Bridge structure toughness dynamic evaluation method and system based on digital twinning

Abstract

The invention relates to the field of bridge engineering, in particular to a bridge structure toughness dynamic evaluation method and system based on digital twinning, and the system firstly collects the environment, geometry and performance information of a bridge, and then establishes an accurate bridge digital twinning model; a bridge structure toughness prediction model is systematically constructed by using a space-time diagram convolutional network, and the performance degradation process of a bridge in extreme weather can be predicted; meanwhile, toughness evaluation indexes are formulated by the system, and accurate performance evaluation of real-time monitoring data is realized; through multi-source data fusion, the system improves the comprehensiveness and accuracy of data acquisition; the prediction model based on the space-time diagram convolutional network significantly improves the accuracy of performance degradation prediction; a scientific and efficient tool is provided for bridge maintenance and management, and safe operation of the bridge in extreme weather is effectively guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of bridge engineering, and particularly to a method and system for dynamically evaluating the resilience of bridge structures based on digital twins, which are used to monitor, evaluate, and predict in real time the performance degradation process and resilience state of bridge structures under extreme weather conditions, providing a scientific basis for bridge maintenance and management. Background Art

[0002] With the global climate change, extreme weather events such as typhoons, heavy rains, floods, etc. occur frequently, bringing huge challenges to the safe operation of bridge structures. Traditional bridge health monitoring methods mainly rely on regular inspections and limited sensing data, making it difficult to evaluate the performance state of bridge structures in real time and dynamically under extreme weather conditions. At the same time, existing bridge structure evaluation methods often only focus on static performance indicators of the structure, lacking dynamic evaluation of the structure's resilience and being difficult to reflect the ability of the bridge to resume normal functions after being impacted by extreme weather.

[0003] As a key technology of the cyber-physical system, digital twin technology has been widely applied in fields such as aerospace and manufacturing. However, the research on applying digital twin technology to the evaluation of bridge structure resilience is still in its infancy. The existing research has the following main problems: First, the sensing data is single, making it difficult to comprehensively reflect the real-time state of the bridge structure; second, the data processing method is simple, unable to fully mine the spatio-temporal correlation information contained in the data; third, the prediction model has low accuracy, making it difficult to accurately predict the performance degradation process of the bridge structure under extreme weather conditions; fourth, there is a lack of a reasonable resilience evaluation index system, making it difficult to quantitatively evaluate the resilience of the bridge structure.

[0004] Therefore, there is an urgent need to develop a method and system for dynamically evaluating the resilience of bridge structures based on digital twin technology to achieve real-time monitoring, accurate prediction, and dynamic evaluation of bridge structures under extreme weather conditions, providing technical support for the intelligent management and maintenance of bridges. Summary of the Invention

[0005] The present invention provides a method and system for dynamically evaluating the resilience of bridge structures based on digital twins, aiming to solve the problems of inaccurate and non-real-time performance evaluation of bridge structures under extreme weather conditions in the prior art. By establishing a digital twin model, the dynamic evaluation of the resilience of bridge structures is realized, providing a scientific basis for bridge maintenance and management.

[0006] The object of the present invention is to provide a method and system for dynamically evaluating the resilience of bridge structures based on digital twins, which realizes the prediction of the performance state and dynamic evaluation of the resilience of bridge structures under extreme weather conditions through multi-source data collection, digital twin model construction, spatio-temporal graph convolutional network prediction, and resilience index evaluation.

[0007] The present invention proposes a dynamic assessment method for the resilience of bridge structures based on digital twins, including:

[0008] A collection step, including obtaining bridge environmental information, geometric information, and performance information;

[0009] A modeling step, including establishing a digital twin model of the bridge based on the environmental information, geometric information, and performance information;

[0010] A prediction step, including constructing a prediction model for the resilience of the bridge structure based on a spatio-temporal graph convolutional network to predict the performance degradation process of the digital twin model of the bridge under extreme weather conditions;

[0011] An evaluation step, including constructing resilience evaluation indicators and performing real-time performance evaluation on real-time monitoring data according to the resilience evaluation indicators.

[0012] Preferably, the collection step specifically includes:

[0013] Collecting surface information and local structure information of different bridge areas through an unmanned aerial vehicle (UAV) visual inspection system;

[0014] Performing three-dimensional reconstruction on the surface information and local structure information and determining the parametric law under different conditions;

[0015] Deploying distributed fiber optic sensors to obtain the spatial distribution of bridge strain;

[0016] Combining the strain data at the deployment points with environmental information for correction to obtain true strain data;

[0017] Comparing and analyzing the true strain data with SAR images to construct a digital twin model that is synchronized and updated with the physical entity.

[0018] Preferably, the specific method for collecting surface information and local structure information of different bridge areas through an unmanned aerial vehicle (UAV) visual inspection system is:

[0019] Based on the geometric characteristics of each part of the bridge, determining the positions of the collection points and the collection angles of the collection points;

[0020] Based on the geometric relationships of the collection points, performing operations such as pixel alignment, scale transformation, image fusion, distortion correction, and image registration on the images obtained at each collection point, and combining photogrammetry technology to obtain the three-dimensional deformation and spatial deformation of different components such as bridge piers, bridge decks, and bridge bearings.

[0021] Preferably, the specific method for deploying distributed fiber optic sensors to obtain the spatial distribution of bridge strain is:

[0022] Calibrating and calibrating the deployed distributed fiber optic sensors to obtain the strain values at each measurement point under different conditions;

[0023] Strain data at different cross-sections and different depths along the line are obtained through fiber Bragg grating strain sensors and Raman optical signal fiber sensors;

[0024] Combined with the real-time nature of the strain data, the correlation between the bridge strain response and traffic loads is determined.

[0025] Preferably, the prediction step specifically includes:

[0026] Based on the three-dimensional and finite element models of the bridge surface and local areas, combined with the strain data of each measurement point, the spatio-temporal evolution law of the strain response is determined;

[0027] Based on the stochastic process model, the relationship between the strain value and the environment and loads is determined, and combined with the Bayesian inference process, the historical data is updated as prior information;

[0028] On the basis of the prior information, based on the spatio-temporal graph convolutional network, the bridge structural performance evolution equation is determined, and the prior mean and covariance matrix of the bridge structural performance evolution equation are obtained;

[0029] Combined with the Kalman filter algorithm, the posterior mean and covariance matrix of the bridge structural performance evolution equation are obtained.

[0030] Preferably, the specific method for determining the spatio-temporal evolution law of the strain response is:

[0031] Combined with the strain data of each measurement point, a bridge dynamic monitoring mechanism model with environmental loads as the input end and strain response as the output end is constructed, and the mathematical relationship between the environmental load spatial relationship matrix and the strain response is established.

[0032] Preferably, the specific method for updating the historical data as prior information is:

[0033] Based on the stochastic process model, the time series of the bridge structural performance is determined and satisfies the normal distribution;

[0034] The correlation degree between any two measurement points is determined through the autocorrelation function. When the distance is less than the correlation length, it is determined to be highly correlated, otherwise it is determined to be uncorrelated.

[0035] Preferably, the specific method for determining the bridge structural performance evolution equation based on the spatio-temporal graph convolutional network is:

[0036] Construct the correlation of traffic load data in time and space, and construct the time correlation and space correlation of traffic load data through the space correlation of historical traffic load data;

[0037] Based on the spatial local attention mechanism, the space correlation of traffic load data is constructed, and the weights of each sensor node and its adjacent nodes are defined;

[0038] Through the fully connected layer, softmax function and regression function, map the residual term to the traffic load data to obtain the residualized result of the traffic load data.

[0039] Preferably, the evaluation step specifically includes:

[0040] Calculate the average relative error of all measurement points according to the predicted values of each measurement point;

[0041] Combine the average relative error and determine the threshold through the average maximum likelihood estimator;

[0042] Based on the average relative error index, establish a resilience evaluation index and determine the relationship between the resilience evaluation index and the strain values of each measurement point;

[0043] Combine Monte Carlo simulation to predict the change process of the bridge structure health state over time under extreme weather conditions, and realize dynamic resilience assessment.

[0044] The dynamic resilience assessment system for bridge structures based on digital twin includes:

[0045] A data acquisition module for obtaining bridge environmental information, geometric information and performance information;

[0046] A digital twin module for establishing a bridge digital twin model based on the environmental information, geometric information and performance information;

[0047] A prediction module for constructing a bridge structure resilience prediction model based on a spatio-temporal graph convolutional network to predict the performance degradation process of the bridge digital twin model under extreme weather conditions;

[0048] An evaluation module for constructing a resilience evaluation index and performing real-time performance evaluation on real-time monitoring data according to the resilience evaluation index;

[0049] A cloud server for storing the digital twin model, prediction model and evaluation results, and providing a remote access interface;

[0050] A decision-making module for generating a resilience improvement plan based on the evaluation results and seeking the optimal maintenance strategy and traffic control plan through a multi-objective optimization algorithm.

[0051] The beneficial effects of the present invention include:

[0052] 1. Through multi-source data fusion, the comprehensiveness and accuracy of data acquisition are improved;

[0053] 2. Based on the prediction model of the spatio-temporal graph convolutional network, the accuracy of performance degradation prediction is improved;

[0054] 3. A reasonable resilience evaluation index system is established to realize the quantitative evaluation of the bridge structure resilience;

[0055] 4. The availability of the evaluation results and the decision-making efficiency are improved through cloud services and remote access. Brief Description of the Drawings

[0056] Figure 1 It is a flowchart of the method for dynamically evaluating the resilience of a bridge structure based on digital twin according to the present invention;

[0057] Figure 2 It is a detailed flowchart of the acquisition step of the present invention;

[0058] Figure 3 It is a schematic diagram of data acquisition by the UAV vision inspection system of the present invention;

[0059] Figure 4 It is a schematic diagram of the layout of distributed optical fiber sensors of the present invention;

[0060] Figure 5 It is an architecture diagram of the bridge structure resilience prediction model based on spatio-temporal graph convolutional network of the present invention;

[0061] Figure 6 It is a flowchart of calculating the resilience evaluation index of the present invention;

[0062] Figure 7 It is an overall architecture diagram of the system of the present invention. Detailed Embodiments

[0063] Please refer to the attached Figures 1-7 , and the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0064] Embodiment 1: Method for Dynamically Evaluating the Resilience of a Bridge Structure Based on Digital Twin

[0065] Referring to Figure 1 , the method for dynamically evaluating the resilience of a bridge structure based on digital twin provided by the present invention includes an acquisition step, a modeling step, a prediction step, and an evaluation step.

[0066] In this embodiment, the acquisition step includes obtaining bridge environmental information, geometric information, and performance information. Preferably, as Figure 2 shown, the acquisition step specifically includes: collecting surface information and local structure information of different bridge areas through a UAV vision inspection system; performing three-dimensional reconstruction on the surface information and local structure information, and determining the parametric law under different conditions; laying out distributed optical fiber sensors to obtain the spatial distribution of bridge strain; correcting the strain data at the layout points in combination with environmental information to obtain real strain data; comparing and analyzing the real strain data with SAR images to construct a digital twin model that is synchronously updated with the physical entity.

[0067] In a specific embodiment of the present invention, asFigure 3 As shown in Figure 3 , the specific method for collecting surface information and local structure information of different bridge areas through the UAV visual inspection system is as follows: Based on the geometric characteristics of each part of the bridge, determine the positions of the collection points and the collection angles of the collection points; Based on the geometric relationships of the collection points, perform operations such as pixel alignment, scale transformation, image fusion, distortion correction, and image registration on the images obtained at each collection point, and combine photogrammetry technology to obtain the three-dimensional deformation and spatial deformation of different components such as bridge piers, bridge decks, and bridge bearings. In practical applications, a DJI Phantom 4 RTK UAV can be used, equipped with a 1-inch CMOS sensor to shoot 4K ultra-high-definition video, thus ensuring the clarity and accuracy of the image data. At the same time, to ensure the comprehensiveness of data collection, the flight trajectory of the UAV should be designed to cover all key parts of the bridge. Typical collection angles include 0°, °, and 90° to obtain all-round structural information.

[0068] As Figure 4 As shown in Figure 4 , the specific method for obtaining the spatial distribution of bridge strain by deploying distributed fiber optic sensors is as follows: Calibrate and calibrate the deployed distributed fiber optic sensors to obtain the strain values of each measurement point under different conditions; Through fiber Bragg grating strain sensors and Raman optical signal fiber sensors, obtain strain data at different cross-sections and different depths along the line; Combine the real-time nature of the strain data to determine the correlation between bridge strain response and traffic load. In actual deployment, the deployment density of fiber optic sensors is usually one measurement point per 5 meters to ensure the spatial resolution of data collection. At the same time, the calibration of the sensors is carried out using standard strain gauges, and the calibration error is controlled within ±2με to ensure the accuracy of the data.

[0069] In a preferred embodiment of the present invention, the method for correcting the strain data at the deployment points in combination with environmental information to obtain real strain data uses the following formula:

[0070]

[0071] Where is the strain value caused by the instantaneous temperature effect at the i-th measurement point, is the initial temperature of the i-th measurement point, is the strain correction value at the i-th measurement point, T max is the moment when the maximum load appears, t is the time variable, and f(t) is a time-related function, usually in the form of a Gaussian function:

[0072]

[0073] Where σ is a time-related parameter, determined according to the measured data, and the typical value is 0.5 hours. This is because the response time of the bridge structure to temperature changes is usually about 1 hour, and taking half of it can better capture the dynamic impact of temperature changes.

[0074] The specific method of comparing the true strain data with the SAR image and constructing a digital twin model that is synchronized and updated with the physical entity is as follows: RGB images at different times of different bridge structures are obtained based on time-frequency conversion, and SAR images at different times are obtained based on image processing. In practical applications, the short-time Fourier transform (STFT) is used for time-frequency conversion, the window length is set to 8 sampling points, and the overlap rate is 50%. Such parameter settings can balance the frequency resolution and time resolution and are suitable for capturing the dynamic response characteristics of the bridge structure.

[0075] In this embodiment, the prediction step includes constructing a bridge structure toughness prediction model based on a spatio-temporal graph convolutional network to predict the performance degradation process of the bridge digital twin model under extreme weather conditions. Preferably, as Figure 5 shown, the prediction step specifically includes: determining the spatio-temporal evolution law of the strain response based on the three-dimensional and finite element models of the bridge surface and local areas, combined with the strain data of each measurement point; determining the relationship between the strain value and the environment and load based on a stochastic process model, and updating the historical data to prior information in combination with the Bayesian inference process; based on the prior information, determining the bridge structure performance evolution equation based on the spatio-temporal graph convolutional network, and obtaining the prior mean and covariance matrix of the bridge structure performance evolution equation; combining the Kalman filter algorithm to obtain the posterior mean and covariance matrix of the bridge structure performance evolution equation.

[0076] In a specific embodiment of the present invention, the specific method for determining the spatio-temporal evolution law of the strain response is: combining the strain data of each measurement point to construct a bridge dynamic monitoring mechanism model with environmental load as the input end and strain response as the output end, and establishing the mathematical relationship between the environmental load spatial relationship matrix and the strain response. This model can be expressed as:

[0077]

[0078] where ε i (t) is the strain response of the i-th measurement point, G(S,t) is the environmental load spatial relationship matrix, and S is the spatial coordinate variable represents the initial strain data of the i-th measurement point of the bridge structure, and t is the time variable. In practical applications, the environmental load spatial relationship matrix G(S,t) is usually represented in polynomial form:

[0079]

[0080] where a j (t) is the time-related coefficient, n is the order of the polynomial, and generally takes values from 3 to 5, which can better fit the spatial distribution law of the environmental load.

[0081] The specific method for updating historical data into prior information is as follows: Based on a stochastic process model, determine the time series of the bridge structural performance, which follows a normal distribution; determine the correlation degree between any two measurement points through the autocorrelation function. When the distance is less than the correlation length, it is determined to be highly correlated, otherwise it is determined to be uncorrelated. The specific stochastic process model can be expressed as:

[0082]

[0083] μ t = f(ε1, ε2,..., ε n , t),

[0084] where P(t) is the prior mean of the bridge structural performance at the t-th moment, is the prior variance of the bridge structural performance at the t-th moment, ε i is the strain response of the i-th measurement point, f is the empirical distribution model obtained from historical data, representing the i-th measurement point of the bridge structure. The autocorrelation function can be expressed as:

[0085]

[0086] where d ij is the distance between any two measurement points, L c is the correlation length. When d ij < L c , it is highly correlated, otherwise it is uncorrelated. In practical applications, the correlation length L c is usually set to 10%-20% of the bridge span length because within this range, the strain responses of the bridge structure often show strong correlations.

[0087] In a preferred embodiment of the present invention, the specific method for determining the bridge structural performance evolution equation based on the spatio-temporal graph convolutional network is as follows: Construct the correlation of traffic load data in time and space, and construct the time correlation and space correlation of traffic load data through the space correlation of historical traffic load data; construct the space correlation of traffic load data based on the spatial local attention mechanism, and define the weights of each sensor node and its adjacent nodes; map the residual term to the traffic load data through the fully connected layer, softmax function, and regression function to obtain the residualized result of the traffic load data.

[0088] The mathematical expression form of the spatio-temporal graph convolutional network is:

[0089]

[0090] where H (l) is the node feature matrix of the l-th layer, is the adjacency matrix with self-loops added, is the degree matrix of, W (l) is the weight matrix of the l-th layer, and σ is the activation function. In practical applications, ReLU is usually adopted as the activation function:

[0091] σ(x) = max(0, x),

[0092] The spatial local attention mechanism can be expressed as:

[0093]

[0094] where α ij represents the attention coefficient between node i and node j, a is the attention vector, W is the weight matrix, and h i is the feature of node i, is the neighbor set of node i, ∥ represents the concatenation operation, and Leaky ReLU is the activation function. In practical applications, the negative slope parameter of LeakyReLU is usually set to 0.2, and this value is widely regarded as the best practice in most deep learning tasks.

[0095] Combined with the Kalman filter algorithm, the formulas for the posterior mean and covariance matrix of the bridge structure performance evolution equation are:

[0096]

[0097] where is the posterior prediction value of the bridge structure performance, Σ t is the prior variance of the bridge structure performance at different measurement points, μ t is the prior mean of the prediction equation, H is the time variable, z t is the observed value, and R is the observed noise covariance matrix. In practical applications, the observed noise covariance matrix R is usually set as a diagonal matrix, and the values of the diagonal elements are determined according to the accuracy of the sensor, and the typical value is 10 -6 to 10 -4 , which can balance the relationship between measurement noise and model prediction.

[0098] In this embodiment, the evaluation step includes constructing a resilience evaluation index and performing real-time performance evaluation on the real-time monitoring data according to the resilience evaluation index. Preferably, as Figure 6 shown, the evaluation step specifically includes: calculating the average relative error of all measurement points according to the predicted values of each measurement point; combining the average relative error and determining the threshold through the average maximum likelihood estimator; establishing a resilience evaluation index based on the average relative error index and determining the relationship between the resilience evaluation index and the strain values of each measurement point; combining Monte Carlo simulation to predict the change process of the bridge structure health state over time under extreme weather conditions to achieve dynamic resilience evaluation.

[0099] In a specific embodiment of the present invention, the formula for calculating the average relative error of all measurement points is as follows:

[0100]

[0101] Wherein, is the posterior prediction value of the bridge structural performance at the i-th measurement point at the t-th moment, and ε i is the strain prediction value of the i-th measurement point, and n is the number of measurement points. In practical applications, the typical threshold of the average relative error is 5%. When the MARE exceeds this value, it indicates that there is a significant difference between the model prediction and the actual measurement, and the reasons need to be further investigated.

[0102] Combined with the average relative error, the formula for determining the threshold through the average maximum likelihood estimator is:

[0103]

[0104] Wherein, represents the posterior variance of the bridge structure, E is caused by the expectation of the prediction model error, λ is the number of strain measurement points at the t-th moment, n is the number of measurement points, is the posterior value of the averaged bridge structural performance, represents the posterior prediction value of the j-th measurement point, is the initial value of the j-th measurement point at the t-th moment, and ε is the predicted strain mean value at the t-th moment. In practical applications, the typical threshold of AMAPE is 3%. This value is obtained based on a large amount of experimental data analysis and can effectively distinguish normal fluctuations and abnormal changes.

[0105] Based on the average relative error index, the formula for establishing the ductility evaluation index and determining the relationship between the ductility evaluation index and the strain values of each measurement point is:

[0106] RI = 1 - MARE,

[0107]

[0108] Wherein, RI is the average relative error index, and ε i is the strain value of the i-th measurement point at the t-th moment, δ s is the error between the bridge structure and the target structure, is the variance between the bridge structure and the target structure, is the posterior prediction value of the bridge structural performance. The value range of the ductility index RI is [0, 1]. When RI is close to 1, it indicates that the bridge structure has high ductility; when RI is close to 0, it indicates that the ductility of the bridge structure is low and reinforcement or other intervention measures are required.

[0109] Combined with Monte Carlo simulation, the specific method for predicting the change process of the health state of a bridge structure over time under extreme weather is as follows: By generating a large number of random samples, simulating the environmental loads and traffic loads under extreme weather conditions, and inputting them into the prediction model, the change trajectory of the bridge structure performance over time can be obtained. In practical applications, usually 10,000 samples are generated to ensure the reliability of the statistical results.

[0110] Embodiment 2: Dynamic Evaluation System for Bridge Structure Resilience Based on Digital Twin

[0111] As Figure 7 shown, the dynamic evaluation system for bridge structure resilience based on digital twin provided by the present invention includes an acquisition module, a digital twin module, a prediction module, an evaluation module, a cloud server, and a decision-making module.

[0112] The acquisition module is used to obtain bridge environmental information, geometric information, and performance information. Preferably, the acquisition module includes an unmanned aerial vehicle (UAV) visual inspection system and a distributed optical fiber sensor network. The UAV visual inspection system is equipped with a high-definition camera for collecting image information of the bridge surface and local structures; the distributed optical fiber sensor network is arranged at key parts of the bridge for collecting strain data.

[0113] The digital twin module is used to establish a bridge digital twin model based on the environmental information, geometric information, and performance information. Preferably, the digital twin module includes a 3D reconstruction unit, a parameterization law determination unit, a strain data correction unit, and a SAR image comparison and analysis unit. The 3D reconstruction unit processes the image information collected by the UAV to construct a 3D model of the bridge; the parameterization law determination unit analyzes the change laws of the bridge structure under different conditions; the strain data correction unit corrects the original strain data according to the environmental information; the SAR image comparison and analysis unit compares and analyzes the corrected strain data with the SAR image to update the digital twin model.

[0114] The prediction module is used to construct a bridge structure resilience prediction model based on a spatio-temporal graph convolutional network to predict the performance degradation process of the bridge digital twin model under extreme weather conditions. Preferably, the prediction module includes a spatio-temporal evolution law determination unit, a Bayesian inference unit, a spatio-temporal graph convolutional network processing unit, and a Kalman filter processing unit. The spatio-temporal evolution law determination unit combines the strain data of each measurement point to determine the spatio-temporal evolution law of the strain response; the Bayesian inference unit updates the historical data as prior information; the spatio-temporal graph convolutional network processing unit constructs the prior mean and covariance matrix of the bridge structure performance evolution equation; the Kalman filter processing unit calculates the posterior mean and covariance matrix.

[0115] The evaluation module is used to construct resilience evaluation indicators and conduct real-time performance evaluation on real-time monitoring data according to the resilience evaluation indicators. Preferably, the evaluation module includes an average relative error calculation unit, a threshold determination unit, a resilience index establishment unit, and a Monte Carlo simulation unit. The average relative error calculation unit calculates the average relative error of all measurement points; the threshold determination unit determines the threshold through the average maximum likelihood estimator; the resilience index establishment unit establishes resilience evaluation indicators; the Monte Carlo simulation unit predicts the change process of the health state of the bridge structure under extreme weather conditions.

[0116] The cloud server is used to store the digital twin model, the prediction model, and the evaluation results, and provide a remote access interface. Preferably, the cloud server includes a database module, a calculation module, and a communication module. The database module stores the collected original data and processing results; the calculation module executes complex calculation tasks; the communication module realizes data exchange with other modules.

[0117] The decision-making module is used to generate a resilience improvement plan based on the evaluation results and seek the optimal maintenance strategy and traffic control plan through a multi-objective optimization algorithm. Preferably, the decision-making module includes a plan generation unit and an optimization algorithm unit. The plan generation unit generates a preliminary maintenance plan according to the evaluation results; the optimization algorithm unit uses a multi-objective genetic algorithm to comprehensively consider factors such as material properties, maintenance costs, and traffic control duration to seek the optimal resilience improvement plan.

[0118] In this embodiment, the working process of the system includes: First, the acquisition module acquires the environmental information, geometric information, and performance information of the bridge; Second, the digital twin module establishes a bridge digital twin model based on the acquired information; Then, the prediction module constructs a prediction model based on the spatio-temporal graph convolutional network to predict the performance degradation process of the bridge under extreme weather conditions; Next, the evaluation module conducts real-time evaluation on the bridge structure according to the resilience evaluation indicators; Finally, the decision-making module generates a resilience improvement plan based on the evaluation results and provides remote access and decision support through the cloud server.

[0119] The method and system for dynamically evaluating the resilience of a bridge structure based on digital twin of the present invention realizes the prediction of the performance state and dynamic evaluation of the resilience of the bridge structure under extreme weather conditions through multi-source data acquisition, digital twin model construction, spatio-temporal graph convolutional network prediction, and resilience index evaluation, providing strong technical support for the intelligent management and maintenance of the bridge. Compared with the prior art, the present invention has the advantages of comprehensive data acquisition, high prediction accuracy, reasonable evaluation indicators, etc., and has important application value in improving the safety and extending the service life of the bridge.

[0120] It should be noted that the above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A dynamic assessment method of bridge structure resilience based on digital twins, characterized in that: The following steps are involved: The acquisition steps include obtaining bridge environment information, geometric information, and performance information; A modeling step includes establishing a digital twin model of the bridge based on the environmental information, geometric information and performance information; A prediction step includes constructing a bridge structure resilience prediction model based on a spatiotemporal graph convolutional network to predict the performance degradation process of the bridge digital twin model under extreme weather conditions; The evaluation step includes constructing a toughness evaluation index and performing real-time performance evaluation on the real-time monitoring data according to the toughness evaluation index.

2. The bridge structure resilience dynamic assessment method based on digital twin according to claim 1 is characterized in that: The collection steps specifically include: The surface information and local structure information of different bridge areas are collected through the UAV visual inspection system; Reconstructing the surface information and local structure information in three dimensions, and determining parameterization rules under different conditions; Distributed optical fiber sensors are deployed to obtain the spatial distribution of bridge strain; The strain data of the deployment points are corrected in combination with the environmental information to obtain the real strain data; The real strain data is compared and analyzed with the SAR image to construct a digital twin model that is updated synchronously with the physical entity.

3. The bridge structure resilience dynamic assessment method based on digital twin according to claim 2 is characterized in that: The specific method of collecting surface information and local structure information of different bridge areas through the UAV visual inspection system is as follows: Determine the location and angle of the collection points based on the geometric features of each part of the bridge; Based on the geometric relationship of the collection points, the images obtained from each collection point are subjected to operations such as pixel alignment, scale transformation, image fusion, distortion correction and image registration. Combined with photogrammetry technology, the three-dimensional deformation and spatial deformation of different components such as piers, bridge deck systems and bridge supports are obtained.

4. The bridge structure resilience dynamic assessment method based on digital twin according to claim 2 is characterized in that: The specific method of deploying distributed optical fiber sensors to obtain the spatial distribution of bridge strain is: The distributed optical fiber sensors are calibrated and calibrated to obtain the strain values ​​of each measuring point under different conditions; The strain data of different sections and depths along the line are obtained through fiber Bragg grating strain sensors and Raman optical signal fiber sensors; Combined with the real-time nature of strain data, the correlation between bridge strain response and traffic load is determined.

5. The bridge structure resilience dynamic assessment method based on digital twin according to claim 1 is characterized in that: The prediction step specifically includes: Based on the three-dimensional and finite element models of the bridge surface and local parts, combined with the strain data of each measuring point, the temporal and spatial evolution law of the strain response is determined; Based on the random process model, the relationship between strain value and environment and load is determined, and combined with the Bayesian reasoning process, the historical data is updated as prior information; Based on the prior information, the bridge structure performance evolution equation is determined based on the spatiotemporal graph convolutional network, and the prior mean and covariance matrix of the bridge structure performance evolution equation are obtained; Combined with the Kalman filter algorithm, the posterior mean and covariance matrix of the bridge structure performance evolution equation are obtained.

6. The bridge structure resilience dynamic assessment method based on digital twin according to claim 5 is characterized in that: The specific method to determine the temporal and spatial evolution law of strain response is: Combined with the strain data of each measuring point, a bridge dynamic monitoring mechanism model is constructed with environmental load as input and strain response as output, and the mathematical relationship between the spatial relationship matrix of environmental load and strain response is established.

7. The bridge structure resilience dynamic assessment method based on digital twin according to claim 5 is characterized in that: The specific method of updating historical data to prior information is: Based on the random process model, the time series of bridge structure performance is determined and satisfies the normal distribution; The degree of correlation between any two measuring points is determined by the autocorrelation function. When the distance is less than the correlation length, it is judged to be highly correlated, otherwise it is judged to be unrelated.

8. The bridge structure resilience dynamic assessment method based on digital twin according to claim 5 is characterized in that: The specific method for determining the evolution equation of bridge structure performance based on spatiotemporal graph convolutional network is: Construct the temporal and spatial correlation of traffic load data, and construct the temporal and spatial correlation of traffic load data through spatial correlation of historical traffic load data; The spatial correlation of traffic load data is constructed based on the spatial local attention mechanism, and the weight of each sensor node and its adjacent nodes is defined; Through the fully connected layer, softmax function and regression function, the residual term is mapped to the traffic load data to obtain the residual result of the traffic load data.

9. The bridge structure resilience dynamic assessment method based on digital twin according to claim 1 is characterized in that: The evaluation steps specifically include: According to the predicted value of each measuring point, the average relative error of all measuring points is calculated; Combined with the mean relative error, the threshold is determined by averaging the maximum likelihood estimator; Based on the average relative error index, the toughness evaluation index is established, and the relationship between the toughness evaluation index and the strain value of each measuring point is determined; Combined with Monte Carlo simulation, the change process of the health status of the bridge structure under extreme weather conditions over time is predicted to achieve dynamic assessment of resilience.

10. Bridge structure resilience dynamic assessment system based on digital twin, including: Acquisition module, used to obtain bridge environment information, geometric information and performance information; A digital twin module, used to establish a digital twin model of the bridge based on the environmental information, geometric information and performance information; A prediction module is used to construct a bridge structure resilience prediction model based on a spatiotemporal graph convolutional network to predict the performance degradation process of the bridge digital twin model under extreme weather conditions; An evaluation module, used to construct a toughness evaluation index and perform real-time performance evaluation on real-time monitoring data according to the toughness evaluation index; A cloud server, used to store the digital twin model, prediction model and evaluation results, and provide a remote access interface; The decision-making module is used to generate a resilience improvement plan based on the evaluation results and seek the optimal maintenance strategy and traffic control plan through a multi-objective optimization algorithm.

Images