of dominant channel formation) remains higher than 40 points within 24 hours.

[0103] Warning (yellow warning): There are at least two spatially adjacent probes whose matching confidence scores with "PEMR-03" are continuously higher than 70 points within 6 hours.

[0104] The complex global topological analysis of the “seepage activity network” will not be conducted for the time being, and the early warning will be mainly based on the risk status of the local monitoring unit.

[0105] Step 4: Presentation of results: Only a simple web dashboard is needed.

[0106] Dashboard interface:

[0107] Left: Schematic diagram showing the eight probe locations.

[0108] Middle: In the form of a line graph, the 24-hour change curve of each probe's key physical parameters (such as excess pore water pressure and resistivity) is displayed in real time.

[0109] Right: Displays the "matching confidence score" between the current system status and the three "seepage evolution pattern records" in real time in the form of a dashboard or bar chart.

[0110] Top: A prominent status bar showing "System Normal," "Blue Attention," or "Yellow Warning."

[0111] Notification method: When a yellow warning is triggered, the system will send a warning notification to the dam safety person in charge via email and text message.

[0112] Running the demo in the scene:

[0113] Deployment and Baseline Learning: The system was deployed and operated for one month during the dry season. The dashboard showed that the confidence score for the system's status matching the "PEMR-01: Stable Seepage Mode" remained consistently above 90, while the scores for the other two modes were extremely low. The operations team confirmed the system's baseline was normal.

[0114] Incident: Entering the flood season, the water level upstream rapidly rose by 15 meters within 3 days.

[0115] System Response and Diagnosis:

[0116] The pore water pressure readings of all eight probes increased significantly, and the resistivity generally decreased.

[0117] On the dashboard, the "PEMR-01" score dropped rapidly, and the "PEMR-02: Saturation Uniform Rising Mode" score rose sharply and then stabilized, which is consistent with the normal response under high water levels.

[0118] Key changes occurred: the resistivity of probes No. 3 and No. 5, located on the downstream side of the corridor, decreased significantly faster than that of other probes, and their acceleration sensors captured weak but continuous, high-frequency acoustic emission signals.

[0119] Diagnostic Engine Analysis: The system matches these features with its knowledge base in real time. Due to the abnormally accelerated decrease in resistivity (indicating the possible formation of a localized water channel) and the acoustic emission signal (indicating soil particle friction or microcrack activity), the system calculated the status of the area around probes 3 and 5. The confidence score for the match with the "PEMR-03: Initial pattern of dominant channel formation along microcracks" indicator rose rapidly within a few hours and stabilized above 75.

[0120] Warning output:

[0121] The status bar at the top of the dashboard turns yellow and flashes to indicate "Yellow Warning".

[0122] Probes 3 and 5 are highlighted red on the schematic.

[0123] The dam safety officer received a text message on his phone: "HYDRA-VIEW Warning: The risk status of the downstream monitoring area (probes 3 and 5) of the right bank corridor is highly consistent with the 'initial pattern of dominant channel formation' (score 78 / 100). Please immediately activate the key focus plan."

[0124] After receiving the early warning, the operation and maintenance team was able to take targeted preventive measures such as curtain grouting, effectively avoiding further deterioration of safety hazards.

[0125] In summary, the real-time evolution prediction and assessment method of sluice risk provided by the present invention realizes dynamic, accurate and forward-looking diagnosis and early warning of hidden seepage risks in deep foundations of sluices through collaborative perception of multiple physical fields, pattern matching based on knowledge base, trend prediction based on data drive and risk assessment based on multi-criteria fusion. It fundamentally overcomes the limitations of existing technologies, realizes the technical effect of transforming sluice safety monitoring from a "blind spot" to a "measurable area" and transforming risk response from "passive response" to "active foresight", and has significant technological progress and great engineering application value.

[0126] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention and are all included in the scope of protection of the present invention.

Claims

1. A method for predicting and evaluating the real-time evolution of sluice risk, characterized in that: include: Through the in-situ multi-parameter collaborative sensing network deployed inside the deep foundation of the sluice, real-time multi-physical field data of the target area inside the foundation are synchronously collected; Extracting a multidimensional spatiotemporal feature sequence from the real-time multi-physics field data, and matching the multidimensional spatiotemporal feature sequence with a seepage evolution pattern record pre-built in a multi-physics field data fusion and knowledge base to identify a collaborative anomaly pattern representing a current seepage state; Based on the identified collaborative anomaly pattern, predict its risk evolution trend; By integrating the collaborative abnormal pattern and the risk evolution trend and integrating the preset physical instability criteria, the sluice risk is assessed in real time, and an early warning message is generated when the assessment result meets the preset conditions; The step of collecting real-time multi-physical field data includes: synchronously collecting data using a plurality of in-situ multi-parameter collaborative sensing probes arranged in a three-dimensional array in the target area inside the foundation; wherein the in-situ multi-parameter collaborative sensing probes are composite detection units, which are integrated with: a pore water pressure sensor, a three-axis broadband acceleration sensor, a temperature sensor, and a resistivity measurement unit; The seepage evolution pattern record includes: a list of associated key physical field characteristic parameters; a quantitative reference interval or statistical distribution model defined for the key characteristic parameters in the list; and a rule or association model for describing the coordinated change relationship between the characteristic parameters of different physical fields under the seepage evolution pattern; The method also includes a step of pre-constructing the seepage evolution pattern record, which includes at least one of the following: by establishing a multi-physical field coupling mathematical model that can reflect the specific geological structure of the target sluice project, and performing multi-working condition dynamic simulation on a preset typical seepage evolution path, recording the full-process multi-physical field response data of the virtual sensor, and summarizing and extracting the seepage evolution pattern record therefrom; or, by implementing a physical model test or in-situ controlled test that can simulate the specific seepage destruction process of the target sluice, deploying a sensor array in the test model and synchronously collecting multi-physical field data, and correlating the response characteristics observed in the test with the actual seepage evolution state to calibrate and generate the seepage evolution pattern record.

2. A method for predicting and evaluating the real-time evolution of sluice risk according to claim 1, characterized in that: The matching step to identify the coordinated abnormal pattern that characterizes the current seepage state includes: Converting the rules or association model of the cooperative change relationship defined in the seepage evolution pattern record into a set of executable logical judgment rules; Calculating the degree of membership of the current feature value in the multidimensional spatiotemporal feature sequence to the logic judgment rule; By adopting weighted rules and fuzzy logic reasoning, the activation strength of all logical judgment rules and their preset weights are integrated to calculate the matching confidence score between the current monitoring area and each of the seepage evolution pattern records to identify the collaborative anomaly pattern.

3. A method for predicting and evaluating the real-time evolution of sluice risk according to claim 2, characterized in that: After the step of matching to identify the collaborative abnormal pattern that characterizes the current seepage state, the method further includes: treating all monitoring units in the target area inside the foundation as nodes in graph theory, and constructing and dynamically updating the seepage activity network that characterizes the spatial distribution and correlation state of the dominant seepage activity area in real time based on the matching confidence score of each monitoring unit; and using a graph theory analysis algorithm to quantitatively analyze the topological structure of the seepage activity network and its evolution characteristics over time.

4. A method for predicting and evaluating the real-time evolution of sluice risk according to claim 3, characterized in that: The step of real-time assessment of the risk of the sluice gate includes: comprehensively assessing the following criteria: whether the risk status of the local monitoring unit is continuously at or rapidly enters a preset danger precursor level, whether the global topological structure of the seepage activity network reaches or approaches a critical penetration threshold, and the degree of proximity between the calculated local hydraulic gradient or energy dissipation index and the physical instability criterion; wherein, the risk status assessment dynamically determines the risk assessment result by weightedly scoring the collaborative abnormal pattern matching confidence score of the local monitoring unit, the connectivity, penetration or expansion rate index of the dominant seepage activity network, and the local hydraulic gradient or energy dissipation index with the classical soil mechanics instability criterion.

5. The method for predicting and evaluating the real-time evolution of sluice risk according to claim 1, characterized in that: The method also includes: before or during the step of collecting the real-time multi-physical field data, actively emitting controllable elastic wave signals through active excitation and auxiliary observation, and receiving them by the in-situ multi-parameter collaborative sensing network to enhance the detection capability of dynamic changes in the foundation medium structure.

6. A method for predicting and evaluating the real-time evolution of sluice risk according to claim 1, characterized in that: The method further includes: based on the real-time multi-physics field data and the physical model constraints in the knowledge base, reconstructing the three-dimensional effective hydraulic conductivity characteristic field and the dominant seepage network topology structure inside the foundation in real time through multi-physics field joint inversion; and presenting the three-dimensional effective hydraulic conductivity characteristic field and the dominant seepage network topology structure through a three-dimensional dynamic visualization interface.

Images