A method, system, device and medium for building whole life cycle management of a water conservancy project

Abstract

This application relates to a method, system, equipment, and medium for the full life-cycle management of water conservancy projects. The method includes: acquiring and spatiotemporally aligning and fusing multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data; inputting the standard spatiotemporal monitoring data into an attention-based spatiotemporal feature extraction network to obtain a multidimensional state feature vector; inputting the multidimensional state feature vector into a digital twin simulation model for virtual-real mapping and error correction to obtain a dynamic health assessment index; performing time-series evolution analysis on the dynamic health assessment index to obtain a full-cycle performance evolution prediction curve; matching a set of candidate maintenance schemes from an engineering maintenance knowledge graph based on the full-cycle performance evolution prediction curve; performing multi-objective optimization decision-making on the set of candidate maintenance schemes; and generating a full life-cycle management strategy. This method can improve the scientific decision-making ability and resource utilization efficiency of water conservancy project management, effectively ensuring the operational safety of the project throughout its entire life-cycle.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy engineering, and in particular relates to a method, system, equipment and medium for the whole life cycle management of water conservancy projects. Background Technology

[0002] With the continuous expansion of the scale of water conservancy projects and the increasing service life, full life-cycle health management of structures has become a key means to ensure the safe operation of these projects. Currently, water conservancy project monitoring technology is widely used. By deploying sensor networks at key locations and combining them with regular manual inspections, multi-source data such as structural response data, environmental hydrological information, and construction imagery can be acquired, providing fundamental information support for project condition assessment. However, existing monitoring methods often employ independent data acquisition systems, resulting in significant differences in time synchronization, spatial coordinate systems, and data structures among various data sources. This makes it difficult to effectively integrate and uniformly represent multi-source information, leading to a one-sided understanding of the true physical state of the project.

[0003] Traditional engineering health assessment methods often rely on threshold judgments for single monitoring indicators or static analysis based on simplified numerical simulations. These methods lack the ability to explore the correlations between multi-dimensional characteristics such as concrete cracking, metal structural stress, and seepage field evolution, failing to reveal the coupling relationship between apparent structural damage and intrinsic mechanical properties. This makes it difficult for health assessment results to comprehensively reflect the overall safety reserves of the project. Furthermore, existing performance prediction models are usually based on empirical formulas or fixed degradation assumptions, failing to fully utilize dynamic monitoring data for real-time model correction, resulting in insufficient accuracy in predicting future structural reliability trends.

[0004] In the maintenance decision-making process, the current common approach is to formulate periodic maintenance plans based on expert experience, or to take remedial measures only when obvious safety hazards occur. This model is highly subjective and lacks a comprehensive consideration of multiple objectives such as life-cycle costs, project duration impact, and implementation risks, making it difficult to optimize resource allocation while ensuring structural safety. Therefore, there is an urgent need for a life-cycle management method for water conservancy projects that can integrate multi-source heterogeneous monitoring data, achieve accurate health assessment and dynamic performance prediction, and support intelligent maintenance decision-making. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, system, equipment, and medium for the full life-cycle management of water conservancy projects to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for the full life-cycle management of water conservancy projects, including:

[0007] S1. Acquire multi-source heterogeneous monitoring data, and perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time; wherein, the standard spatiotemporal monitoring data is used to characterize the physical state information of water conservancy projects in different spatiotemporal dimensions.

[0008] S2. Input the standard spatiotemporal monitoring data into the spatiotemporal feature extraction network based on the attention mechanism to obtain the multidimensional state feature vector at each monitoring sampling time.

[0009] S3. Input the multidimensional state feature vector into the digital twin simulation model to obtain the dynamic health assessment index at each monitoring sampling time; whereby the dynamic health assessment index is used to characterize the structural safety reserve coefficient of the water conservancy project under the current working conditions.

[0010] S4. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a time-series evolution analysis is performed on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve; wherein, the full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a future preset time span.

[0011] S5. Based on the full-cycle performance evolution prediction curve, retrieve the set of candidate maintenance schemes from the engineering maintenance knowledge graph, and make multi-objective optimization decisions based on the set of candidate maintenance schemes to generate a full life cycle management strategy.

[0012] In one embodiment, S1 includes:

[0013] S11. Collect structural response sensor data, environmental hydrological monitoring data and construction progress image data during the water conservancy project process to obtain multi-source heterogeneous monitoring data;

[0014] S12. Perform time synchronization calibration on the multi-source heterogeneous monitoring data to obtain time-aligned monitoring data;

[0015] S13. Spatial coordinate unification is performed on the time alignment monitoring data to obtain spatiotemporal alignment monitoring data;

[0016] S14. Perform heterogeneous data fusion on the spatiotemporal alignment monitoring data to obtain standard spatiotemporal monitoring data.

[0017] In one embodiment, the spatiotemporal feature extraction network based on the attention mechanism includes a spatiotemporal location encoding layer, a multi-head attention mechanism layer, a temporal convolutional network layer, and a gated recurrent unit layer, S2 including:

[0018] S21. Input the standard spatiotemporal monitoring data into the spatiotemporal location coding layer to generate a feature vector sequence with spatiotemporal location information;

[0019] S22. The spatial dependency relationship of the feature vector sequence is modeled through a multi-head attention mechanism layer to obtain a spatial attention feature map;

[0020] S23. Extract temporal dimension features from the spatial attention feature map through a temporal convolutional network layer to obtain a spatiotemporal joint feature matrix;

[0021] S24. Based on the gated cyclic unit layer, perform correlation analysis of the spatiotemporal joint feature matrix on cracks, stress and seepage, and output multidimensional state feature vectors; among which, the multidimensional state feature vectors are used to characterize the crack development trend, stress distribution state and seepage evolution characteristics of hydraulic engineering.

[0022] In one embodiment, S3 includes:

[0023] S31. An initial digital twin simulation model is constructed based on finite element numerical simulation. The initial digital twin simulation model includes geometric parameters, material parameters, and boundary conditions.

[0024] S32. Using the multidimensional state feature vectors at each monitoring sampling time as observation data, the material parameters and boundary conditions of the initial digital twin simulation model are corrected online to obtain the corrected digital twin simulation model.

[0025] S33. Based on the modified digital twin model, calculate the ratio of the current ultimate bearing capacity of the structure to the actual load effect to obtain the dynamic health assessment index at each monitoring sampling time.

[0026] In one embodiment, S4 includes:

[0027] S41. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a material performance degradation function is fitted based on the dynamic health assessment index sequence to obtain a degradation model; wherein, the expression of the material performance degradation function is:

[0028]

[0029] In the formula, express Dynamic health assessment index at any time, This represents the initial dynamic health assessment index. Represents the degradation rate coefficient. Indicates the degenerate shape parameter, Indicates time;

[0030] S42. Based on the degradation model, predict the dynamic health assessment index sequence and calculate the predicted health indicator sequence within a future preset time span.

[0031] S43. Based on the predicted health indicator sequence and the preset failure threshold, identify the critical time node when the predicted health indicator is lower than the failure threshold, and obtain potential failure mode information.

[0032] S44. Based on potential failure mode information, curve fitting and smoothing are performed on the predicted health index sequence to generate a full-cycle performance evolution prediction curve.

[0033] In one embodiment, S5 includes:

[0034] S51. Based on the full-cycle performance evolution prediction curve, extract the key performance degradation time nodes and the degradation type labels corresponding to the full-cycle performance evolution prediction curve, and construct the maintenance requirement features based on the key performance degradation time nodes and degradation type labels.

[0035] S52. Using maintenance requirement characteristics as an index, retrieve historical maintenance cases from the engineering maintenance knowledge graph to obtain a set of candidate maintenance solutions;

[0036] S53. Based on a multi-objective optimization function aimed at minimizing the total lifecycle maintenance cost and maximizing structural reliability, the set of candidate maintenance schemes is optimized to obtain the optimal maintenance strategy parameters; wherein, the expression of the multi-objective optimization function is:

[0037]

[0038] in, This represents the objective function value of a multi-objective optimization function. Represents decision variables, This represents the total cost of maintenance throughout the entire lifecycle. Indicates structural reliability. This represents the cost weighting coefficient. This represents the reliability weighting coefficient;

[0039] S54. Generate a full lifecycle management strategy based on the optimal maintenance strategy parameters.

[0040] In one embodiment, the gated loop unit layer includes a crack feature extraction branch, a stress feature extraction branch, and a seepage feature extraction branch, and S24 includes:

[0041] S241. Input the spatiotemporal joint feature matrix into the crack feature extraction branch, stress feature extraction branch and seepage feature extraction branch respectively, and generate crack feature vector, stress feature vector and seepage feature vector respectively;

[0042] S242. Using crack feature vectors, stress feature vectors, and seepage feature vectors as node features, an adjacency matrix is ​​constructed based on the spatial adjacency relationship of hydraulic engineering structural units. Information transmission and feature fusion are performed through a graph convolutional network to obtain a coupled feature graph.

[0043] S243. Perform global feature aggregation on the coupled feature map to generate a multi-dimensional state feature vector.

[0044] Secondly, this application also provides a building lifecycle management system for water conservancy projects, including:

[0045] The data preprocessing module is used to acquire multi-source heterogeneous monitoring data, and to perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time; among which, the standard spatiotemporal monitoring data is used to characterize the physical state information of water conservancy projects in different spatiotemporal dimensions.

[0046] The feature extraction module is used to input standard spatiotemporal monitoring data into a spatiotemporal feature extraction network based on an attention mechanism to obtain multidimensional state feature vectors for each monitoring sampling time.

[0047] The health assessment module is used to input multidimensional state feature vectors into the digital twin simulation model to obtain the dynamic health assessment index at each monitoring sampling time; the dynamic health assessment index is used to characterize the structural safety reserve coefficient of the water conservancy project under the current working conditions.

[0048] The evolution prediction module is used to construct a dynamic health assessment index sequence based on the dynamic health assessment index at each monitoring sampling time, and to perform time-series evolution analysis on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve. The full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a preset time span in the future.

[0049] The strategy decision-making module is used to retrieve a set of candidate maintenance solutions from the engineering maintenance knowledge graph based on the full life cycle performance evolution prediction curve, and to make multi-objective optimization decisions based on the set of candidate maintenance solutions to generate a full life cycle management strategy.

[0050] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0051] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0052] The aforementioned method, system, equipment, and medium for the full life-cycle management of water conservancy projects, by integrating multi-source heterogeneous monitoring data and unifying spatiotemporal benchmarks, can achieve comprehensive perception and standardized characterization of the physical state of water conservancy projects. Employing an attention-based spatiotemporal feature extraction network to deeply mine and correlate multi-dimensional features such as cracks, stress, and seepage overcomes the limitations of traditional methods that rely on single indicators and struggle to reveal multi-field coupling patterns. Furthermore, it leverages online correction and dynamic health assessment using digital twin simulation models to achieve a leap from static assessment to real-time accurate diagnosis. By constructing a performance degradation model through temporal evolution analysis of health indices, it can scientifically predict future reliability trends of structures, providing a basis for proactive maintenance. Finally, combining engineering maintenance knowledge graphs and multi-objective optimization decision-making, it generates a full life-cycle management strategy that balances cost, schedule, and risk, improving the intelligence level, scientific decision-making capabilities, and resource utilization efficiency of water conservancy project management, effectively ensuring the operational safety of the project throughout its entire life-cycle. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating a method for the full life-cycle management of a water conservancy project, as described in one embodiment.

[0055] Figure 2 This is a schematic diagram of the structure of a building life cycle management system for a water conservancy project in one embodiment;

[0056] Figure 3 This is a schematic diagram of a computer device in one embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] In one embodiment, reference Figure 1The document presents a flowchart illustrating a method for the full life-cycle management of water conservancy projects provided in this application. This embodiment uses the application of this method to an engineering management terminal (hereinafter referred to as the terminal) as an example. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] S1. Acquire multi-source heterogeneous monitoring data, and perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time.

[0060] For example, the engineering management terminal can connect to various monitoring devices and data storage terminals deployed at the water conservancy project site through a preset data interface. After obtaining data acquisition authorization, it can acquire multi-source heterogeneous monitoring data. Specifically, the multi-source heterogeneous monitoring data may include, but is not limited to: structural response sensor data collected by strain gauges, joint gauges, and rebar gauges embedded inside concrete dams or metal structures; environmental hydrological monitoring data collected by water level gauges, flow meters, rain gauges, and piezometers; and construction progress image data and surface defect image data periodically collected by drones, fixed cameras, or panoramic cameras.

[0061] For example, after acquiring multi-source heterogeneous monitoring data, the engineering management terminal can perform spatiotemporal benchmark unification on the multi-source heterogeneous monitoring data to obtain spatiotemporally aligned monitoring data. The time benchmark unification can adopt the Network Time Protocol (NTP). The engineering management terminal can calibrate the timestamps of all monitoring data to the same time benchmark based on NTP, eliminating data misalignment problems caused by sampling time differences between different monitoring devices. The spatial benchmark unification uses the geodetic coordinate system determined during the water conservancy project construction design stage as the standard, transforming and calibrating the spatial coordinates of data collected by different monitoring devices to ensure the consistency of various data in spatial location.

[0062] For example, after unifying the spatiotemporal reference of multi-source heterogeneous monitoring data, the engineering management terminal can perform feature-level fusion on the spatiotemporally aligned monitoring data, associating high-dimensional image features with low-dimensional sensor numerical features to generate structured standard spatiotemporal monitoring data with a unified spatiotemporal reference. This standard spatiotemporal monitoring data can be used to characterize the physical state information of water conservancy projects in different time and spatial dimensions.

[0063] S2. Input the standard spatiotemporal monitoring data into the spatiotemporal feature extraction network based on the attention mechanism to obtain the multidimensional state feature vector at each monitoring sampling time.

[0064] For example, after completing the data standardization process, the engineering management terminal can input the standard spatiotemporal monitoring data into a pre-trained deep learning network. The deep learning network can be a spatiotemporal feature extraction network based on an attention mechanism. The deep learning network can include a spatiotemporal location encoding layer, a multi-head attention mechanism layer, a temporal convolutional network layer, and a gated recurrent unit layer. The deep learning network can be used to automatically mine deep features related to structural health from the fused data.

[0065] Specifically, the attention-based spatiotemporal feature extraction network can map the input standardized data to a high-dimensional feature space through a spatiotemporal location encoding layer, generating a sequence of feature vectors with spatiotemporal location information. The multi-head attention mechanism in the attention-based spatiotemporal feature extraction network can simultaneously learn the stress-strain correlation between two monitoring points that are spatially far apart, obtaining spatial dependencies. After processing by the attention mechanism, a spatial attention feature map is obtained. The Temporal Convolutional Network (TCN) processes the spatial attention feature map, effectively expanding the receptive field without increasing the computational load through a dilated causal convolution structure, capturing the evolutionary patterns over long time sequences, extracting the temporal dimension variation features, and outputting a spatiotemporal joint feature matrix. The Gated Recurrent Unit (GRU) performs sequence modeling on the spatiotemporal joint feature matrix and performs collaborative analysis and feature decoupling on the three key physical processes of crack development, stress distribution, and seepage evolution, outputting a multidimensional state feature vector. Among these, the monitoring points can include, but are not limited to, the dam crest and dam foundation.

[0066] S3. Input the multidimensional state feature vector into the digital twin simulation model to obtain the dynamic health assessment index at each monitoring sampling time.

[0067] For example, the engineering management terminal can use multi-dimensional state feature vectors as key inputs to drive the operation of a digital twin simulation model, obtaining dynamic health assessment indices at each monitoring sampling time. The digital twin simulation model can be constructed based on finite element numerical simulation, according to engineering design drawings and geological survey reports. The digital twin simulation model can include geometric parameters, material parameters, and reasonable boundary conditions. Material parameters can include elastic modulus, Poisson's ratio, and permeability coefficient, etc.; boundary conditions can include reservoir water level pressure, temperature load, and foundation constraints, etc.

[0068] For example, the engineering management terminal uses the multidimensional state feature vector at the current moment as observation data and employs data assimilation algorithms, such as ensemble Kalman filtering, to perform online correction of the material parameters and boundary conditions of the initial model.

[0069] For example, if the observed stress feature vector shows an abnormal increase in stress in a certain area, the algorithm can automatically adjust the material stiffness parameters of that area to make the response of the simulation model approximate the real monitoring data, thereby obtaining a corrected digital twin simulation model that better reflects the current actual working conditions of the project.

[0070] For example, based on the modified digital twin simulation model, the engineering management terminal can perform nonlinear static and dynamic analysis to calculate the ultimate bearing capacity of the structure under the current material properties and load conditions. At the same time, the engineering management terminal can calculate the current actual load effect based on the actual load information contained in the multidimensional state feature vector. The ratio of the ultimate bearing capacity to the actual load effect is defined as the dynamic health assessment index, which can be used to quantify the safety reserve coefficient of the structure under the current working conditions.

[0071] S4. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a time-series evolution analysis is performed on the dynamic health assessment index sequence to obtain the full-cycle performance evolution prediction curve.

[0072] For example, the engineering management terminal can construct a dynamic health assessment index sequence based on the dynamic health assessment index at each monitoring sampling time, and perform time-series evolution analysis on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve. This full-cycle performance evolution prediction curve can be used to characterize the structural reliability change trend of a water conservancy project within a predetermined time span. The dynamic health assessment index sequence is time-series data obtained by arranging the dynamic health assessment indices at each monitoring sampling time in chronological order. This sequence can reflect the changing history of the engineering structure's safety status. The engineering management terminal can employ time-series analysis algorithms to perform stationarity checks and trend extraction on the index sequence, eliminating the influence of random noise on the analysis results. Then, based on the material performance degradation law, a material performance degradation function is fitted, which describes the degradation trend of the engineering materials over time. Based on this degradation function, the engineering management terminal predicts the engineering health status within a predetermined time span, obtaining a predicted health index sequence. Combined with a predetermined failure threshold, it identifies critical time nodes and potential failure modes where the project may fail. Finally, through curve fitting and smoothing, a full-cycle performance evolution prediction curve is generated.

[0073] S5. Based on the full-cycle performance evolution prediction curve, retrieve the set of candidate maintenance schemes from the engineering maintenance knowledge graph, and make multi-objective optimization decisions based on the set of candidate maintenance schemes to generate a full life cycle management strategy.

[0074] For example, the engineering maintenance knowledge graph is a pre-built knowledge base of the engineering management terminal. It can cover various maintenance cases, technologies, and costs related to water conservancy projects. The knowledge graph can be stored in a graph structure for easy retrieval and correlation analysis. The engineering management terminal can extract key performance degradation time nodes and corresponding degradation type labels from the full-cycle performance evolution prediction curve to construct standardized maintenance requirement features. Using these features as a retrieval index, a semantic matching algorithm is used to search for similar historical maintenance cases in the engineering maintenance knowledge graph, resulting in a set of candidate maintenance schemes that meet the current engineering maintenance requirements. Multi-objective optimization decision-making focuses on minimizing the full lifecycle maintenance cost and maximizing structural reliability. A multi-objective optimization function is constructed, and a non-dominated sorting genetic algorithm is used to optimize the candidate maintenance scheme set, balancing maintenance costs and structural safety to determine the optimal maintenance strategy parameters. This generates a full lifecycle management strategy including maintenance time, maintenance methods, and maintenance resource allocation, achieving intelligent and optimized engineering maintenance.

[0075] In the aforementioned method for the full life-cycle management of water conservancy projects, by unifying and fusing spatiotemporal benchmarks of multi-source monitoring data, effective integration of multi-dimensional data can be achieved, avoiding data misalignment and bias, and providing data support for project status assessment. By leveraging attention-based spatiotemporal feature extraction and digital twin simulation models, the correlation of multi-dimensional structural features can be accurately mined, enabling dynamic health assessment of the project and improving the comprehensiveness and accuracy of the assessment. By generating performance prediction curves through temporal evolution analysis, the accuracy of predicting future structural reliability can be improved. Based on knowledge graph retrieval and multi-objective optimization decision-making, the optimal full life-cycle management strategy can be generated, balancing maintenance costs and structural safety, achieving optimal resource allocation, thereby comprehensively ensuring the safe and stable operation of water conservancy projects and improving the intelligence and refinement of full life-cycle management.

[0076] In an optional embodiment, S1 includes:

[0077] S11. Collect structural response sensor data, environmental hydrological monitoring data, and construction progress image data during the water conservancy project process to obtain multi-source heterogeneous monitoring data.

[0078] Optionally, structural response sensor data can be collected by devices such as strain sensors, displacement sensors, and vibration sensors deployed on-site. This data can characterize the mechanical response states of the engineering structure, including stress, displacement, and vibration. The data is uploaded to the engineering management terminal in real time via a wireless transmission module. Environmental hydrological monitoring data is collected by devices such as hydrological monitoring stations, rain gauges, and water level gauges. This data can cover environmental parameters such as rainfall, water level, water temperature, and water quality, providing data support for analyzing the impact of environmental factors on the engineering structure. Construction progress image data can be collected and recorded by on-site monitoring cameras and drone aerial photography equipment, recording the entire construction process. This data can be used to trace construction quality and progress. The engineering management terminal can receive various types of data synchronously through a unified data acquisition interface, generating multi-source heterogeneous monitoring data to ensure data integrity and real-time performance.

[0079] S12. Perform time synchronization calibration on the multi-source heterogeneous monitoring data to obtain time-aligned monitoring data.

[0080] Optionally, the engineering management terminal can adopt the NTP protocol and use the system time of the engineering management terminal as a reference to calibrate the timestamps of various monitoring data. For data with inconsistent sampling frequencies, linear interpolation can be used to supplement missing data points to ensure that all monitoring data are synchronized in the time dimension. Each time node corresponds to complete multi-type monitoring data, thereby obtaining time-aligned monitoring data.

[0081] S13. Spatial coordinate unification is performed on the time alignment monitoring data to obtain spatiotemporal alignment monitoring data.

[0082] Optionally, the engineering management terminal can use the geodetic coordinate system determined during the construction and design phase of the water conservancy project as the standard. Through coordinate transformation algorithms, the spatial coordinates of various monitoring data can be transformed to the same coordinate system. At the same time, the deviation of the sensor installation position can be corrected to ensure that each set of monitoring data corresponds to the specific location of the engineering structure, thereby achieving dual alignment of monitoring data in time and space, and thus obtaining spatiotemporally aligned monitoring data.

[0083] S14. Perform heterogeneous data fusion on the spatiotemporal alignment monitoring data to obtain standard spatiotemporal monitoring data.

[0084] Optionally, the engineering management terminal can first perform standardized preprocessing on various types of data, using the min-max normalization method to map the data to the [0,1] interval to eliminate dimensional differences; then, an adaptive weighted fusion algorithm is used to assign weights according to indicators such as monitoring accuracy and data integrity of various types of data, and to perform fusion calculation on multiple types of data at the same spatiotemporal node to generate a single, comprehensive standard spatiotemporal monitoring data, which can comprehensively and accurately characterize the physical state of the project at that spatiotemporal node.

[0085] In an optional embodiment, the attention-based spatiotemporal feature extraction network includes a spatiotemporal location encoding layer, a multi-head attention mechanism layer, a temporal convolutional network layer, and a gated recurrent unit layer, where S2 includes:

[0086] S21. Input the standard spatiotemporal monitoring data into the spatiotemporal location coding layer to generate a feature vector sequence with spatiotemporal location information.

[0087] Optionally, the engineering management terminal can input standard spatiotemporal monitoring data into the spatiotemporal location coding layer, add spatiotemporal location coding to the data, transform spatiotemporal location information into learnable features, and fuse it with the features of the original monitoring data to generate a feature vector sequence with spatiotemporal location information, ensuring that the network can accurately capture the spatiotemporal correlation of the data.

[0088] S22. Spatial dependency modeling is performed on the feature vector sequence through a multi-head attention mechanism layer to obtain a spatial attention feature map.

[0089] Optionally, the multi-head attention mechanism layer can employ multiple attention heads for parallel computation. Each attention head can capture the spatial dependencies in the feature vector sequence from different perspectives. By calculating the correlation between features at different spatial locations, the multi-head attention mechanism layer can assign higher attention weights to features at key spatial locations, suppressing interference from irrelevant features. The engineering management terminal can use the computation of the multi-head attention mechanism layer to transform the feature vector sequence into a spatial attention feature map. This spatial attention feature map can be used to characterize the relationships between different spatial locations of the engineering structure, highlighting the feature information of key parts.

[0090] S23. Extract temporal dimension features from the spatial attention feature map through a temporal convolutional network layer to obtain a spatiotemporal joint feature matrix.

[0091] Optionally, the temporal convolutional network layer uses a one-dimensional convolutional kernel to perform convolution operations on the spatial attention feature map in the temporal dimension. This effectively captures the dynamic changes of features in the temporal dimension and extracts the temporal correlation features between different monitoring times. The engineering management terminal can input the spatial attention feature map into the temporal convolutional network layer to organically fuse spatial and temporal features, generating a spatiotemporal joint feature matrix. This spatiotemporal joint feature matrix can be used to reflect the state change characteristics of the engineering structure in the spatiotemporal dimension.

[0092] S24. Based on the gated cyclic unit layer, perform correlation analysis of cracks, stress and seepage on the spatiotemporal joint feature matrix, and output multidimensional state feature vector.

[0093] Optionally, the engineering management terminal can perform correlation analysis on features related to cracks, stress, and seepage in the spatiotemporal joint feature matrix through a gated cyclic unit layer, uncovering the inherent coupling relationship among the three, and transforming the complex spatiotemporal joint feature matrix into a low-dimensional, highly representative multidimensional state feature vector. This multidimensional state feature vector can be used to characterize the crack development trend, stress distribution state, and seepage evolution characteristics of hydraulic engineering projects.

[0094] In an optional embodiment, S3 includes:

[0095] S31. An initial digital twin simulation model is constructed based on finite element numerical simulation. The initial digital twin simulation model includes geometric parameters, material parameters, and boundary conditions.

[0096] Optionally, finite element numerical simulation involves discretizing the hydraulic engineering structure into multiple finite elements, establishing mechanical equilibrium equations between the elements based on theories such as elasticity and fluid mechanics, and then simulating the physical state of the engineering structure. The engineering management terminal can combine basic information such as construction design drawings, material testing reports, and on-site survey data to determine the geometric parameters, material parameters, and boundary conditions of the model. An initial digital twin simulation model is then constructed using finite element analysis software, ensuring that the model accurately replicates the geometric shape and mechanical properties of the engineering entity. Geometric parameters may include structural dimensions and shape, material parameters may include elastic modulus, Poisson's ratio, and tensile strength, and boundary conditions may include constraint conditions and load conditions.

[0097] S32. Using the multidimensional state feature vectors at each monitoring sampling time as observation data, the material parameters and boundary conditions of the initial digital twin simulation model are corrected online to obtain the corrected digital twin simulation model.

[0098] Optionally, the engineering management terminal can use multi-dimensional state feature vectors as observation data and employ a Bayesian estimation algorithm to dynamically adjust the material parameters and boundary conditions of the initial model, minimizing the deviation between the simulation results output by the model and the actual monitoring data. This enables real-time synchronization between the model and the physical project, ensuring that the corrected digital twin simulation model can accurately reflect the actual physical state of the project.

[0099] S33. Based on the modified digital twin model, calculate the ratio of the current ultimate bearing capacity of the structure to the actual load effect to obtain the dynamic health assessment index at each monitoring sampling time.

[0100] Optionally, ultimate bearing capacity refers to the maximum load that an engineering structure can withstand, while actual load effect refers to the mechanical response of the engineering structure under the load under current working conditions. The ratio of the two is used to quantitatively characterize the safety reserve level of the engineering structure. The engineering management terminal can simulate and calculate the actual load effect of the engineering structure under current working conditions by modifying the digital twin simulation model. At the same time, based on the material and geometric parameters of the model, it calculates the current ultimate bearing capacity of the structure, and calculates the ratio of the two to obtain the dynamic health assessment index. This dynamic health assessment index is used to reflect the safety status of the engineering structure under current working conditions in real time, providing a quantitative basis for engineering safety assessment.

[0101] In an optional embodiment, S4 includes:

[0102] S41. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a material performance degradation function is fitted based on the dynamic health assessment index sequence to obtain a degradation model.

[0103] Alternatively, the expression for the material property degradation function can be:

[0104]

[0105] In the formula, express Dynamic health assessment index at any time, This represents the initial dynamic health assessment index. Represents the degradation rate coefficient. Indicates the degenerate shape parameter, Indicates time.

[0106] Schematic, in the above material property degradation function expression, The dynamic health assessment index at time t represents the target value predicted by the model. This represents the initial dynamic health assessment index, which is the baseline safety reserve when the structure is built or repaired. This represents the degradation rate coefficient, which reflects how quickly the structural performance deteriorates; the larger the value, the faster the degradation. Represents the degenerate shape parameter, when Time indicates rapid early degradation followed by gradual degradation later. The time indicates that the degradation process is accelerating, when This is simply exponential degradation; This represents the time calculated from the reference time. The material property degradation function describes the entire process of material properties undergoing slow degradation, accelerated degradation, and eventual failure under environmental erosion and repeated loading. By minimizing the error between the fitted curve and historical data points, the degradation rate coefficient most suitable for the current project is solved. and degenerate shape parameters This allows us to determine a degradation model that describes the specific performance indicators of the structure.

[0107] For example, the engineering management terminal can continuously record and store the dynamic health assessment index calculated at each moment, generating a dataset sorted by time, namely the dynamic health assessment index sequence; the engineering management terminal can use a numerical optimization algorithm to fit a preset degradation function model to obtain the material performance degradation function.

[0108] S42. Based on the degradation model, predict the dynamic health assessment index sequence and calculate the predicted health indicator sequence within a preset time span in the future.

[0109] Optionally, the degradation model can be used to reflect the degradation trend of engineering material performance. The engineering management terminal can input time parameters within a preset time span into the degradation model to calculate the predicted health indicators corresponding to each time node, and generate a sequence of predicted health indicators arranged in chronological order. The preset time span can be flexibly set according to factors such as the design service life and maintenance needs of the water conservancy project. The prediction process fully considers potential impacts such as environmental factors and load changes to ensure the accuracy and reliability of the prediction results.

[0110] S43. Based on the predicted health indicator sequence and the preset failure threshold, identify the critical time node when the predicted health indicator is lower than the failure threshold, and obtain potential failure mode information.

[0111] Optionally, the failure threshold is a critical value of a health indicator pre-determined by the engineering management terminal based on the design standards, safety specifications, and actual operational needs of the water conservancy project. When the dynamic health assessment index falls below this threshold, it indicates a potential safety hazard in the engineering structure, suggesting possible failure. The engineering management terminal can compare each indicator in the predicted health indicator sequence with the failure threshold, identify the time points when the indicators fall below the threshold (i.e., critical time points), and analyze the possible failure types based on the corresponding engineering state characteristics to obtain potential failure mode information. These failure types include crack propagation, structural deformation, and abnormal seepage.

[0112] S44. Based on potential failure mode information, curve fitting and smoothing are performed on the predicted health index sequence to generate a full-cycle performance evolution prediction curve.

[0113] Optionally, the engineering management terminal uses a moving average method to smooth the predicted health indicator sequence, eliminating the influence of random noise. Then, a multinomial fitting algorithm is used to perform curve fitting on the smoothed sequence to generate a full-cycle performance evolution prediction curve. This full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a predetermined time span, clarifying the health degradation patterns and potential failure risks of the engineering structure.

[0114] In an optional embodiment, S5 includes:

[0115] S51. Based on the full-cycle performance evolution prediction curve, extract the key performance degradation time nodes and the degradation type labels corresponding to the full-cycle performance evolution prediction curve, and construct the maintenance requirement features based on the key performance degradation time nodes and degradation type labels.

[0116] Optionally, the critical performance degradation time node refers to the time node when the health status of the engineering structure shows obvious degradation or is about to reach the failure threshold. The degradation type label is a label determined based on the degradation trend and potential failure modes of the prediction curve. The engineering management terminal automatically extracts the critical performance degradation time node and degradation type label through curve analysis algorithm, and merges the two to construct a standardized maintenance requirement feature. The degradation type label can include crack degradation, stress degradation, and seepage degradation, etc. The maintenance requirement feature can be used to characterize the maintenance requirements of the project.

[0117] S52. Using maintenance requirement characteristics as an index, retrieve historical maintenance cases from the engineering maintenance knowledge graph to obtain a set of candidate maintenance solutions.

[0118] Optionally, the engineering maintenance knowledge graph is a pre-built structured knowledge base in the engineering management terminal, containing information such as various historical maintenance cases, maintenance technologies, maintenance processes, and maintenance costs for water conservancy projects. Each case is labeled with corresponding maintenance requirement characteristics, maintenance solutions, and implementation effects. The engineering management terminal can use semantic matching algorithms to match the constructed maintenance requirement characteristics with the case characteristics in the knowledge graph, retrieve historical maintenance cases similar to the current engineering maintenance needs, extract maintenance solutions from these cases, and generate a set of candidate maintenance solutions, ensuring that the candidate solutions are targeted and feasible.

[0119] S53. Based on a multi-objective optimization function that aims to minimize the total lifecycle maintenance cost and maximize the structural reliability, the set of candidate maintenance schemes is optimized to obtain the optimal maintenance strategy parameters.

[0120] Alternatively, the expression for the multi-objective optimization function can be:

[0121]

[0122] in, This represents the objective function value of a multi-objective optimization function. Represents decision variables, This represents the total cost of maintenance throughout the entire lifecycle. Indicates structural reliability. This represents the cost weighting coefficient. This represents the reliability weighting coefficient.

[0123] Schematic, in the above multi-objective optimization function expression, This represents the objective function value of a multi-objective optimization function. The decision variables can be a specific combination of maintenance measures and their implementation time. Indicated in decision variables The following is the total of all direct and indirect maintenance costs involved from now until the end of the project's life cycle; This indicates the average or minimum structural reliability that the project can achieve throughout its entire life cycle after implementing this decision; and These are weighting coefficients used to adjust the relative importance of cost and reliability according to managerial preferences. Operators are used to find the balance between the lowest cost and the highest reliability. Total lifecycle maintenance cost includes all maintenance-related costs such as maintenance material costs, labor costs, equipment costs, and downtime losses. Structural reliability refers to the probability that an engineering structure will operate normally within a preset time. and The determination is based on factors such as the project's safety level and economic budget, and is used to balance the relationship between maintenance costs and structural reliability.

[0124] For example, the engineering management terminal uses a genetic algorithm to perform optimization calculations on the set of candidate maintenance schemes based on a multi-objective optimization function that aims to minimize the total life cycle maintenance cost and maximize the structural reliability, and selects the scheme with the smallest objective function value to obtain the optimal maintenance strategy parameters.

[0125] S54. Generate a full lifecycle management strategy based on the optimal maintenance strategy parameters.

[0126] Optionally, the optimal maintenance strategy parameters may include maintenance time, maintenance method, maintenance frequency, and maintenance resource allocation. Based on these parameters, the project management terminal can generate a complete lifecycle management strategy by combining the project's actual operating status, full-cycle performance evolution prediction curves, and potential failure mode information. This lifecycle management strategy can cover the entire process management requirements of the project's operation, maintenance, and decommissioning phases, clearly defining the maintenance priorities, implementation steps, and resource allocation schemes for each phase. This enables intelligent and refined management of the entire lifecycle of the water conservancy project, achieving optimal resource allocation while ensuring the structural safety of the project.

[0127] In an optional embodiment, the gated cyclic unit layer includes a crack feature extraction branch, a stress feature extraction branch, and a seepage feature extraction branch, and S24 includes:

[0128] S241. Input the spatiotemporal joint feature matrix into the crack feature extraction branch, stress feature extraction branch and seepage feature extraction branch respectively to generate crack feature vector, stress feature vector and seepage feature vector respectively.

[0129] Optionally, the gated loop unit layer can be designed with a multi-branch structure, which may include a crack feature extraction branch, a stress feature extraction branch, and a seepage feature extraction branch. The crack feature extraction branch can extract features related to crack development in the engineering structure, such as crack width, length, and propagation rate; the stress feature extraction branch can extract features related to stress distribution in the structure, such as peak stress and stress distribution uniformity; and the seepage feature extraction branch can extract features related to seepage field evolution, such as seepage flow rate, seepage velocity, and seepage pressure. The engineering management terminal, through parallel computation of the crack feature extraction branch, stress feature extraction branch, and seepage feature extraction branch, obtains crack feature vectors, stress feature vectors, and seepage feature vectors that can accurately characterize the state of a single physical field, respectively.

[0130] S242. Using crack feature vectors, stress feature vectors, and seepage feature vectors as node features, an adjacency matrix is ​​constructed based on the spatial adjacency relationship of hydraulic engineering structural units. Information transmission and feature fusion are performed through a graph convolutional network to obtain a coupled feature graph.

[0131] Optionally, the spatial adjacency relationship of structural units in hydraulic engineering is determined based on the engineering geometry. Adjacent structural units have mechanical connections and mutual influences, and the adjacency matrix can be used to characterize this spatial relationship. Graph convolutional networks, based on the adjacency matrix, can achieve information transfer and fusion between different node features, uncovering the intrinsic coupling relationship between cracks, stress, and seepage. They fuse three independent feature vectors into a unified coupled feature graph, which comprehensively reflects the mutual influence patterns among the three.

[0132] S243. Perform global feature aggregation on the coupled feature map to generate a multi-dimensional state feature vector.

[0133] Optionally, the coupled feature map can include local features and interrelated features of cracks, stress, and seepage. Global feature aggregation extracts key global features from the coupled feature map through pooling operations, eliminates local redundant information, and transforms the two-dimensional coupled feature map into a one-dimensional feature vector, i.e., a multi-dimensional state feature vector. This multi-dimensional state feature vector integrates crack development trends, stress distribution states, seepage evolution characteristics, and the coupling relationship between the three, and is used to characterize the structural state of hydraulic engineering projects.

[0134] In the aforementioned method for the full life-cycle management of water conservancy projects, the unified spatiotemporal benchmark and heterogeneous data fusion enable the effective integration of multi-source data, avoiding data misalignment and bias, and providing data support for project status assessment. By leveraging attention-based spatiotemporal feature extraction and digital twin simulation models, multi-dimensional feature correlations are accurately mined, enabling dynamic health assessment of the project and improving the comprehensiveness and accuracy of the assessment. Performance prediction curves generated through temporal evolution analysis improve the accuracy of future structural reliability predictions. Based on knowledge graph retrieval and multi-objective optimization decision-making, scientific screening of maintenance schemes and optimized resource allocation are achieved, avoiding the subjectivity of experience-based decisions. Thus, while ensuring the safe operation of the project, intelligent and refined full life-cycle management can be realized, improving the efficiency and scientific nature of water conservancy project management.

[0135] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0136] Based on the same inventive concept, this application also provides a building lifecycle management system for implementing the above-mentioned method for building lifecycle management of a water conservancy project. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the building lifecycle management system for a water conservancy project provided below can be found in the above-described limitations of the building lifecycle management method for a water conservancy project, and will not be repeated here.

[0137] In one exemplary embodiment, such as Figure 2 As shown, a structural diagram of a building lifecycle management system 10 for water conservancy projects is provided, including:

[0138] The data preprocessing module 11 can be used to acquire multi-source heterogeneous monitoring data, and perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time; wherein, the standard spatiotemporal monitoring data is used to characterize the physical state information of water conservancy projects in different spatiotemporal dimensions.

[0139] The feature extraction module 12 can be used to input standard spatiotemporal monitoring data into a spatiotemporal feature extraction network based on an attention mechanism to obtain multidimensional state feature vectors at each monitoring sampling time.

[0140] The health assessment module 13 can be used to input multi-dimensional state feature vectors into the digital twin simulation model to obtain the dynamic health assessment index at each monitoring sampling time; wherein, the dynamic health assessment index is used to characterize the structural safety reserve coefficient of the water conservancy project under the current working conditions;

[0141] The evolution prediction module 14 can be used to construct a dynamic health assessment index sequence based on the dynamic health assessment index at each monitoring sampling time, and perform time-series evolution analysis on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve; wherein, the full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a future preset time span.

[0142] The strategy decision module 15 is used to retrieve a set of candidate maintenance schemes from the engineering maintenance knowledge graph based on the full life cycle performance evolution prediction curve, and to make multi-objective optimization decisions based on the set of candidate maintenance schemes to generate a full life cycle management strategy.

[0143] Furthermore, the data preprocessing module 11 can also be used for:

[0144] S11. Collect structural response sensor data, environmental hydrological monitoring data and construction progress image data during the water conservancy project process to obtain multi-source heterogeneous monitoring data;

[0145] S12. Perform time synchronization calibration on the multi-source heterogeneous monitoring data to obtain time-aligned monitoring data;

[0146] S13. Spatial coordinate unification is performed on the time alignment monitoring data to obtain spatiotemporal alignment monitoring data;

[0147] S14. Perform heterogeneous data fusion on the spatiotemporal alignment monitoring data to obtain standard spatiotemporal monitoring data.

[0148] Furthermore, the attention-based spatiotemporal feature extraction network includes a spatiotemporal location encoding layer, a multi-head attention mechanism layer, a temporal convolutional network layer, and a gated recurrent unit layer. The feature extraction module 12 can also be used for:

[0149] S21. Input the standard spatiotemporal monitoring data into the spatiotemporal location coding layer to generate a feature vector sequence with spatiotemporal location information;

[0150] S22. The spatial dependency relationship of the feature vector sequence is modeled through a multi-head attention mechanism layer to obtain a spatial attention feature map;

[0151] S23. Extract temporal dimension features from the spatial attention feature map through a temporal convolutional network layer to obtain a spatiotemporal joint feature matrix;

[0152] S24. Based on the gated cyclic unit layer, perform correlation analysis of the spatiotemporal joint feature matrix on cracks, stress and seepage, and output multidimensional state feature vectors; among which, the multidimensional state feature vectors are used to characterize the crack development trend, stress distribution state and seepage evolution characteristics of hydraulic engineering.

[0153] Furthermore, the health assessment module 13 can also be used for:

[0154] S31. An initial digital twin simulation model is constructed based on finite element numerical simulation. The initial digital twin simulation model includes geometric parameters, material parameters, and boundary conditions.

[0155] S32. Using the multidimensional state feature vectors at each monitoring sampling time as observation data, the material parameters and boundary conditions of the initial digital twin simulation model are corrected online to obtain the corrected digital twin simulation model.

[0156] S33. Based on the modified digital twin model, calculate the ratio of the current ultimate bearing capacity of the structure to the actual load effect to obtain the dynamic health assessment index at each monitoring sampling time.

[0157] Furthermore, the evolution prediction module 14 can also be used for:

[0158] S41. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a material performance degradation function is fitted based on the dynamic health assessment index sequence to obtain a degradation model.

[0159] S42. Based on the degradation model, predict the dynamic health assessment index sequence and calculate the predicted health indicator sequence within a future preset time span.

[0160] S43. Based on the predicted health indicator sequence and the preset failure threshold, identify the critical time node when the predicted health indicator is lower than the failure threshold, and obtain potential failure mode information.

[0161] S44. Based on potential failure mode information, curve fitting and smoothing are performed on the predicted health index sequence to generate a full-cycle performance evolution prediction curve.

[0162] Furthermore, the strategy decision module 15 can also be used for:

[0163] S51. Based on the full-cycle performance evolution prediction curve, extract the key performance degradation time nodes and the degradation type labels corresponding to the full-cycle performance evolution prediction curve, and construct the maintenance requirement features based on the key performance degradation time nodes and degradation type labels.

[0164] S52. Using maintenance requirement characteristics as an index, retrieve historical maintenance cases from the engineering maintenance knowledge graph to obtain a set of candidate maintenance solutions;

[0165] S53. Based on a multi-objective optimization function that aims to minimize the total life-cycle maintenance cost and maximize the structural reliability, the set of candidate maintenance schemes is optimized to obtain the optimal maintenance strategy parameters.

[0166] S54. Generate a full lifecycle management strategy based on the optimal maintenance strategy parameters.

[0167] Furthermore, the gated loop unit layer includes a crack feature extraction branch, a stress feature extraction branch, and a seepage feature extraction branch. The feature extraction module 12 can also be used for:

[0168] S241. Input the spatiotemporal joint feature matrix into the crack feature extraction branch, stress feature extraction branch and seepage feature extraction branch respectively, and generate crack feature vector, stress feature vector and seepage feature vector respectively;

[0169] S242. Using crack feature vectors, stress feature vectors, and seepage feature vectors as node features, an adjacency matrix is ​​constructed based on the spatial adjacency relationship of hydraulic engineering structural units. Information transmission and feature fusion are performed through a graph convolutional network to obtain a coupled feature graph.

[0170] S243. Perform global feature aggregation on the coupled feature map to generate a multi-dimensional state feature vector.

[0171] In one embodiment, such as Figure 3 A computer device 300 is provided, comprising:

[0172] At least one processor 301, and at least one memory 302 communicatively connected to said processor 301; said memory stores application code executable by said processor, said application code being executed by said processor to enable said processor to perform the steps of a construction life-cycle management method for a water conservancy project as described above;

[0173] The computer device may also include: sensor 303;

[0174] The processor 301, memory 302, and sensor 303 can be connected via bus 304 or other means. The figure shows an example of connection via bus 304. Figure 3 The character is represented by a single thick line, but this does not mean that there is only one bus or a type of bus.

[0175] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0176] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0177] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for managing the entire life cycle of a waterworks construction, characterized by, The method includes: S1. Acquire multi-source heterogeneous monitoring data, and perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time; wherein, the standard spatiotemporal monitoring data is used to characterize the physical state information of water conservancy projects in different spatiotemporal dimensions; S2. Input the standard spatiotemporal monitoring data into a spatiotemporal feature extraction network based on an attention mechanism to obtain a multidimensional state feature vector for each monitoring sampling time. S3. Input the multidimensional state feature vector into the digital twin simulation model to obtain the dynamic health assessment index at each of the monitoring sampling times; wherein, the dynamic health assessment index is used to characterize the structural safety reserve coefficient of the water conservancy project under the current working conditions; S4. Based on the dynamic health assessment index at each monitoring sampling time, a dynamic health assessment index sequence is constructed, and a time-series evolution analysis is performed on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve; wherein, the full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a future preset time span. S5. Based on the full-cycle performance evolution prediction curve, retrieve a set of candidate maintenance schemes from the engineering maintenance knowledge graph, and perform multi-objective optimization decision-making based on the set of candidate maintenance schemes to generate a full life cycle management strategy.

2. The method of claim 1, wherein, S1 includes: S11. Collect structural response sensor data, environmental hydrological monitoring data, and construction progress image data during the water conservancy project process to obtain the multi-source heterogeneous monitoring data; S12. Perform time synchronization calibration on the multi-source heterogeneous monitoring data to obtain time-aligned monitoring data; S13. Spatial coordinate unification is performed on the time alignment monitoring data to obtain spatiotemporal alignment monitoring data; S14. Perform heterogeneous data fusion on the spatiotemporal alignment monitoring data to obtain the standard spatiotemporal monitoring data.

3. The method of claim 2, wherein, The attention-based spatiotemporal feature extraction network includes a spatiotemporal location encoding layer, a multi-head attention mechanism layer, a temporal convolutional network layer, and a gated recurrent unit layer. S2 includes: S21. Input the standard spatiotemporal monitoring data into the spatiotemporal location coding layer to generate a feature vector sequence with spatiotemporal location information; S22. The spatial dependency relationship of the feature vector sequence is modeled through the multi-head attention mechanism layer to obtain a spatial attention feature map; S23. Extract the temporal dimension features from the spatial attention feature map through the temporal convolutional network layer to obtain the spatiotemporal joint feature matrix; S24. Based on the gated cyclic unit layer, perform correlation analysis on the spatiotemporal joint feature matrix of cracks, stress and seepage, and output a multidimensional state feature vector; wherein, the multidimensional state feature vector is used to characterize the crack development trend, stress distribution state and seepage evolution characteristics of the hydraulic engineering project.

4. The method of claim 3, wherein, S3 includes: S31. An initial digital twin simulation model is constructed based on finite element numerical simulation, wherein the initial digital twin simulation model includes geometric parameters, material parameters and boundary conditions; S32. Using the multidimensional state feature vectors at each of the monitoring sampling times as observation data, the material parameters and boundary conditions of the initial digital twin simulation model are corrected online to obtain the corrected digital twin simulation model. S33. Based on the modified digital twin model, calculate the ratio of the current ultimate bearing capacity of the structure to the actual load effect, and obtain the dynamic health assessment index at each of the monitoring sampling times.

5. The method of claim 4, wherein, S4 includes: S41. Based on the dynamic health assessment index at each of the monitoring sampling times, a dynamic health assessment index sequence is constructed, and a material performance degradation function is fitted based on the dynamic health assessment index sequence to obtain a degradation model; wherein, the expression of the material performance degradation function is: wherein represents a dynamic health assessment index at a time instant, represents an initial dynamic health assessment index, represents a degeneration rate coefficient, represents a degeneration shape parameter, represents time; S42. Based on the degradation model, predict the dynamic health assessment index sequence to calculate the predicted health indicator sequence within a future preset time span; S43. Based on the predicted health indicator sequence and the preset failure threshold, identify the critical time node when the predicted health indicator is lower than the failure threshold, and obtain potential failure mode information. S44. Based on the potential failure mode information, perform curve fitting and smoothing on the predicted health index sequence to generate the full-cycle performance evolution prediction curve.

6. The method of claim 5, wherein, S5 includes: S51. Based on the full-cycle performance evolution prediction curve, extract the key performance degradation time nodes and the degradation type labels corresponding to the full-cycle performance evolution prediction curve, and construct maintenance requirement features based on the key performance degradation time nodes and the degradation type labels. S52. Using the maintenance requirement features as an index, retrieve historical maintenance cases from the engineering maintenance knowledge graph to obtain a set of candidate maintenance solutions; S53. Based on a multi-objective optimization function aimed at minimizing the total lifecycle maintenance cost and maximizing structural reliability, the candidate maintenance scheme set is optimized to obtain the optimal maintenance strategy parameters; wherein, the expression of the multi-objective optimization function is: wherein, represents a target function value of the multi-objective optimization function, represents a decision variable, represents the total life cycle maintenance cost, represents the structural reliability, represents a cost weight coefficient, represents a reliability weight coefficient; S54. Generate the full lifecycle management strategy based on the optimal maintenance strategy parameters.

7. The method according to claim 3, characterized in that, The gated loop unit layer includes a crack feature extraction branch, a stress feature extraction branch, and a seepage feature extraction branch. S24 includes: S241. Input the spatiotemporal joint feature matrix into the crack feature extraction branch, the stress feature extraction branch and the seepage feature extraction branch respectively to generate crack feature vector, stress feature vector and seepage feature vector respectively; S242. Using the crack feature vector, the stress feature vector, and the seepage feature vector as node features, an adjacency matrix is ​​constructed based on the spatial adjacency relationship of the hydraulic engineering structural units. Information transmission and feature fusion are performed through a graph convolutional network to obtain a coupled feature graph. S243. Perform global feature aggregation on the coupled feature map to generate the multidimensional state feature vector.

8. A building lifecycle management system for water conservancy projects, characterized in that, The system includes: The data preprocessing module is used to acquire multi-source heterogeneous monitoring data, and to perform spatiotemporal benchmark unification and heterogeneous data fusion on the multi-source heterogeneous monitoring data to obtain standard spatiotemporal monitoring data at each monitoring sampling time; wherein, the standard spatiotemporal monitoring data is used to characterize the physical state information of water conservancy projects in different spatiotemporal dimensions. The feature extraction module is used to input the standard spatiotemporal monitoring data into a spatiotemporal feature extraction network based on an attention mechanism to obtain a multidimensional state feature vector for each monitoring sampling time. The health assessment module is used to input the multidimensional state feature vector into the digital twin simulation model to obtain the dynamic health assessment index at each of the monitoring sampling times; wherein, the dynamic health assessment index is used to characterize the structural safety reserve coefficient of the water conservancy project under the current working conditions; The evolution prediction module is used to construct a dynamic health assessment index sequence based on the dynamic health assessment index at each monitoring sampling time, and to perform time-series evolution analysis on the dynamic health assessment index sequence to obtain a full-cycle performance evolution prediction curve; wherein, the full-cycle performance evolution prediction curve is used to characterize the structural reliability change trend of the water conservancy project within a future preset time span. The strategy decision module is used to retrieve a set of candidate maintenance schemes from the engineering maintenance knowledge graph based on the full-cycle performance evolution prediction curve, and to make multi-objective optimization decisions based on the set of candidate maintenance schemes to generate a full life cycle management strategy.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images