Structural safety assessment method and system for hydraulic structures based on multi-source information fusion

Abstract

The present invention discloses a method and system for assessing the structural safety of hydraulic structures based on multi-source information fusion. The method comprises the following steps: first, obtaining a monitoring data set of the hydraulic structure to be assessed, dividing it into a structural sub-region monitoring data set, and extracting the safety impact features of the sub-regions; second, dividing the features into feature sets of a first impact dimension and a second impact dimension; then, constructing a first impact association network based on the degree of association, and a second impact association network based on the physical spatial location; coupling the features with the associated entities to obtain comprehensive assessment features of the two dimensions; finally, fusing the two-dimensional features of the same sub-region, determining the safety level, and outputting the assessment results. With this design, through multi-dimensional feature association analysis and coupled fusion, a comprehensive and accurate assessment of the structural safety of hydraulic structures is achieved, providing a scientific basis for engineering safety management.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a method and system for evaluating the structural safety of hydraulic structures based on multi-source information fusion. Background Art

[0002] As the core facilities of water conservancy projects, the structural safety of hydraulic structures is directly related to public safety and regional economic development. Therefore, it is crucial to accurately assess their operating status. In the existing technology, traditional safety assessment methods are mostly based on single or limited monitoring parameters, focusing only on local characteristics or isolated factors, and ignoring the dynamic correlation between multiple physical fields and multiple spatial location factors. Although some methods attempt to integrate multiple factors, they do not systematically distinguish the characteristic action mechanisms of different influencing dimensions, resulting in the assessment results being difficult to fully reflect the true state of the structure. In addition, the existing methods do not adequately explore the spatial heterogeneity of monitoring data and the coupling relationship between features, which can easily lead to one-sided or lagging assessments. Therefore, there is an urgent need for a structural safety assessment method for hydraulic structures that can integrate multiple influencing factors and integrate multi-dimensional correlation features to improve the accuracy and reliability of the assessment. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for evaluating the structural safety of hydraulic structures based on multi-source information fusion.

[0004] In a first aspect, an embodiment of the present invention provides a method for assessing the structural safety of hydraulic structures based on multi-source information fusion, comprising:

[0005] Acquiring a hydraulic structure structure monitoring data set of the hydraulic structure to be evaluated, dividing the hydraulic structure structure monitoring data set to obtain a monitoring data group for each structural sub-area, and extracting features of each monitoring data group to obtain a safety impact feature of each sub-area;

[0006] Dividing the security impact characteristics of each sub-region respectively, obtaining first impact dimension characteristics corresponding to each sub-region security impact characteristic, forming a first impact dimension feature set, and obtaining second impact dimension characteristics corresponding to each sub-region security impact characteristic, forming a second impact dimension feature set;

[0007] Establishing a first influence association network corresponding to the first influence dimension feature set according to the degree of association between each first influence dimension feature in the first influence dimension feature set, and establishing a second influence association network corresponding to the second influence dimension feature set according to the physical spatial position of each monitoring data group;

[0008] Perform feature coupling based on the first influence dimension feature and the associated entity feature corresponding to the first influence dimension feature in the first influence association network to obtain a first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set; and perform feature coupling based on the second influence dimension feature and the associated entity feature corresponding to the second influence dimension feature in the second influence association network to obtain a second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set;

[0009] The first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-area safety impact feature are fused to obtain the target safety assessment feature corresponding to each sub-area safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-area safety impact feature to obtain the safety assessment result corresponding to the hydraulic structure structure monitoring data set.

[0010] In a second aspect, an embodiment of the present invention provides a server system, including a server, wherein the server is configured to execute the method described in the first aspect.

[0011] Compared with the existing technology, the beneficial effects provided by the present invention include: using a hydraulic structure structural safety assessment method and system based on multi-source information fusion disclosed by the present invention, by obtaining the monitoring data set of the hydraulic structure to be assessed, dividing it into structural sub-area monitoring data groups and extracting the sub-area safety impact characteristics; secondly, dividing the characteristics into feature sets of the first impact dimension and the second impact dimension; then constructing a first impact association network based on the degree of correlation, and constructing a second impact association network based on the physical space location; through the coupling of features and associated entities, respectively obtaining comprehensive assessment features of the two dimensions; finally, fusing the two-dimensional features of the same sub-area, determining the safety level, and outputting the assessment results. With this design, through multi-dimensional feature correlation analysis and coupled fusion, a comprehensive and accurate assessment of the structural safety of hydraulic structures is achieved, providing a scientific basis for engineering safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly describes the drawings required for use in the embodiments. It should be understood that the following drawings illustrate only certain embodiments of the present invention and should not be construed as limiting the scope of the present invention. Those skilled in the art can, without inventive effort, derive other relevant drawings from these drawings.

[0013] Figure 1 A schematic flow chart of the steps of a hydraulic structure structural safety assessment method based on multi-source information fusion provided by an embodiment of the present invention;

[0014] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present invention. It should be understood that the described embodiments are only a portion of the embodiments of the present invention, not all of them. Generally, the components of the embodiments of the present invention described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations.

[0016] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0017] In order to solve the technical problems in the above background technology, Figure 1 This is a flow chart of a hydraulic structure structural safety assessment method based on multi-source information fusion provided in an embodiment of the present disclosure. The hydraulic structure structural safety assessment method based on multi-source information fusion is introduced in detail below.

[0018] Step S201: obtaining a hydraulic structure monitoring data set of the hydraulic structure to be evaluated, dividing the hydraulic structure monitoring data set to obtain a monitoring data group for each structural sub-area, and extracting features of each monitoring data group to obtain a safety impact feature of each sub-area;

[0019] Step S202: Divide the security impact characteristics of each sub-region respectively, obtain the first impact dimension characteristics corresponding to each sub-region security impact characteristic, and form a first impact dimension feature set; and obtain the second impact dimension characteristics corresponding to each sub-region security impact characteristic, and form a second impact dimension feature set;

[0020] Step S203: establishing a first influence association network corresponding to the first influence dimension feature set according to the degree of association between each first influence dimension feature in the first influence dimension feature set, and establishing a second influence association network corresponding to the second influence dimension feature set according to the physical spatial position of each monitoring data group;

[0021] Step S204: Feature coupling is performed based on the first influence dimension feature and the associated entity feature corresponding to the first influence dimension feature in the first influence association network to obtain a first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set, and feature coupling is performed based on the second influence dimension feature and the associated entity feature corresponding to the second influence dimension feature in the second influence association network to obtain a second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set;

[0022] Step S205: The first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-region safety impact feature are integrated to obtain the target safety assessment feature corresponding to each sub-region safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-region safety impact feature to obtain the safety assessment result corresponding to the hydraulic structure structure monitoring data set.

[0023] In an embodiment of the present invention, the server first exchanges data with the concrete gravity dam's automated monitoring system to obtain a data set of hydraulic structure monitoring data covering different structural sections of the dam (e.g., dam sections 1 through 10). This data includes dam stress time history data collected by stress sensors, displacement change data at various points on the dam recorded by displacement meters, seepage rate and cumulative seepage data collected by seepage monitoring equipment, dam and surrounding temperature data collected by temperature sensors, and upstream and downstream water level data provided by water level monitoring stations. Based on the gravity dam's structural design drawings and the sub-region division specifications for hydraulic engineering projects, the server divides the entire dam structure into several structural sub-regions, including the upstream dam surface region, downstream dam surface region, dam foundation region, dam crest region, and dam heel region. Each sub-region corresponds to a monitoring data set consisting of data collected by all monitoring equipment within that region. For each monitoring data group, the server first extracts basic monitoring features: taking the monitoring data group of a sub-area on the upstream dam surface as an example, the signal processing algorithm is used to extract the time domain features such as peak value, valley value, and change rate of stress data, the cumulative change of displacement data, the periodic interval change trend and other features, the real-time value of seepage data and the deviation characteristics compared with the historical period; at the same time, based on the position of the sub-area in the three-dimensional coordinate system of the dam body (such as the spatial coordinates are (x=15m, y=25m, z=8m)), the server uses a spatial quantization coding algorithm (such as one-hot encoding or vector embedding algorithm based on coordinate normalization) to convert the physical spatial position into a computable spatial feature vector; then, the server integrates the basic monitoring features and spatial features into the sub-area safety impact feature of the sub-area through feature splicing and dimensional alignment operations, so that the feature not only contains monitoring information such as mechanics and seepage, but also incorporates the potential impact information of spatial position on safety assessment.

[0024] After extracting safety impact features for all sub-regions, the server performs dimensional decomposition on each sub-region. Taking the dam foundation sub-region as an example, its safety impact features are divided into two categories: one is the first-dimensional features focusing on structural mechanical response, including the uniformity of dam foundation stress distribution, the gradient of foundation settlement deformation, and acoustic monitoring of crack development within concrete. These first-dimensional features from all sub-regions collectively constitute the first-dimensional feature set. The other is the second-dimensional features centered on environmental factors and geological conditions, encompassing the distribution of seepage pressure under the dam foundation, electromagnetic induction quantification of the degree of weathering of the surrounding rock mass, and the impact of upstream and downstream water level differences on dam foundation loads. These second-dimensional features from all sub-regions collectively constitute the second-dimensional feature set. The server uses pre-defined dimensional decomposition rules (such as expert knowledge-based feature label classification or machine learning-based feature clustering) to decompose the safety impact features of all sub-regions, forming two feature sets covering the entire hydraulic structure.

[0025] Next, the server constructs a first-impact association network and a second-impact association network. For the first-impact dimension feature set, the server uses a mutual information algorithm to calculate the degree of correlation between any two first-impact dimension features. Taking the upstream dam surface stress feature and the dam crest stress feature as an example, by analyzing the time-varying numerical sequences of the two, the server calculates the mutual information entropy to quantify the degree of correlation. Based on the correlation threshold (e.g., a strong correlation is determined when the mutual information entropy is greater than 0.6), the correlation relationship (strong, weak, or no correlation) between the features is determined. Each first-impact dimension feature is considered a "physical unit" in the network, and directed or undirected "action links" are constructed based on the correlation relationship. If the upstream dam surface stress feature and the dam crest stress feature are strongly correlated, a bidirectional action link is established in the network to characterize their mutual influence. The resulting first-impact association network can intuitively present the interaction pattern of structural mechanics influencing factors between different sub-regions. For the second impact association network, the server determines the spatial adjacency relationship of the second impact dimension features through spatial topological analysis based on the physical spatial position of each monitoring data group: for example, the seepage pressure characteristics of the underground water at the dam base and the seepage characteristics of the adjacent downstream dam surface are judged to be "strong spatial correlation" because the two are adjacent to each other in the dam body space and have hydraulic connections; similarly, each second impact dimension feature is regarded as a "physical unit" and an action link is constructed according to the spatial adjacency relationship (for example, a one-way link is established between the seepage characteristics of the dam foundation and the downstream dam surface to characterize the transmission direction of the seepage), thereby forming a second impact association network, which reflects the mutual transmission relationship of environmental geological influencing factors in the spatial dimension.

[0026] In the feature coupling phase, the server calculates the statistics and variance of the associated entity features for each first-influence dimension feature in the first-influence association network. For example, a stress feature on the upstream dam surface (denoted as feature (A)) is associated with other stress features in the first-influence association network that interact with feature (A), such as the dam crest stress feature (B) and the adjacent dam section stress feature (C). The server first calculates the mean of these associated entity features to obtain the first-association region mean feature (reflecting the average stress level within the associated region). The server then calculates the difference between feature (A) and each associated entity feature and takes the variance of the difference sequence to obtain the first difference feature (reflecting the degree of difference between feature (A) and the associated region). The server then inputs feature (A), the first difference feature, and the first-association region mean feature into a fusion model constructed using a multi-layer perceptron. Through nonlinear activation and dimensionality conversion, the first fused feature is obtained. This fused feature is then linearly converted into a first, unified, comprehensive assessment feature using a fully connected layer. This feature comprehensively considers the synergistic effects of the structural mechanical characteristics and those of the associated regions, and is used to quantify the safety level of the structural mechanical dimension. The operating logic of the second impact association network is consistent: taking the seepage pressure characteristics of the dam base (denoted as feature (D)) as an example, the second associated regional mean characteristics and second difference characteristics of its associated entity characteristics (such as the seepage characteristics of the downstream dam surface (E) and the seepage characteristics of the adjacent dam section (F)) are calculated, and the second comprehensive assessment characteristics are obtained after fusion to quantify the safety level of the environmental geological dimension.

[0027] Finally, the server performs feature fusion and safety level determination. For the first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the safety impact feature of the same sub-region (e.g., the upstream dam surface sub-region has both the first comprehensive assessment feature (S1) of the mechanical dimension and the second comprehensive assessment feature (S2) of the environmental dimension), the server adopts a weighted fusion strategy (the weights are set by water conservancy engineering experts based on the importance of the project and historical accident cases, such as 60% for structural safety and 40% for environmental impact), and the weights are calculated by formula (S target =0.6×S1+0.4×S2) to calculate the target safety assessment characteristics of the sub-area (S target Then, the server calls the pre-trained security level determination model (such as a classification model based on a support vector machine, or a threshold rule set according to industry standards): If (S target∈[0,0.3)) is judged as "safe", ([0.3,0.7)) is judged as "warning", and ([0.7,1]) is judged as "dangerous"). The target safety assessment characteristics of each sub-area are graded. After integrating the judgment results of all sub-areas, if key sub-areas such as the dam foundation and dam heel are all "safe" and there are no "warning" or "dangerous" levels in non-critical sub-areas, the server outputs the safety assessment result corresponding to the concrete gravity dam hydraulic structure monitoring dataset as "overall safe". If the dam foundation sub-area has a "warning" level and multiple sub-areas on the downstream dam surface are "dangerous", the assessment result is "there are safety hazards in the structure, which require immediate investigation", providing accurate structural safety decision-making basis for engineering management departments.

[0028] In the embodiment of the present invention, the extraction of the features of each monitoring data group and the acquisition of the safety impact features of each sub-region can be implemented through the following examples.

[0029] Extracting the basic monitoring features of each monitoring data group to obtain the basic monitoring features of each monitoring data group;

[0030] Obtaining the physical spatial position of each monitoring data group, quantizing and encoding the physical spatial position of each monitoring data group, and obtaining spatial features of each monitoring data group;

[0031] The basic monitoring features of each monitoring data group and the corresponding spatial features of the monitoring data group are integrated to obtain the safety impact features of each sub-region.

[0032] In an embodiment of the present invention, exemplarily, the server performs feature extraction operations on the monitoring data groups of each structural sub-area of ​​the concrete gravity dam to be evaluated. Taking the monitoring data group of a sub-area on the upstream dam surface of the dam as an example, the server first extracts basic monitoring features: in the time-course data collected by the stress sensor in the sub-area, the server calculates the peak value of the stress value (such as the stress reaching 35MPa at a certain moment is the peak value of the day), the valley value (such as the stress drops to 28MPa at 2 am) and the hourly stress change rate (such as the stress rises from 30MPa to 32MPa from 7:00 to 8:00, and the change rate is 2MPa / h) through the sliding window algorithm; in the dam surface displacement data recorded by the displacement meter, the server calculates the daily cumulative displacement increment (such as the difference between the current day and the previous day). The server then extracts the real-time seepage rate (e.g., the current seepage rate is 0.5 L / s) and the deviation from the historical mean for the same period (e.g., the historical mean for the same period is 0.4 L / s, and the current deviation is +0.1 L / s). The server then integrates the characteristic parameters of these multiple types of monitoring data to form the basic monitoring feature vector for the sub-region monitoring data set, covering key information in dimensions such as mechanics, deformation, and seepage. Next, the server obtains the physical spatial location information of the sub-region. Based on the three-dimensional design model of the dam body, the corresponding coordinates of the upstream dam face sub-region in the spatial coordinate system are (x=10m, y=20m, z=5m) ((x) along the dam axis, (y) perpendicular to the dam axis, and (z) vertical height). The server adopts a spatial quantization coding strategy, first normalizing the three-dimensional coordinates (dividing the coordinate value by the total length of the dam body in the corresponding direction. For example, the total length of the dam axis is 200m, so (x) is normalized to (10 / 200=0.05); the width perpendicular to the dam axis is 50m, (y) is normalized to (20 / 50=0.4); the dam height is 80m, (z) is normalized to (5 / 80=0.0625)), and then converting the normalized coordinates into a one-hot encoding or a low-dimensional embedding vector (for example, using an autoencoder to learn a compact representation of spatial position) to generate a dimensionally adapted spatial feature vector for the monitoring data group, so that the spatial position information can be used in subsequent numerical calculations and feature fusion. Finally, the server performs a feature integration operation: taking the upstream dam surface sub-area as an example, its basic monitoring feature vector is 12-dimensional (including multiple indicator features such as stress, displacement, and seepage), and its spatial feature vector is 3-dimensional (normalized coordinate embedding). The server uses feature splicing technology (such as splicing in dimensional order) to merge the two types of features into a 15-dimensional vector. This vector not only retains the structural response information reflected by the monitoring data, but also incorporates the spatial constraint information of the spatial position on the safety assessment, thereby forming the sub-area safety impact characteristics of the sub-area.The server repeats the above process for the monitoring data groups of all structural sub-areas such as the dam foundation, dam top, and dam heel, ensuring that each sub-area generates safety impact features containing "monitoring information + spatial information", providing basic feature support for the subsequent fusion of multi-dimensional influencing factors.

[0033] In an embodiment of the present invention, the first influence association network corresponding to the first influence dimension feature set is established according to the degree of association between each first influence dimension feature in the first influence dimension feature set, which can be implemented through the following examples.

[0034] Calculating the degree of feature correlation between each of the first impact dimension features, and determining the correlation relationship between each of the first impact dimension features according to the degree of feature correlation;

[0035] Each of the first impact dimension features is taken as an entity unit, and an action link is constructed for each of the first impact dimension features according to the association relationship to obtain the first impact association network.

[0036] In an embodiment of the present invention, exemplarily, the server first calculates the degree of feature correlation for the first influencing dimension feature set (covering the structural mechanics features of each sub-region of the concrete gravity dam, such as upstream dam surface stress features, dam top stress features, dam foundation settlement deformation gradient features, dam heel crack acoustic monitoring features, etc.). Taking the upstream dam surface stress characteristics (denoted as Feature (A)) and the dam crest stress characteristics (denoted as Feature (B)) as examples, the server extracted time-varying monitoring sequence data for both (e.g., a 30-day daily stress peak sequence) and quantified their correlation using a mutual information algorithm. By calculating the entropy difference between the joint probability distribution and the marginal probability distribution of the two sequences, the mutual information entropy value was 0.7 (with a preset strong correlation threshold of 0.5), indicating a strong correlation between Feature (A) and Feature (B). Furthermore, taking the upstream dam surface stress characteristics (A) and the dam foundation settlement deformation gradient characteristics (denoted as Feature (C)) as examples, the monthly mean series of the two (e.g., 12 months of data on the mean stress and mean settlement gradient) were extracted, and the mutual information entropy value was calculated to be 0.3 (below the strong correlation threshold). Combined with engineering knowledge (the indirect influence mechanism of stress on dam foundation settlement), the relationship was determined to be weakly correlated. For features without obvious physical correlation (e.g., the upstream dam surface stress characteristics (A) and the slope displacement characteristics far from the dam body, which have no direct connection to spatial and mechanical conduction), the calculated mutual information entropy value was close to 0, indicating no correlation. The server traverses all feature pairs in the first impact dimension feature set and completes the annotation of the associated action relationships of all features. Subsequently, the server uses each first impact dimension feature as an entity unit and constructs an action link based on the associated action relationship. For strongly associated features (A) and (B), a bidirectional directed link is established in the network (arrows in both directions indicate the mutual influence of mechanical responses, such as changes in stress on the upstream dam surface will be transmitted to the dam top, and vice versa, adjustments to the dam top load will also react to the upstream dam surface); for weakly associated features (A) and (C), a unidirectional directed link is established (arrows point from (A) to (C), representing the indirect driving of the upstream dam surface stress on the dam foundation settlement through the dam body force transmission path); no links are constructed between unrelated features. By analogy, the correlation between all characteristics, such as the dam foundation settlement characteristics and the settlement characteristics of the adjacent dam sections (strong correlation, bidirectional link), the dam heel crack characteristics and the dam foundation stress characteristics (weak correlation, unidirectional link), is converted into a network topological structure, and finally a first-influence correlation network is formed. This network uses structural mechanics characteristics as nodes and correlation relationships as directed edges, intuitively presenting the interaction patterns of the mechanical response characteristics of different sub-regions in the time and space dimensions, and providing a topological correlation basis for subsequent feature coupling.

[0037] In an embodiment of the present invention, the establishing of the second influence association network corresponding to the second influence dimension feature set according to the physical spatial position of each monitoring data group can be implemented through the following examples.

[0038] Determine the spatial topology information of the second impact dimension features corresponding to each sub-region security impact feature according to the physical spatial position of each monitoring data group, and determine the structural adjacency relationship between each second impact dimension feature in the second impact dimension feature set according to the spatial topology information;

[0039] Each of the second impact dimension features is taken as an entity unit, and an action link is constructed for each of the second impact dimension features according to the structural adjacency relationship to obtain the second impact association network.

[0040] In an exemplary embodiment of the present invention, the server first determines spatial topology information based on the physical spatial location of each monitoring data set for the second impact dimension feature set (covering environmental geological characteristics of each sub-region of the concrete gravity dam, such as the seepage pressure characteristics of the dam base, the seepage characteristics of the downstream dam surface, the electromagnetic induction characteristics of the dam heel rock weathering, and the upstream water level load impact characteristics). For example, for the seepage pressure characteristics of the dam base (denoted as Feature (D)), the monitoring equipment is installed in the rock layer below the dam foundation (spatial coordinates (x=12m, y=22m, z=-5m, with the negative direction (z) indicating the depth of the bedrock); the monitoring equipment for the seepage characteristics of the downstream dam surface (denoted as Feature (E)) is located on the downstream surface of the dam body (coordinates (x=12m, y=22m, z=3m)). The server calculated the vertical distance between the two using spatial coordinates (the difference in the (z) direction is 8m). Combined with the dam's anti-seepage design drawings (which include hydraulic channels such as drainage holes between the dam foundation and the downstream dam surface), it determined that features (D) and (E) exhibited a spatial topological relationship of "vertical adjacency and hydraulic connectivity." Furthermore, using the electromagnetic induction characteristic of rock weathering at the dam heel (denoted as feature (F), coordinates (x=8m, y=18m, z=2m), corresponding to the upstream contact surface between the dam body and the bedrock) and the rock weathering characteristic within the dam foundation (denoted as feature (G), coordinates (x=8m, y=18m, z=-3m), located within the dam foundation rock mass) as examples, the server determined that features (F) and (G) exhibited a spatial topological relationship of "front-to-back adjacency and rock weathering conduction" based on the spatial coordinate relationship (continuity between the dam heel and the dam foundation rock mass along the (z) direction) and the engineering geological report (weathering at the dam heel easily permeates and diffuses into the dam foundation rock mass). Based on this spatial topology information, the server further determines structural adjacency relationships: Features (D) and (E) are hydraulically connected and spatially close, resulting in a "strong spatial adjacency." Features (F) and (G) are moderately connected due to the mutual weathering effect of rock mass continuity. The dam crest temperature monitoring feature (denoted as feature (H), located at the dam crest (x=15m, y=25m, z=80m)) and dam foundation seepage feature (D) are considered "no spatial adjacency" due to the vertical distance exceeding 70m and the lack of a direct hydraulic or geological transmission path. The server traverses all feature pairs in the second-influence dimension feature set and, combining rules such as spatial distance calculation (e.g., Euclidean distance) and engineering structural connectivity (hydraulic channels, rock mass integrity), labels the adjacency relationships of all features (strong, moderate, weak, or none). Subsequently, the server treats each second-dimension feature as a physical unit and constructs an action link according to the structural adjacency relationship: for strongly correlated features (D) and (E), a unidirectional directed link is established in the network (the arrow points from (D) to (E)) because the seepage pressure is transmitted from the dam foundation to the downstream dam surface; for moderately correlated features (F) and (G), a bidirectional directed link is established (the arrow bidirectionally connects (F) and (G)) because rock weathering diffuses in both directions on the contact surface; no link is constructed between unrelated features.Similarly, the upstream water level load impact characteristic (denoted as characteristic (I), corresponding to the mechanical response of the upstream water level acting on the dam foundation) and the dam foundation seepage characteristic (D) are determined to be strongly adjacent due to the changes in the seepage field driven by the water level load, and a one-way link ((ItoD)) is established. Ultimately, all second-dimension impact characteristics exist as "nodes," and structural adjacency relationships are connected as "directed / undirected edges," forming a second-dimension impact association network. This network intuitively presents the transmission paths and interaction patterns of environmental geological characteristics in the spatial dimension, providing topological support for spatial associations for subsequent feature coupling.

[0041] In an embodiment of the present invention, performing feature coupling based on the first influence dimension feature and the associated entity feature corresponding to the first influence dimension feature in the first influence association network to obtain the first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set includes:

[0042] Calculating the associated region mean feature of the associated entity feature corresponding to the first impact dimension feature to obtain a first associated region mean feature, and calculating the difference between the first impact dimension feature and the associated entity feature corresponding to the first impact dimension feature to obtain a first difference feature;

[0043] Fusing the first impact dimension feature, the first difference feature, and the first associated region mean feature to obtain a first fused feature, and performing a linear transformation based on the first fused feature to obtain a first comprehensive evaluation feature corresponding to the first impact dimension feature;

[0044] For each first influence dimension feature in the first influence association network, a first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set is obtained.

[0045] In an embodiment of the present invention, for example, when the server performs feature coupling operations on the upstream dam surface stress characteristics (denoted as the first impact dimension feature (A)) in the first impact association network, it first focuses on its associated entity features: the dam crest stress characteristics (B) and the adjacent dam section stress characteristics (C), which have a strong correlation link with (A) in the first impact association network. The server extracts the real-time monitoring data of (B) and (C): the mean of the stress series of (B) on that day is (32MPa), and the mean of the stress series of (C) on that day is (30MPa). By calculating the mean of these two associated entity features through arithmetic averaging, the first associated region mean feature is obtained as ((32+30) / 2=31MPa), which quantifies the average level of mechanical response in the associated region (A). Next, the server obtains the real-time stress value of (A) itself (e.g., (35 MPa)), and calculates the stress differences between (A) and (B), and (A) and (C), respectively: (35-32=3 MPa), (35-30=5 MPa); the arithmetic mean of the difference sequence is taken, and the first difference feature is obtained as ((3+5) / 2=4 MPa), which intuitively reflects the degree of deviation between (A) and the mechanical characteristics of the associated area. The server then concatenates the original stress feature (35 MPa, i.e., the first impact dimension feature), the first difference feature (4 MPa), and the first associated regional mean feature (31 MPa) of (A) to form a (1×3) feature vector ([35, 4, 31]). This vector is then input into a pre-set multi-layer perceptron model (with a hidden layer using a ReLU activation function and a linear output layer). The model first performs nonlinear interaction on the fused features through the hidden layer (e.g., extracting the synergistic influence of stress level, difference, and regional mean), and then linearly converts them into a single-dimensional numerical value through the output layer (e.g., the output quantitative score is (2.8), corresponding to the mechanical dimension safety level). This numerical value is the first comprehensive assessment feature corresponding to (A). The server repeats the above process for other features in the first impact dimension feature set: taking the dam foundation settlement deformation gradient feature (D) as an example, its associated entity features are the settlement features (E) and (F) of adjacent dam sections that are strongly associated in the first impact association network. The server calculates the mean of (E) (sedimentation gradient (0.02 mm / m)) and (F) (sedimentation gradient (0.015 mm / m)) to obtain the mean feature of the first associated region (0.0175 mm / m); calculates the difference between (D)'s own sedimentation gradient (0.03 mm / m) and (E) and (F) (0.01 mm / m, 0.015 mm / m), with an average difference of (0.0125 mm / m) (the first difference feature); and fuses the original gradient feature, difference feature, and mean feature of (D) into the model to output the first comprehensive evaluation feature of (D).By traversing all features in the first impact dimension feature set, the server finally generates a corresponding first comprehensive evaluation feature for each first impact dimension feature, providing a quantitative basis for the mechanical dimension for subsequent multi-dimensional safety assessment.

[0046] In an embodiment of the present invention, feature coupling is performed based on the second influence dimension feature and the associated entity feature corresponding to the second influence dimension feature in the second influence association network to obtain the second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set, which can be implemented through the following examples.

[0047] Calculating the associated region mean feature of the associated entity feature corresponding to the second impact dimension feature to obtain a second associated region mean feature, and calculating the difference between the second impact dimension feature and the associated entity feature corresponding to the second impact dimension feature to obtain a second difference feature;

[0048] Fusing the second impact dimension feature, the second difference feature, and the second associated region mean feature to obtain a second fused feature, and performing a linear transformation based on the second fused feature to obtain a second comprehensive evaluation feature corresponding to the second impact dimension feature;

[0049] For each second influence dimension feature in the second influence association network, obtain the second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set.

[0050] In an embodiment of the present invention, for example, when the server performs feature coupling on the seepage pressure characteristics of the dam base (denoted as the second impact dimension feature (D)) in the second impact association network, it first focuses on its associated entity features: the downstream dam surface seepage characteristics (E) and the adjacent dam section seepage characteristics (F), which have a strong spatial adjacency association with (D) in the second impact association network. The server extracts the real-time seepage pressure monitoring value of (E) (e.g., (20kPa)) and the real-time seepage pressure monitoring value of (F) (e.g., (18kPa)), and calculates the mean of these two associated entity features through arithmetic averaging, resulting in the second associated region mean feature of ((20+18) / 2=19kPa), which quantifies the average level of seepage pressure in the associated region (D). Next, the server obtains (D)'s real-time seepage pressure value (e.g., 22 kPa) and calculates the pressure differences between (D) and (E), and (D) and (F): (22 - 20 = 2 kPa) and (22 - 18 = 4 kPa, respectively). The arithmetic mean of these differences is calculated, yielding a second difference feature ((2 + 4) / 2 = 3 kPa), which intuitively reflects the degree of deviation between (D) and the seepage characteristics of the associated region. The server then concatenates (D)'s original seepage pressure feature (22 kPa, the second influencing dimension feature), the second difference feature (3 kPa), and the second associated region mean feature (19 kPa) into a (1 × 3) feature vector ([22, 3, 19]), which is then fed into a multilayer perceptron model with a sigmoid activation function. The model first uses the hidden layer to nonlinearly interact with the fused features (extracting the synergistic influence of seepage pressure level, variability, and regional mean), and then linearly converts them into a single-dimensional numerical value (e.g., a quantitative score of (3.2)) through the output layer. This is the second comprehensive assessment feature corresponding to (D). The server repeats this process for other features in the second impact dimension feature set: For example, the electromagnetic induction feature (F) of the dam heel rock weathering is associated with the weathering feature (G) within the dam foundation (50mV) and the weathering feature (H) of the upstream slope (45mV). The second associated regional mean feature ((50+45) / 2=47.5mV) is calculated, and the difference between (F)'s own value (55mV) and the average of (G) and (H) (7.5mV) is calculated (the second difference feature). This fusion is then input into the model output, which is the second comprehensive assessment feature (F). After traversing all features, the server generates corresponding assessment features for each second impact dimension feature to support subsequent multi-dimensional safety assessments.

[0051] In an embodiment of the present invention, the first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-region safety impact feature are fused to obtain the target safety assessment feature corresponding to each sub-region safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-region safety impact feature to obtain the safety assessment result corresponding to the hydraulic structure structure monitoring data set, which can be implemented through the following examples.

[0052] Obtaining a first feature enhancement weight, and weightedly enhancing the first comprehensive evaluation feature corresponding to each first impact dimension feature in the first impact dimension feature set according to the first feature enhancement weight, to obtain a first weighted enhancement feature corresponding to each first impact dimension feature in the first impact dimension feature set;

[0053] Perform weighted enhancement on the second comprehensive evaluation feature corresponding to each second impact dimension feature in the second impact dimension feature set according to the first feature enhancement weight, to obtain a second weighted enhancement feature corresponding to each second impact dimension feature in the second impact dimension feature set;

[0054] The first weighted enhancement feature and the second weighted enhancement feature corresponding to the same sub-area safety impact feature are fused to obtain the enhanced safety impact features corresponding to each sub-area safety impact feature, and the structural safety level is determined based on the enhanced safety impact features corresponding to each sub-area safety impact feature to obtain the target safety assessment result corresponding to the hydraulic structure structure monitoring data set.

[0055] In an exemplary embodiment of the present invention, the server first obtains the first feature enhancement weight from the water conservancy project safety assessment configuration module. This weight is set by domain experts based on engineering failure cases and regulatory requirements. The structural mechanics dimension enhancement weight corresponding to the first impact dimension feature is 0.6, and the environmental geology dimension enhancement weight corresponding to the second impact dimension feature is 0.4. For the generation of the first weighted enhancement feature, the first impact dimension feature of the upstream dam face sub-region (e.g., the upstream dam face stress feature (A)) is used as an example: the first comprehensive assessment feature obtained through previous coupling is (2.8) (a larger value indicates a higher mechanical safety hazard). The server weights and enhances this feature according to the first feature enhancement weight (0.6), calculating (2.8 × 0.6 = 1.68), which is the first weighted enhancement feature corresponding to (A). Similarly, for the dam foundation settlement deformation gradient feature (D) (the first comprehensive assessment feature is (3.0)), the server calculates (3.0 × 0.6 = 1.8). Weighted enhancement of all features in the first impact dimension feature set is completed, forming the first weighted enhancement feature set. When processing the second weighted enhancement feature simultaneously, the second impact dimension feature of the upstream dam face sub-region (e.g., the seepage pressure feature at the dam base (D)) is taken as an example. Its second comprehensive assessment feature is (3.2) (a larger value indicates a higher environmental safety hazard). The server weights it according to the first feature enhancement weight (0.4), calculating (3.2 × 0.4 = 1.28), which is the second weighted enhancement feature corresponding to (D). For the electromagnetic induction feature of the dam heel rock weathering (F) (the second comprehensive assessment feature is (2.5)), the server calculates (2.5 × 0.4 = 1.0), completing the weighted enhancement of the second impact dimension feature set and forming the second weighted enhancement feature set. Subsequently, the weighted features of the same sub-region are fused: For the upstream dam face sub-region, its first weighted enhancement feature (1.68) (mechanical dimension) and second weighted enhancement feature (1.28) (environmental dimension) need to be fused. The server uses a weighted summation strategy (with weights consistent with the enhanced weights, i.e., 0.6 for mechanical and 0.4 for environmental) to calculate (1.68 × 0.6 + 1.28 × 0.4 = 1.008 + 0.512 = 1.52). This represents the enhanced safety impact characteristic of the upstream dam face sub-region. This process is repeated for other sub-regions, such as the dam foundation and heel, to generate enhanced safety impact characteristics for each sub-region. Finally, the server determines the structural safety level by applying pre-set rules (e.g., ([0, 0.5)) for "safe," ([0.5, 1.0)) for "warning," and ([1.0, 2.0]) for "dangerous").If the enhanced safety impact characteristics of the upstream dam surface (1.52) and the enhanced safety impact characteristics of the dam foundation sub-area (1.59) (assumed values) are both "dangerous", and key sub-areas such as the dam heel also reach the "dangerous" level, the server will integrate the results and output the target safety assessment result of the concrete gravity dam as "the structure has serious safety hazards and requires immediate maintenance." If most sub-areas are "safe" and only a few are "warning", the output is "local attention is required, the overall safety is temporarily safe", providing a decision-making basis for project management.

[0056] In an embodiment of the present invention, the first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set is weightedly enhanced according to the first feature enhancement weight to obtain the first weighted enhancement feature corresponding to each first influence dimension feature in the first influence dimension feature set. This can be implemented through the following examples.

[0057] Perform feature mapping on the first comprehensive evaluation feature corresponding to each first impact dimension feature in the first impact dimension feature set according to the first feature enhancement weight, obtain a first mapping conversion feature corresponding to each first impact dimension feature in the first impact dimension feature set, and calculate a feature deviation value corresponding to the first mapping conversion feature to obtain a first feature deviation value;

[0058] Perform weighted fusion on the first comprehensive evaluation features corresponding to each first impact dimension feature in the first impact dimension feature set to obtain the first weighted fusion features corresponding to each first impact dimension feature in the first impact dimension feature set;

[0059] Calculate the multiplication result of the first weighted fusion feature and the first feature deviation value to obtain the first weighted enhancement feature corresponding to each first impact dimension feature in the first impact dimension feature set.

[0060] In an embodiment of the present invention, for example, when the server performs weighted enhancement on the upstream dam surface stress feature (denoted as the first impact dimension feature (A)) in the first impact dimension feature set, it first obtains the first feature enhancement weight (set to (0.6) by water conservancy engineering experts based on the impact weight of structural mechanics on safety). For the first comprehensive evaluation feature of (A) (assuming the value after preliminary coupling is (2.8)), the server first performs feature mapping: linearly scales the first comprehensive evaluation feature of (A) according to the first feature enhancement weight (0.6), calculating (2.8×0.6=1.68), which is the first mapping conversion feature corresponding to (A). Subsequently, the server counts the mapping values ​​of all the first comprehensive evaluation features in the sub-area within the first impact dimension feature set (e.g., the dam top stress feature (B) is mapped to (1.5), and the adjacent dam section stress feature (C) is mapped to (1.32)), calculates the mean of these mapping values ​​((1.68+1.5+1.32)÷3=1.5), and takes the difference between the first mapping conversion feature of (A) and the mean as the first feature deviation value, i.e., (1.68-1.5=0.18). Next, the server performs weighted fusion on the first comprehensive evaluation feature of (A) and other first-influence dimension features in the same subregion (such as the first comprehensive evaluation feature (2.5) of (B) and the first comprehensive evaluation feature (2.2) of (C)). The server assigns feature weights within the subregion based on the first feature enhancement weight ((A) accounts for (0.6), (B) accounts for (0.2), and (C) accounts for (0.2)). The result is (2.8×0.6+2.5×0.2+2.2×0.2=1.68+0.5+0.44=2.62), which is the first weighted fusion feature corresponding to (A). Finally, the server multiplies the first weighted fusion feature by the first feature deviation value, that is, (2.62×0.18≈0.4716). This result is the first weighted enhancement feature corresponding to (A). For other features in the first impact dimension feature set (such as the dam foundation settlement deformation gradient feature (D), whose first comprehensive assessment feature is (3.0)), the server repeats the above process: mapping (0.6) to (1.8), calculating the mean of the sub-region mapping values ​​(assuming (1.6)) to obtain a deviation of (0.2); weighted fusion ((D) accounts for (0.7), and other features in the same sub-region account for (0.3)) to obtain (3.0 × 0.7 + 2.8 × 0.3 = 2.1 + 0.84 = 2.94); multiplying (2.94 × 0.2 = 0.588) to generate the first weighted enhanced feature of (D). By traversing all features in the first impact dimension feature set, the server completes the generation of the full first weighted enhanced feature, providing the basis for the enhanced mechanical dimension for the subsequent sub-region safety feature fusion.

[0061] In an embodiment of the present invention, the second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set is weightedly enhanced according to the first feature enhancement weight to obtain the second weighted enhancement feature corresponding to each second influence dimension feature in the second influence dimension feature set. This can be implemented through the following examples.

[0062] Perform feature mapping on the second comprehensive evaluation feature corresponding to each second impact dimension feature in the second impact dimension feature set according to the first feature enhancement weight, obtain a second mapping conversion feature corresponding to each second impact dimension feature in the second impact dimension feature set, and calculate a feature deviation value corresponding to the second mapping conversion feature to obtain a second feature deviation value;

[0063] Perform weighted fusion on the second comprehensive evaluation features corresponding to each second impact dimension feature in the second impact dimension feature set to obtain second weighted fusion features corresponding to each second impact dimension feature in the second impact dimension feature set;

[0064] Calculate the multiplication result of the second weighted fusion feature and the second feature deviation value to obtain the second weighted enhancement feature corresponding to each second impact dimension feature in the second impact dimension feature set.

[0065] In this embodiment of the present invention, for example, when the server performs weighted enhancement on the seepage pressure characteristic of the dam base (denoted as the second impact dimension feature (D)) in the second impact dimension feature set, it first obtains the first feature enhancement weight (set to (0.4) by water conservancy engineering experts based on the weight of the impact of environmental geology on safety). For the second comprehensive assessment feature of (D) (the value after preliminary coupling is (3.2)), the server first performs feature mapping: the second comprehensive assessment feature of (D) is linearly scaled according to the first feature enhancement weight (0.4), and the calculated value is (3.2×0.4=1.28), which is the second mapping conversion feature corresponding to (D). Subsequently, the server counts all the second mapping conversion features in the sub-area of ​​the second influencing dimension feature set (for example, the downstream dam surface seepage feature (E) is mapped to (3.0×0.4=1.2), and the adjacent dam section seepage feature (F) is mapped to (2.8×0.4=1.12)), calculates the mean of these mapping values ​​((1.28+1.2+1.12)÷3=1.2), and takes the difference between the second mapping conversion feature of (D) and the mean as the second feature deviation value, that is, (1.28-1.2=0.08). Next, the server performs weighted fusion on the second comprehensive evaluation feature of (D) and other second-influence dimension features in the same sub-region (such as the second comprehensive evaluation feature of (E) (3.0) and the second comprehensive evaluation feature of (F) (2.8)). The server assigns feature weights within the sub-region based on the first feature enhancement weight ((D) accounts for (0.5), (E) accounts for (0.3), and (F) accounts for (0.2)). The result is (3.2×0.5+3.0×0.3+2.8×0.2=1.6+0.9+0.56=3.06), which is the second weighted fusion feature corresponding to (D). Finally, the server multiplies the second weighted fusion feature by the second feature deviation value, that is, (3.06×0.08≈0.2448). This result is the second weighted enhancement feature corresponding to (D). For other features in the second impact dimension feature set (such as the electromagnetic induction feature of the dam heel rock weathering (F), whose second comprehensive evaluation feature is (2.5)), the server repeats the above process: according to (0.4), it is mapped to (2.5×0.4=1.0), and the average of the sub-region mapping values ​​is calculated (assuming that the weathering feature of the dam foundation interior (G) in the same sub-region is mapped to (2.7×0.4=1.08), and the weathering feature of the upstream slope (H) is mapped to (2.3×0.4 =0.92), the mean is ((1.0+1.08+0.92)÷3=1.0)) to get the deviation (1.0-1.0=0); weight fusion ((F) accounts for (0.4), (G) accounts for (0.3), (H) accounts for (0.3)) to get (2.5×0.4+2.7×0.3+2.3×0.3=1.0+0.81+0.69=2.5); multiplication (2.5×0=0) generates the second weighted enhancement feature of (F).By traversing all the features in the second impact dimension feature set, the server completes the generation of the full second weighted enhanced features, providing an enhanced environmental geological dimension basis for the subsequent sub-regional safety feature fusion.

[0066] In an embodiment of the present invention, the structural safety level is determined according to the enhanced safety impact characteristics corresponding to the safety impact characteristics of each sub-area, and the target safety assessment result corresponding to the hydraulic structure structure monitoring data set is obtained, which can be implemented through the following examples.

[0067] Dividing the enhanced security impact features corresponding to each sub-region security impact feature to obtain a first enhanced impact dimension feature set, a second enhanced impact dimension feature set, and a third enhanced impact dimension feature set, wherein the sum of the number of dimensions of the second enhanced impact dimension feature in the second enhanced impact dimension feature set and the number of dimensions of the third enhanced impact dimension feature in the third enhanced impact dimension feature set is the same as the number of dimensions of the second impact dimension feature;

[0068] Establish a first reinforcement influence association network corresponding to the first reinforcement influence dimension feature set according to the degree of correlation between each first reinforcement influence dimension feature in the first reinforcement influence dimension feature set, and establish a second reinforcement influence association network corresponding to the second reinforcement influence dimension feature set according to the physical spatial position of each monitoring data group;

[0069] Determine, according to the physical spatial position of each monitoring data group, the adjacent strengthening influence dimension features corresponding to each third strengthening influence dimension feature in the third strengthening influence dimension feature set, and establish a third strengthening influence association network corresponding to the third strengthening influence dimension feature set according to the degree of correlation between the adjacent strengthening influence dimension features corresponding to each third strengthening influence dimension feature in the third strengthening influence dimension feature set;

[0070] Perform feature coupling based on the first reinforcement influence dimension feature and the associated entity feature corresponding to the first reinforcement influence dimension feature in the first reinforcement influence association network to obtain first coupled reinforcement features corresponding to each first reinforcement influence dimension feature in the first reinforcement influence dimension feature set;

[0071] Perform feature coupling based on the associated entity features corresponding to the second reinforcement influence dimension features and the second reinforcement influence dimension features in the second reinforcement influence association network, to obtain second coupled reinforcement features corresponding to each second reinforcement influence dimension feature in the second reinforcement influence dimension feature set;

[0072] Perform feature coupling based on the third reinforcement influence dimension feature and the associated entity features corresponding to the third reinforcement influence dimension features in the third reinforcement influence association network, to obtain third coupling reinforcement features corresponding to each third reinforcement influence dimension feature in the third reinforcement influence dimension feature set;

[0073] The first coupling enhancement feature, the second coupling enhancement feature, and the third coupling enhancement feature corresponding to the same sub-region security impact feature are fused to obtain the target enhanced security impact feature corresponding to each sub-region security impact feature;

[0074] A structural safety level is determined according to the target enhanced safety impact characteristics corresponding to each sub-region safety impact characteristic, and an enhanced safety assessment result corresponding to the hydraulic structure structure monitoring data set is obtained.

[0075] In the embodiment of the present invention, illustratively, the server takes a concrete gravity dam as an evaluation object and performs a structural safety level determination process on the enhanced safety impact characteristics of each sub-area as follows.

[0076] Enhancement Feature Set Division: For the enhanced safety impact characteristics of the upstream dam face sub-region (assuming a 10-dimensional fusion vector), the server, based on the dimensional division rule of "structural mechanics-environmental geology foundation-environmental geology extension," splits them into: a first enhanced impact dimension feature set (focusing on structural mechanics depth characteristics, such as the upstream dam face stress gradient enhancement characteristics and the dam crest stress synergistic enhancement characteristics, totaling three dimensions and reflecting the spatiotemporal evolution of the mechanical response); a second enhanced impact dimension feature set (environmental geology foundation characteristics, such as the dam foundation seepage pressure foundation enhancement characteristics and the downstream dam face seepage conduction enhancement characteristics, totaling three dimensions and corresponding to the core path of hydraulic connectivity); and a third enhanced impact dimension feature set (environmental geology extension characteristics, such as the dam heel rock weathering diffusion enhancement characteristics and the upstream slope weathering linkage enhancement characteristics, totaling four dimensions and covering the edge effects of geological degradation). Because the original number of dimensions of the second impact dimension is 7 (3 + 4), the constraint that the sum of the second and third enhanced dimensions equals the original number of the second impact dimension is satisfied. The server repeats this operation for all sub-regions, including the dam foundation, dam heel, and dam crest, to form three types of enhanced feature sets for the entire dam body. Construction of the enhanced impact correlation network: 1. The first enhanced impact correlation network: for the first enhanced impact dimension feature set (taking the upstream dam surface stress gradient enhancement feature (A1) and the dam top stress synergistic enhancement feature (B1) as examples), the server extracts the time series data of the two for 30 consecutive days, calculates the degree of correlation through the mutual information algorithm (for example, the mutual information entropy is 0.8, which is higher than the strong correlation threshold of 0.6), and determines the "strong correlation-bidirectional action" relationship; takes each first enhanced feature as an entity unit, and constructs a directed link according to the association relationship (a bidirectional arrow is set between (A1) and (B1) to represent the cross-regional transmission of stress response), forming a network to quantify the interaction pattern of the deep characteristics of structural mechanics. 2. Second Enhanced Impact Association Network: For the second enhanced impact dimension feature set (taking the dam foundation seepage pressure basic enhancement feature (D2) and the downstream dam surface seepage conduction enhancement feature (E2) as examples), the server determines the "strong hydraulic adjacency-unidirectional conduction" relationship based on the physical spatial location of the monitoring data set ((D2) coordinate (z=-5m) corresponds to the dam foundation rock layer, and (E2) coordinate (z=3m) corresponds to the downstream dam surface) and the dam body anti-seepage design drawings (a group of drainage holes is set between the dam foundation and the downstream dam surface). Each second enhanced feature is regarded as a physical unit, and a directed link is constructed according to the spatial adjacency relationship ((D2toE2) has a unidirectional arrow to represent the transmission of seepage pressure from the dam foundation to the downstream dam surface), forming a network to present the spatial conduction law of the basic characteristics of environmental geology.3. Third Enhancement Impact Association Network: For the third enhancement impact dimension feature set (taking the dam heel rock weathering diffusion enhancement feature (F3) and the dam foundation internal weathering linkage enhancement feature (G3) as examples), the server first identifies adjacent enhancement features using spatial coordinates ((F3) coordinates (x=8m, y=18m, z=2m) correspond to the dam heel, and (G3) coordinates (x=8m, y=18m, z=-3m) correspond to the dam foundation interior). It then extracts electromagnetic induction time history data for both rock weathering features and calculates the Pearson correlation coefficient (e.g., 0.75 indicates a strong correlation) to determine the "strong weathering diffusion-bidirectional interaction" relationship. Each third enhancement feature is treated as an entity unit, and directed links are constructed based on the adjacent association relationship ((F3-G3) with bidirectional arrows representing the cross-interface diffusion of rock weathering). This network is formed to characterize the adjacent interaction mechanism of environmental geological expansion features. Enhancement Feature Coupling and Fusion: 1. First Coupling Enhancement Feature Generation: For (A1) of the first enhancement impact association network, its associated entity feature is (B1), etc. The server calculates the mean of the associated entity features (e.g., the mean of (B1) is (1.5)) to obtain the first associated region mean feature; calculates the difference between (A1) and the associated entity features (e.g., (A1) is (1.8) and the difference is (0.3)) to obtain the first difference feature; inputs (A1), the first difference feature, and the first associated region mean feature into the multilayer perceptron, and after nonlinear fusion and linear transformation, outputs the first coupled enhancement feature corresponding to (A1) (e.g., (1.6)). 2. Second coupled enhancement feature generation: For (D2), which has a second enhanced impact on the associated network, its associated entity feature is (E2), etc. The server calculates the mean of the associated entity features (e.g., the mean of (E2) is (1.2)) to obtain the second associated region mean feature; calculates the difference between (D2) and the associated entity features (e.g., (D2) is (1.4) and the difference is (0.2)) to obtain the second difference feature; inputs (D2), the second difference feature, and the second associated region mean feature into the fusion model, outputting the second coupled enhancement feature corresponding to (D2) (e.g., (1.3)). 3. Generating the Third Coupling Enhancement Feature: For the third enhanced impact association network (F3), its associated entity feature is (G3), etc. The server calculates the mean of the associated entity features (e.g., the mean of (G3) is (1.1)) to obtain the third associated region mean feature; calculates the difference between (F3) and the associated entity feature (e.g., if (F3) is (1.3), the difference is (0.2)) to obtain the third difference feature; inputs (F3), the third difference feature, and the third associated region mean feature into the coupling model, outputting the third coupling enhancement feature corresponding to (F3) (e.g., (1.2)). 4. Fusion of Target Enhanced Safety Impact Features: Taking the upstream dam face sub-area as an example, its first coupling enhancement feature (1.6) (structural mechanics depth), second coupling enhancement feature (1.3) (environmental geology foundation), and third coupling enhancement feature (1.2) (environmental geology extension) need to be fused.The server performs a weighted summation based on the engineering weights (structural mechanics 0.5, environmental geology foundation 0.3, and environmental geology extension 0.2): (1.6×0.5+1.3×0.3+1.2×0.2=0.8+0.39+0.24=1.43), which is the target enhanced safety impact characteristic for this sub-area. Safety Level Determination and Result Output: The server invokes preset threshold rules (([0,0.5)) for "safe," ([0.5,1.0)) for "warning," and ([1.0,2.0]) for "dangerous") to evaluate the safety impact of the target enhanced characteristics of each sub-region. If the upstream dam face target enhanced characteristic (1.43) is considered "dangerous," the dam foundation sub-region target enhanced characteristic (1.55) (assumed value) also reaches "dangerous," and key sub-regions such as the dam heel and dam crest are all at the "dangerous" level, the server integrates the results of all sub-regions of the dam body and outputs an enhanced safety assessment result: "The structure presents extremely serious safety hazards, requiring immediate activation of the emergency maintenance plan." If most sub-regions are at "warning" and only a few peripheral sub-regions are at "dangerous," the server outputs "local hazards require focused investigation, the overall situation is at warning, and increased monitoring frequency is recommended." This in-depth coupling of multi-dimensional enhanced characteristics and level determination provides more refined and targeted decision-making for the structural safety management of hydraulic structures.

[0077] In an embodiment of the present invention, the third reinforcement influence association network corresponding to the third reinforcement influence dimension feature set is established according to the degree of association between the adjacent reinforcement influence dimension features corresponding to each third reinforcement influence dimension feature in the third reinforcement influence dimension feature set, which can be implemented through the following examples.

[0078] Selecting a feature to be analyzed and a reference feature from each of the third reinforcement impact dimension features;

[0079] Determine, from each third enhanced impact dimension feature according to the physical spatial position of each monitoring data group, each adjacent feature to be analyzed corresponding to the feature to be analyzed, and perform feature integration on each adjacent feature to be analyzed to obtain a fused adjacent feature to be analyzed;

[0080] Determine each target neighboring feature corresponding to the reference feature from each third enhanced impact dimension feature according to the physical spatial position of each monitoring data group, perform feature integration on each target neighboring feature, and obtain a target integrated neighboring feature;

[0081] Calculating the degree of correlation between the to-be-analyzed fused adjacent feature and the target integrated adjacent feature, and obtaining the degree of correlation between the to-be-analyzed feature and the reference feature;

[0082] For each of the third reinforcement influence dimension features, obtain the degree of correlation between the adjacent reinforcement influence dimension features corresponding to each of the third reinforcement influence dimension features, and use the degree of correlation between the adjacent reinforcement influence dimension features corresponding to each of the third reinforcement influence dimension features as the target degree of correlation between each of the third reinforcement influence dimension features;

[0083] According to the target association closeness, the standard action relationship between each third reinforcement influence dimension feature is determined, and each third reinforcement influence dimension feature is used as an entity unit. According to the standard action relationship, each third reinforcement influence dimension feature is constructed into an action link according to the relationship to obtain the third reinforcement influence association network.

[0084] In an embodiment of the present invention, the server exemplarily executes the following process for constructing a third reinforcement impact association network for a concrete gravity dam's third reinforcement impact dimension feature set (covering extended environmental geological features such as the dam heel rock weathering diffusion enhancement feature (F3), the dam foundation internal weathering linkage enhancement feature (G3), the upstream slope weathering linkage enhancement feature (H3), and the downstream dam surface weathering penetration enhancement feature (I3)): Selecting features to be analyzed and reference features: From the third reinforcement impact dimension feature set, the server selects the dam heel rock weathering diffusion enhancement feature (F3) as the feature to be analyzed and the dam foundation internal weathering linkage enhancement feature (G3) as the reference feature. (F3) corresponds to the rock weathering electromagnetic induction enhancement value in the dam heel region (spatial coordinates (x=8m, y=18m, z=2m)), and (G3) corresponds to the rock weathering linkage enhancement value in the dam foundation internal region (coordinates (x=8m, y=18m, z=-3m)). Identify and integrate adjacent features: 1. Integration of adjacent features to be analyzed: Based on the physical location of (F3), the server searches for third enhancement features within the same subregion with a spatial distance less than 5m. (G3) (inside the dam foundation, vertical distance 5m) and the upstream slope weathering linkage enhancement feature (H3) (coordinates (x=5m, y=15m, z=1m), horizontal distance 3m) are identified as adjacent features to be analyzed. The enhancement values ​​(1.1) and (1.0) of (G3) are extracted, and the arithmetic average integration is performed to obtain the fused adjacent features to be analyzed: ((1.1+1.0)÷2=1.05). 2. Integration of target adjacent features: Based on the physical location of (G3), the server searches for third enhancement features within a spatial distance less than 5m. (F3) (dam heel, vertical distance 5m) and the downstream dam surface weathering penetration enhancement feature (I3) (coordinates (x=12m, y=22m, z=3m), horizontal distance 4m) are identified as target adjacent features. Extract the reinforcement value (1.2) of (F3) and the reinforcement value (1.3) of (I3), and integrate them through arithmetic averaging to obtain the target integrated neighboring features: ((1.2 + 1.3) ÷ 2 = 1.25). Calculate the degree of association and construct the network: 1. Calculate the correlation between feature pairs: The server uses the Pearson correlation coefficient to calculate the degree of association between (F3) (the feature to be analyzed) and (G3) (the reference feature). The historical reinforcement value sequence of (F3) (e.g., ([1.2, 1.3, 1.1, 1.4])) and the historical reinforcement value sequence of (G3) (e.g., ([1.1, 1.2, 1.0, 1.3])) are fed into the Pearson algorithm. The calculated correlation coefficient is (0.8), which is greater than the strong association threshold (0.7), indicating that (F3) and (G3) are strongly associated.2. Summary of all feature correlations: The server repeats the above operation for all feature pairs in the third enhancement impact dimension feature set (e.g., a correlation coefficient of 0.6 is calculated between (F3) and (H3), indicating a moderate correlation; a correlation coefficient of 0.75 is calculated between (G3) and (I3), indicating a strong correlation). The correlation strength of each pair of adjacent enhancement features is used as the target correlation strength. 3. Action Link and Network Construction: The server determines a standard action relationship based on the target correlation strength (strong correlations are assigned bidirectional directed links, moderate correlations are assigned unidirectional directed links, and weak correlations are assigned no links). For example, for (F3) and (G3), a bidirectional link is assigned due to the strong correlation (the bidirectional arrows indicate the cross-interface diffusion of rock weathering); for (F3) and (H3), a unidirectional link is assigned due to the moderate correlation (the arrow points from (F3) to (H3), indicating the driving force of dam heel weathering on the upstream slope). Taking each third-enhancement feature as an entity unit, an action link is constructed according to the standard action relationship, and finally a third-enhancement impact association network is formed. This network uses environmental geological extension features as nodes and the degree of association as directed edges, intuitively presenting the interactive conduction mode of extended environmental factors such as weathering and infiltration in adjacent sub-regions, providing topological support for subsequent feature coupling.

[0085] In an embodiment of the present invention, the first coupling enhancement feature, the second coupling enhancement feature and the third coupling enhancement feature corresponding to the same sub-area security impact feature are fused to obtain the target enhanced security impact feature corresponding to each sub-area security impact feature, which can be implemented through the following examples.

[0086] Obtaining a second feature enhancement weight, and weightedly enhancing the first coupling enhancement feature corresponding to each first enhancement influence dimension feature in the first enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a first enhancement gain feature corresponding to each first enhancement influence dimension feature in the first enhancement influence dimension feature set;

[0087] Perform weighted enhancement on the second coupling enhancement feature corresponding to each second enhancement influence dimension feature in the second enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a second enhancement gain feature corresponding to each second enhancement influence dimension feature in the second enhancement influence dimension feature set;

[0088] Perform weighted enhancement on the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a third enhancement gain feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set;

[0089] The first enhanced gain feature, the second enhanced gain feature, and the third enhanced gain feature corresponding to the same sub-region security impact feature are fused to obtain the target enhanced security impact feature corresponding to each sub-region security impact feature.

[0090] In this embodiment of the present invention, the server illustratively retrieves the second feature enhancement weight from the water conservancy project safety assessment rule library. These weights, set by domain experts based on multi-dimensional impact priorities, include a weight of 0.4 for the structural mechanics depth dimension (first enhancement), a weight of 0.3 for the environmental geology foundation dimension (second enhancement), and a weight of 0.3 for the environmental geology extension dimension (third enhancement). These weights guide the weighted enhancement of each enhancement dimension. For the upstream dam surface stress gradient enhancement feature (A1) in the first enhancement impact dimension feature set (which, after coupling, yields a first coupling enhancement feature value of 1.6, reflecting the synergistic safety level of the structural mechanics depth response), the server performs weighted enhancement based on the second feature enhancement weight of 0.4: (1.6 × 0.4 = 0.64), which is the first enhancement gain feature corresponding to (A1). This process is repeated for other features in the first enhancement impact dimension feature set (e.g., the dam crest stress synergistic enhancement feature (B1), with a first coupling enhancement feature value of 1.5), yielding (1.5 × 0.4 = 0.6), completing the generation of all first enhancement gain features and forming the first enhancement gain feature set. For the dam foundation seepage pressure enhancement feature (D2) in the second enhancement dimension feature set (the second coupling enhancement feature value is (1.4) after coupling, reflecting the safety level of environmental geological foundation conduction), the server performs weighted enhancement using the second feature enhancement weight (0.3): (1.4 × 0.3 = 0.42), which is the second enhancement gain feature corresponding to (D2). Repeating this process for other features in the second enhancement dimension feature set (such as the downstream dam face seepage conduction enhancement feature (E2), with a second coupling enhancement feature value of (1.3)), yields (1.3 × 0.3 = 0.39), completing the generation of all second enhancement gain features and forming the second enhancement gain feature set. For the dam heel rock mass weathering diffusion enhancement feature (F3) in the third enhancement dimension feature set (the third coupling enhancement feature value is (1.2) after coupling, reflecting the safety level of environmental geological extension interaction), the server performs weighted enhancement using the second feature enhancement weight (0.3): (1.2 × 0.3 = 0.36), which is the third enhancement gain feature corresponding to (F3). Repeat this process for other features in the third enhancement impact dimension feature set (e.g., upstream slope weathering linkage enhancement feature (H3) and third coupling enhancement feature value (1.1)), obtaining (1.1 × 0.3 = 0.33), completing the generation of the full set of third enhancement gain features and forming the third enhancement gain feature set. Taking the upstream dam face sub-region as an example, the corresponding first enhancement gain feature (0.64) (structural mechanics depth), second enhancement gain feature (0.39) (environmental geology foundation, corresponding to the gain value of the downstream dam face seepage conduction enhancement feature (E2)), and third enhancement gain feature (0.36) (environmental geology extension, corresponding to the gain value of the dam heel rock weathering diffusion enhancement feature (F3)) need to be integrated.The server uses a weighted summation strategy, directly adding the three types of gain features according to their weighted proportions (because the weights are already embedded in the enhancement phase, this summation is equivalent to multi-dimensional collaborative fusion): (0.64 + 0.39 + 0.36 = 1.39). This result is the target enhanced safety impact characteristic of the upstream dam face sub-region. The server repeats this process for all sub-regions, including the dam foundation, dam heel, and dam crest. Through a "weighted enhancement-cross-dimensional summation" approach, it deeply integrates the structural mechanics depth, environmental geological foundation, and enhanced safety information from the extended dimension to generate the target enhanced safety impact characteristic for each sub-region, providing a more accurate multi-dimensional safety quantification basis for subsequent structural safety level determination.

[0091] In an embodiment of the present invention, the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set is weightedly enhanced according to the second feature enhancement weight to obtain the third enhancement gain feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set. This can be implemented through the following examples.

[0092] Perform feature mapping on the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set according to the second feature enhancement weight, obtain a third mapping conversion feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set, and calculate a feature deviation value corresponding to the third mapping conversion feature to obtain a third feature deviation value;

[0093] Perform weighted fusion on the second coupling enhancement features corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set to obtain third weighted fusion features corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set;

[0094] Calculate the multiplication result of the third weight fusion feature and the third feature deviation value to obtain the third enhanced gain feature corresponding to each third enhanced influence dimension feature in the third enhanced influence dimension feature set.

[0095] In an embodiment of the present invention, for example, when the server performs weighted enhancement on the dam heel rock mass weathering diffusion enhancement feature (F3) in the third enhancement impact dimension feature set, it first obtains the second feature enhancement weight (set to (0.3) by water conservancy engineering experts based on the safety impact priority of the environmental geology expansion dimension). For the third coupling enhancement feature of (F3) (the value after the initial coupling is (1.2), reflecting the safety level of the interaction between the dam heel weathering and the adjacent area), the server performs feature mapping based on the second feature enhancement weight (0.3): linearly scaling the third coupling enhancement feature of (F3) results in a calculated value of (1.2 × 0.3 = 0.36), which is the third mapping conversion feature corresponding to (F3). Subsequently, the server counted all the third mapping conversion features in the sub-area within the third enhancement impact dimension feature set (e.g., the weathering linkage enhancement feature (G3) inside the dam foundation is mapped to (1.1×0.3=0.33), and the weathering linkage enhancement feature (H3) on the upstream slope is mapped to (1.0×0.3=0.3)), calculated the mean of these mapping values ​​((0.36+0.33+0.3)÷3=0.33), and took the difference between the third mapping conversion feature of (F3) and the mean as the third feature deviation value, i.e., (0.36-0.33=0.03). Next, the server performs weighted fusion on the third coupling enhancement feature of (F3) and other third enhancement impact dimension features in the same subregion (e.g., the third coupling enhancement feature (1.1) of (G3) and the third coupling enhancement feature (1.0) of (H3)). Based on the priority of the environmental geological extension features within the subregion ((F3) accounts for (0.5), (G3) accounts for (0.3), and (H3) accounts for (0.2)), the server calculates (1.2 × 0.5 + 1.1 × 0.3 + 1.0 × 0.2 = 0.6 + 0.33 + 0.2 = 1.13), which is the third weighted fusion feature corresponding to (F3). Finally, the server multiplies the third weighted fusion feature by the third feature deviation value, that is, (1.13 × 0.03 ≈ 0.0339), which is the third enhancement gain feature corresponding to (F3). For the downstream dam surface weathering and penetration enhancement feature (I3) in the third enhancement impact dimension feature set (the third coupling enhancement feature is (1.3)), the server repeats the process: mapping by (0.3) to obtain (1.3×0.3=0.39), statistically calculating the mean of the sub-region mapping values ​​(assuming the same sub-region (G3) mapping (0.33), (I3) itself (0.39), and other features (0.36)) to calculate the deviation (0.39-0.36=0.03); according to the sub-region weight ((I3) accounts for (0.4), the same region feature accounts for (0.6)), the weight fusion is (1.3×0.4+1.2×0.6=0.52+0.72=1.24); multiplying (1.24×0.03=0.0372) to generate the third enhancement gain feature of (I3).By traversing all third-enhancement features, the server completes the generation of full-scale third-enhancement gain features, providing an enhanced basis for the environmental geological extension dimension for the fusion of target enhancement safety impact features.

[0096] In the embodiments of the present invention, the following implementation modes are also provided.

[0097] The target enhanced security impact feature is used as the enhanced security impact feature, and the enhanced security impact feature corresponding to each sub-region security impact feature is divided to obtain a first enhanced impact dimension feature set, a second enhanced impact dimension feature set, and a third enhanced impact dimension feature set. The number of dimensions of the second enhanced impact dimension feature in the second enhanced impact dimension feature set is increased according to a preset number of dimensions, and the number of dimensions of the third enhanced impact dimension feature in the third enhanced impact dimension feature set is decreased according to the preset number of dimensions.

[0098] Until the preset termination conditions are met, the final sub-region safety impact characteristics corresponding to each sub-region safety impact characteristic are obtained, and the structural safety level is determined based on the final sub-region safety impact characteristics corresponding to each sub-region safety impact characteristic, so as to obtain the final safety assessment result corresponding to the hydraulic structure structure monitoring data set.

[0099] In an embodiment of the present invention, the server illustratively adjusts dimensionality allocation according to rules: the number of dimensions of the second enhanced impact dimension feature set is "pre-set increased" (e.g., from the initial 3 dimensions to 4 dimensions, adding the dynamic fluctuation feature of the dam foundation seepage pressure), and the number of dimensions of the third enhanced impact dimension feature set is "pre-set decreased" (e.g., from the initial 4 dimensions to 3 dimensions, retaining the core dam heel-dam foundation weathering linkage feature), ensuring that the constraint of "second enhanced + third enhanced dimension = original second impact dimension feature dimension" persists. Taking the upstream dam face as an example, the newly enhanced safety impact features are re-split into: a first enhanced impact dimension feature set focusing on structural mechanics depth (3 dimensions, including time-varying stress gradient features); a second enhanced impact dimension feature set (4 dimensions, adding seepage pressure spectrum features) based on an expanded environmental geology foundation; and a third enhanced impact dimension feature set (3 dimensions, retaining cross-interface weathering conduction features) based on a streamlined environmental geology expansion. The server repeatedly executes the entire process of "strengthening the impact association network construction - feature coupling - multi-dimensional fusion to generate target enhancement features" until the preset termination condition is met (such as the number of iterations reaches 5, or the change rate of the target enhancement features between two consecutive rounds is less than 0.01). Assume that after 3 iterations, the target enhancement feature of the upstream dam surface stabilizes at (1.42), and the final features of the dam foundation and dam heel sub-regions converge to (1.45) and (1.41), respectively. The server performs a level assessment on the final sub-region safety impact features of the entire dam body sub-region: if the majority of key sub-region features are in the "dangerous" range (e.g., ([1.0, 2.0])), the final safety assessment result is output as "After multiple rounds of iteration verification, the hydraulic structure structure has persistent serious safety hazards and requires immediate comprehensive overhaul and reinforcement." If the features fall back to the "warning" range (e.g., ([0.5, 1.0))) after iteration, the output is "The structural safety hazard is in a controllable warning state. It is recommended to increase the monitoring frequency and formulate a preventive maintenance plan." Through iterative dimension adjustment and feature optimization, the accuracy of safety assessments can be progressively improved and the reliability of conclusions can be verified.

[0100] In an embodiment of the present invention, determining the structural safety level according to the target safety assessment characteristics corresponding to the safety impact characteristics of each sub-region and obtaining the safety assessment result corresponding to the hydraulic structure structure monitoring data set can be implemented through the following examples.

[0101] Dividing the target security assessment features corresponding to the sub-region security impact features respectively to obtain a first target impact dimension feature set, a second target impact dimension feature set, and a third target impact dimension feature set, wherein the sum of the number of dimensions of the second target impact dimension features in the second target impact dimension feature set and the number of dimensions of the third target impact dimension features in the third target impact dimension feature set is the same as the number of dimensions of the second impact dimension features;

[0102] Establishing a first target impact association network corresponding to the first target impact dimension feature set according to the degree of association between each first target impact dimension feature in the first target impact dimension feature set, and establishing a second target impact association network corresponding to the second target impact dimension feature set according to the physical spatial position of each monitoring data group;

[0103] Determine the adjacent target influence dimension features corresponding to each third target influence dimension feature in the third target influence dimension feature set according to the physical spatial position of each monitoring data group, and establish a third target influence association network corresponding to the third target influence dimension feature set according to the degree of association between the adjacent target influence dimension features corresponding to each third target influence dimension feature in the third target influence dimension feature set;

[0104] Perform feature coupling based on the first target influence dimension feature and the associated entity feature corresponding to the first target influence dimension feature in the first target influence dimension influence association network, to obtain first coupled target features corresponding to each first target influence dimension feature in the first target influence dimension feature set;

[0105] Perform feature coupling based on the associated entity features corresponding to the second target influence dimension features and the second target influence dimension features in the second target influence association network, to obtain second coupled target features corresponding to each second target influence dimension feature in the second target influence dimension feature set;

[0106] Perform feature coupling based on the third target influence dimension feature and the associated entity features corresponding to the third target influence dimension features in the third target influence association network, to obtain third coupled target features corresponding to each third target influence dimension feature in the third target influence dimension feature set;

[0107] fusing the first coupling target feature, the second coupling target feature, and the third coupling target feature corresponding to the same sub-region security impact feature to obtain the current sub-region security impact feature corresponding to each sub-region security impact feature;

[0108] A structural safety level is determined according to the current sub-region safety impact characteristics corresponding to each sub-region safety impact characteristic, and a current safety assessment result corresponding to the hydraulic structure structure monitoring data set is obtained.

[0109] In this embodiment of the present invention, taking the target safety assessment features of the upstream dam face sub-region (assuming a 10-dimensional fusion vector covering comprehensive safety information from structural mechanics and environmental geology) as an example, the server, following the dimensional allocation rule of "structural core - environmental foundation - environmental extension," splits them into: a first target impact dimension feature set (focusing on the core structural mechanics response, such as the upstream dam face stress coordination target feature and the dam crest stress conduction target feature, totaling 3 dimensions and reflecting the critical path of mechanical safety); a second target impact dimension feature set (environmental-geological foundation interaction, such as the dam foundation seepage pressure target feature and the downstream dam face seepage conduction target feature, totaling 4 dimensions and corresponding to the core interface of hydraulic connectivity); and a third target impact dimension feature set (environmental-geological extension effects, such as the dam heel rock weathering target feature and the dam foundation internal weathering linkage target feature, totaling 3 dimensions and covering the edge conduction of geological degradation). These three feature sets meet the constraint that "the number of second target dimensions + the number of third target dimensions = the number of original second impact dimension feature dimensions (4 + 3 = 7)." The server repeats this division for all sub-regions, including the dam foundation and dam heel, to form a target impact dimension feature set for the entire dam body. First target impact association network: For the upstream dam surface stress coordination target feature (A1') and the dam crest stress conduction target feature (B1') of the first target impact dimension feature set, the server extracts the historical safety quantification sequences of the two (such as the (A1') sequence ([1.2, 1.3, 1.1]) and the (B1') sequence ([1.1, 1.2, 1.0])). The degree of association is calculated using the mutual information algorithm (the mutual information entropy is (0.7), which is higher than the strong association threshold (0.6)). The "strong association-bidirectional action" relationship is determined. Each first target feature is taken as an entity unit, and a directed link is constructed according to the association relationship (a bidirectional arrow is set at (A1'-B1') to represent the cross-regional conduction of stress safety response). This forms a network to quantify the interaction pattern of the core characteristics of structural mechanics. Second target impact association network: For the second target impact dimension feature set of the dam foundation seepage pressure target feature (D2') and the downstream dam surface seepage conduction target feature (E2'), the server determines the "strong hydraulic adjacency-unidirectional conduction" relationship based on the physical spatial location of the monitoring data group ((D2') corresponds to the dam foundation rock layer (z=-5m), (E2') corresponds to the downstream dam surface (z=3m)) and the dam body anti-seepage design (drainage hole group connection); each second target feature is regarded as an entity unit, and a directed link is constructed according to the spatial adjacency relationship ((D2'toE2') sets a unidirectional arrow to represent the spatial transmission of seepage safety pressure), forming a network to present the safety conduction law of the basic characteristics of environmental geology.The third target impact association network: For the third target impact dimension feature set, the dam heel rock weathering target feature (F3') and the dam foundation interior weathering linkage target feature (G3'), the server first identifies adjacent target features using spatial coordinates ((F3') dam heel (z=2m), (G3') dam foundation interior (z=-3m)). The server then extracts the safety quantitative series of rock weathering for both features and calculates the degree of correlation using the Pearson correlation coefficient (a correlation coefficient of 0.8 indicates a strong correlation), determining a "strong weathering linkage-bidirectional interaction" relationship. Each third target feature is treated as an entity unit, and directed links are constructed based on the adjacent association relationship (with bidirectional arrows from (F3'-G3') representing the cross-interface diffusion of weathering safety effects). This network is formed to characterize the safety interaction mechanism of environmental geological extension characteristics. First coupled target feature generation: For (A1') of the first target impact association network, its associated entity feature is (B1'), etc. The server calculates the mean of the associated entity features (e.g., (B1') mean (1.1)) to obtain the first associated region mean feature. It calculates the difference between (A1') and the associated entity features (e.g., (A1') value (1.2), difference (0.1)) to obtain the first difference feature. (A1'), the first difference feature, and the first associated region mean feature are input into the fusion model, outputting the first coupled target feature corresponding to (A1') (e.g., (1.15)). Second coupled target feature generation: For the second target impact associated network (D2'), its associated entity feature is (E2'), etc. The server calculates the mean of the associated entity features (e.g., (E2') mean (1.2)) to obtain the second associated region mean feature. It calculates the difference between (D2') and the associated entity features (e.g., (D2') value (1.3), difference (0.1)) to obtain the second difference feature. (D2'), the second difference feature, and the second associated region mean feature are input into the coupling model, outputting the second coupled target feature corresponding to (D2') (e.g., (1.25)). Third coupled target feature generation: For (F3') of the third target impact association network, its associated entity feature is (G3'), etc. The server calculates the mean of the associated entity features (e.g., (G3') mean (1.0)) to obtain the third associated region mean feature; calculates the difference between (F3') and the associated entity features (e.g., (F3') value (1.1), difference (0.1)) to obtain the third difference feature; inputs (F3'), the third difference feature, and the third associated region mean feature into the interaction model, outputting the third coupling target feature corresponding to (F3') (e.g., (1.05)). Fusion of current sub-region safety impact features: Taking the upstream dam face sub-region as an example, its first coupling target feature (1.15) (structural core), second coupling target feature (1.25) (environmental foundation), and third coupling target feature (1.05) (environmental extension) need to be fused.The server performs a weighted summation based on the engineering weights (structural core (0.4), environmental foundation (0.3), and environmental extension (0.3)): (1.15×0.4+1.25×0.3+1.05×0.3=0.46+0.375+0.315=1.15), which is the current sub-area security impact characteristic of the sub-area. The server uses preset threshold rules (([0,0.5)) "safe", ([0.5,1.0)) "warning", (1.0,2.0]) "dangerous") to evaluate the current safety impact characteristics of each sub-area one by one. If the current characteristic of the upstream dam surface (1.15) is "warning", the current characteristic of the dam foundation sub-area (1.22) (hypothetical value) reaches "danger", and most key sub-areas such as the dam heel and dam crest are in the "warning-dangerous" range, the server integrates the results of the entire dam body and outputs the current safety assessment result as "the structural safety status of the hydraulic structure is in the critical range of warning-danger, and it is recommended to initiate special safety inspections and intensified monitoring." If most sub-areas fall back into the "safe-warning" range, the output is "the structural safety risk is controllable; maintain regular monitoring frequency and strengthen inspections in key areas." Through the deep coupling and iterative judgment of multi-dimensional target characteristics, the accuracy and reliability of the safety assessment conclusions are improved.

[0110] In the embodiments of the present invention, the following implementation modes are also provided.

[0111] Inputting the hydraulic structure structure monitoring data set into a structural safety level determination model, dividing the hydraulic structure structure monitoring data set using the structural safety level determination model to obtain each monitoring data group, and extracting features of each monitoring data group to obtain safety impact features of each sub-region;

[0112] The safety impact characteristics of each sub-region are divided respectively by the structural safety level determination model to obtain first impact dimension characteristics corresponding to each sub-region safety impact characteristic, forming a first impact dimension feature set, and second impact dimension characteristics corresponding to each sub-region safety impact characteristic are obtained to form a second impact dimension feature set;

[0113] Establishing a first impact association network corresponding to the first impact dimension feature set using the correlation density between each first impact dimension feature in the first impact dimension feature set through the structural safety level determination model, and establishing a second impact association network corresponding to the second impact dimension feature set based on the physical spatial position of each monitoring data group;

[0114] The structural safety level determination model uses the first impact dimension feature in the first impact association network and the associated entity feature corresponding to the first impact dimension feature to perform feature coupling to obtain a first comprehensive evaluation feature corresponding to each first impact dimension feature in the first impact dimension feature set, and performs feature coupling based on the second impact dimension feature in the second impact association network and the associated entity feature corresponding to the second impact dimension feature to obtain a second comprehensive evaluation feature corresponding to each second impact dimension feature in the second impact dimension feature set;

[0115] The structural safety level determination model is used to fuse the first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-region safety impact feature to obtain the target safety assessment feature corresponding to each sub-region safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-region safety impact feature to obtain the output safety assessment result corresponding to the hydraulic structure structure monitoring data set.

[0116] In an embodiment of the present invention, illustratively, the server inputs a data set of hydraulic structure monitoring of a concrete gravity dam (including multi-source monitoring data such as stress, displacement, seepage, temperature, and water level of various parts of the dam body) into a pre-trained structural safety level determination model, and the model performs evaluation according to a modular process: the model calls the "data division and feature extraction" module, and divides the monitoring data of the entire dam body into sub-region data groups such as the upstream dam surface monitoring data group (including stress sensors, displacement meters, and seepage flow meter data in this area), and the dam foundation monitoring data group (including bedrock stress gauges, seepage pressure gauges, and settlement meter data) based on the dam structure design drawings (such as dam section division and anti-seepage area boundaries). Taking the upstream dam surface monitoring data set as an example, the model extracts basic monitoring features such as peak stress time history (e.g., stress reaching 35 MPa on a certain day), daily cumulative displacement increment (e.g., displacement increased by 0.2 mm compared to the previous day), and seepage rate (e.g., real-time seepage (0.5 L / s)). Furthermore, the "spatial encoding" submodule quantifies the position of the upstream dam surface in a three-dimensional coordinate system (e.g., (x = 10 m, y = 20 m, z = 5 m)) into a spatial feature vector. Finally, the basic monitoring features are combined with the spatial features to generate a sub-regional safety impact feature for the upstream dam surface (a 15-dimensional vector encompassing mechanical, deformation, seepage, and spatial information). This process is repeated for all sub-regions, including the dam foundation, heel, and crest, outputting a set of sub-regional safety impact features for the entire dam body. The model utilizes the "Dimension Splitting" module. Based on predefined "structural mechanics-environmental geology" labels (e.g., stress and settlement are classified as structural mechanics, and seepage, weathering, and water level are classified as environmental geology), the model splits the safety impact characteristics of each sub-area into: first-dimension features (structural mechanics, such as upstream dam face stress gradient and dam foundation settlement and deformation characteristics), which form the first-dimension feature set; and second-dimension features (environmental geology, such as dam foundation seepage pressure and dam heel rock weathering characteristics), which form the second-dimension feature set. For the upstream dam face sub-area, for example, its safety impact characteristics are split into three-dimensional first-dimension features (stress synergy, displacement gradient, and fracture acoustic characteristics) and four-dimensional second-dimension features (seepage pressure, temperature load, rock mass moisture, and water level response characteristics), ensuring that both feature dimensions cover the full dimensionality of the original safety impact characteristics. The model calls the "network construction" module and executes it in two steps: the first impact association network: for the first impact dimension feature set (such as the upstream dam surface stress gradient feature (A) and the dam top stress conduction feature (B)), the model uses the "mutual information algorithm" to calculate the degree of correlation between feature pairs (such as the mutual information entropy (0.7) between (A) and (B), which determines a strong correlation), and determines the "strong correlation-two-way action" relationship; each first impact dimension feature is taken as an entity unit, and a directed link is constructed according to the correlation relationship ((A-B) is set with a two-way arrow), forming a first impact association network to quantify the interaction pattern of structural mechanical characteristics.Second impact association network: For the second impact dimension feature set (such as the dam foundation seepage pressure characteristics (D) and the downstream dam surface seepage conduction characteristics (E)), the model determines the "strong hydraulic adjacency-unidirectional conduction" relationship through the "spatial coordinate distance + engineering connectivity rule"; each second impact dimension feature is regarded as an entity unit, and a directed link is constructed according to the spatial adjacency relationship ((DtoE) with a unidirectional arrow), forming a second impact association network, which presents the spatial conduction law of environmental geological characteristics. The model calls the "feature coupling" module and processes the two types of networks separately: the first comprehensive evaluation feature: for (A) of the first influencing association network, the model calculates the mean of its associated entity features ((B) etc.) (such as the mean of (B) (32MPa), that is, the mean feature of the first associated area), the difference between (A) and the associated entity (such as (A) value (35MPa), the difference (3MPa), that is, the first difference feature); (A), the first difference feature, and the mean feature of the first associated area are input into the "fusion-linear conversion" submodule, and the first comprehensive evaluation feature corresponding to (A) is output (such as the quantitative score is (2.8)). Second comprehensive assessment feature: For (D) in the second impact association network, the model calculates the mean of its associated entity features ((E), etc.) (e.g., the mean of (E) is 20kPa, i.e., the mean feature of the second associated region), and the difference between (D) and its associated entities (e.g., the value of (D) is 22kPa, the difference is 2kPa, i.e., the second difference feature). (D), the second difference feature, and the mean feature of the second associated region are input into the isomorphic submodule, and the second comprehensive assessment feature corresponding to (D) is output (e.g., the quantitative score is (0.8)). The model calls the "Multi-dimensional Fusion and Judgment" module: Target safety assessment feature fusion: Taking the upstream dam face sub-region as an example, its first comprehensive assessment feature (2.8) (structural mechanics) and the second comprehensive assessment feature (0.8) (environmental geology) are weighted and fused according to the engineering weights (structure (0.6), environment (0.4)), and the calculated value is (2.8×0.6+0.8×0.4=2), generating the target safety assessment feature. Safety level determination: The model calls the “threshold classification” submodule and determines that the upstream dam surface target feature (2) is “dangerous” according to the preset rules (([0,0.5)) “safe”, ([0.5,1.0)) “warning”, ([1.0,2.0]) “dangerous”); after repeated determination of the dam foundation, dam heel and other sub-areas, the results of the entire dam body are integrated to output the safety assessment results (such as “the concrete gravity dam structure has serious safety hazards and emergency maintenance needs to be started immediately”).

[0117] Through the modular pipeline operation of the structural safety level determination model, the server realizes the automatic mapping from multi-source monitoring data to safety level conclusions, providing efficient and accurate technical support for the structural safety management of hydraulic structures.

[0118] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned hydraulic structure safety assessment method based on multi-source information fusion. Figure 2 As shown, Figure 2 This is a block diagram of the structure of a computer device 100 provided in an embodiment of the present invention. Computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or exchange, memory 111, processor 112, and communication unit 113 are electrically connected to each other, directly or indirectly. For example, these components can be electrically connected via one or more communication buses or signal lines.

[0119] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in light of the above teachings. These embodiments have been selected and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the present disclosure and to utilize various embodiments with various modifications as appropriate for the specific application contemplated.

Claims

1. A hydraulic structure structural safety assessment method based on multi-source information fusion, characterized in that: include: Acquiring a hydraulic structure structure monitoring data set of the hydraulic structure to be evaluated, dividing the hydraulic structure structure monitoring data set to obtain a monitoring data group for each structural sub-area, and extracting features of each monitoring data group to obtain a safety impact feature of each sub-area; Dividing the security impact characteristics of each sub-region respectively, obtaining first impact dimension characteristics corresponding to each sub-region security impact characteristic, forming a first impact dimension feature set, and obtaining second impact dimension characteristics corresponding to each sub-region security impact characteristic, forming a second impact dimension feature set; Establishing a first influence association network corresponding to the first influence dimension feature set according to the degree of association between each first influence dimension feature in the first influence dimension feature set, and establishing a second influence association network corresponding to the second influence dimension feature set according to the physical spatial position of each monitoring data group; Perform feature coupling based on the first influence dimension feature and the associated entity feature corresponding to the first influence dimension feature in the first influence association network to obtain a first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set; and perform feature coupling based on the second influence dimension feature and the associated entity feature corresponding to the second influence dimension feature in the second influence association network to obtain a second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set; The first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-area safety impact feature are fused to obtain the target safety assessment feature corresponding to each sub-area safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-area safety impact feature to obtain the safety assessment result corresponding to the hydraulic structure structure monitoring data set.

2. The method according to claim 1, characterized in that Extracting the features of each monitoring data group to obtain the safety impact features of each sub-region includes: Extracting the basic monitoring features of each monitoring data group to obtain the basic monitoring features of each monitoring data group; Obtaining the physical spatial position of each monitoring data group, quantizing and encoding the physical spatial position of each monitoring data group, and obtaining spatial features of each monitoring data group; The basic monitoring features of each monitoring data group and the corresponding spatial features of the monitoring data group are integrated to obtain the safety impact features of each sub-region.

3. The method according to claim 1, characterized in that The step of establishing a first influence association network corresponding to the first influence dimension feature set according to the degree of association between each first influence dimension feature in the first influence dimension feature set includes: Calculating the degree of feature correlation between each of the first impact dimension features, and determining the correlation relationship between each of the first impact dimension features according to the degree of feature correlation; Each of the first impact dimension features is taken as an entity unit, and an action link is constructed for each of the first impact dimension features according to the association relationship to obtain the first impact association network.

4. The method according to claim 1, wherein The establishing of a second impact association network corresponding to the second impact dimension feature set according to the physical spatial position of each monitoring data group includes: Determine the spatial topology information of the second impact dimension features corresponding to each sub-region security impact feature according to the physical spatial position of each monitoring data group, and determine the structural adjacency relationship between each second impact dimension feature in the second impact dimension feature set according to the spatial topology information; Each of the second impact dimension features is taken as an entity unit, and an action link is constructed for each of the second impact dimension features according to the structural adjacency relationship to obtain the second impact association network.

5. The method according to claim 1, characterized in that The performing feature coupling based on the first influence dimension feature and the associated entity feature corresponding to the first influence dimension feature in the first influence association network to obtain the first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set includes: Calculating the associated region mean feature of the associated entity feature corresponding to the first impact dimension feature to obtain a first associated region mean feature, and calculating the difference between the first impact dimension feature and the associated entity feature corresponding to the first impact dimension feature to obtain a first difference feature; Fusing the first impact dimension feature, the first difference feature, and the first associated region mean feature to obtain a first fused feature, and performing a linear transformation based on the first fused feature to obtain a first comprehensive evaluation feature corresponding to the first impact dimension feature; For each first influence dimension feature in the first influence association network, a first comprehensive evaluation feature corresponding to each first influence dimension feature in the first influence dimension feature set is obtained.

6. The method according to claim 1, characterized in that The step of performing feature coupling based on the second influence dimension feature and the associated entity feature corresponding to the second influence dimension feature in the second influence association network to obtain the second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set includes: Calculating the associated region mean feature of the associated entity feature corresponding to the second impact dimension feature to obtain a second associated region mean feature, and calculating the difference between the second impact dimension feature and the associated entity feature corresponding to the second impact dimension feature to obtain a second difference feature; Fusing the second impact dimension feature, the second difference feature, and the second associated region mean feature to obtain a second fused feature, and performing a linear transformation based on the second fused feature to obtain a second comprehensive evaluation feature corresponding to the second impact dimension feature; For each second influence dimension feature in the second influence association network, obtain the second comprehensive evaluation feature corresponding to each second influence dimension feature in the second influence dimension feature set.

7. The method according to claim 1, characterized in that The first comprehensive assessment feature and the second comprehensive assessment feature corresponding to the same sub-region safety impact feature are fused to obtain the target safety assessment feature corresponding to each sub-region safety impact feature, and the structural safety level is determined based on the target safety assessment feature corresponding to each sub-region safety impact feature to obtain the safety assessment result corresponding to the hydraulic structure structure monitoring data set, including: Obtain a first feature enhancement weight, perform feature mapping on the first comprehensive evaluation feature corresponding to each first impact dimension feature in the first impact dimension feature set according to the first feature enhancement weight, obtain a first mapping conversion feature corresponding to each first impact dimension feature in the first impact dimension feature set, and calculate a feature deviation value corresponding to the first mapping conversion feature to obtain a first feature deviation value; Perform weighted fusion on the first comprehensive evaluation features corresponding to each first impact dimension feature in the first impact dimension feature set to obtain the first weighted fusion features corresponding to each first impact dimension feature in the first impact dimension feature set; Calculate the multiplication result of the first weighted fusion feature and the first feature deviation value to obtain the first weighted enhancement feature corresponding to each first impact dimension feature in the first impact dimension feature set; Perform feature mapping on the second comprehensive evaluation feature corresponding to each second impact dimension feature in the second impact dimension feature set according to the first feature enhancement weight, obtain a second mapping conversion feature corresponding to each second impact dimension feature in the second impact dimension feature set, and calculate a feature deviation value corresponding to the second mapping conversion feature to obtain a second feature deviation value; Perform weighted fusion on the second comprehensive evaluation features corresponding to each second impact dimension feature in the second impact dimension feature set to obtain second weighted fusion features corresponding to each second impact dimension feature in the second impact dimension feature set; Calculate the multiplication result of the second weighted fusion feature and the second feature deviation value to obtain the second weighted enhancement feature corresponding to each second impact dimension feature in the second impact dimension feature set; The first weighted enhancement feature and the second weighted enhancement feature corresponding to the same sub-area safety impact feature are fused to obtain the enhanced safety impact features corresponding to each sub-area safety impact feature, and the structural safety level is determined based on the enhanced safety impact features corresponding to each sub-area safety impact feature to obtain the target safety assessment result corresponding to the hydraulic structure structure monitoring data set.

8. The method according to claim 7, characterized in that The determining of the structural safety level according to the enhanced safety impact characteristics corresponding to the safety impact characteristics of each sub-region, and obtaining the target safety assessment result corresponding to the hydraulic structure structure monitoring data set, includes: Dividing the enhanced security impact features corresponding to each sub-region security impact feature to obtain a first enhanced impact dimension feature set, a second enhanced impact dimension feature set, and a third enhanced impact dimension feature set, wherein the sum of the number of dimensions of the second enhanced impact dimension feature in the second enhanced impact dimension feature set and the number of dimensions of the third enhanced impact dimension feature in the third enhanced impact dimension feature set is the same as the number of dimensions of the second impact dimension feature; Establish a first reinforcement influence association network corresponding to the first reinforcement influence dimension feature set according to the degree of correlation between each first reinforcement influence dimension feature in the first reinforcement influence dimension feature set, and establish a second reinforcement influence association network corresponding to the second reinforcement influence dimension feature set according to the physical spatial position of each monitoring data group; Determine, according to the physical spatial position of each monitoring data group, adjacent strengthening influence dimension features corresponding to each third strengthening influence dimension feature in the third strengthening influence dimension feature set, and select features to be analyzed and reference features from each third strengthening influence dimension feature; Determine, from each third enhanced impact dimension feature according to the physical spatial position of each monitoring data group, each adjacent feature to be analyzed corresponding to the feature to be analyzed, and perform feature integration on each adjacent feature to be analyzed to obtain a fused adjacent feature to be analyzed; Determine each target neighboring feature corresponding to the reference feature from each third enhanced impact dimension feature according to the physical spatial position of each monitoring data group, perform feature integration on each target neighboring feature, and obtain a target integrated neighboring feature; Calculating the degree of correlation between the to-be-analyzed fused adjacent feature and the target integrated adjacent feature, and obtaining the degree of correlation between the to-be-analyzed feature and the reference feature; For each of the third reinforcement influence dimension features, obtain the degree of correlation between the adjacent reinforcement influence dimension features corresponding to each of the third reinforcement influence dimension features, and use the degree of correlation between the adjacent reinforcement influence dimension features corresponding to each of the third reinforcement influence dimension features as the target degree of correlation between each of the third reinforcement influence dimension features; Determine the standard action relationship between each of the third reinforcement influence dimension features according to the target association closeness, and use each of the third reinforcement influence dimension features as an entity unit, and construct an action link for each of the third reinforcement influence dimension features according to the standard action relationship to obtain a third reinforcement influence association network; Perform feature coupling based on the first reinforcement influence dimension feature and the associated entity feature corresponding to the first reinforcement influence dimension feature in the first reinforcement influence association network to obtain first coupled reinforcement features corresponding to each first reinforcement influence dimension feature in the first reinforcement influence dimension feature set; Perform feature coupling based on the associated entity features corresponding to the second reinforcement influence dimension features and the second reinforcement influence dimension features in the second reinforcement influence association network, to obtain second coupled reinforcement features corresponding to each second reinforcement influence dimension feature in the second reinforcement influence dimension feature set; Perform feature coupling based on the third reinforcement influence dimension feature and the associated entity features corresponding to the third reinforcement influence dimension features in the third reinforcement influence association network, to obtain third coupling reinforcement features corresponding to each third reinforcement influence dimension feature in the third reinforcement influence dimension feature set; Obtaining a second feature enhancement weight, and weightedly enhancing the first coupling enhancement feature corresponding to each first enhancement influence dimension feature in the first enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a first enhancement gain feature corresponding to each first enhancement influence dimension feature in the first enhancement influence dimension feature set; Perform weighted enhancement on the second coupling enhancement feature corresponding to each second enhancement influence dimension feature in the second enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a second enhancement gain feature corresponding to each second enhancement influence dimension feature in the second enhancement influence dimension feature set; Perform weighted enhancement on the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set according to the second feature enhancement weight, to obtain a third enhancement gain feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set; Fusing the first enhancement gain feature, the second enhancement gain feature, and the third enhancement gain feature corresponding to the same sub-region security impact feature to obtain the target enhancement security impact feature corresponding to each sub-region security impact feature; A structural safety level is determined according to the target enhanced safety impact characteristics corresponding to each sub-region safety impact characteristic, and an enhanced safety assessment result corresponding to the hydraulic structure structure monitoring data set is obtained.

9. The method according to claim 8, characterized in that The weighted enhancement of the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set according to the second feature enhancement weight to obtain the third enhancement gain feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set includes: Perform feature mapping on the third coupling enhancement feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set according to the second feature enhancement weight, obtain a third mapping conversion feature corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set, and calculate a feature deviation value corresponding to the third mapping conversion feature to obtain a third feature deviation value; Perform weighted fusion on the second coupling enhancement features corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set to obtain third weighted fusion features corresponding to each third enhancement influence dimension feature in the third enhancement influence dimension feature set; Calculate the multiplication result of the third weight fusion feature and the third feature deviation value to obtain the third enhanced gain feature corresponding to each third enhanced influence dimension feature in the third enhanced influence dimension feature set.

10. A server system, characterized in that: The method comprises a server, wherein the server is configured to execute the method according to any one of claims 1 to 9.

Images