Construction Method of Unity3D Bridge Digital Twin Platform Integrated with Midas-Civil Calculation Function

Abstract

The present invention relates to the technical field of the construction of a digital twin platform for bridges, and specifically discloses a method for constructing a Unity3D digital twin platform for bridges integrating Midas-Civil calculation functions. Through digital twin technology, real-time monitoring and evaluation of bridges are carried out to improve the structural safety and maintenance efficiency of bridges. Stress distribution data of bridges are collected in real time through bridge monitoring devices, and stress concentration areas are analyzed in the digital twin platform; the crack propagation rate and stress gradient change values are extracted to construct a comprehensive feature vector, and a machine learning model is used to evaluate the current state of the bridge, and high-risk areas are identified according to the evaluation results; by predicting the future crack propagation trend and stress distribution changes in low-risk areas, the maintenance strategy of the bridge is dynamically optimized; based on the risk assessment results, a resource optimization allocation model is constructed to achieve the optimal allocation and dynamic adjustment of bridge maintenance resources, ensuring the accurate implementation of the maintenance strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of constructing a digital twin platform for bridges, and particularly to a method for constructing a Unity3D bridge digital twin platform integrating the calculation function of Midas-Civil. Background Art

[0002] With the continuous increase of modern traffic demands, bridges, as key infrastructure, their safety and stability are crucial for the operation of the traffic system. To ensure that bridges maintain good performance during long-term use, traditional bridge inspection methods mainly rely on manual inspections and regular evaluations. However, traditional manual inspection methods have certain limitations, such as uncertain periodicity, dependence on personnel experience, and low inspection efficiency, and cannot accurately reflect the health status of the bridge structure in real time. Therefore, with the development of technology, more and more bridge monitoring and health assessment methods have started to introduce modern digital technologies, especially digital twin technology and virtual simulation technology. Digital twin technology can realize real-time monitoring and simulation analysis of physical entities by constructing virtual models of real physical entities, and is widely used in the health monitoring of engineering structures. Combining big data, the Internet of Things, and artificial intelligence technologies, the digital twin platform can not only obtain and analyze data in real time, but also perform structural prediction and optimization in the virtual platform, providing data support for maintenance decisions. At the same time, Midas-Civil, as a structural analysis and design software widely used in the field of civil engineering, has powerful calculation and simulation functions, and can provide important support in the process of bridge structure design, analysis, and optimization.

[0003] The existing technologies have the following deficiencies:

[0004] Firstly, many existing bridge digital twin platforms fail to effectively integrate the calculation functions of advanced structural analysis software, such as Midas-Civil, resulting in the platform being unable to provide high-precision structural analysis and dynamic optimization. Without the support of accurate calculations, the platform often has difficulty dealing with complex bridge health assessment and structural optimization problems, especially in predicting and evaluating complex situations such as crack propagation and stress concentration. Secondly, existing platforms usually lack cross-platform real-time interaction capabilities, making it difficult to closely combine real-time data with virtual models, so that decision-makers lack sufficient visual support and real-time feedback when making bridge maintenance decisions. Moreover, many existing technologies are difficult to dynamically adjust maintenance strategies and resource allocation, usually relying on static rules for risk assessment and maintenance decisions, and cannot fully respond to the rapid changes in the bridge health status. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for constructing a Unity3D bridge digital twin platform integrating the calculation function of Midas-Civil to solve the problems in the above background.

[0006] The object of the present invention can be achieved by the following technical solutions:

[0007] A method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil calculation function, comprising the following steps:

[0008] S1: Real-time collect the stress distribution characteristic data of the bridge through bridge monitoring devices, and analyze the stress distribution characteristics in the digital twin platform, identify and label the stress concentration areas;

[0009] S2: Based on the determined stress concentration areas, extract the crack change rate characteristics and stress distribution gradient characteristics, and calculate the crack propagation rate and stress gradient change value respectively;

[0010] S3: Construct a comprehensive feature vector with the crack propagation rate and stress gradient change value as input data, train and establish a machine learning model to evaluate the current state of the bridge;

[0011] S4: According to the model evaluation results, divide the bridge into high-risk areas and low-risk areas, and highlight the high-risk areas in the digital twin platform;

[0012] S5: Predict the future crack propagation trend and stress distribution change in the low-risk areas, and dynamically optimize the bridge maintenance strategy according to the prediction results.

[0013] As a further solution of the present invention: The analysis of the stress distribution characteristics specifically includes:

[0014] Standardize the collected stress data, calculate the relative position characteristics and stress differences of each measurement point, where the relative position characteristic is the spatial Euclidean distance between two measurement points, and the stress difference is the absolute value of the stress difference;

[0015] Based on the relative position characteristics and stress differences, construct a distance matrix, comprehensively considering the spatial distance and stress differences.

[0016] As a further solution of the present invention: Identifying and labeling the stress concentration areas specifically includes:

[0017] Use the K-Means clustering algorithm to perform clustering analysis on the distance matrix, set the number of clusters to 2, respectively representing the normal area and the stress concentration area, and the clustering result outputs the category label of each measurement point. If the label is 2, it means the corresponding area is the stress concentration area, and if the label is 1, it means the corresponding area is the normal area.

[0018] As a further solution of the present invention: The extraction of the crack change rate characteristics and stress distribution gradient characteristics specifically includes:

[0019] Apply wavelet transform to the standardized stress data in the stress concentration area, select Haar wavelet for wavelet decomposition, and decompose the stress signal into multiple frequency bands;

[0020] The crack change rate is the ratio of the stress change in the area. Calculate the ratio of the maximum stress fluctuation in the stress concentration area to the average stress change in the stress concentration area to obtain the crack change rate;

[0021] Output the calculated crack change rate feature;

[0022] Calculate the stress distribution gradient through spatial derivatives. The calculation expression is: ; where, is the gradient of the stress in space, represents the spatial derivative of the stress in the axis direction, represents the spatial derivative of the stress in the axis direction, represents the spatial derivative of the stress in the axis direction;

[0023] In the high-frequency detail coefficients after wavelet transform, obtain the stress distribution gradient by performing spatial derivative processing on the wavelet detail coefficients. The stress gradient is expressed as the local gradient of the wavelet detail coefficients. The calculation expression is: ; where, represents the gradient of the layer wavelet, representing the local change of the stress distribution, represents the number of layers of wavelet transform, represents the total number of layers of wavelet transform, represents the stress distribution gradient;

[0024] Output the stress gradient feature according to the calculated stress distribution gradient feature.

[0025] As a further solution of the present invention: Calculate the crack propagation rate according to the crack change rate feature, specifically including:

[0026] Perform wavelet reconstruction on the crack change rate feature data, extract the components of the stress signal at different scales to obtain wavelet components , where, is the spatial position parameter, represents the number of layers of wavelet transform. Sum up the wavelet components of all wavelet layers to obtain the wavelet reconstruction signal ;

[0027] Based on the wavelet component , apply local weighted regression to calculate the crack propagation rate. The calculation expression is: ; where, represents the crack propagation rate, represents the total number of layers of wavelet transform, represents the weighting coefficient of the wavelet component of the represents the total number of wavelet components, represents the number of wavelet components, represents the derivative in space of the signal after wavelet reconstruction.

[0028] As a further solution of the present invention: According to the stress distribution gradient characteristics, calculate the stress gradient change value, specifically including:

[0029] Represent the stress gradient distribution characteristic signal as a three-dimensional space signal, perform wavelet decomposition on this signal, decompose it into multiple wavelet detail components of different scales, and extract the wavelet detail coefficients of each layer as the expression form of the multi-scale stress gradient signal;

[0030] Combined with the gradient information of each layer of wavelet decomposition, calculate the stress gradient change value, and the calculation expression is: ;

[0031] In the formula, represents the stress gradient change value, represents the volume of the stress concentration area, represents the spatial coordinate, represents the gradient modulus value of the wavelet detail coefficient of the represents the weight of the represents the number of layers of wavelet transform, represents the total number of layers of wavelet transform.

[0032] As a further solution of the present invention: Construct a comprehensive feature vector from the crack propagation rate and the stress gradient change value, and use it as input data to train and establish a machine learning model, specifically including:

[0033] Construct a comprehensive feature vector from the crack propagation rate and the stress gradient change value, use the comprehensive feature vector as the input of the machine learning model, the machine learning model takes predicting the risk coefficient of the bridge with each group of comprehensive feature vectors as the prediction target, takes minimizing the sum of squared errors between the predicted risk coefficients and the true risk coefficients of all samples as the training target, trains the machine learning model until the sum of squared prediction errors reaches convergence and then stops the model training, determines the risk coefficient of the bridge according to the model output result and evaluates the safety of the bridge structure, where the machine learning model is a polynomial regression model.

[0034] As a further solution of the present invention: Divide the bridge into high-risk areas and low-risk areas according to the model evaluation results, specifically including:

[0035] Based on the output of the model, determine whether the risk coefficient of the bridge in the current monitoring period is greater than or equal to the preset threshold. If so, it is recorded as a high-risk area; if not, it is recorded as a low-risk area.

[0036] As a further solution of the present invention: predicting the future crack propagation trend and stress distribution change in the low-risk area specifically includes:

[0037] Use the edge detection algorithm to extract the lace area in the bridge stress distribution, generate a binary image of the stress lace, and calculate the lace texture features such as edge smoothness, lace density, and edge direction gradient distribution; based on multi-frame stress distribution data, perform time series analysis on the extracted lace texture features, and construct time series features through the sliding window technique, including average gradient, maximum change rate, and frequency distribution; construct a prediction model, use the time series features and texture features of the stress lace as inputs to predict the crack propagation rate and stress distribution change in the low-risk area; train the deep learning model with historical stress distribution data and actual observed values of crack propagation, and optimize the model parameters using the mean square error loss function and the structural similarity loss function until the prediction error converges; based on the trained model, predict the future crack propagation trend and stress distribution change in the low-risk area, calculate the risk coefficient of the bridge in the future monitoring period based on the predicted future crack propagation trend and stress distribution change, and compare the risk coefficient of the bridge in the future monitoring period with the preset threshold to determine whether the risk coefficient of the bridge in the future monitoring period is greater than or equal to the preset threshold. If so, it is recorded as a high-risk area; if not, it is recorded as a low-risk area, and mark the high-risk area according to the judgment result.

[0038] As a further solution of the present invention: dynamically optimizing the bridge maintenance strategy according to the prediction result specifically includes:

[0039] Obtain the risk coefficient of the bridge, construct a resource optimization allocation model based on the risk coefficient, with the goal of minimizing the maintenance cost, and the constraints include: total resources, regional priority, and minimum maintenance requirements, to obtain the optimal resource allocation for each area; based on real-time data feedback, automatically adjust the resource allocation strategy when the crack propagation or stress change exceeds the threshold; generate a visual scheme of resource allocation through the digital twin platform, highlight the high-risk areas, and provide support for decision-making.

[0040] The beneficial effects of the present invention:

[0041] (1) By collecting the stress distribution data of the bridge in real time and applying wavelet transform for multi-scale analysis of the stress signal, the present invention can effectively extract the crack propagation rate and the stress gradient change value, and these features can accurately reflect the health state of the bridge structure. Combining with machine learning models, especially polynomial regression models, the present invention comprehensively evaluates the crack propagation rate and the stress gradient change value, so as to achieve accurate identification and evaluation of the high-risk areas of the bridge. Compared with traditional evaluation methods, the technical means of the present invention can more comprehensively capture the dynamic stress evolution process of the bridge under the influence of multiple factors such as load and environmental changes, providing higher timeliness and accuracy for risk prediction. Through such accurate risk assessment results, early warning can be realized in bridge health monitoring, providing a scientific basis for optimizing bridge maintenance strategies, ensuring that high-risk areas can be maintained in a timely and priority manner, significantly improving the reliability of maintenance decisions, and minimizing the safety hazards caused by crack propagation and stress concentration, effectively extending the service life of the bridge.

[0042] (2) The present invention realizes the optimal allocation of bridge maintenance resources by constructing a resource optimization allocation model and combining with a dynamic adjustment mechanism. This model takes into account multiple factors, including the risk coefficient of the bridge, the maintenance budget, and the priority of each area, so as to ensure that high-risk areas can obtain resource support first, and at the same time maximize the maintenance effect under limited resource conditions. By introducing a dynamic adjustment mechanism, the maintenance plan can be adjusted according to the real-time changes in the health state of the bridge structure. Especially when the crack propagation rate or the stress distribution gradient changes significantly, the system will automatically optimize the resource allocation to ensure that key areas receive timely and sufficient attention. This flexible and efficient resource allocation strategy not only significantly reduces the maintenance cost, but also improves the economy and sustainability of bridge maintenance, ensuring the long-term safety and functionality of the bridge. Through scientific data support and decision optimization, the present invention effectively realizes the precise allocation of maintenance resources, enabling limited resources to play the greatest role, improving the efficiency and effect of bridge maintenance work, reducing the situation of over-maintenance or under-maintenance, and promoting the modernization and refinement of bridge maintenance management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below with reference to the accompanying drawings.

[0044] Figure 1 It is a specific step flow block diagram of the construction method of the Unity3D bridge digital twin platform integrating the Midas-Civil calculation function of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0046] Please refer to Figure 1 As shown, the present invention is a method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil calculation functions, including the following steps:

[0047] S1: Real-time collect the stress distribution characteristic data of the bridge through bridge monitoring devices, and analyze the stress distribution characteristics in the digital twin platform to identify and mark the stress concentration areas;

[0048] S2: Based on the determined stress concentration areas, extract the crack change rate characteristics and stress distribution gradient characteristics, and calculate the crack propagation rate and stress gradient change value respectively;

[0049] S3: Use the crack propagation rate and stress gradient change value to construct a comprehensive feature vector as input data, train and establish a machine learning model to evaluate the current state of the bridge;

[0050] S4: According to the model evaluation results, divide the bridge into high-risk areas and low-risk areas, and highlight the high-risk areas in the digital twin platform;

[0051] S5: Predict the future crack propagation trend and stress distribution change in the low-risk areas, and dynamically optimize the bridge maintenance strategy according to the prediction results.

[0052] In S1, real-time collect the stress distribution characteristic data of the bridge through bridge monitoring devices, and analyze the stress distribution characteristics in the digital twin platform to identify and mark the stress concentration areas, specifically including:

[0053] The bridge monitoring devices include strain gauges (stress sensors) and displacement sensor devices. The strain gauges are installed on the components of the bridge and can measure the minute deformations on the surface of the structure caused by load changes in real time and convert them into stress values. These stress data are transmitted to the data center in real time through the acquisition system connected to the sensors. These sensors transmit the collected data to the digital twin platform through wireless or wired networks to obtain the stress distribution characteristics in real time.

[0054] Real-time collect the stress distribution data of each part of the bridge through bridge monitoring devices. The data includes the stress values and corresponding spatial coordinates of multiple stress measurement points;

[0055] Standardize the collected stress data so that stress data at different scales are comparable. The standardized stress values are convenient for further analysis and eliminate the deviation of data under different magnitudes.

[0056] Based on the standardized stress data, calculate the relative position characteristics and stress differences of each measurement point. The relative position characteristic of each measurement point is the spatial Euclidean distance between two measurement points, and the stress difference is the absolute value of the difference in stress values between two measurement points.

[0057] Construct a distance matrix according to the relative position characteristics and stress differences of each measurement point: ; where and represent two measurement points, and are weight coefficients that adjust the influence of spatial distance and stress difference in distance calculation;

[0058] Use the K-Means clustering algorithm to perform clustering analysis on the distance matrix. The goal is to classify data points with similar stresses into the same cluster. The number of clusters , respectively representing the normal area and the stress concentration area. The clustering result is output as the category of each measurement point. If the output is 1, it is recorded as the normal area. If the output is 2, it is recorded as the stress concentration area.

[0059] The process of using the K-Means clustering algorithm to perform clustering analysis on the distance matrix is as follows:

[0060] Initialize cluster centers. Usually, the cluster centers are selected by random selection or using the K-Means++ algorithm. Assign each measurement point to the nearest cluster center, calculate the new center point of each cluster, that is, the mean value of all points within the cluster, reassign the points and repeat the clustering step until the cluster centers no longer change or change very little, reaching the convergence condition. After clustering is completed, output the category labels of each measurement point.

[0061] In S2, based on the determined stress concentration area, extract the crack change rate characteristics and stress distribution gradient characteristics, and calculate the crack propagation rate and stress gradient change value respectively, specifically including:

[0062] Apply wavelet transform to the standardized stress data in the stress concentration area, select Haar wavelet for wavelet decomposition, and decompose the stress signal into multiple frequency bands. Among them, the high-frequency component reflects the rapid change of stress and can capture the stress change during the crack propagation process;

[0063] The calculation expression for selecting Haar wavelet for wavelet decomposition is: ;

[0064] In the formula, represents the standardized stress data, is the mother wavelet, is the scale factor that controls the scalability of the wavelet function, is the spatial position parameter, represents the translation variable during the integration process;

[0065] The crack change rate is the ratio of the stress change within the region, and the calculation expression is: ; where is the maximum stress fluctuation within the stress concentration region, is the average stress change within the stress concentration region, represents the crack change rate;

[0066] It should be noted that: the crack change rate refers to the relationship between the stress fluctuation in the stress concentration region and the crack propagation;

[0067] Using the high-frequency part after wavelet transform, by calculating the standard deviation of the high-frequency detail coefficients, the crack change rate characteristics are further extracted;

[0068] Output the calculated crack change rate characteristics, and based on the crack change rate and the health status of the stress concentration region, conduct subsequent structural health assessments.

[0069] The stress distribution gradient is calculated through spatial derivatives, and the calculation expression is: ; where is the gradient of the stress in space, represents the spatial derivative of the stress in the axis direction, represents the spatial derivative of the stress in the axis direction, represents the spatial derivative of the stress in the axis direction;

[0070] It should be noted that: the stress distribution gradient represents the rate of change of the stress in space.

[0071] In the high-frequency detail coefficients after wavelet transform, by performing spatial derivative processing on the wavelet detail coefficients, the stress distribution gradient is obtained. The stress gradient is expressed as the local gradient of the wavelet detail coefficients, and the calculation expression is: ; where represents the gradient of the layer wavelet, representing the local change of the stress distribution, represents the number of layers of the wavelet transform, represents the total number of layers of the wavelet transform, represents the stress distribution gradient;

[0072] Based on the calculated stress distribution gradient characteristics and combined with the changes in the stress concentration area, evaluate the stress distribution of the structure and the risk of crack propagation, output the stress gradient characteristics, and provide data support for structural health assessment.

[0073] Perform wavelet reconstruction on the crack change rate characteristic data to extract the components of the stress signal at different scales and obtain the wavelet components , where is the spatial position parameter, represents the number of layers of wavelet transform. Sum up the wavelet components of all wavelet layers to obtain the wavelet reconstruction signal ;

[0074] Based on the wavelet component , calculate the crack propagation rate using local weighted regression. The calculation expression is: ; where represents the crack propagation rate, represents the total number of layers of wavelet transform, represents the weighted coefficient of the wavelet component of the th layer, represents the total number of wavelet components, represents the number of wavelet components, represents the derivative of the signal after wavelet reconstruction in space; among them, the weighted coefficient is calculated according to the local weighted regression formula. The calculation expression is: ; where represents the size of the weighted window;

[0075] It should be noted that: through the weighted regression result, the crack propagation rate obtained represents the crack propagation rate at each spatial point and can accurately evaluate the risk of crack propagation based on the stress data and local regression trend.

[0076] Represent the stress gradient distribution characteristic signal as a three-dimensional space signal, perform wavelet decomposition on this signal, decompose it into multiple wavelet detail components at different scales, and extract the wavelet detail coefficients of each layer as the expression form of the multi-scale stress gradient signal;

[0077] Combined with the gradient information of each layer of wavelet decomposition, establish a stress gradient change model, quantify the characteristics of the gradient signal, and obtain the stress gradient change value to represent the overall change of the stress distribution;

[0078] Perform local weighted regression processing on the calculated stress gradient change value, use the kernel function to perform weighted smoothing on the spatial points to improve the robustness and accuracy of the calculation results, and output the optimized stress gradient change value;

[0079] Output the optimized stress gradient change value as the final result for bridge health assessment and judgment of high-risk areas;

[0080] The specific calculation expression of the stress gradient change value is: ;

[0081] In the formula, represents the stress gradient change value, represents the volume of the stress concentration area, represents the spatial coordinates, represents the gradient modulus value of the wavelet detail coefficient of the th layer, represents the weight of the th layer, indicating the contribution degree of different wavelet layers to the stress gradient change, represents the number of layers of wavelet transform,

[0082] In S3, construct a comprehensive feature vector from the crack propagation rate and the stress gradient change value as input data, train and establish a machine learning model to evaluate the current state of the bridge, specifically including:

[0083] Based on the stress concentration area of the bridge, extract the crack propagation rate and the stress gradient change value, eliminate the influence between different feature magnitudes through data standardization processing, and combine the standardized crack propagation rate and stress gradient change value into a comprehensive feature vector, representing the key feature parameters of the bridge structure health;

[0084] Select the polynomial regression model as the machine learning model, take the risk coefficient of the bridge as the prediction target, and establish a prediction model of the mapping feature vector and the risk coefficient by minimizing the error between the model prediction value and the true risk coefficient;

[0085] Take the comprehensive feature vector as the input sample and the risk coefficient as the output label, define the loss function as the minimization target of the sum of the squares of the errors between the predicted risk coefficients and the true values of all samples; use the gradient descent method to iteratively optimize the model parameters, update the weight and bias parameters in the polynomial regression model until the error function converges or reaches the preset threshold to stop training;

[0086] Adopt the cross-validation method to evaluate the performance of the trained model, calculate the mean square error between the model prediction value and the true risk coefficient with the validation set data, and ensure the prediction accuracy and generalization ability of the model on the test set;

[0087] Use the trained model to predict the comprehensive feature vector generated from the real-time monitoring data of the bridge, and output the risk coefficient of the bridge; evaluate the safety of the overall structure of the bridge according to the output result, and set corresponding maintenance measures in combination with the risk coefficient.

[0088] In S4, according to the model evaluation results, the bridge is divided into high-risk areas and low-risk areas, and the high-risk areas are highlighted in the digital twin platform, specifically including:

[0089] According to the output of the model, it is judged whether the risk coefficient of the bridge in the current monitoring period is greater than or equal to the preset threshold. If so, it is recorded as a high-risk area; if not, it is recorded as a low-risk area, and the high-risk area is highlighted in the digital twin platform.

[0090] In S5, the future crack propagation trend and stress distribution change in the low-risk area are predicted, and according to the prediction results, the bridge maintenance strategy is dynamically optimized, specifically including:

[0091] Use the edge detection algorithm to extract the lace area in the bridge stress distribution, generate a binary image of the stress lace, and calculate lace texture features such as edge smoothness, lace density, and edge direction gradient distribution; based on multi-frame stress distribution data, perform time series analysis on the extracted lace texture features, and construct time series features through the sliding window technique, including average gradient, maximum change rate, and frequency distribution; construct a prediction model, which is a joint model combining a multi-scale convolutional neural network and a long short-term memory network; use the time series features and texture features of the stress lace as inputs to predict the crack propagation rate and stress distribution change in the low-risk area; train the deep learning model with historical stress distribution data and actual crack propagation observations, and optimize the model parameters using the mean square error loss function and the structural similarity loss function until the prediction error converges; based on the trained model, predict the future crack propagation trend and stress distribution change in the low-risk area, calculate the risk coefficient of the bridge in the future monitoring period based on the predicted future crack propagation trend and stress distribution change, and compare the risk coefficient of the bridge in the future monitoring period with the preset threshold to judge whether the risk coefficient of the bridge in the future monitoring period is greater than or equal to the preset threshold. If so, it is recorded as a high-risk area; if not, it is recorded as a low-risk area, and mark the high-risk area according to the judgment result.

[0092] According to the prediction results, calculate the optimal allocation plan of bridge maintenance resources and dynamically adjust the bridge maintenance plan, specifically including:

[0093] Based on the risk coefficient, construct a resource optimization allocation model with the goal of minimizing the bridge maintenance cost. The constraint conditions include that the total resource volume does not exceed the budget, resources are preferentially allocated to high-risk areas, and the resource allocation in each area meets the minimum maintenance requirements. By solving this optimization model, the resource allocation amount for each high-risk area is obtained.

[0094] Dynamic adjustment strategy generation: According to the calculation results of the resource optimization allocation model, obtain the optimal resource allocation plan for each high-risk area, and set the dynamic adjustment threshold. When the crack propagation rate or the stress distribution gradient changes exceed the preset threshold, the system will automatically update the resource allocation strategy again to give priority to ensuring the maintenance of high-risk areas.

[0095] Real-time feedback optimization: After implementing bridge maintenance, collect bridge status data in real time, evaluate the maintenance effect, recalculate the comprehensive risk score according to the evaluation results, and dynamically optimize the resource allocation strategy in combination with the reinforcement learning algorithm to improve the resource allocation efficiency.

[0096] Result visualization output: Through the bridge digital twin platform, generate the visualization results of the maintenance plan, display the optimized resource allocation plan and the dynamic adjustment plan, highlight the high-risk areas, and provide support for scientific decision-making. The present invention includes a digital twin model of Unity3D, a digital twin system backend server based on Flask and Midas-Civil, a human-computer interaction interface implemented based on the basic components and windows in Unity3D, and a visualization interaction platform based on Unity WebGL and Vue.js. The digital twin system backend server based on Flask and Midas-Civil communicates with the bridge monitoring device to obtain the monitoring data during the bridge construction or operation and maintenance process, and stores it in the Mysql database through the corresponding script; after obtaining the change in the monitored bridge load size, through the GUI interaction module in the Web-based digital twin platform, after selecting through the interface, simultaneously transmit the calculation instruction to the Midas-Civil software, and automatically export the relevant calculation data and store it in the Mysql database after the calculation is completed; finally, on the Web side based on Vue.js, display the monitored stress data and the real-time calculated data on the Echart icon at the same time, realize the modification of the Midas-Civil calculation parameters and real-time calculation controlled by the Web side, and the real-time comparison of the calculation data and the monitoring data.

[0097] Working principle of the present invention: The stress distribution characteristic data of the bridge are collected in real time through strain gauges and displacement sensors on the bridge, and analyzed in the digital twin platform to identify and mark the stress concentration areas. Then, based on the standardized stress data of the stress concentration areas, the wavelet transform is used to extract the crack change rate characteristics, and the stress gradient change value is calculated through spatial derivatives, so as to obtain the crack propagation rate and the stress gradient change value. Next, these two characteristic data are constructed into a comprehensive characteristic vector as the input sample, and a polynomial regression model is used for machine learning training to predict the current risk coefficient of the bridge and evaluate the bridge health status. According to the risk assessment results, the bridge is divided into high-risk and low-risk areas, and the high-risk areas will be highlighted in the digital twin platform for timely corresponding maintenance measures. For the low-risk areas, combined with the edge detection algorithm and time series analysis method, the future crack propagation trend and stress distribution changes are predicted, and the maintenance strategy of the bridge is dynamically optimized based on the prediction results. Finally, by establishing a resource optimization allocation model, the maintenance resources are reasonably allocated, and the optimization is fed back and the maintenance plan is adjusted in real time to ensure that the high-risk areas are given priority. Through reinforcement learning and the visualization platform, the resource allocation scheme is dynamically adjusted and optimized to improve the efficiency and accuracy of bridge maintenance and provide a scientific basis for decision-making.

[0098] The above has described in detail one embodiment of the present invention, but the content described is only the preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions, characterized in that: The following steps are involved: S1: The stress distribution characteristic data of the bridge is collected in real time through the bridge monitoring equipment, and the stress distribution characteristics are analyzed in the digital twin platform to identify and mark the stress concentration areas; S2: Based on the determined stress concentration area, the crack change rate characteristics and stress distribution gradient characteristics are extracted, and the crack extension rate and stress gradient change value are calculated respectively; S3: The crack propagation rate and stress gradient change values ​​are used to construct a comprehensive feature vector as input data to train and establish a machine learning model to evaluate the current state of the bridge, including: The crack propagation rate and stress gradient change value are constructed as a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model predicts the risk factor of the bridge with each group of comprehensive feature vectors as the prediction target, and minimizes the sum of square errors between the predicted risk coefficients of all samples and the true risk coefficients as the training target. The machine learning model is trained until the sum of squares of the prediction errors reaches convergence, and the model training is stopped. The risk factor of the bridge is determined according to the output results of the model and the safety of the bridge structure is evaluated. Among them, the machine learning model is a polynomial regression model; According to the output of the model, it is determined whether the risk coefficient of the bridge in the current monitoring period is greater than or equal to the preset threshold. If so, it is recorded as a high-risk area; if not, it is recorded as a low-risk area; S4: Based on the model assessment results, the bridge is divided into high-risk areas and low-risk areas, and the high-risk areas are highlighted in the digital twin platform; S5: Predict future crack growth trends and stress distribution changes in low-risk areas, and dynamically optimize bridge maintenance strategies based on the prediction results.

2. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 1, characterized in that: The analysis of stress distribution characteristics specifically includes: The collected stress data are standardized and the relative position characteristics and stress difference of each measuring point are calculated, where the relative position characteristics are the spatial Euclidean distance between two measuring points and the stress difference is the absolute value of the stress difference; Based on the relative position characteristics and stress differences, a distance matrix is ​​constructed, comprehensively considering the spatial distance and stress differences.

3. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 2, characterized in that: Identify and annotate areas of stress concentration, including: The K-Means clustering algorithm is used to perform cluster analysis on the distance matrix. The number of clusters is set to 2, representing the normal area and the stress concentration area respectively. The clustering result outputs the category label of each measurement point. If the label is 2, it means that the corresponding area is the stress concentration area. If the label is 1, it means that the corresponding area is the normal area.

4. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 1, characterized in that: The extraction of crack change rate characteristics and stress distribution gradient characteristics specifically includes: Wavelet transform is applied to the standardized stress data of the stress concentration area, and Haar wavelet is selected for wavelet decomposition to decompose the stress signal into multiple frequency bands; The crack change rate is the ratio of the stress change in the region. The crack change rate is obtained by calculating the ratio of the maximum stress fluctuation in the stress concentration region to the average stress change in the stress concentration region. Output the calculated crack change rate characteristics; The stress distribution gradient is calculated by spatial derivative, and the calculation expression is: ;in, is the gradient of stress in space, Indicates stress in The spatial derivative in the axial direction, Indicates stress in The spatial derivative in the axial direction, Indicates stress in The spatial derivative in the axial direction; In the high-frequency detail coefficients after wavelet transformation, the stress distribution gradient is obtained by performing spatial derivative processing on the wavelet detail coefficients. The stress gradient is expressed as the local gradient of the wavelet detail coefficient. The calculation expression is: ;in, Indicates The gradient of the layer wavelet represents the local variation of stress distribution, represents the number of wavelet transform layers, represents the total number of wavelet transform layers, represents the stress distribution gradient; According to the calculated stress distribution gradient characteristics, the stress gradient characteristics are output.

5. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 4, characterized in that: According to the crack change rate characteristics, the crack extension rate is calculated, including: The crack change rate characteristic data is reconstructed by wavelet, and the components of the stress signal at different scales are extracted to obtain the wavelet components. ,in, is the spatial position parameter, Represents the number of layers of wavelet transform, and the wavelet components of all wavelet layers are Perform sum calculation to obtain the wavelet reconstructed signal ; Based on wavelet components , local weighted regression is used to calculate the crack growth rate, and the calculation expression is: ;in, represents the crack growth rate, represents the total number of wavelet transform layers, Indicates The weighting coefficients of the layer wavelet components, represents the total number of wavelet components, represents the number of wavelet components, Represents the spatial derivative of the signal after wavelet reconstruction.

6. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 4, characterized in that: According to the stress distribution gradient characteristics, the stress gradient change value is calculated, including: The characteristic signal of stress gradient distribution is expressed as a three-dimensional spatial signal, and the signal is decomposed by wavelet into multiple wavelet detail components of different scales, and the wavelet detail coefficients of each layer are extracted as the expression form of multi-scale stress gradient signal; Combined with the gradient information of each layer decomposed by wavelet, the stress gradient change value is calculated, and the calculation expression is: ; In the formula, represents the stress gradient change value, represents the volume of the stress concentration area, represents the spatial coordinates, Indicates The gradient modulus of the layer wavelet detail coefficient, Indicates The weight of the layer, represents the number of wavelet transform layers, Indicates the total number of wavelet transform layers.

7. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 1, characterized in that: The prediction of future crack expansion trends and stress distribution changes in low-risk areas specifically includes: The edge detection algorithm is used to extract the lace area in the bridge stress distribution, generate a binary image of the stress lace, and calculate the edge smoothness, lace density and lace texture features of the edge direction gradient distribution; based on multi-frame stress distribution data, the extracted lace texture features are analyzed in time series, and the time series features are constructed through sliding window technology, including average gradient, maximum change rate and frequency distribution; a prediction model is constructed, and the time series features and texture features of the stress lace are used as input to predict the crack propagation rate and stress distribution changes in low-risk areas; the deep learning model is tested through historical stress distribution data and actual observations of crack propagation. Conduct training and use mean square error loss function and structural similarity loss function to optimize model parameters until the prediction error converges; predict future crack expansion trends and stress distribution changes in low-risk areas based on the trained model; calculate the risk coefficient of bridges in the future monitoring period based on the predicted future crack expansion trends and stress distribution changes; compare the risk coefficient of bridges in the future monitoring period with the preset threshold to determine whether the risk coefficient of bridges in the future monitoring period is greater than or equal to the preset threshold; if so, record it as a high-risk area; if not, record it as a low-risk area; mark the high-risk area based on the judgment result.

8. The method for constructing a Unity3D bridge digital twin platform integrating Midas-Civil computing functions according to claim 1, characterized in that: The bridge maintenance strategy is dynamically optimized according to the prediction results, specifically including: The risk coefficient of the bridge is obtained, and a resource optimization allocation model is constructed based on the risk coefficient. The goal is to minimize the maintenance cost. The constraints include: total resources, regional priority, and minimum maintenance requirements, so as to obtain the optimal resource allocation for each area. Based on real-time data feedback, when crack expansion or stress change exceeds the threshold, the resource allocation strategy is automatically adjusted. A visualization plan for resource allocation is generated through the digital twin platform, and high-risk areas are highlighted to provide support for decision-making.

Images