Dam leakage diagnosis method and system based on AI technology

Abstract

The application relates to the dam body detection technical field, and in particular to a dam body leakage diagnosis method and system based on AI technology.The method comprises the following steps: acquiring a sensor array data set, analyzing the sensor array data set, determining a unified space-time detection data set; analyzing the unified space-time detection data set, determining a freeze-thaw conversion time axis, and according to the freeze-thaw conversion time axis, tracking and screening corresponding detection data in the unified space-time detection data set to determine a target period detection data set; according to the target period detection data set, performing evolution simulation on the leakage development trend of a dam foundation curtain, and determining and outputting dam foundation curtain leakage trend information.The application can accurately capture the promoting effect of the freeze-thaw cycle on the dam foundation curtain leakage condition, and provide the dam foundation curtain leakage trend information for intuitively representing the dam foundation curtain leakage development state to corresponding dam body maintenance personnel, so that the hysteresis of the dam foundation curtain leakage detection result can be effectively eliminated.

Description

Technical Field

[0001] This application relates to the field of dam detection technology, and in particular to a dam leakage diagnosis method and system based on AI technology. Background Technology

[0002] Leakage is an early sign of dam instability. Detecting dam leakage is a key step in extending the service life of dams and reducing the risk of dam safety accidents. It can effectively inhibit the aging effect of dam leakage on dam materials and is one of the core requirements of water conservancy project safety management.

[0003] However, existing dam seepage detection technologies are insufficient to accurately and timely monitor the curtain seepage of dam foundations in high-altitude and cold regions. As a result, the detection results of curtain seepage of dam foundations often lag behind the development trend of curtain seepage of the dam body, leading to unsatisfactory dam maintenance results based on the curtain seepage detection results. Summary of the Invention

[0004] This application provides a method and system for diagnosing dam leakage based on AI technology to solve the above-mentioned technical problems.

[0005] Firstly, this application provides a dam leakage diagnosis method based on AI technology, the method comprising:

[0006] Acquire the sensor array dataset, analyze the sensor array dataset, and determine a unified spatiotemporal detection dataset;

[0007] Analyze the unified spatiotemporal detection dataset, determine the freeze-thaw transition time axis, and based on the freeze-thaw transition time axis, track and filter the corresponding detection data within the unified spatiotemporal detection dataset to determine the target period detection dataset;

[0008] Based on the target periodic detection dataset, the evolutionary trend of seepage development in the dam foundation curtain is simulated, and the seepage trend information of the dam foundation curtain is determined and output.

[0009] This solution performs spatiotemporal alignment processing on the data within the sensor array dataset to obtain a unified spatiotemporal detection dataset. This effectively eliminates leakage analysis errors caused by spatiotemporal data deviations, ensuring that subsequent leakage analysis is based on a solid and reliable data foundation. Based on the unified spatiotemporal detection dataset, focusing on the freeze-thaw transition time axis of the dam foundation curtain, the corresponding detection data within the unified spatiotemporal detection dataset is tracked and filtered to obtain a target cycle detection dataset. This reduces the negative impact of data blind spots on the accuracy of leakage analysis. On this basis, the evolutionary development trend of leakage in the dam foundation curtain is simulated, accurately capturing the driving effect of freeze-thaw cycles on the leakage situation of the dam foundation curtain. The leakage trend information of the dam foundation curtain, which is used to intuitively characterize the leakage development status of the dam foundation curtain, is provided to the corresponding dam maintenance personnel, effectively eliminating the lag in the dam foundation curtain leakage detection results.

[0010] Optionally, the sensor array dataset includes a subset of temperature data, a subset of osmotic pressure data, and a subset of deformation data;

[0011] The temperature data subset was acquired by distributed fiber optic temperature sensors arranged in a three-dimensional mesh topology on the dam foundation curtain.

[0012] The subset of seepage pressure data is acquired by distributed fiber optic micro-pressure sensors arranged in the form of the three-dimensional grid topology on the dam foundation curtain.

[0013] The deformation data subset is acquired by distributed fiber optic stress sensors deployed on the dam foundation curtain in the form of the three-dimensional mesh topology.

[0014] The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are all deployed on the side of the dam foundation curtain facing the upstream of the dam body.

[0015] The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are connected in series to the same fiber optic path using optical wavelength division multiplexing technology.

[0016] This scheme deploys distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors on the dam foundation curtain in a three-dimensional mesh topology, forming a three-dimensional monitoring network. This allows for timely detection of leakage at both horizontal and vertical joints. Furthermore, wavelength division multiplexing (WDM) technology enables single-fiber transmission of multi-sensor data, eliminating signal delay differences caused by traditional multi-cable deployments and reducing sensor deployment costs. By confining the sensors to the upstream side of the dam, the system can promptly capture the characteristics exhibited in the initial stages of leakage development, reducing the lag in subsequent leakage analysis.

[0017] Optionally, the analysis of the sensor array dataset to determine a unified spatiotemporal detection dataset includes:

[0018] Based on the three-dimensional grid topology of the distributed optical fiber sensors on the dam foundation curtain, a spatial grid model of the dam foundation curtain is constructed.

[0019] Based on the dam foundation curtain space grid model, the topology of the three-dimensional grid is analyzed to determine the relative position coordinates of each sensor and construct a unified spatial position coordinate set.

[0020] Analyze the unified spatial coordinate set and evaluate the signal processing delay of each sensor based on the relative distance between each sensor and its corresponding signal demodulator;

[0021] Based on the signal processing delay, timestamp correction processing is performed on the data collected by each sensor in the sensor array dataset to determine a unified timestamp dataset.

[0022] Based on the unified spatial coordinate set and the unified timestamp dataset, each data point in the sensor array dataset is marked with a spatial location and a timestamp, respectively. After the spatiotemporal unification is completed, the temperature data synchronization subset, the osmotic pressure data synchronization subset, and the deformation data synchronization subset are determined, and the unified spatiotemporal detection dataset is constructed.

[0023] This scheme performs rigorous spatiotemporal alignment processing on temperature, pressure, and stress data collected from different sensors. Based on the unified spatial coordinate set and unified timestamp dataset obtained from the spatiotemporal alignment, each data point in the sensor array dataset is marked with its spatial location and timestamp. By integrating the synchronized subsets of temperature, pressure, and deformation data after spatiotemporal alignment, a unified spatiotemporal detection dataset is constructed. This prevents misjudgments of data coupling relationships caused by spatiotemporal differences in the data, providing high-confidence data for subsequent freeze-thaw-leakage coupling analysis and significantly improving the reliability of leakage trend prediction and the scientific nature of maintenance decisions.

[0024] Optionally, the analysis of the unified spatiotemporal detection dataset to determine the freeze-thaw transition time axis includes:

[0025] Based on the aforementioned spatial grid model of the dam foundation curtain, and according to the synchronized subset of temperature data in the aforementioned unified spatiotemporal detection dataset, a curtain temperature gradient distribution map is constructed.

[0026] Based on the aforementioned dam foundation curtain spatial grid model, and according to the synchronized subset of seepage pressure data in the aforementioned unified spatiotemporal detection dataset, a seepage pressure change rate distribution map is constructed.

[0027] Spatiotemporal correlation analysis is performed on the curtain temperature gradient distribution map and the permeation pressure change rate distribution map. When the distribution of extreme points of temperature gradient in the curtain temperature gradient distribution map and the inflection point of permeation pressure change rate in the permeation pressure change rate distribution map form a spatiotemporal coupling relationship, it is determined that the region corresponding to the current data has entered the critical state of freeze-thaw transition, and the timestamp of the corresponding data is extracted as the freeze-thaw start time node of the region where the current sensor is located.

[0028] The curtain depth of the dam foundation curtain is determined based on the spatial grid model of the dam foundation curtain.

[0029] Based on the curtain depth, and according to the freeze-thaw start time node corresponding to the area where each sensor is located, the freeze-thaw state is deduced by analyzing the sensor deployment blind zone outside the area where the sensor is located, and the freeze-thaw start time node of the sensor deployment blind zone is determined.

[0030] The freeze-thaw transition time axis is constructed based on the freeze-thaw start time nodes corresponding to the area where the sensor is located and the blind zone where the sensor is deployed.

[0031] This scheme analyzes synchronized subsets of temperature and seepage pressure data to derive curtain temperature gradient distribution maps and seepage pressure change rate distribution maps, respectively, reflecting the characteristics of temperature and seepage pressure changes. Based on the spatiotemporal coupling relationship between the extreme points of the temperature gradient and the inflection points of the seepage pressure change rate, the time point when the sensor-corresponding area enters the critical state of freeze-thaw transition is captured as the freeze-thaw initiation time node. On this basis, combined with the depth of the dam foundation curtain where the sensor coverage blind zone is located, the freeze-thaw state of the sensor deployment blind zone is deduced, thereby obtaining the freeze-thaw initiation time node of the sensor deployment blind zone. This allows the analysis of the freeze-thaw time cycle to cover the entire area of ​​the dam foundation curtain, thus improving the comprehensiveness and accuracy of subsequent leakage analysis.

[0032] Optionally, based on the curtain depth, and according to the freeze-thaw initiation time node corresponding to the area where each sensor is located, the step of analyzing the sensor deployment blind zone outside the sensor's location area to deduce the freeze-thaw state and determine the freeze-thaw initiation time node of the sensor deployment blind zone includes:

[0033] Based on the spatial grid model of the dam foundation curtain and the curtain depth, according to the preset unit depth interval, the dam foundation curtain is divided into several temperature conduction levels along the vertical depth direction;

[0034] Based on the freeze-thaw initiation time nodes corresponding to several sensors located in adjacent temperature conduction layers, the propagation rate of the freeze-thaw transition critical state between adjacent temperature conduction layers is determined, and an inter-layer freeze-thaw state transmission chain is constructed.

[0035] Based on the inter-level freeze-thaw state transmission chain, the freeze-thaw start time node corresponding to each sensor deployment blind zone is determined according to the freeze-thaw start time node corresponding to several sensors located in adjacent temperature conduction levels.

[0036] This scheme divides the entire area corresponding to the dam foundation curtain into several temperature conduction levels based on a preset unit depth range, further refining the granularity of the analysis. Based on the freeze-thaw initiation time nodes corresponding to several sensors within adjacent temperature conduction levels, the propagation rate of the freeze-thaw transition critical state between temperature conduction levels is analyzed, and a freeze-thaw state transmission chain between levels is constructed. In this way, the freeze-thaw initiation time node corresponding to the sensor deployment blind zone is derived, so that the subsequent leakage analysis process can cover the entire area of ​​the dam foundation curtain while based on a precise time axis, preventing misjudgment of leakage caused by monitoring blind zones.

[0037] Optionally, the step of tracking and filtering the corresponding detection data within the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target period detection dataset includes:

[0038] Based on the freeze-thaw transition time axis, the temperature data synchronization subset, the osmotic pressure data synchronization subset, and the deformation data synchronization subset that are temporally correlated with the freeze-thaw start time node are extracted from the unified spatiotemporal detection dataset to construct a sensor coverage area synchronization dataset;

[0039] Based on the propulsion rate in the inter-level freeze-thaw state transmission chain, and combined with the relative spatial distance between the sensor deployment blind zone and adjacent sensors, a spatiotemporal interpolation function model is established.

[0040] Based on the spatiotemporal interpolation function model, and according to the synchronous dataset of the sensor coverage area, the temperature data, osmotic pressure data, and deformation data of the sensor deployment blind zone on the freeze-thaw transition time axis are spatiotemporally continuous interpolated to generate a blind zone interpolation data sequence.

[0041] The blind zone interpolation data sequence is spatiotemporally aligned with the data sequence in the sensor coverage area synchronization dataset, and the target periodic detection dataset covering the entire dam foundation curtain area is constructed based on the sensor coverage area synchronization dataset and the spatiotemporally aligned blind zone interpolation data sequence.

[0042] This scheme, based on the synchronous dataset of the sensor coverage area, and according to the advancement rate in the freeze-thaw state transmission chain between layers, combined with the relative spatial distance between the sensor blind zone and adjacent sensors, utilizes a blind zone interpolation strategy driven by the physical conduction mechanism to generate a blind zone interpolation data sequence. Based on the synchronous dataset of the sensor coverage area and the spatiotemporally aligned blind zone interpolation data sequence, a target periodic detection dataset is jointly constructed, significantly improving the spatial coverage of the target periodic detection dataset, effectively eliminating monitoring blind zones, and improving the accuracy and comprehensiveness of leakage analysis results.

[0043] Optionally, the step of performing an evolutionary simulation of the seepage development trend of the dam foundation curtain based on the target periodic detection dataset, and determining and outputting the dam foundation curtain seepage trend information, includes:

[0044] Analyze the spatial mesh model of the dam foundation curtain to identify and mark the joint areas of the cylindrical structure;

[0045] Based on the material leakage test data, the temperature data, osmotic pressure data, and deformation data in the target period detection dataset are assigned basic leakage influence weights respectively;

[0046] Based on the joint leakage test data, the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the joint area of ​​the cylindrical structure in the target period detection dataset are subjected to key area influence correction processing to determine the leakage influence weights of the joint area.

[0047] Based on the target period detection dataset, according to the basic leakage influence weight and the joint area leakage influence weight, the temperature data, seepage pressure data and deformation data corresponding to different areas in the target period detection dataset are weighted and fused to determine the real-time leakage risk corresponding to different areas.

[0048] Based on the real-time leakage risk and corresponding deformation data of different regions, several high-risk leakage areas are identified, and a leakage risk distribution heat map is constructed based on the several high-risk leakage areas.

[0049] The heat map of leakage risk distribution is used as a visual output of the leakage trend information of the dam foundation curtain.

[0050] This scheme analyzes the spatial grid model of the dam foundation curtain wall, identifies the joint areas of the cylindrical structure, and corrects the impact of the data corresponding to the joint areas of the cylindrical structure based on the foundation leakage influence weight, determining the leakage influence weight of the joint areas. On this basis, according to the foundation leakage influence weight and the joint area leakage influence weight, the temperature data, seepage pressure data, and deformation data under different areas are weighted and fused for analysis, resulting in the real-time leakage risk corresponding to different areas. A corresponding leakage risk distribution heat map is generated as a visualization output of the leakage trend information of the dam foundation curtain wall, making the assessment of leakage risk in the cylindrical structure joint area, which has a high leakage risk, more accurate.

[0051] Optionally, based on the joint leakage test data, the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the cylindrical structure joint area within the target periodic detection dataset are subjected to key area influence correction processing to determine the joint area leakage influence weights, including:

[0052] Based on the joint leakage test data, the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface were extracted.

[0053] The data on the change in joint width is positively correlated with the weight of the leakage impact in the joint area;

[0054] The thermal conductivity of the joint filling material is negatively correlated with the weight of leakage in the joint area.

[0055] The shear strength degradation coefficient of the joint interface is positively correlated with the weight of leakage influence in the joint area;

[0056] Based on the aforementioned basic leakage influence weights, and according to the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface, a joint area influence correction function is constructed.

[0057] Based on the joint area influence correction function, the influence weight of the basic leakage is adjusted by focusing on key areas, and the influence weight of the joint area leakage is quantified.

[0058] This scheme establishes the correlation between the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface in the joint leakage test data and the leakage influence weight of the joint area. A joint area influence correction function is constructed, and based on the joint area influence correction function, the basic leakage influence weight is modified for key areas. The leakage influence weight of the joint area is quantified, and a multi-parameter coupling correction mechanism is used to reduce the assessment error of the leakage risk of the joint area.

[0059] Optionally, based on the real-time leakage risk and corresponding deformation data of different areas, several high-risk leakage areas are determined, including:

[0060] The real-time leakage risk corresponding to different regions is compared with the risk precursor benchmark value. Based on the comparison results, regions where the real-time leakage risk exceeds the risk precursor benchmark value are marked as key analysis regions.

[0061] Based on the target period detection dataset, and taking the key analysis area as a benchmark, the deformation data corresponding to each key analysis area and its adjacent temperature conduction level are extracted from the target period detection dataset to construct a key area deformation dataset.

[0062] Based on the deformation dataset of the key areas, analyze the deformation development direction of the current key analysis areas and determine the target tracking areas;

[0063] The current tracking target area and the key analysis area are designated as the high-risk leakage area.

[0064] This solution integrates key analysis areas and tracking target areas into high-risk leakage areas through a dual assessment mechanism of real-time leakage risk and deformation development direction during the evaluation of high-risk leakage areas. By analyzing areas with potential leakage spread risks, the lag in risk identification is reduced.

[0065] Secondly, this application provides a dam leakage diagnosis system based on AI technology, the system comprising:

[0066] The data alignment module is used to acquire the sensor array dataset, analyze the sensor array dataset, and determine a unified spatiotemporal detection dataset.

[0067] The data tracking module is used to analyze the unified spatiotemporal detection dataset, determine the freeze-thaw transition time axis, and track and filter the corresponding detection data in the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target period detection dataset.

[0068] The risk analysis module is used to simulate the evolution of the leakage development trend of the dam foundation curtain based on the target periodic detection dataset, and to determine and output the leakage trend information of the dam foundation curtain. Attached Figure Description

[0069] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0070] Figure 1 This is a schematic diagram illustrating an application scenario provided in one embodiment of this application;

[0071] Figure 2 A flowchart illustrating an AI-based dam leakage diagnosis method provided in one embodiment of this application;

[0072] Figure 3 This is a schematic diagram of a dam leakage diagnosis system based on AI technology, provided as an embodiment of this application. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0074] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0075] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0076] Existing dam seepage detection technologies are insufficient for timely and accurate control of curtain seepage in dam foundations in high-altitude and cold regions. As a result, the detection results of curtain seepage in dam foundations often lag behind the development trend of curtain seepage in the dam body, leading to unsatisfactory dam maintenance results based on the curtain seepage detection results.

[0077] Based on this, this application provides a dam leakage diagnosis method and system based on AI technology. The data within the sensor array dataset is spatiotemporally aligned to obtain a unified spatiotemporal detection dataset, effectively eliminating leakage analysis errors caused by data spatiotemporal deviations and ensuring that subsequent leakage analysis is based on a solid and reliable data foundation. Based on the unified spatiotemporal detection dataset, focusing on the freeze-thaw transition time axis of the dam foundation curtain, the corresponding detection data within the unified spatiotemporal detection dataset is tracked and filtered to obtain a target cycle detection dataset, reducing the negative impact of data blind spots on the accuracy of leakage analysis. On this basis, the evolutionary development trend of the dam foundation curtain leakage is simulated, accurately capturing the driving effect of freeze-thaw cycles on the dam foundation curtain leakage situation, and providing dam foundation curtain leakage trend information, which intuitively characterizes the leakage development status of the dam foundation curtain, to the corresponding dam maintenance personnel, effectively eliminating the lag in dam foundation curtain leakage detection results.

[0078] Figure 1 This is a schematic diagram illustrating an application scenario provided by this application. In the process of detecting leakage in the dam foundation curtain wall, the method provided in this application is used to accurately capture the driving effect of freeze-thaw cycles on the leakage of the dam foundation curtain wall, effectively eliminating the lag in the detection results.

[0079] Specifically, the method of this application is applied to any server that communicates with the sensor array to obtain the sensor array dataset provided by the sensor array. The data within the sensor array dataset is then spatiotemporally aligned to obtain a unified spatiotemporal detection dataset, effectively eliminating leakage analysis errors caused by data spatiotemporal deviations. This ensures that subsequent leakage analysis is based on a solid and reliable data foundation. Based on the unified spatiotemporal detection dataset, the focus is on the freeze-thaw transition time axis of the dam foundation curtain. The corresponding detection data within the unified spatiotemporal detection dataset is tracked and filtered to obtain a target cycle detection dataset, thereby reducing the negative impact of data blind spots on the accuracy of leakage analysis. On this basis, the evolutionary development trend of leakage in the dam foundation curtain is simulated to accurately capture the driving effect of freeze-thaw cycles on the leakage situation of the dam foundation curtain. The leakage trend information of the dam foundation curtain, which is used to intuitively characterize the leakage development status of the dam foundation curtain, is provided to the corresponding dam maintenance personnel, effectively eliminating the lag in the leakage detection results of the dam foundation curtain.

[0080] For specific implementation details, please refer to the following examples.

[0081] Figure 2 This is a flowchart illustrating a dam leakage diagnosis method based on AI technology, provided as an embodiment of this application. The method of this embodiment can be applied to servers in the above scenarios. Figure 2 As shown, the method includes:

[0082] S201. Obtain the sensor array dataset, analyze the sensor array dataset, and determine the unified spatio-temporal detection dataset.

[0083] The sensor array dataset can be a set of multi-dimensional physical parameters of the dam foundation curtain collected in real time by the sensor array, including subsets of temperature, seepage pressure, and deformation data.

[0084] The unified spatio-temporal detection dataset can be a set of sensor data after spatio-temporal calibration, a data set containing unified timestamp and spatial coordinate marks, eliminating the data asynchrony problem caused by differences in sensor layout positions.

[0085] Specifically, the core purpose of building the dam foundation curtain is to block or control the seepage path of groundwater in the dam foundation rock or soil through an artificially constructed anti-seepage barrier, thereby ensuring the safe and stable operation of the dam; the dam foundation curtain supporting the dam in the alpine region is vulnerable to the influence of soil freeze-thaw cycles. The repeated freeze-thaw action will accelerate the aging and cracking of the dam curtain material, thereby exacerbating the development of the leakage situation. Moreover, due to the large scale and underground location of the dam foundation curtain, it is difficult to ensure the detection accuracy of the leakage situation of the dam foundation curtain during the freeze-thaw cycle in the existing technology and there is an obvious lag. By deploying a sensor array, the large-scale dam foundation curtain is divided into different monitoring areas based on the sensor positions. Through the spatio-temporal alignment algorithm, the detection data collected in the sensor array dataset is spatio-temporally unified to obtain the unified spatio-temporal detection dataset, providing scientific data in the same spatio-temporal dimension for subsequent leakage analysis and reducing the analysis error caused by data asynchrony.

[0086] S202. Analyze the unified spatio-temporal detection dataset, determine the freeze-thaw conversion timeline, and track and screen the corresponding detection data in the unified spatio-temporal detection dataset according to the freeze-thaw conversion timeline to determine the target period detection dataset.

[0087] The freeze-thaw conversion timeline can be a time series used to represent the time when different regions of the dam foundation curtain enter the freeze-thaw critical state.

[0088] The target period detection dataset can be a data set composed of the data in the unified spatio-temporal detection dataset that is on the freeze-thaw conversion timeline.

[0089] Specifically, due to the nonlinear characteristics of changes in key influencing factors such as temperature and seepage pressure during the freeze-thaw process, and the inability of sensors to completely cover the entire area of ​​the dam foundation curtain, existing technologies are unable to effectively analyze the continuous impact of freeze-thaw contraction on the leakage of the dam foundation curtain. Based on a unified spatiotemporal detection dataset, by combining extreme values ​​of temperature gradients (reflecting the release of latent heat of phase change) and seepage pressure inflection points (reflecting the sudden change in water pressure caused by the phase change of ice and water), the spatiotemporal boundary of freeze-thaw transition is accurately identified, and a freeze-thaw transition time axis is derived. On this basis, the changing trends of different influencing factors within the freeze-thaw transition time axis are tracked, and the changing state of influencing factors in areas not covered by sensors is deduced. In this way, a data analysis framework that can cover the entire dam foundation curtain area is constructed, and a target periodic detection dataset for mapping the leakage of each area of ​​the dam foundation curtain is obtained.

[0090] S203. Based on the target periodic detection dataset, perform evolution simulation of the leakage development trend of the dam foundation curtain, determine and output the leakage trend information of the dam foundation curtain.

[0091] The leakage development trend can refer to the development status of leakage in the dam foundation curtain.

[0092] Evolutionary simulation can be a process of simulating the progress of seepage in the dam foundation curtain based on current detection data.

[0093] Information on the leakage trend of the dam foundation curtain can be used to visualize and characterize the leakage situation and trend of the dam foundation curtain.

[0094] Specifically, after obtaining the target period detection dataset, based on a series of data within the target period detection dataset that directly or indirectly map the leakage situation in different areas of the dam foundation curtain, a weighted fusion strategy is used to assess the leakage risk exhibited by different areas of the dam foundation curtain. Then, based on the spatial transmission characteristics of leakage, the diffusion path of high leakage risk areas is tracked, generating dam foundation curtain leakage trend information for visual characterizing the development trend of dam foundation curtain leakage. This dam foundation curtain leakage trend information is then provided to the corresponding dam maintenance personnel through human-computer interaction hardware devices, such as high-definition displays, as intuitive reference data for subsequent dam maintenance personnel to formulate maintenance strategies.

[0095] This solution performs spatiotemporal alignment processing on the data within the sensor array dataset to obtain a unified spatiotemporal detection dataset. This effectively eliminates leakage analysis errors caused by spatiotemporal data deviations, ensuring that subsequent leakage analysis is based on a solid and reliable data foundation. Based on the unified spatiotemporal detection dataset, focusing on the freeze-thaw transition time axis of the dam foundation curtain, the corresponding detection data within the unified spatiotemporal detection dataset is tracked and filtered to obtain a target cycle detection dataset. This reduces the negative impact of data blind spots on the accuracy of leakage analysis. On this basis, the evolutionary development trend of leakage in the dam foundation curtain is simulated, accurately capturing the driving effect of freeze-thaw cycles on the leakage situation of the dam foundation curtain. The leakage trend information of the dam foundation curtain, which is used to intuitively characterize the leakage development status, is provided to the corresponding dam maintenance personnel, effectively eliminating the lag in the leakage detection results of the dam foundation curtain.

[0096] In some embodiments, the sensor array dataset includes temperature data subsets, seepage pressure data subsets, and deformation data subsets. The temperature data subset is acquired by distributed fiber optic temperature sensors arranged in a three-dimensional mesh topology on the dam foundation curtain. The seepage pressure data subset is acquired by distributed fiber optic micro-pressure sensors arranged in a three-dimensional mesh topology on the dam foundation curtain. The deformation data subset is acquired by distributed fiber optic stress sensors arranged in a three-dimensional mesh topology on the dam foundation curtain. The distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors are all arranged on the side of the dam foundation curtain facing upstream of the dam body. Using wavelength division multiplexing (WDM) technology, the distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors are connected in series to the same fiber optic path.

[0097] A subset of temperature data can be a time-series dataset reflecting temperature changes in different areas of the dam foundation curtain.

[0098] A subset of seepage pressure data can be a time-series dataset reflecting changes in pore water pressure within the dam foundation curtain.

[0099] A subset of deformation data can be a time-series dataset that reflects the deformation of the dam foundation curtain structure.

[0100] The three-dimensional mesh topology can be a three-dimensional monitoring network topology in which sensors are arranged at equal intervals as cubic nodes on the surface and in the depth of the dam foundation curtain.

[0101] A distributed fiber optic grating temperature sensor can be a sensor that achieves continuous temperature measurement along an optical fiber based on the wavelength shift of the grating reflection.

[0102] A distributed fiber optic micro-pressure sensor can be a sensor that realizes distributed sensing of micro-pressure based on changes in fiber micro-bending loss.

[0103] A distributed fiber optic stress sensor can be a sensor that realizes distributed stress sensing based on the time-domain reflection of polarized light from an optical fiber.

[0104] Optical wavelength division multiplexing (WDM) technology is a technique that enables parallel transmission of data from multiple sensors by allocating different wavelength channels within the same optical fiber.

[0105] Specifically, in the process of leakage detection during the freeze-thaw cycle of dam foundation curtain walls in high-altitude and cold regions, temperature data reveals the migration of the freeze-thaw interface, seepage pressure data reflects sudden changes in pore water pressure, and deformation data captures the trend of structural cracking. The spatiotemporal correlation of these three can map the leakage situation. Existing technologies have the following shortcomings in the data collection stage for leakage detection of dam foundation curtain walls in high-altitude and cold regions: Traditional point sensors can only acquire local temperature or pressure data and cannot capture the three-dimensional expansion characteristics of the leakage path caused by freeze-thaw cycles; discrete sensors have data timestamp deviations due to independent power supply and communication lines, making it difficult to establish the causal relationship between temperature gradient, sudden changes in seepage pressure, and deformation response; conventional electronic sensors are prone to drift or failure in low-temperature and high-humidity environments; dam foundation curtain walls are usually composed of several concrete cylinders spliced ​​together, and there are joints between the concrete cylinders, so the surface in contact with the soil is not a plane. Instead, the dam foundation is a wavy curved surface, which existing sensor deployment methods cannot adapt to. This solution forms a three-dimensional monitoring network by deploying distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors on the surface of the dam foundation curtain. This allows for timely detection of leakage at both horizontal and vertical joints. Furthermore, wavelength division multiplexing (WDM) technology enables single-fiber transmission of multi-sensor data, eliminating signal delay differences caused by traditional multi-cable deployments and reducing sensor deployment costs. The distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors are all deployed on the upstream side of the dam foundation curtain, enabling the sensors to capture the characteristics exhibited in the initial stage of leakage development. Deploying these sensors on the downstream side of the dam would significantly enhance the lag in leakage analysis.

[0106] This scheme deploys distributed fiber optic temperature sensors, distributed fiber optic micro-pressure sensors, and distributed fiber optic stress sensors on the dam foundation curtain in a three-dimensional mesh topology, forming a three-dimensional monitoring network. This allows for timely detection of leakage at both horizontal and vertical joints. Furthermore, wavelength division multiplexing (WDM) technology enables single-fiber transmission of multi-sensor data, eliminating signal delay differences caused by traditional multi-cable deployments and reducing sensor deployment costs. By confining the sensors to the upstream side of the dam, the system can promptly capture the characteristics exhibited in the initial stages of leakage development, reducing the lag in subsequent leakage analysis.

[0107] In some embodiments, a spatial grid model of the dam foundation curtain is constructed based on the three-dimensional grid topology of the distributed optical fiber sensors on the dam foundation curtain. Based on the spatial grid model, the three-dimensional grid topology is analyzed to determine the relative position coordinates of each sensor, thus constructing a unified spatial position coordinate set. The unified spatial position coordinate set is analyzed, and the signal processing delay of each sensor is evaluated based on the relative distance between each sensor and its corresponding signal demodulator. Based on the signal processing delay, the data collected by each sensor in the sensor array dataset is timestamped to determine a unified timestamped dataset. Based on the unified spatial position coordinate set and the unified timestamped dataset, each data point in the sensor array dataset is marked with a spatial position and a timestamp, respectively. After spatiotemporal unification, a synchronized subset of temperature data, a synchronized subset of seepage pressure data, and a synchronized subset of deformation data are determined, thus constructing a unified spatiotemporal detection dataset.

[0108] The dam foundation curtain space grid model can be a digital three-dimensional space model used to characterize the three-dimensional monitoring area, based on the three-dimensional grid topology layout of distributed optical fiber sensors on the dam foundation curtain.

[0109] The relative position coordinates can be the three-dimensional coordinates of the sensor in the spatial grid model of the dam foundation curtain.

[0110] A unified spatial location coordinate set can be a data set that records the location coordinates of all sensors in the spatial grid model of the dam foundation curtain.

[0111] A signal demodulator can be a component that converts the modulated optical signal output from an optical fiber sensor into an electrical signal.

[0112] Signal processing delay can be the time it takes for sensor-acquired data to be converted from physical signals to digital signals and transmitted to a signal demodulator.

[0113] Timestamp correction processing can calibrate the timestamps of raw sensor data based on signal processing delays, eliminating time asynchrony issues caused by differences in transmission paths.

[0114] A unified timestamp dataset can be a time-series dataset formed by timestamping all sensor data, ensuring that data from different sensors are strictly aligned on the timeline.

[0115] The temperature data synchronization subset can be a set of temperature data that has undergone spatiotemporal synchronization processing.

[0116] The synchronized subset of osmotic pressure data can be a set of osmotic pressure data that has undergone spatiotemporal synchronization processing.

[0117] The deformation data synchronization subset can be a set of deformation data that has undergone spatiotemporal synchronization processing.

[0118] Specifically, sensor data is prone to timestamp deviations due to differences in transmission distance (time differences caused by spatial differences, spatiotemporal inconsistency), failing to accurately reflect the dynamic coupling relationship of temperature-pressure-deformation during freeze-thaw cycles. For example, pressure changes caused by a sudden drop in temperature may be incorrectly correlated to subsequent time periods due to signal delays in spatial transmission processing, leading to delayed leakage warnings. By using a finite element mesh generation algorithm, a three-dimensional mesh is divided based on the three-dimensional mesh topology of the distributed fiber optic sensors on the dam foundation curtain, constructing a spatial mesh model of the dam foundation curtain. Based on the coordinate system of the spatial mesh model, the relative position coordinates of each sensor are marked, constructing a unified spatial position coordinate set. This ensures that the measurement standards of each sensor in the spatial dimension are consistent, achieving spatial alignment. Furthermore, the transmission distance of each sensor signal to the demodulator via the fiber optic path is determined. Combining the optical signal propagation speed and the inherent delay of the demodulator signal processing, the signal processing delay of each sensor is evaluated. Based on this signal processing delay, the timestamps corresponding to the original sensor data are subtracted and corrected, constructing a unified timestamp dataset. This ensures that the measurement standards of each sensor data in the time dimension are consistent, achieving time alignment.

[0119] This scheme performs rigorous spatiotemporal alignment processing on temperature, pressure, and stress data collected from different sensors. Based on the unified spatial coordinate set and unified timestamp dataset obtained from the spatiotemporal alignment, each data point in the sensor array dataset is marked with its spatial location and timestamp. By integrating the synchronized subsets of temperature, pressure, and deformation data after spatiotemporal alignment, a unified spatiotemporal detection dataset is constructed. This prevents misjudgments of data coupling relationships caused by spatiotemporal differences in the data, providing high-confidence data for subsequent freeze-thaw-leakage coupling analysis and significantly improving the reliability of leakage trend prediction and the scientific nature of maintenance decisions.

[0120] In some embodiments, based on the dam foundation curtain spatial grid model, a curtain temperature gradient distribution map is constructed according to a synchronized subset of temperature data in the unified spatiotemporal detection dataset; based on the dam foundation curtain spatial grid model, a seepage pressure change rate distribution map is constructed according to a synchronized subset of seepage pressure data in the unified spatiotemporal detection dataset; spatiotemporal correlation analysis is performed on the curtain temperature gradient distribution map and the seepage pressure change rate distribution map. When the distribution of extreme points of temperature gradient in the curtain temperature gradient distribution map and the inflection point of seepage pressure change rate in the seepage pressure change rate distribution map form a spatiotemporal coupling relationship, it is determined that the region corresponding to the current data has entered the critical state of freeze-thaw transition, and the timestamp of the corresponding data is extracted as the freeze-thaw start time node of the region where the current sensor is located; the curtain depth of the dam foundation curtain is determined according to the dam foundation curtain spatial grid model; based on the curtain depth, according to the freeze-thaw start time node corresponding to the region where each sensor is located, the freeze-thaw state is deduced by analyzing the sensor deployment blind zone outside the sensor's region, and the freeze-thaw start time node of the sensor deployment blind zone is determined; a freeze-thaw transition time axis is constructed according to the freeze-thaw start time nodes corresponding to the sensor's region and the sensor deployment blind zone respectively.

[0121] The curtain temperature gradient distribution map can be a three-dimensional temperature change heat map generated based on a synchronous subset of temperature data, reflecting the temperature gradient distribution characteristics of different areas of the dam foundation curtain.

[0122] The permeability change rate distribution map can be a dynamic pressure change trend map generated based on a synchronous subset of permeability data, used to characterize the rate of change of permeability inside the curtain over time.

[0123] Temperature gradient extreme points can be spatial coordinates of points in the temperature gradient distribution map where the gradient value exceeds a set threshold, reflecting the dynamic migration location of the freeze-thaw interface.

[0124] The inflection point of the rate of change of osmotic pressure can be a spatiotemporal node where the second derivative is zero in the distribution map of the rate of change of osmotic pressure, indicating the key position where the osmotic pressure trend reverses.

[0125] Spatiotemporal correlation analysis can be a process of jointly analyzing the coupling relationship between extreme points of temperature gradient and inflection points of osmotic pressure change rate in time series and spatial location.

[0126] Spatiotemporal coupling can be a correlation between the extreme points of the temperature gradient and the inflection point of the rate of change of osmotic pressure in the time and space dimensions.

[0127] The freeze-thaw transition critical state can be the transitional stage in which the dam foundation curtain material changes from a frozen state to a thawed state (or vice versa).

[0128] The freeze-thaw start time can be the timestamp when a region enters the critical state of freeze-thaw transition.

[0129] The curtain depth can be the vertical distance between the top and bottom of the curtain at the dam foundation.

[0130] Specifically, since the dam foundation curtain in high-altitude and cold regions is directly affected by freeze-thaw cycles, before analyzing the leakage performance of the dam foundation curtain during freeze-thaw cycles, it is necessary to first determine whether the current dam foundation curtain is within a freeze-thaw transition cycle. Because the dam foundation curtain is large in scale and freeze-thaw transition is a long-term process, the time when different areas of the dam foundation curtain enter the freeze-thaw transition cycle varies, requiring regional discussion to ensure the accuracy of the leakage analysis process. Based on a synchronized subset of temperature data, the temperature performance of different regions at different times is analyzed, the temperature change gradient between regions corresponding to different temperature sensors is quantified, a curtain temperature gradient distribution map is constructed, and a synchronized subset of seepage pressure data is extracted simultaneously to analyze the temperature performance of different regions at different times. The permeability value changes, and the permeability change rate between the corresponding areas of different permeability sensors is quantified to generate a permeability change rate distribution map. Based on the curtain temperature gradient distribution map and the permeability change rate distribution map, the extreme points of the temperature gradient (the extreme point of the temperature gradient is determined by fitting the temperature gradient extreme value threshold through experimental data, such as 3℃ / m) and the inflection point of the permeability change rate (the inflection point of the permeability change rate is determined by fitting the permeability change rate inflection point threshold through data, such as 10Pa / min) are extracted respectively. The extreme points of the temperature gradient and the permeability inflection point are aligned with the time axis. When the two overlap in space and the time difference is less than the corresponding difference benchmark (such as 30s), it is determined that the area has entered the critical state of freeze-thaw transition. The timestamps corresponding to the data points that meet the above conditions are extracted as the freeze-thaw start time node of the grid cell where the sensor is located.

[0131] Based on this, since the dam foundation curtain is vertically buried underground, there is a significant height difference between its top and bottom. The freeze-thaw cycle is affected by ground temperature conduction, and the starting time of the freeze-thaw cycle at the top of the dam foundation curtain is significantly earlier than that at the bottom. Furthermore, the starting time of the freeze-thaw cycle shows a downward increasing trend based on the depth of the dam foundation curtain, resulting in different freeze-thaw cycles exhibiting misalignment characteristics on the same time axis. Therefore, the analysis of the freeze-thaw start time of the sensor coverage blind area can be based on the sensor data and the depth of the dam foundation curtain where the sensor coverage blind area is located. This allows the analysis of the freeze-thaw cycle to cover the entire area of ​​the dam foundation curtain, thereby improving the comprehensiveness and accuracy of subsequent leakage analysis.

[0132] This scheme analyzes synchronized subsets of temperature and seepage pressure data to derive curtain temperature gradient distribution maps and seepage pressure change rate distribution maps, respectively, reflecting the characteristics of temperature and seepage pressure changes. Based on the spatiotemporal coupling relationship between the extreme points of the temperature gradient and the inflection points of the seepage pressure change rate, the time point when the sensor-corresponding area enters the critical state of freeze-thaw transition is captured as the freeze-thaw initiation time node. On this basis, combined with the depth of the dam foundation curtain where the sensor coverage blind zone is located, the freeze-thaw state of the sensor deployment blind zone is deduced, thereby obtaining the freeze-thaw initiation time node of the sensor deployment blind zone. This allows the analysis of the freeze-thaw time cycle to cover the entire area of ​​the dam foundation curtain, thus improving the comprehensiveness and accuracy of subsequent leakage analysis.

[0133] In some embodiments, based on the spatial grid model of the dam foundation curtain and the curtain depth, the dam foundation curtain is divided into several temperature conduction layers along the vertical depth direction according to a preset unit depth interval; based on the freeze-thaw initiation time nodes corresponding to several sensors in adjacent temperature conduction layers, the advancement rate of the freeze-thaw transition critical state between adjacent temperature conduction layers is determined, and a freeze-thaw state transmission chain between layers is constructed; based on the freeze-thaw state transmission chain between layers, the freeze-thaw initiation time node corresponding to the freeze-thaw initiation time nodes corresponding to several sensors in adjacent temperature conduction layers is determined.

[0134] The preset unit depth range can be the unit depth referenced in the process of dividing the temperature conduction layers. The preset unit depth range is set according to the sensor layout spacing and the overall layout depth of the dam foundation curtain.

[0135] Temperature conduction levels can be a hierarchical structure in which the dam foundation curtain is divided along the vertical depth direction according to a preset unit depth (such as 2 meters).

[0136] Spatiotemporal correlation can be the degree of matching between the extreme points of temperature gradients and the inflection points of osmotic pressure at adjacent sensor nodes in terms of time series and spatial distribution.

[0137] The propulsion rate can be the average speed at which the critical state of freeze-thaw transition migrates between adjacent temperature conduction levels.

[0138] The inter-level freeze-thaw state transfer chain can be a logical model describing the migration path of freeze-thaw fronts between temperature conduction levels.

[0139] Specifically, based on the preset unit depth range, the boundaries of each level are marked in the spatial grid model of the dam foundation curtain. The dam foundation curtain is divided into several temperature conduction levels along the vertical depth direction. The freeze-thaw initiation time nodes of all sensors in adjacent levels (such as level 1 and level 2) are extracted. The time difference between levels for each sensor pair is calculated. Based on the ratio of the interlayer spacing to the interlayer time difference, and the local advance rate corresponding to each sensor, the median of all current local advance rates is taken as the current interlayer advance rate. Based on the advance rates between different levels, the freeze-thaw state transmission chain between levels is constructed. On this basis, based on the relative distance between different sensor blind zones and their neighboring sensors, and combined with the freeze-thaw initiation time nodes of neighboring sensors, the freeze-thaw initiation time node corresponding to the sensor blind zone is derived.

[0140] This scheme divides the entire area corresponding to the dam foundation curtain into several temperature conduction levels based on a preset unit depth range, further refining the granularity of the analysis. Based on the freeze-thaw initiation time nodes corresponding to several sensors within adjacent temperature conduction levels, the propagation rate of the freeze-thaw transition critical state between temperature conduction levels is analyzed, and a freeze-thaw state transmission chain between levels is constructed. In this way, the freeze-thaw initiation time node corresponding to the sensor deployment blind zone is derived, so that the subsequent leakage analysis process can cover the entire area of ​​the dam foundation curtain while based on a precise time axis, preventing misjudgment of leakage caused by monitoring blind zones.

[0141] In some embodiments, based on the freeze-thaw transition time axis, a synchronized subset of temperature data, a synchronized subset of seepage pressure data, and a synchronized subset of deformation data that are temporally correlated with the freeze-thaw start time node are extracted from the unified spatiotemporal detection dataset to construct a synchronized dataset for the sensor coverage area. Based on the advancement rate in the inter-level freeze-thaw state transmission chain and combined with the relative spatial distance between the sensor deployment blind zone and adjacent sensors, a spatiotemporal interpolation function model is established. Based on the spatiotemporal interpolation function model, according to the synchronized dataset for the sensor coverage area, spatiotemporal continuity interpolation is performed on the temperature data, seepage pressure data, and deformation data of the sensor deployment blind zone on the freeze-thaw transition time axis to generate a blind zone interpolation data sequence. The blind zone interpolation data sequence is spatiotemporally aligned with the data sequence in the synchronized dataset for the sensor coverage area, and a target periodic detection dataset covering the entire dam foundation curtain area is constructed based on the synchronized dataset for the sensor coverage area and the spatiotemporally aligned blind zone interpolation data sequence.

[0142] The sensor coverage area synchronization dataset can be a data set composed of a subset of temperature, osmotic pressure, and deformation data collected by sensors that are directly associated with the freeze-thaw start time node, extracted from a unified spatiotemporal detection dataset.

[0143] Relative spatial distance can be the relative distance between the sensor deployment blind zone and adjacent sensors under the temperature conduction level division.

[0144] The spatiotemporal interpolation function model can be a data interpolation model based on the relationship between the propagation rate of the freeze-thaw state transmission chain and the spatial distance, used to derive the changes in physical quantities in the sensor blind zone.

[0145] Blind zone interpolation data sequences can be temperature, seepage pressure, and deformation data sequences generated through a spatiotemporal interpolation function model within the sensor deployment blind zone, filling the spatiotemporal data gaps in unmonitored areas.

[0146] Spatiotemporal continuity interpolation is an interpolation method that maintains data continuity in both time and space dimensions, avoiding abrupt distortions caused by traditional interpolation.

[0147] Specifically, after deriving the freeze-thaw initiation time node corresponding to the sensor blind zone, since the temperature, pressure, and deformation data corresponding to the sensor blind zone are missing, further derivation of the corresponding data parameters within the sensor blind zone is required to ensure the continuity and accuracy of leakage analysis and prevent high leakage analysis errors caused by data fragmentation. This involves analyzing synchronized subsets of temperature, pressure, and deformation data that have a temporal correlation with the freeze-thaw initiation time node in the unified spatiotemporal detection dataset. Since the temperature, pressure, and deformation data corresponding to the current sensor blind zone are missing, these synchronized subsets represent the data directly collected by all current sensors. This is used to construct a synchronized dataset for the sensor coverage area. Simultaneously, based on the advancement rate and relative spatial distance, the impact corresponding to the current data blind zone is quantified. The impact coefficient characterizes the adjustment required for data within the current sensor coverage blind zone based on data collected from adjacent sensors. Specifically, the advancement rate is positively correlated with the impact coefficient, while the relative spatial distance is negatively correlated. The corresponding correlation coefficients are fitted using experimental data. Based on the impact coefficients within different sensor coverage blind zones, a spatiotemporal difference function model is constructed. On this basis, according to the corresponding impact coefficients mapped within the spatiotemporal difference function model, the corresponding data in the synchronous dataset of the sensor coverage area are adjusted proportionally to generate a blind zone interpolation data sequence. The timestamps of the blind zone interpolation data sequence are uniformly calibrated to the freeze-thaw transition time axis, and the spatial coordinates are mapped to the corresponding temperature conduction level to achieve spatiotemporal alignment. The synchronous dataset of the sensor coverage area and the spatiotemporally aligned blind zone interpolation data sequence are integrated into a single dataset, namely the target periodic detection dataset.

[0148] This scheme, based on the synchronous dataset of the sensor coverage area, and according to the advancement rate in the freeze-thaw state transmission chain between layers, combined with the relative spatial distance between the sensor blind zone and adjacent sensors, utilizes a blind zone interpolation strategy driven by the physical conduction mechanism to generate a blind zone interpolation data sequence. Based on the synchronous dataset of the sensor coverage area and the spatiotemporally aligned blind zone interpolation data sequence, a target periodic detection dataset is jointly constructed, significantly improving the spatial coverage of the target periodic detection dataset, effectively eliminating monitoring blind zones, and improving the accuracy and comprehensiveness of leakage analysis results.

[0149] In some embodiments, the spatial grid model of the dam foundation curtain is analyzed to identify and mark the cylindrical structural joint areas; based on material leakage test data, basic leakage influence weights are assigned to the temperature data, seepage pressure data, and deformation data in the target period detection dataset; based on joint leakage test data, the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the cylindrical structural joint areas in the target period detection dataset are corrected for key areas to determine the leakage influence weights of the joint areas; based on the target period detection dataset, according to the basic leakage influence weights and the joint area leakage influence weights, the temperature data, seepage pressure data, and deformation data corresponding to different areas in the target period detection dataset are weighted and fused to determine the real-time leakage risk corresponding to different areas; based on the real-time leakage risk and corresponding deformation data of different areas, several high-risk leakage areas are identified, and a leakage risk distribution heatmap is constructed based on these high-risk leakage areas; the leakage risk distribution heatmap is then visualized as leakage trend information of the dam foundation curtain.

[0150] The cylindrical structural joint area can be the joint area formed by splicing precast concrete cylinders in the dam foundation curtain, and its leakage risk is significantly higher than that of the homogeneous structure area.

[0151] The weight of the basic leakage impact can be the proportion of the basic impact of temperature, seepage pressure, and deformation data on leakage risk assessment based on the inherent properties of the dam concrete material.

[0152] The joint leakage test data can be obtained from the accelerated aging leakage test of concrete column joints.

[0153] The key area impact correction process can be a process of adjusting the basic impact weights of each impact parameter in the joint area based on the characteristics of leakage analysis in the cylindrical structure joint area that is significantly higher than that in the mean structure area.

[0154] The weighting of leakage impact in the joint area can be a weighting of leakage risk assessment adjusted for high risk in the joint area.

[0155] Weighted fusion analysis can be a process of analyzing the comprehensive leakage risk of a region by superimposing temperature, seepage pressure, and deformation data according to their weights.

[0156] Real-time leakage risk can be a quantitative indicator that represents the probability of leakage occurring in a certain area of ​​the dam foundation curtain at the current moment.

[0157] High-risk leakage areas can be the set of dam curtain areas where high leakage analysis is performed.

[0158] A leakage risk distribution heatmap can be a visual representation of the spatial distribution of real-time leakage risk, marked with different colors to map gradients.

[0159] Specifically, existing technologies for leakage analysis of dam foundation curtain typically treat the dam foundation curtain as a homogeneous whole, failing to consider the amplified impact of influencing factors in the joint areas, where the multi-cylindrical splicing structure of the dam foundation curtain itself presents a significantly higher risk of leakage. This is addressed by loading a spatial mesh model of the dam foundation curtain, extracting the spatial coordinates and diameter parameters of all cylindrical components, generating joint centerlines along the cylindrical axis, and extending them to both sides to form joint influence zones (the joint influence zones exhibit hyperbolic paraboloid characteristics). These joint influence zones are then highlighted in the 3D model to identify the joint areas of the cylindrical structures. Furthermore, material leakage experimental data is retrieved, and based on the material characteristics of the dam foundation curtain, basic leakage influence weights are assigned to the temperature, seepage pressure, and deformation data within the target periodic detection dataset. Further, based on the basic leakage influence weights and combined with joint leakage experimental data, the data influence corresponding to the joint area of ​​the cylindrical structure is corrected to obtain the leakage influence weight of the joint area. The temperature, seepage pressure, and deformation data corresponding to each area are multiplied by the corresponding weight value and risk conversion coefficient, and the sum is used to obtain the real-time leakage risk corresponding to that area. All real-time leakage risk values ​​are mapped to the HSV color space (high risk: hue 0° red, low risk: hue 240° blue), and the corresponding color gradient is rendered on the surface of the three-dimensional dam foundation curtain model to generate a leakage risk distribution heat map. At the same time, a time axis control can be integrated into the leakage risk distribution heat map to support the playback of the risk evolution process according to the freeze-thaw stage. The above leakage risk distribution heat map can be visualized and output through human-computer interaction devices, such as high-definition displays.

[0160] This scheme analyzes the spatial grid model of the dam foundation curtain wall, identifies the joint areas of the cylindrical structure, and corrects the impact of the data corresponding to the joint areas of the cylindrical structure based on the foundation leakage influence weight, determining the leakage influence weight of the joint areas. On this basis, according to the foundation leakage influence weight and the joint area leakage influence weight, the temperature data, seepage pressure data, and deformation data under different areas are weighted and fused for analysis, resulting in the real-time leakage risk corresponding to different areas. A corresponding leakage risk distribution heat map is generated as a visualization output of the leakage trend information of the dam foundation curtain wall, making the assessment of leakage risk in the cylindrical structure joint area, which has a high leakage risk, more accurate.

[0161] In some embodiments, based on joint leakage test data, joint width variation data, joint filler thermal conductivity, and joint interface shear strength degradation coefficient are extracted. Joint width variation data is positively correlated with the weight of leakage impact in the joint area; the joint filler thermal conductivity is negatively correlated with the weight of leakage impact in the joint area; and the joint interface shear strength degradation coefficient is positively correlated with the weight of leakage impact in the joint area. Based on the weight of basic leakage impact, a joint area impact correction function is constructed according to the joint width variation data, joint filler thermal conductivity, and joint interface shear strength degradation coefficient. Based on the joint area impact correction function, the weight of basic leakage impact is adjusted for key areas to quantify the weight of leakage impact in the joint area.

[0162] Joint width variation data can be monitoring data that reflects the dynamic changes in joint width under freeze-thaw cycles.

[0163] The thermal conductivity of joint filler material can be a physical parameter characterizing the thermal conductivity of joint filler material.

[0164] The shear strength degradation coefficient of the joint interface can be used as an indicator to quantify the degree of reduction in shear performance of the concrete joint interface due to freeze-thaw damage.

[0165] The joint area influence correction function can be a mathematical function that associates joint characteristic parameters with leakage risk weights.

[0166] Specifically, based on the positive correlation conversion ratio (positive number) between joint width variation data and joint area leakage influence weight, the negative correlation conversion ratio (negative number) between joint filling material thermal conductivity and joint area leakage influence weight, and the exponential correlation conversion ratio (positive number) between joint interface shear strength degradation coefficient and joint area leakage influence weight, a comprehensive influence factor is obtained by summing the above three conversion ratios. Using the basic leakage influence weight as the benchmark value, a joint area influence correction function is constructed based on the product between the basic leakage influence weight and the comprehensive influence factor. This achieves key area influence correction processing for the basic leakage influence weight, thus deriving the joint area leakage influence weight.

[0167] This scheme establishes the correlation between the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface in the joint leakage test data and the leakage influence weight of the joint area. A joint area influence correction function is constructed, and based on the joint area influence correction function, the basic leakage influence weight is modified for key areas. The leakage influence weight of the joint area is quantified, and a multi-parameter coupling correction mechanism is used to reduce the assessment error of the leakage risk of the joint area.

[0168] In some embodiments, the real-time leakage risk corresponding to different regions is compared with the risk precursor benchmark value. Based on the comparison results, regions whose real-time leakage risk exceeds the risk precursor benchmark value are marked as key analysis regions. Based on the target period detection dataset, and taking the key analysis regions as the benchmark, deformation data corresponding to each key analysis region and its adjacent temperature conduction level are extracted from the target period detection dataset to construct a key region deformation dataset. Based on the key region deformation dataset, the deformation development direction of the current key analysis region is analyzed to determine the tracking target region. The current tracking target region and the key analysis region are designated as high-risk leakage regions.

[0169] Risk precursor benchmarks can be pre-set leakage risk thresholds used to distinguish between normal risk states and abnormal risk states that require special attention. Risk precursor benchmarks can be obtained by fitting experimental data.

[0170] The key analysis area can be the dam foundation curtain sub-area where the real-time leakage risk exceeds the risk precursor benchmark value.

[0171] A key area deformation dataset can be a collection containing deformation data of the key analysis area and its adjacent temperature conduction layers.

[0172] The target region for tracking can be the spatial region corresponding to the current maximum deformation value in the deformation dataset.

[0173] Specifically, in assessing high-risk leakage areas, since leakage can dynamically spread and develop, the judgment of high-risk leakage areas cannot solely rely on the level of real-time leakage risk. It is necessary to include adjacent areas along the deformation development direction of the corresponding high-real-time leakage risk area within the high-risk leakage area. By analyzing areas with potential leakage spread risk, the lag in risk identification is reduced. Based on the numerical comparison results of different real-time leakage risks with risk precursor benchmark values, key analysis areas are selected. Then, based on the deformation dataset of key areas, the stress concentration direction of the key analysis areas (the relative direction between the key analysis area and areas in adjacent layers where stress increases exceed the average value) is analyzed. This stress concentration direction is the deformation development direction, and areas along this direction are identified as tracking target areas. Finally, by integrating the current tracking target areas and key analysis areas, high-risk leakage areas are derived.

[0174] This solution integrates key analysis areas and tracking target areas into high-risk leakage areas through a dual assessment mechanism of real-time leakage risk and deformation development direction during the evaluation of high-risk leakage areas. By analyzing areas with potential leakage spread risks, the lag in risk identification is reduced.

[0175] Figure 3 A schematic diagram of a dam leakage diagnosis system based on AI technology provided in one embodiment of this application is shown below. Figure 3 As shown, the dam leakage diagnosis system 300 based on AI technology in this embodiment includes: a data alignment module 301, a data tracking module 302, and a risk analysis module 303.

[0176] Data alignment module 301 is used to acquire sensor array dataset, analyze sensor array dataset, and determine unified spatiotemporal detection dataset;

[0177] The data tracking module 302 is used to analyze the unified spatiotemporal detection dataset, determine the freeze-thaw transition time axis, and track and filter the corresponding detection data in the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target period detection dataset.

[0178] The risk analysis module 303 is used to simulate the evolution of the leakage development trend of the dam foundation curtain based on the target periodic detection dataset, and to determine and output the leakage trend information of the dam foundation curtain.

[0179] Optionally, in the data alignment module 301, the sensor array dataset includes a subset of temperature data, a subset of osmotic pressure data, and a subset of deformation data.

[0180] The temperature data subset was acquired by distributed fiber optic temperature sensors arranged in a three-dimensional mesh topology on the dam foundation curtain.

[0181] The subset of seepage pressure data is acquired by distributed fiber optic micro-pressure sensors arranged in the form of the three-dimensional grid topology on the dam foundation curtain.

[0182] The deformation data subset is acquired by distributed fiber optic stress sensors deployed on the dam foundation curtain in the form of the three-dimensional mesh topology.

[0183] The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are all deployed on the side of the dam foundation curtain facing the upstream of the dam body.

[0184] The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are connected in series to the same fiber optic path using optical wavelength division multiplexing technology.

[0185] Optionally, the data alignment module 301 is specifically used for:

[0186] Based on the three-dimensional grid topology of the distributed optical fiber sensors on the dam foundation curtain, a spatial grid model of the dam foundation curtain is constructed.

[0187] Based on the dam foundation curtain space grid model, the topology of the three-dimensional grid is analyzed to determine the relative position coordinates of each sensor and construct a unified spatial position coordinate set.

[0188] Analyze the unified spatial coordinate set and evaluate the signal processing delay of each sensor based on the relative distance between each sensor and its corresponding signal demodulator;

[0189] Based on the signal processing delay, timestamp correction processing is performed on the data collected by each sensor in the sensor array dataset to determine a unified timestamp dataset.

[0190] Based on the unified spatial coordinate set and the unified timestamp dataset, each data point in the sensor array dataset is marked with a spatial location and a timestamp, respectively. After the spatiotemporal unification is completed, the temperature data synchronization subset, the osmotic pressure data synchronization subset, and the deformation data synchronization subset are determined, and the unified spatiotemporal detection dataset is constructed.

[0191] Optionally, when analyzing the unified spatiotemporal detection dataset and determining the freeze-thaw transition time axis, the data tracking module 302 is specifically used for:

[0192] Based on the aforementioned spatial grid model of the dam foundation curtain, and according to the synchronized subset of temperature data in the aforementioned unified spatiotemporal detection dataset, a curtain temperature gradient distribution map is constructed.

[0193] Based on the aforementioned dam foundation curtain spatial grid model, and according to the synchronized subset of seepage pressure data in the aforementioned unified spatiotemporal detection dataset, a seepage pressure change rate distribution map is constructed.

[0194] Spatiotemporal correlation analysis is performed on the curtain temperature gradient distribution map and the permeation pressure change rate distribution map. When the distribution of extreme points of temperature gradient in the curtain temperature gradient distribution map and the inflection point of permeation pressure change rate in the permeation pressure change rate distribution map form a spatiotemporal coupling relationship, it is determined that the region corresponding to the current data has entered the critical state of freeze-thaw transition, and the timestamp of the corresponding data is extracted as the freeze-thaw start time node of the region where the current sensor is located.

[0195] The curtain depth of the dam foundation curtain is determined based on the spatial grid model of the dam foundation curtain.

[0196] Based on the curtain depth, and according to the freeze-thaw start time node corresponding to the area where each sensor is located, the freeze-thaw state is deduced by analyzing the sensor deployment blind zone outside the area where the sensor is located, and the freeze-thaw start time node of the sensor deployment blind zone is determined.

[0197] The freeze-thaw transition time axis is constructed based on the freeze-thaw start time nodes corresponding to the area where the sensor is located and the blind zone where the sensor is deployed.

[0198] Optionally, when the data tracking module 302, based on the curtain depth and according to the freeze-thaw start time node corresponding to the area where each sensor is located, analyzes the sensor deployment blind zone outside the sensor's location area to deduce the freeze-thaw state and determine the freeze-thaw start time node of the sensor deployment blind zone, it is specifically used for:

[0199] Based on the spatial grid model of the dam foundation curtain and the curtain depth, according to the preset unit depth interval, the dam foundation curtain is divided into several temperature conduction levels along the vertical depth direction;

[0200] Based on the freeze-thaw initiation time nodes corresponding to several sensors located in adjacent temperature conduction layers, the propagation rate of the freeze-thaw transition critical state between adjacent temperature conduction layers is determined, and an inter-layer freeze-thaw state transmission chain is constructed.

[0201] Based on the inter-level freeze-thaw state transmission chain, the freeze-thaw start time node corresponding to each sensor deployment blind zone is determined according to the freeze-thaw start time node corresponding to several sensors located in adjacent temperature conduction levels.

[0202] Optionally, when the data tracking module 302 tracks and filters the corresponding detection data within the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target period detection dataset, it is specifically used for:

[0203] Based on the freeze-thaw transition time axis, the temperature data synchronization subset, the osmotic pressure data synchronization subset, and the deformation data synchronization subset that are temporally correlated with the freeze-thaw start time node are extracted from the unified spatiotemporal detection dataset to construct a sensor coverage area synchronization dataset;

[0204] Based on the propulsion rate in the inter-level freeze-thaw state transmission chain, and combined with the relative spatial distance between the sensor deployment blind zone and adjacent sensors, a spatiotemporal interpolation function model is established.

[0205] Based on the spatiotemporal interpolation function model, and according to the synchronous dataset of the sensor coverage area, the temperature data, osmotic pressure data, and deformation data of the sensor deployment blind zone on the freeze-thaw transition time axis are spatiotemporally continuous interpolated to generate a blind zone interpolation data sequence.

[0206] The blind zone interpolation data sequence is spatiotemporally aligned with the data sequence in the sensor coverage area synchronization dataset, and the target periodic detection dataset covering the entire dam foundation curtain area is constructed based on the sensor coverage area synchronization dataset and the spatiotemporally aligned blind zone interpolation data sequence.

[0207] Optionally, the risk analysis module 303 is specifically used for:

[0208] Analyze the spatial mesh model of the dam foundation curtain to identify and mark the joint areas of the cylindrical structure;

[0209] Based on the material leakage test data, the temperature data, osmotic pressure data, and deformation data in the target period detection dataset are assigned basic leakage influence weights respectively;

[0210] Based on the joint leakage test data, the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the joint area of ​​the cylindrical structure in the target period detection dataset are subjected to key area influence correction processing to determine the leakage influence weights of the joint area.

[0211] Based on the target period detection dataset, according to the basic leakage influence weight and the joint area leakage influence weight, the temperature data, seepage pressure data and deformation data corresponding to different areas in the target period detection dataset are weighted and fused to determine the real-time leakage risk corresponding to different areas.

[0212] Based on the real-time leakage risk and corresponding deformation data of different regions, several high-risk leakage areas are identified, and a leakage risk distribution heat map is constructed based on the several high-risk leakage areas.

[0213] The heat map of leakage risk distribution is used as a visual output of the leakage trend information of the dam foundation curtain.

[0214] Optionally, the risk analysis module 303, based on the joint leakage test data, performs key area influence correction processing on the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the cylindrical structure joint area within the target periodic detection dataset. Specifically, when determining the joint area leakage influence weights, it is used for:

[0215] Based on the joint leakage test data, the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface were extracted.

[0216] The data on the change in joint width is positively correlated with the weight of the leakage impact in the joint area;

[0217] The thermal conductivity of the joint filling material is negatively correlated with the weight of leakage in the joint area.

[0218] The shear strength degradation coefficient of the joint interface is positively correlated with the weight of leakage influence in the joint area;

[0219] Based on the aforementioned basic leakage influence weights, and according to the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface, a joint area influence correction function is constructed.

[0220] Based on the joint area influence correction function, the influence weight of the basic leakage is adjusted by focusing on key areas, and the influence weight of the joint area leakage is quantified.

[0221] Optionally, when the risk analysis module 303 determines several high-risk leakage areas based on the real-time leakage risk and corresponding deformation data of different areas, it is specifically used for:

[0222] The real-time leakage risk corresponding to different regions is compared with the risk precursor benchmark value. Based on the comparison results, regions where the real-time leakage risk exceeds the risk precursor benchmark value are marked as key analysis regions.

[0223] Based on the target period detection dataset, and taking the key analysis area as a benchmark, the deformation data corresponding to each key analysis area and its adjacent temperature conduction level are extracted from the target period detection dataset to construct a key area deformation dataset.

[0224] Based on the deformation dataset of the key areas, analyze the deformation development direction of the current key analysis areas and determine the target tracking areas;

[0225] The current tracking target area and the key analysis area are designated as the high-risk leakage area.

[0226] The system in this embodiment can be used to execute the methods of any of the above embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

Claims

1. A method for diagnosing dam leakage based on AI technology, characterized in that, include: Acquire the sensor array dataset, analyze the sensor array dataset, and determine a unified spatiotemporal detection dataset; Analyze the unified spatiotemporal detection dataset, determine the freeze-thaw transition time axis, and based on the freeze-thaw transition time axis, track and filter the corresponding detection data within the unified spatiotemporal detection dataset to determine the target period detection dataset; Based on the target periodic detection dataset, the evolutionary trend of seepage development of the dam foundation curtain is simulated, and the seepage trend information of the dam foundation curtain is determined and output. The analysis of the sensor array dataset to determine a unified spatiotemporal detection dataset includes: Based on the three-dimensional grid topology of the distributed optical fiber sensors on the dam foundation curtain, a spatial grid model of the dam foundation curtain is constructed. Based on the dam foundation curtain space grid model, the topology of the three-dimensional grid is analyzed to determine the relative position coordinates of each sensor and construct a unified spatial position coordinate set. Analyze the unified spatial coordinate set and evaluate the signal processing delay of each sensor based on the relative distance between each sensor and its corresponding signal demodulator; Based on the signal processing delay, timestamp correction processing is performed on the data collected by each sensor in the sensor array dataset to determine a unified timestamp dataset. Based on the unified spatial location coordinate set and the unified timestamp dataset, each data in the sensor array dataset is marked with a spatial location and a timestamp, respectively, and the temperature data synchronization subset, osmotic pressure data synchronization subset, and deformation data synchronization subset are determined after the spatiotemporal unification is completed, and the unified spatiotemporal detection dataset is constructed. The analysis of the unified spatiotemporal detection dataset to determine the freeze-thaw transition time axis includes: Based on the aforementioned spatial grid model of the dam foundation curtain, and according to the synchronized subset of temperature data in the aforementioned unified spatiotemporal detection dataset, a curtain temperature gradient distribution map is constructed. Based on the aforementioned dam foundation curtain spatial grid model, and according to the synchronized subset of seepage pressure data in the aforementioned unified spatiotemporal detection dataset, a seepage pressure change rate distribution map is constructed. Spatiotemporal correlation analysis is performed on the curtain temperature gradient distribution map and the permeation pressure change rate distribution map. When the distribution of extreme points of temperature gradient in the curtain temperature gradient distribution map and the inflection point of permeation pressure change rate in the permeation pressure change rate distribution map form a spatiotemporal coupling relationship, it is determined that the region corresponding to the current data has entered the critical state of freeze-thaw transition, and the timestamp of the corresponding data is extracted as the freeze-thaw start time node of the region where the current sensor is located. The curtain depth of the dam foundation curtain is determined based on the spatial grid model of the dam foundation curtain. Based on the curtain depth, and according to the freeze-thaw start time node corresponding to the area where each sensor is located, the freeze-thaw state is deduced by analyzing the sensor deployment blind zone outside the area where the sensor is located, and the freeze-thaw start time node of the sensor deployment blind zone is determined. The freeze-thaw transition time axis is constructed based on the freeze-thaw start time nodes corresponding to the area where the sensor is located and the blind zone where the sensor is deployed.

2. The method according to claim 1, characterized in that, The sensor array dataset includes temperature data subsets, osmotic pressure data subsets, and deformation data subsets; The temperature data subset was acquired by distributed fiber optic temperature sensors arranged in a three-dimensional mesh topology on the dam foundation curtain. The subset of seepage pressure data is acquired by distributed fiber optic micro-pressure sensors arranged in the form of the three-dimensional grid topology on the dam foundation curtain. The deformation data subset is acquired by distributed fiber optic stress sensors deployed on the dam foundation curtain in the form of the three-dimensional mesh topology. The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are all deployed on the side of the dam foundation curtain facing the upstream of the dam body. The distributed fiber optic temperature sensor, the distributed fiber optic micro-pressure sensor, and the distributed fiber optic stress sensor are connected in series to the same fiber optic path using optical wavelength division multiplexing technology.

3. The method according to claim 2, characterized in that, Based on the curtain depth, and according to the freeze-thaw initiation time node corresponding to the area where each sensor is located, the freeze-thaw state is deduced by analyzing the sensor deployment blind zone outside the sensor's location area, and the freeze-thaw initiation time node of the sensor deployment blind zone is determined, including: Based on the spatial grid model of the dam foundation curtain and the curtain depth, according to the preset unit depth interval, the dam foundation curtain is divided into several temperature conduction levels along the vertical depth direction; Based on the freeze-thaw initiation time nodes corresponding to several sensors located in adjacent temperature conduction layers, the propagation rate of the freeze-thaw transition critical state between adjacent temperature conduction layers is determined, and an inter-layer freeze-thaw state transmission chain is constructed. Based on the inter-level freeze-thaw state transmission chain, the freeze-thaw start time node corresponding to each sensor deployment blind zone is determined according to the freeze-thaw start time node corresponding to several sensors located in adjacent temperature conduction levels.

4. The method according to claim 3, characterized in that, The step of tracking and filtering the corresponding detection data within the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target periodic detection dataset includes: Based on the freeze-thaw transition time axis, the temperature data synchronization subset, the osmotic pressure data synchronization subset, and the deformation data synchronization subset that are temporally correlated with the freeze-thaw start time node are extracted from the unified spatiotemporal detection dataset to construct a sensor coverage area synchronization dataset; Based on the propulsion rate in the inter-level freeze-thaw state transmission chain, and combined with the relative spatial distance between the sensor deployment blind zone and adjacent sensors, a spatiotemporal interpolation function model is established. Based on the spatiotemporal interpolation function model, and according to the synchronous dataset of the sensor coverage area, the temperature data, osmotic pressure data, and deformation data of the sensor blind zone on the freeze-thaw transition time axis are spatiotemporally continuous interpolated to generate a blind zone interpolation data sequence. The blind zone interpolation data sequence is spatiotemporally aligned with the data sequence in the sensor coverage area synchronization dataset, and the target periodic detection dataset covering the entire dam foundation curtain area is constructed based on the sensor coverage area synchronization dataset and the spatiotemporally aligned blind zone interpolation data sequence.

5. The method according to claim 4, characterized in that, The process of simulating the evolution of seepage development in the dam foundation curtain based on the target periodic detection dataset, determining and outputting dam foundation curtain seepage trend information, includes: Analyze the spatial mesh model of the dam foundation curtain to identify and mark the joint areas of the cylindrical structure; Based on the material leakage test data, the temperature data, osmotic pressure data, and deformation data in the target period detection dataset are assigned basic leakage influence weights respectively; Based on the joint leakage test data, the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the joint area of ​​the cylindrical structure in the target period detection dataset are subjected to key area influence correction processing to determine the leakage influence weights of the joint area. Based on the target period detection dataset, according to the basic leakage influence weight and the joint area leakage influence weight, the temperature data, seepage pressure data and deformation data corresponding to different areas in the target period detection dataset are weighted and fused to determine the real-time leakage risk corresponding to different areas. Based on the real-time leakage risk and corresponding deformation data of different regions, several high-risk leakage areas are identified, and a leakage risk distribution heat map is constructed based on the several high-risk leakage areas. The heat map of leakage risk distribution is used as a visual output of the leakage trend information of the dam foundation curtain.

6. The method according to claim 5, characterized in that, The step involves performing a key area influence correction process on the basic leakage influence weights of the temperature data, seepage pressure data, and deformation data corresponding to the joint area of ​​the cylindrical structure within the target periodic detection dataset, based on the joint leakage test data, to determine the leakage influence weights of the joint area, including: Based on the joint leakage test data, the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface were extracted. The data on the change in joint width is positively correlated with the weight of the leakage impact in the joint area; The thermal conductivity of the joint filling material is negatively correlated with the weight of leakage in the joint area. The shear strength degradation coefficient of the joint interface is positively correlated with the weight of leakage influence in the joint area; Based on the aforementioned basic leakage influence weights, and according to the joint width variation data, the thermal conductivity of the joint filling material, and the shear strength degradation coefficient of the joint interface, a joint area influence correction function is constructed. Based on the joint area influence correction function, the influence weight of the basic leakage is adjusted by focusing on key areas, and the influence weight of the joint area leakage is quantified.

7. The method according to claim 6, characterized in that, Based on the real-time leakage risk and corresponding deformation data of different areas, several high-risk leakage areas are identified, including: The real-time leakage risk corresponding to different regions is compared with the risk precursor benchmark value. Based on the comparison results, regions where the real-time leakage risk exceeds the risk precursor benchmark value are marked as key analysis regions. Based on the target period detection dataset, and taking the key analysis area as a benchmark, the deformation data corresponding to each key analysis area and its adjacent temperature conduction level are extracted from the target period detection dataset to construct a key area deformation dataset. Based on the deformation dataset of the key areas, analyze the deformation development direction of the current key analysis areas and determine the target tracking areas; The current tracking target area and the key analysis area are designated as the high-risk leakage area.

8. A dam leakage diagnosis system based on AI technology, characterized in that, Applied to the method as described in any one of claims 1-7, comprising: The data alignment module is used to acquire the sensor array dataset, analyze the sensor array dataset, and determine a unified spatiotemporal detection dataset. The data tracking module is used to analyze the unified spatiotemporal detection dataset, determine the freeze-thaw transition time axis, and track and filter the corresponding detection data in the unified spatiotemporal detection dataset according to the freeze-thaw transition time axis to determine the target period detection dataset. The risk analysis module is used to simulate the evolution of the leakage development trend of the dam foundation curtain based on the target periodic detection dataset, and to determine and output the leakage trend information of the dam foundation curtain.

Images