Dam structure digital twinning sensing method and system based on sensor network optimization

Abstract

The invention provides a dam structure digital twinning sensing method and system based on sensor network optimization, and relates to the technical field of dam safety monitoring, and the method comprises the steps: constructing a three-dimensional finite element model of a dam structure; carrying out compressed sensing on the observation process of the dam structure to obtain a digital twinning sensing model of the dam structure; constructing quantitative indexes and optimization conditions of digital twinning perception; optimizing the sensor network of the dam structure by combining the three-dimensional finite element model based on the quantitative index and the optimization condition to obtain an optimal sensor network; and performing training optimization on the digital twinborn perception model through monitoring data of the optimal sensor network to obtain a digital twinborn body of the dam structure, the digital twinborn body being used for obtaining global response data of the dam structure. The problems that the global structure state of the dam is difficult to effectively perceive under the limited monitoring condition, the digital twinning perception mechanism lacks theoretical support, and the monitoring network perception capability is difficult to quantitatively evaluate are solved.

Description

Technical Field

[0001] This invention relates to the field of dam safety monitoring technology, and more specifically, to a digital twin sensing method and system for dam structures based on sensor network optimization. Background Technology

[0002] Dams, as important hydraulic structures, are widely used in flood control, water supply, power generation, and water resource allocation projects. Their operational safety is directly related to the safety of people's lives and property and the stable development of the regional economy and society. In recent years, with the continuous expansion of water conservancy projects, dams have shown characteristics such as increased dam height, complex structure, and diversified operating conditions. The dam body and foundation have been in service for a long time under high head, high stress, and multi-field coupling conditions. Their structural response and damage evolution processes have obvious nonlinear and time-varying characteristics, which bring great challenges to dam safety monitoring and operation management.

[0003] Currently, dam structural health monitoring primarily relies on permanent monitoring facilities deployed according to relevant specifications, such as measuring points for displacement, stress, strain, and seepage, and then constructs monitoring and evaluation models based on the data from these measuring points. While this type of point-based monitoring model is mature in engineering practice, its number and spatial distribution are typically limited by factors such as project cost, construction conditions, and operation and maintenance costs, making it difficult to achieve a comprehensive understanding of the overall state of the dam structure. In complex operating conditions or under localized damage, relying solely on a small amount of monitoring data is insufficient to accurately reflect the dam's overall response characteristics, resulting in inadequate perception capabilities. To improve monitoring and analysis capabilities, various model-based safety monitoring methods have been proposed in existing technologies.

[0004] One type of method is based on the theories of continuum mechanics and structural mechanics, which analyze the deformation, stress, and stability of dam structures by establishing numerical models. This type of method has good physical meaning and engineering interpretability, but it is highly dependent on model parameters, boundary conditions, and operating condition assumptions. The model calibration process is complex, the computational cost is high, and it is difficult to meet the needs of rapid updates and real-time analysis during operation.

[0005] Another approach uses historical monitoring data as a basis to model and predict the dam structure response using statistical or machine learning models. Among these, statistical models based on empirical assumptions and shallow machine learning methods have been applied in engineering, but their ability to characterize complex nonlinear relationships and multi-factor coupling effects is limited, and their adaptability to changes in operating conditions or extreme conditions is insufficient.

[0006] In recent years, deep learning methods have been introduced into the field of dam safety monitoring due to their strong nonlinear modeling capabilities. However, these methods heavily rely on limited measurement point data, making it difficult to effectively represent the overall structural state of the dam. Furthermore, they still have shortcomings in terms of physical interpretability and engineering reliability. With the development of intelligent sensing and information technology, digital twin technology has been used for dam safety monitoring and analysis. By integrating physical entities, numerical models, and monitoring data, it achieves a mapping between virtual models and the actual operating state of the dam.

[0007] However, the application of existing digital twin technology in dam engineering is still in its early stages. Its sensing capabilities largely depend on existing monitoring networks. It has failed to systematically characterize the perceptibility of dam structures under limited observation conditions from a theoretical perspective. The relationship between sensor deployment and sensing effect lacks quantitative evaluation methods, which restricts the application effect of digital twin models under complex working conditions.

[0008] In summary, existing dam safety monitoring and digital twin technologies generally suffer from problems such as sparse monitoring data, insufficient ability to perceive the overall structural state, and a lack of systematic theoretical support for the perception mechanism. These limitations make it difficult to conduct high-precision, quantifiable assessments of the overall structural safety status of dams under limited monitoring conditions. Therefore, it is necessary to propose a new technical solution that, from the perspective of perception modeling and monitoring network optimization, improves the structural state perception capability and operational safety assurance level of dams under sparse observation conditions. Summary of the Invention

[0009] The purpose of this invention is to provide a digital twin sensing method and system for dam structures based on sensor network optimization, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows: In a first aspect, this application provides a digital twin sensing method for dam structures based on sensor network optimization, including: Construct a three-dimensional finite element model of the dam structure; Compressed sensing was used to observe the dam structure, resulting in a digital twin sensing model of the dam structure. Quantitative indicators and optimization conditions for constructing digital twin perception; Based on the aforementioned quantitative indicators and optimization conditions, and combined with the modal matrix provided by the three-dimensional finite element model, the sensor network of the dam structure is optimized to obtain the optimal sensor network. The digital twin sensing model is trained and optimized using monitoring data from the optimal sensor network to obtain a digital twin of the dam structure. This digital twin is used to acquire global response data of the dam structure.

[0010] Secondly, this application also provides a digital twin sensing system for dam structures based on sensor network optimization, comprising: The first building module is used to construct a three-dimensional finite element model of the dam structure; The sensing module is used to perform compressed sensing of the dam structure observation process to obtain a digital twin sensing model of the dam structure. The second building module is used to construct the quantitative indicators and optimization conditions for digital twin perception; The optimization module is used to optimize the sensor network of the dam structure based on the quantitative indicators and optimization conditions, combined with the modal matrix provided by the three-dimensional finite element model, to obtain the optimal sensor network. The training module is used to train and optimize the digital twin perception model using the monitoring data of the optimal sensor network to obtain a digital twin of the dam structure. The digital twin is used to acquire the global response data of the dam structure.

[0011] The beneficial effects of this invention are as follows: (1) This method integrates compressed sensing theory into the construction of a digital twin sensing model. It constructs a mathematical mapping relationship between the high-dimensional state of the dam structure and sparse observations through modal decomposition. Simultaneously, it constructs cross-domain adaptation constraints and information integrity satisfaction conditions based on effective measurement operators, enabling the model to possess both data-driven capabilities and physical theory support. During the model training and optimization phase, a strategy of pre-training with simulated data combined with fine-tuning using measured data from the optimal sensor network is adopted. The model deviation is corrected by combining cross-domain differences between real and simulated effective measurement operators, effectively solving the inherent differences between numerical simulation and the engineering entity. This ensures that the model output both closely matches the actual operating state of the dam and conforms to the laws of structural dynamic characteristics. Compared to traditional purely data-driven modeling methods, the digital twin constructed by this method significantly improves the reconstruction accuracy of the dam structure's full-domain response data, accurately restoring key state information such as displacement and stress at each node of the dam.

[0012] (2) This invention also uses modal observability and the number of effective measurement operators as core quantitative indicators, and combines the QR principal component method for initial selection with the greedy algorithm for iterative expansion as a combined optimization strategy. At the same time, it incorporates the dual spatial distribution constraints of maximizing the information content of the Fisher information matrix and minimizing the node spacing to achieve accurate screening and optimized deployment of sensor nodes. Compared with the traditional sensor deployment method based on engineering experience, it effectively avoids the problems of local concentration of measurement points, information redundancy, or missed capture of key modes. Under the constraint of quantity, the optimized sensor network can maximize the capture of the core modal information of the dam structure and ensure the overall distinguishability of multimodal information. This greatly improves the sensor network's ability to perceive the overall structural state of the dam and lays a high-quality monitoring data foundation for subsequent digital twin modeling.

[0013] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the digital twin sensing method for dam structures based on sensor network optimization as described in this embodiment of the invention; Figure 2 This is a schematic diagram of a three-dimensional finite element model of the dam structure in an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the quantitative indicators of different methods in the embodiments of the present invention; Figure 4 This is a comparison chart of twin accuracy before optimization in an embodiment of the present invention; Figure 5 This is a comparison chart of the optimized twin accuracy in an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0017] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0018] Example 1: This embodiment provides a digital twin sensing method for dam structures based on sensor network optimization.

[0019] It should be noted that existing dam safety monitoring systems primarily rely on sparsely deployed monitoring points, resulting in significant spatial incompleteness of monitoring information. This leads to considerable uncertainty in accurately reconstructing and assessing the overall structural state of the dam based on limited observation data. Especially under complex operating conditions or during the evolution of localized damage, traditional monitoring methods struggle to quantify the coverage of the monitoring network on the dam's structural state and lack clear criteria for determining whether monitoring is sufficient to reflect the overall structural condition. Therefore, how to quantitatively characterize the perceptibility of the dam's overall structural state under sparse observation conditions is a primary technical problem that current technologies urgently need to address.

[0020] Secondly, existing digital twin models largely rely on the empirical fusion of numerical models and monitoring data. Their sensing capabilities are significantly constrained by existing sensor deployment schemes, lacking a unified theoretical framework to describe the intrinsic relationship between sensor deployment, structural modal characteristics, and state reconfigurability. Under limited measurement points, the sensing capability of digital twin models for potential structural damage or abnormal states is difficult to quantify and assess, making it difficult to guarantee the model's sensing effectiveness and reliability. Therefore, how to establish a dam digital twin sensing modeling method with a clear theoretical foundation and quantitatively characterize its observability of structural states is another key issue in existing technologies.

[0021] Furthermore, with the emergence of new monitoring methods such as remote sensing, drones, and intelligent inspection, current technologies lack effective methods for unified modeling and collaborative optimization of these new monitoring methods with traditional fixed monitoring facilities. Existing monitoring network designs are mostly based on experience or local optimization, making it difficult to systematically configure different types of sensors from an overall perception perspective. This results in low utilization efficiency of monitoring resources and limits the perception accuracy and stability of the digital twin model. Therefore, how to collaboratively configure and optimize multi-source monitoring methods within a unified theoretical framework to improve the overall perception capability of the dam's digital twin model is one of the urgent technical problems to be solved.

[0022] See Figure 1 The figure shows that the method includes steps S1, S2, S3, S4 and S5.

[0023] Step S1: Construct a three-dimensional finite element model of the dam structure; Step S2: Compressed sensing is performed on the dam structure observation process to obtain a digital twin sensing model of the dam structure; Step S2 includes: Step S21: Perform compressed sensing on the dam structure observation process to obtain a compressed sensing model of the dam structure; This step treats the dam structure as a high-dimensional dynamic system from a cybernetics perspective. It constructs a compressed sensing model of the observation process through mode decomposition and linear projection, thereby realizing the mapping of the high-dimensional structural state to sparse observation vectors.

[0024] Step S21 includes: Step S211: Based on cybernetics, the dam structure state is decomposed into a modal combination form to obtain a high-dimensional state vector of the dam structure state. The high-dimensional state vector is composed of the modal matrix and time-varying modal coefficients of the dam structure. In this step, the high-dimensional state vector is: In the formula, express The high-dimensional state vector of the dam structure at any given time. express Time of the first The time-varying modal coefficients corresponding to each mode The first part representing the dam structure One modality, Indicates the number of modes. The modal matrix represents the dam structure. Indicates in The time-varying modal coefficient vector at time t.

[0025] in, , , .

[0026] Step S212: Perform linear projection on the observation process of the dam structure to obtain the linear relationship between the sparse observation vector and the high-dimensional state vector; In this step, the physical observation process of the dam structure is abstracted into a linear projection process, taking into account random errors in the observation process (such as sensor measurement errors and environmental interference), and a system is constructed. The linear relationship between the time-sparse observation vector and the high-dimensional state vector: In the formula, express The sparse observation vector at time step, Represents the observation matrix of the sensing network. express The first error at any given time.

[0027] Among them, the perception network observation matrix This represents the placement of the sensors within the dam structure and is a 0-1 sparse matrix.

[0028] Step S213: Substitute the modal combination form of the high-dimensional state vector into the linear relationship to obtain the compressed sensing model of the dam structure.

[0029] In this step, the modal combination form of the high-dimensional state vector obtained in step S211 is... Substitute the linear relationship from step S212 to eliminate the high-dimensional state vector. The compressed sensing model of the dam structure observation process is obtained: In the formula, express The second error at time.

[0030] Step S22: Map the compressed sensing model into a neural network with sparse observation vectors as input and high-dimensional state vectors as prediction output; In this step, based on the mapping logic of the compressed sensing model, a neural network is built (Long Short-Time Memory (LSTM) neural network is preferred for high arch dam projects to adapt to the time-varying characteristics of structural response). The neural network is defined as follows: ,in, These represent the learnable parameters of a neural network, including weights, biases, and hidden layer dimensions.

[0031] Step S23: Based on the numerical simulation results of the three-dimensional finite element model, obtain the global state simulation value and the sparse observation simulation value; Step S24: Construct the objective function of the neural network based on the global state simulation values ​​and sparse observation simulation values; In this step, the core objective is to minimize the mean square error between the neural network predictions and the finite element simulations. The objective function of the neural network is as follows: In the formula, This indicates taking the minimum value. Indicates the length of the time series. express The sparse observation simulation matrix at time points. express The global state simulation matrix of the dam structure at any given time. This represents the square of the Euclidean norm.

[0032] in, and All data were obtained from multi-condition numerical simulations of the dam's three-dimensional finite element model.

[0033] Step S25: Construct an effective measurement operator using the modal matrix of the dam structure and the observation matrix of the sensing network; In this step, the effective measurement operator is defined as follows: , , Represents the observation matrix of the sensing network. This represents the modal matrix of the dam structure.

[0034] Based on the differences between numerical simulation and engineering entities, effective measurement operators for simulation are constructed respectively. and real effective measurement operators : In the formula, The actual modal matrix representing the dam project entity (inferred from real monitoring data). This represents the modal matrix in finite element numerical simulation.

[0035] Step S26: Construct cross-domain adaptation constraints and information integrity conditions by using effective measurement operators; In this step, considering the inherent error between the modal matrix of the finite element numerical simulation and the actual modal matrix of the dam entity, this error is quantified and its range is limited through cross-domain adaptation constraints. The cross-domain adaptation constraints are as follows: In the formula, This represents the cross-domain mode matrix error. Describe the Frobenius norm. This indicates the upper limit of the error.

[0036] In the digital twin prediction framework, the effective measurement operator is required to approximately maintain its Euclidean norm for any time-varying modal coefficient vector, ensuring that the sparse observation process does not lose or over-amplify the modal information of the dam structure. This information integrity satisfies the restricted isometry condition (RIP) of compressed sensing theory, as shown in the formula: In the formula, Represents a restricted isochronous constant. This represents the time-varying modal coefficient vector.

[0037] Step S27: Based on the neural network, cross-domain adaptation constraints, and information integrity satisfaction conditions, a digital twin perception model is obtained.

[0038] Step S3: Construct quantitative indicators and optimization conditions for digital twin perception; This step constructs two core quantitative indicators—single-modal observability and multi-modal overall distinguishability—based on the dam structure's sensing network observation matrix, modal matrix, and effective measurement operators. Combined with preset thresholds based on engineering practice, it forms the optimization conditions for digital twin sensing, providing quantitative evaluation basis and constraint criteria for subsequent sensor network optimization.

[0039] Step S3 includes: Step S31: Construct an observability index for each mode using the perception network observation matrix and the mode matrix, wherein the perception network observation matrix is ​​constructed using a sensor network; This step quantitatively characterizes the detectability of a single structural mode under the current sensor layout by using the observation matrix and modality matrix of the sensing network, reflecting the proportion of energy retained by modal information in the observation space.

[0040] Specifically, the observability index is: In the formula, Indicates the first The observability index of each modality Represents the observation matrix of the sensing network. The first part representing the dam structure One modality, Indicates the number of modes. This represents the Euclidean norm.

[0041] When the observability index is close to zero, it indicates that the mode is almost undetectable under the current sensor layout.

[0042] Step S32: Perform singular value decomposition on the effective measurement operator to construct the condition number of the effective measurement operator; In this step, singular value decomposition is performed on the effective measurement operator to construct the condition number index, which quantitatively characterizes the overall distinguishability of multi-modes in the observation space and avoids the overlap of observation information of different modes.

[0043] Specifically, for effective measurement operators Perform singular value decomposition to obtain Filter out the largest singular value from all singular values. and minimum singular value Through the maximum singular value and minimum singular value The condition number for calculating the effective measurement operator is: In the formula, Indicates the effective measurement operator condition number.

[0044] When the condition number is closer to 1, it indicates that the distribution of singular values ​​of the effective measurement operator is more uniform, the information overlap of multi-modes in the observation space is lower, the overall distinguishability is stronger, and the global reconstruction stability of the sensor network is better. When the condition number is larger, it indicates that the observation information of multi-modes is more easily confused, and the reconstruction stability is worse.

[0045] Step S33: Use the observability index and the condition number of the effective measurement operator for each modality as the quantitative index of digital twin perception; Step S34: Combine preset thresholds to construct optimized conditions for digital twin perception through quantitative indicators.

[0046] In this step, the optimization conditions are: In the formula, This indicates taking the minimum value. This represents the theoretical limit of minimum modal observability. It represents the theoretical limit of the condition number.

[0047] Step S4: Based on the quantitative indicators and optimization conditions, and combined with the modal matrix provided by the three-dimensional finite element model, optimize the sensor network of the dam structure to obtain the optimal sensor network; Step S4 includes: Step S41: Define the optimization objective function for the sensor network; In this step, given a number of sensors, sensor locations are selected to maximize minimum modal observability while minimizing the number of conditions. The objective function is expressed as: In the formula, This indicates taking the maximum value. Represents the selected set of sensor nodes. This represents the total number of nodes in the 3D finite element model of the dam. Represents a set The number of sensors, This indicates the total number of sensors required. Indicates in set Next The observability index of each modality Represents the regularity coefficient. express The condition number of Represents a set An effective measurement operator.

[0048] Step S42: Perform modal analysis on the three-dimensional finite element model to obtain the modal matrix of the dam structure; Step S43: Perform QR decomposition on the transpose of the modality matrix, combine the column pivot selection method and the optimization objective function to obtain the first preset number of initial sensor nodes, and obtain the current sensor set; In this step, the transpose of the modality matrix is ​​decomposed using QR decomposition with column pivoting: In the formula, express The transpose of the matrix, Represents the QR decomposition matrix. Represents an orthogonal matrix. This represents an upper triangular matrix.

[0049] Set the first preset quantity ,from Before the election The finite element nodes corresponding to each principal element are used as initial sensor nodes, and these initial sensor nodes are integrated into the current sensor set. ,Right now ,in for The Middle The node number corresponding to each principal element. .

[0050] Step S44: Construct a candidate node set by excluding the remaining nodes after excluding the current sensor set; In this step, the current sensor set is excluded from the set of all nodes in the 3D finite element model of the dam. The initial selection of sensor nodes, and the remaining nodes constitute the candidate node set. The candidate node set is the pool of candidate nodes that is subsequently expanded by the greedy algorithm iteration.

[0051] Step S45: Using quantitative indicators and optimization conditions, a greedy algorithm is used to expand the sensor nodes, and a second preset number of expanded sensor nodes are selected from the candidate node set. Step S45 includes: Step S451: For each candidate sensor node, construct a temporary sensor set using the current sensor set and the candidate sensor node, where the candidate sensor node is any sensor node in the candidate node set; In this step, the current sensor set is initialized, and the candidate node set is traversed. Each candidate sensor node in , to the current sensor set With candidate nodes Combine and construct temporary sensor sets ,Right now .

[0052] Step S452: Calculate the corresponding sensing network observation matrix and effective measurement operator based on the temporary sensor set; Step S453: Calculate the quantization index of each temporary sensor set using the sensing network observation matrix and effective measurement operators; Step S454: Substitute the quantification index into the optimization objective function to calculate the optimization objective function value of the temporary sensor set corresponding to each candidate sensor node; Step S455: Based on maximizing the objective function value and spatial distribution constraints, select a candidate sensor node as an extended sensor node, add the extended sensor node to the current sensor set, and remove the extended sensor node from the candidate node set; In this step, the candidate sensor node with the largest objective function value is selected as the candidate expanded sensor node. For candidate expanded sensor nodes Perform spatial distribution constraint verification when If the spatial distribution constraint (minimum distance between nodes) is met, then the candidate extended sensor node is determined. To expand the number of sensor nodes, if a node does not meet the spatial constraints, it is discarded, and a candidate node with the second largest objective function value is selected and re-verified until a node that meets the constraints is found.

[0053] Therefore, the expanded expression is: In the formula, This represents the candidate node corresponding to the maximum value. , This represents taking the logarithm and determinant. Represents the Fisher information matrix. express transpose, Represents the regularity coefficient. Represents a set The perception network observation matrix The modal matrix represents the dam structure. Represents the identity matrix.

[0054] Step S456: Select the next expanded sensor node using the updated current sensor set and candidate node set, until a second preset number of expanded sensor nodes are obtained.

[0055] Step S46: Update the perception network observation matrix and effective measurement operators by initially selecting sensor nodes and expanding sensor nodes; Step S47: Determine whether the optimization conditions are met by using the updated perception network observation matrix and effective measurement operators; Step S48: If the optimization conditions are not met, remove the sensor nodes that do not meet the optimization conditions and reselect the expanded sensor nodes. Step S49: If the optimization conditions are met, then the current sensor set is taken as the optimal sensor network.

[0056] Step S5: Train and optimize the digital twin sensing model using the monitoring data from the optimal sensor network to obtain a digital twin of the dam structure. The digital twin is used to acquire global response data of the dam structure.

[0057] In this step, simulated data under the optimal sensor network is obtained through a three-dimensional finite element model. The neural network is then trained using the simulated data to obtain the trained neural network. Next, sensors are deployed on the dam structure using the optimal sensor network to acquire corresponding actual monitoring data, actual modal matrices, and corresponding real effective measurement operators. The trained neural network is then fine-tuned using the real effective measurement operators and the corresponding simulated effective measurement operators to correct deviations caused by differences between numerical simulation and the actual engineering entity. This ensures that the model output more closely reflects the actual operating state of the dam, resulting in a digital twin used to acquire the full-domain response data of the dam structure.

[0058] Example 2: In this embodiment, the DS hydropower station is used as an example to illustrate the method of the present invention. Specifically, the DS hydropower station is a large-scale hydropower project primarily for power generation, with additional benefits in flood control, sediment retention, and other comprehensive utilization aspects. The dam is a hyperbolic concrete arch dam with a maximum height of approximately 210m. Its large structural scale and complex stress distribution make it representative. On September 5, 2022, a strong earthquake occurred near the dam site, with the epicenter approximately 21km from the dam. The earthquake significantly impacted the dam structure, providing a practical engineering background for conducting structural response analysis of the dam under complex operating conditions. Based on the dam's operation and seismic conditions, monitoring information such as horizontal displacement, stress, and seismic acceleration of the dam body is comprehensively selected as the research object. Displacement monitoring employs a combination of vertical surveying and external geodetic measurements, deploying monitoring points in multiple typical dam sections and grouting galleries on both banks to form a monitoring network covering the overall deformation characteristics of the dam. Stress monitoring follows the arch-beam combination arrangement principle, installing multi-directional strain gauges in representative dam sections to reflect the key stress states of the dam. Seismic monitoring instruments are mainly deployed on the dam crest, internal galleries, and both banks, covering the main structural units. Based on the above monitoring system and combined with a finite element numerical model, a digital twin reconstruction analysis of the dam's structural deformation and stress response under seismic conditions is performed.

[0059] Specifically, based on the structural characteristics, topographic and geological conditions, and dynamic load characteristics of the DS dam, a three-dimensional finite element model of the arch dam-reservoir-foundation was established in the finite element software ABAQUS. The Dagangshan arch dam has 28 transverse joints. The finite element mesh for each dam section was divided based on the actual locations of these joints, and a contact force model was used to establish the contact relationships between the joints. The discretized dam body contains 3941 hexahedral elements and 4902 nodes. To consider the ground radiation damping effect, spring and damping elements were set at the truncated boundary of the dam foundation to form a viscoelastic artificial boundary. The discretized dam foundation contains 64895 elements and 69378 nodes. Furthermore, to consider the interaction between the dam and the reservoir under dynamic conditions, the reservoir area was discretized and modeled using acoustic elements in ABAQUS. The reservoir area contains 21528 three-dimensional acoustic elements and 24516 nodes. Corresponding boundary conditions were set at the free surface of the reservoir water, the reservoir tail, and the fluid-structure interaction interface. Finally, a three-dimensional dynamic analysis model of the Dagangshan arch dam was established, as shown below. Figure 2 As shown.

[0060] A method for calculating quantitative indicators of digital twins was developed. The calculations were performed using PyCharm IDE and PyTorch 2.5.1 based on Python 3.10 in an Ubuntu 16.04 environment. The server configuration was as follows: CPU: Intel Core i7-1700; GPU: Nvidia RTX 4070 SUPER.

[0061] Then, a sensor network optimization algorithm was established. The dam deformation and stress time series calculated using finite element analysis, along with the existing monitoring network layout, were input. The number of additional sensing points to be optimized was set, and sensing optimization calculations were performed. The calculations were conducted using PyCharm IDE and PyTorch 2.5.1 based on Python 3.10 in an Ubuntu 16.04 environment. The server configuration was as follows: CPU: Intel Core i7-1700; GPU: Nvidia RTX 4070 SUPER. Comparison of quantification metrics for different methods is provided. Figure 3 As shown.

[0062] Finally, based on the optimized sensor network, a Long Short-Term Memory (LSTM) neural network is used. The input is the calculated value corresponding to the network nodes of the source domain sensor arrangement, and the output is the calculated value corresponding to all finite element network nodes of the dam body. Training is performed using PyCharm IDE and PyTorch 2.5.1 based on Python 3.10 in an Ubuntu 16.04 environment. The server configuration is as follows: CPU: Intel Core i7-1700, GPU: Nvidia RTX 4070 SUPER. The learning rate is set to 0.001, the batch size is 64, the Adam optimizer is selected to update network parameters, the number of iterations is set to 500, and the model has 192 hidden layers. A comparison of the twin accuracy before and after optimization using different methods is provided. Figure 4 and 5 As shown.

[0063] Therefore, the sensor network-optimized digital twin method for dam structures provided in this embodiment, when applied to high arch dam projects, can effectively optimize the sensor network perception of the dam's digital twin sensing model. Compared with traditional technical specifications, it can significantly improve twin accuracy, increasing deformation twin accuracy by 45.76% and stress twin accuracy by 40.96% during earthquakes. Furthermore, the method of this invention can also improve the modal observability of the digital twin sensing model for dam structures. Compared with traditional technical specifications, it can significantly improve the average modal observability by 43.30%, enabling a more accurate and real-time reflection of the dam's deformation and stress twinning processes.

[0064] In summary, the method of the present invention has significant advantages in terms of the overall perception capability of dam structure, the efficiency of monitoring information utilization, and the reliability of digital twin perception model. Its good technical effect mainly comes from the proposed key technical solutions and their synergistic effects.

[0065] It should be noted that current dam structural safety monitoring still mainly relies on measurement point layout methods based on engineering experience and statistical analysis or local data-driven models. These methods, under conditions of limited spatial distribution of measurement points, struggle to fully characterize the overall structural state of the dam. This is especially true in complex stress structures such as high arch dams, where the dam body and foundation are subjected to long-term high stress, multi-field coupling, and strongly nonlinear conditions. Traditional point-based monitoring cannot effectively perceive the evolution of potential damage. Furthermore, existing digital twin methods often focus on repeated calibration of numerical models, lacking a unified description of the perception mechanism. This makes it difficult to quantitatively assess the coverage of the digital twin model to the actual structural state under limited monitoring conditions, thus restricting its reliability and stability in engineering safety assessments.

[0066] To address the aforementioned issues, this invention treats the dam as a high-dimensional dynamic system from the perspective of systems and control theory, and introduces compressed sensing theory to uniformly model the "sensing-reconstruction" process of digital twins. By mining the low-dimensional structural features of the dam's structural response under suitable conditions, stable reconstruction of key state quantities such as dam deformation and stress is achieved under constraints on the number and deployment of monitoring points. This theoretical framework provides a rigorous mathematical foundation for the sensing capabilities of digital twin models, transforming the global structural sensing under limited monitoring conditions from an empirical problem into a quantifiable and optimizable engineering problem.

[0067] Furthermore, this invention constructs a perception index system for dam digital twins, using modal observability and reconstructive stability as quantitative indicators. This enables a quantitative characterization of the ability of different perception network configurations to represent the overall structural state of the dam. Based on this, by combining QR decomposition with a greedy search strategy, key locations of monitoring points are selected and progressively optimized, achieving the synergistic integration of traditional monitoring facilities and novel perception methods (such as remote sensing and unmanned inspection). Without significantly increasing monitoring costs, this invention effectively improves the perception network's ability to capture key structural modes and potential damage information, fundamentally improving the perception accuracy of the digital twin model in responding to complex spatial conditions.

[0068] Compared to traditional pure numerical simulation methods, this invention effectively reduces the model's sensitivity to uncertainties in material parameters and boundary conditions by incorporating actual monitoring information and applying constraints within a compressed sensing framework. This avoids large-scale, high-frequency numerical calculations and significantly improves the real-time performance and stability of the digital twin model during engineering operation. Compared to data-driven methods that rely solely on monitoring point data, the proposed method achieves high-precision reconstruction of the entire dam's structural state under sparse observation conditions, exhibiting stronger generalization ability and adaptability to extreme conditions. Therefore, this invention realizes the transformation of dam structural safety monitoring from traditional limited measurement point monitoring to full-domain digital twin perception, which can more accurately and stably reflect the deformation and stress evolution process of the dam under normal operation, extreme working conditions and seismic action. It provides a well-founded, highly feasible and economically efficient intelligent monitoring solution for dam structural safety assessment and risk early warning, and has significant engineering application value and promotion prospects.

[0069] Example 3: This embodiment provides a digital twin sensing system for dam structures based on sensor network optimization. The system includes: The first building module is used to construct a three-dimensional finite element model of the dam structure; The sensing module is used to perform compressed sensing of the dam structure observation process to obtain a digital twin sensing model of the dam structure. The second building module is used to construct the quantitative indicators and optimization conditions for digital twin perception; The optimization module is used to optimize the sensor network of the dam structure based on the quantitative indicators and optimization conditions, combined with the modal matrix provided by the three-dimensional finite element model, to obtain the optimal sensor network. The training module is used to train and optimize the digital twin perception model using the monitoring data of the optimal sensor network to obtain a digital twin of the dam structure. The digital twin is used to acquire the global response data of the dam structure.

[0070] The sensing module includes: The compressed sensing unit is used to perform compressed sensing on the observation process of the dam structure to obtain a compressed sensing model of the dam structure. The mapping unit is used to map the compressed sensing model into a neural network that takes sparse observation vectors as input and high-dimensional state vectors as prediction outputs. The acquisition unit is used to obtain global state simulation values ​​and sparse observation simulation values ​​based on the numerical simulation results of the three-dimensional finite element model. The first building unit is used to construct the objective function of the neural network based on global state simulation values ​​and sparse observation simulation values; The second building unit is used to construct an effective measurement operator using the modal matrix of the dam structure and the observation matrix of the sensing network. The third building unit is used to construct cross-domain adaptation constraints and information integrity satisfaction conditions through effective measurement operators; The fourth building block is used to obtain the digital twin perception model based on neural networks, cross-domain adaptation constraints, and information integrity requirements.

[0071] The second building module includes: The fifth construction unit is used to construct the observability index of each mode through the perception network observation matrix and the mode matrix, wherein the perception network observation matrix is ​​constructed through a sensor network; The decomposition unit is used to perform singular value decomposition on the effective measurement operator and construct the condition number of the effective measurement operator; The sixth building block is used to use the observability index of each modality and the condition number of the effective measurement operator as quantitative indicators of digital twin perception. The seventh building unit is used to construct optimized conditions for digital twin perception by combining preset thresholds and using quantitative indicators.

[0072] The optimization module includes: Define the unit, which is used to define the optimization objective function of the sensor network; The analysis unit is used to perform modal analysis on the three-dimensional finite element model to obtain the modal matrix of the dam structure; The decomposition unit is used to perform QR decomposition on the transpose of the mode matrix. Combining the column pivot selection method and the optimization objective function, it obtains a first preset number of initial sensor nodes to obtain the current sensor set. The eighth building unit is used to construct a candidate node set by excluding the remaining nodes after excluding the current sensor set; The expansion unit is used to expand the sensor nodes by using a greedy algorithm based on quantitative indicators and optimization conditions, and to select a second preset number of expanded sensor nodes from the candidate node set. The update unit is used to update the observation matrix and effective measurement operators of the sensing network by initially selecting sensor nodes and expanding sensor nodes; The judgment unit is used to determine whether the optimization conditions are met by using the updated perception network observation matrix and effective measurement operators; The first processing unit is used to remove sensor nodes that do not meet the optimization conditions and reselect expanded sensor nodes if the optimization conditions are not met. The second processing unit is used to select the current sensor set as the optimal sensor network if the optimization conditions are met.

[0073] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0074] Example 4: Corresponding to the above method embodiments, this embodiment also provides a digital twin sensing device for dam structures based on sensor network optimization. The digital twin sensing device for dam structures based on sensor network optimization described below and the digital twin sensing method for dam structures based on sensor network optimization described above can be referred to in correspondence.

[0075] The sensor network-optimized dam structure digital twin sensing device 800 may include a processor 801 and a memory 802. The sensor network-optimized dam structure digital twin sensing device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0076] The processor 801 controls the overall operation of the sensor network-optimized dam structure digital twin sensing device 800 to complete all or part of the steps in the sensor network-optimized dam structure digital twin sensing method described above. The memory 802 stores various types of data to support the operation of the sensor network-optimized dam structure digital twin sensing device 800. This data may include, for example, instructions for any application or method operating on the sensor network-optimized dam structure digital twin sensing device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the sensor network-optimized dam structure digital twin sensing device 800 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, and an NFC module.

[0077] In an exemplary embodiment, the dam structure digital twin sensing device 800 based on sensor network optimization can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned dam structure digital twin sensing method based on sensor network optimization.

[0078] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the sensor network-optimized dam structure digital twin sensing method described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above. These program instructions may be executed by the processor 801 of the sensor network-optimized dam structure digital twin sensing device 800 to complete the sensor network-optimized dam structure digital twin sensing method described above.

[0079] Example 5: Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the above-described digital twin sensing method for dam structures based on sensor network optimization.

[0080] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the sensor network-optimized digital twin sensing method for dam structures described in the above method embodiments.

[0081] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0083] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A digital twin sensing method for dam structures based on sensor network optimization, characterized in that, include: Construct a three-dimensional finite element model of the dam structure; Compressed sensing was used to observe the dam structure, resulting in a digital twin sensing model of the dam structure. Quantitative indicators and optimization conditions for constructing digital twin perception; Based on the aforementioned quantitative indicators and optimization conditions, and combined with the modal matrix provided by the three-dimensional finite element model, the sensor network of the dam structure is optimized to obtain the optimal sensor network. The digital twin sensing model is trained and optimized using monitoring data from the optimal sensor network to obtain a digital twin of the dam structure. This digital twin is used to acquire global response data of the dam structure.

2. The digital twin sensing method for dam structures based on sensor network optimization according to claim 1, characterized in that, The compressed sensing process of observing the dam structure yields a digital twin sensing model of the dam structure, including: Compressed sensing was used to observe the dam structure, resulting in a compressed sensing model of the dam structure. The compressed sensing model is mapped to a neural network that takes sparse observation vectors as input and high-dimensional state vectors as prediction outputs. Numerical simulation results based on the three-dimensional finite element model are used to obtain global state simulation values ​​and sparse observation simulation values. The objective function of the neural network is constructed based on global state simulation values ​​and sparse observation simulation values. An effective measurement operator is constructed using the modal matrix of the dam structure and the observation matrix of the sensing network. Construct cross-domain adaptation constraints and information integrity conditions by using effective measurement operators; Based on neural networks, cross-domain adaptation constraints, and information integrity requirements, a digital twin perception model is obtained.

3. The digital twin sensing method for dam structures based on sensor network optimization according to claim 2, characterized in that, The compressed sensing process of observing the dam structure yields a compressed sensing model of the dam structure, including: Based on cybernetics, the state of the dam structure is decomposed into a modal combination form to obtain a high-dimensional state vector of the dam structure state. The high-dimensional state vector is composed of the modal matrix and time-varying modal coefficients of the dam structure. By linearly projecting the observation process of the dam structure, a linear relationship between the sparse observation vector and the high-dimensional state vector is obtained; By substituting the modal combination form of the high-dimensional state vector into the linear relationship, a compressed sensing model of the dam structure is obtained.

4. The digital twin sensing method for dam structures based on sensor network optimization according to claim 1, characterized in that, The quantitative indicators and optimization conditions for constructing digital twin perception include: The observability index of each mode is constructed by means of a sensor network observation matrix and a mode matrix, wherein the sensor network observation matrix is ​​constructed by means of a sensor network. Singular value decomposition is performed on the effective measurement operator to construct the condition number of the effective measurement operator; The observability index and the condition number of effective measurement operators for each modality are used as quantitative indicators for digital twin perception. By combining preset thresholds, optimization conditions for digital twin perception are constructed through quantitative indicators.

5. The digital twin sensing method for dam structures based on sensor network optimization according to claim 1, characterized in that, The process of optimizing the sensor network of the dam structure based on the quantitative indicators and optimization conditions, combined with the modal matrix provided by the three-dimensional finite element model, to obtain the optimal sensor network includes: Define the optimization objective function for the sensor network; Modal analysis was performed on the three-dimensional finite element model to obtain the modal matrix of the dam structure; The transpose of the modality matrix is ​​decomposed using QR decomposition. Combined with the column pivot selection method and the optimization objective function, a first preset number of initial sensor nodes are obtained to obtain the current sensor set. A candidate node set is constructed by excluding the remaining nodes after excluding the current sensor set; By using quantitative indicators and optimization conditions, a greedy algorithm is employed to expand the number of sensor nodes, and a second preset number of expanded sensor nodes are obtained from the candidate node set. The observation matrix and effective measurement operators of the sensing network are updated by initially selecting sensor nodes and expanding sensor nodes; Determine whether the optimization conditions are met by using the updated perception network observation matrix and effective measurement operators; If the optimization conditions are not met, remove the sensor nodes that do not meet the optimization conditions and reselect the expanded sensor nodes. If the optimization conditions are met, the current sensor set will be taken as the optimal sensor network.

6. The digital twin sensing method for dam structures based on sensor network optimization according to claim 5, characterized in that, The process involves expanding sensor nodes using a greedy algorithm based on quantified indicators and optimization conditions, selecting a second preset number of expanded sensor nodes from the candidate node set, including: For each candidate sensor node, a temporary sensor set is constructed by combining the current sensor set and the candidate sensor node, where the candidate sensor node is any sensor node in the candidate node set. The corresponding sensor network observation matrix and effective measurement operators are calculated based on a temporary sensor set. The quantitative indicators for each temporary sensor set are calculated using the perception network observation matrix and effective measurement operators; Substitute the quantitative indicators into the optimization objective function to calculate the optimization objective function value of the temporary sensor set corresponding to each candidate sensor node; Based on maximizing the objective function value and spatial distribution constraints, a candidate sensor node is selected as an extended sensor node, which is then added to the current sensor set and removed from the candidate node set. By using the updated current sensor set and candidate node set, the next expanded sensor node is selected until a second preset number of expanded sensor nodes are obtained.

7. A digital twin sensing system for dam structures based on sensor network optimization, characterized in that, include: The first building module is used to construct a three-dimensional finite element model of the dam structure; The sensing module is used to perform compressed sensing of the dam structure observation process to obtain a digital twin sensing model of the dam structure. The second building module is used to construct the quantitative indicators and optimization conditions for digital twin perception; The optimization module is used to optimize the sensor network of the dam structure based on the quantitative indicators and optimization conditions, combined with the modal matrix provided by the three-dimensional finite element model, to obtain the optimal sensor network. The training module is used to train and optimize the digital twin perception model using the monitoring data of the optimal sensor network to obtain a digital twin of the dam structure. The digital twin is used to acquire the global response data of the dam structure.

8. The digital twin sensing system for dam structures based on sensor network optimization according to claim 7, characterized in that, The sensing module includes: The compressed sensing unit is used to perform compressed sensing on the observation process of the dam structure to obtain a compressed sensing model of the dam structure. The mapping unit is used to map the compressed sensing model into a neural network with sparse observation vectors as input and high-dimensional state vectors as prediction outputs; The acquisition unit is used to obtain global state simulation values ​​and sparse observation simulation values ​​based on the numerical simulation results of the three-dimensional finite element model. The first building unit is used to construct the objective function of the neural network based on global state simulation values ​​and sparse observation simulation values; The second building unit is used to construct an effective measurement operator using the modal matrix of the dam structure and the observation matrix of the sensing network. The third building unit is used to construct cross-domain adaptation constraints and information integrity satisfaction conditions through effective measurement operators; The fourth building block is used to obtain the digital twin perception model based on neural networks, cross-domain adaptation constraints, and information integrity requirements.

9. The digital twin sensing system for dam structures based on sensor network optimization according to claim 7, characterized in that, The second building module includes: The fifth construction unit is used to construct the observability index of each mode through the perception network observation matrix and the mode matrix, wherein the perception network observation matrix is ​​constructed through the sensor network; The decomposition unit is used to perform singular value decomposition on the effective measurement operator and construct the condition number of the effective measurement operator; The sixth building block is used to use the observability index of each modality and the condition number of the effective measurement operator as quantitative indicators of digital twin perception. The seventh building unit is used to construct optimized conditions for digital twin perception by combining preset thresholds and using quantitative indicators.

10. The digital twin sensing system for dam structures based on sensor network optimization according to claim 7, characterized in that, The optimization module includes: Define the unit, which is used to define the optimization objective function of the sensor network; The analysis unit is used to perform modal analysis on the three-dimensional finite element model to obtain the modal matrix of the dam structure; The decomposition unit is used to perform QR decomposition on the transpose of the mode matrix. Combining the column pivot selection method and the optimization objective function, it obtains a first preset number of initial sensor nodes to obtain the current sensor set. The eighth building unit is used to construct a candidate node set by excluding the remaining nodes after excluding the current sensor set; The expansion unit is used to expand the sensor nodes by using a greedy algorithm based on quantitative indicators and optimization conditions, and to select a second preset number of expanded sensor nodes from the candidate node set. The update unit is used to update the observation matrix and effective measurement operators of the sensing network by initially selecting sensor nodes and expanding sensor nodes; The judgment unit is used to determine whether the optimization conditions are met by using the updated perception network observation matrix and effective measurement operators; The first processing unit is used to remove sensor nodes that do not meet the optimization conditions and reselect expanded sensor nodes if the optimization conditions are not met. The second processing unit is used to select the current sensor set as the optimal sensor network if the optimization conditions are met.

Images