Cross-iron swivel bridge construction monitoring method and system based on digital twinning

Abstract

The invention discloses the technical field of civil engineering and bridge construction, and particularly discloses a method and a system for monitoring construction of a cross-iron swivel bridge based on digital twinning. The method comprises the following steps: S1, acquiring real-time monitoring data of a cross-iron swivel bridge; the monitoring data comprises stress, displacement, temperature and cable force; s2, preprocessing the monitoring data, wherein the preprocessing comprises data cleaning, standardization and de-noising processing; s3, inputting the preprocessed monitoring data into a monitoring model of a digital twin platform, processing the monitoring data by the monitoring model in combination with a time sequence algorithm, generating a change trend of the real-time monitoring data and current state data, and outputting the data to an early warning module; and S4, comparing the current state data and the change trend data generated by the monitoring model with a preset early warning threshold value, and generating an early warning notification. The invention further comprises a system, intelligent management of bridge construction can be achieved, the construction risk is reduced, and the construction safety and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of civil engineering and bridge construction, and in particular to a construction monitoring method and system for a cross-railway swivel bridge based on digital twins. Background Art

[0002] With the continuous development of technology in the field of civil engineering, bridge construction monitoring has gradually become an important part of ensuring structural safety and construction quality. The construction monitoring platform provides a scientific basis for civil engineering construction by monitoring various parameters (such as displacement, stress, temperature, etc.) during the construction process. However, the existing technology still faces many challenges in practical application and urgently needs to be improved and optimized.

[0003] Information technology is increasingly used in construction monitoring platforms, including front-end and back-end code programming, database management, and application development. By collecting, storing, processing, and visualizing construction data, information technology improves the intelligence level of monitoring. In addition, the introduction of Web development technology enables users to access the monitoring platform through a browser. The combination of front-end technologies such as HTML, CSS, and JavaScript with server-side languages ​​such as Python and Java and databases such as MySQL and MongoDB provides flexibility for platform development and deployment.

[0004] Building Information Modeling (BIM) technology, as an important innovation following CAD technology, has triggered a revolution in the field of engineering construction. BIM technology has significantly improved the efficiency of project design, construction, and management through digital modeling software. However, the existing bridge construction monitoring platform still has obvious deficiencies in data integration, real-time, and ease of operation. The data formats and types generated by different sensors and devices vary, which leads to challenges in data integration and processing, affecting the uniformity and availability of monitoring results. In addition, many monitoring platforms are unable to provide real-time data, limiting timely responses to potential risks.

[0005] The existing system also has problems such as complex operation and unfriendly user interface, which not only affects the user experience, but also may lead to human errors and data omissions. The high initial investment, including the cost of sensors, data acquisition equipment, related software and cloud services, also hinders the application and promotion of small and medium-sized enterprises. At the same time, the lack of data analysis and prediction capabilities limits the monitoring platform's support for decision-making and risk assessment.

[0006] Although BIM technology has performed well in the bridge design and construction phase, its application in the operation and maintenance phase is still insufficient. The compatibility issues of BIM models, the cumbersomeness of maintenance and updates, and the risk of possible distortion of model information further limit its potential in bridge construction monitoring. In addition, the sensitive data of BIM models faces security risks of cyber attacks when stored in the cloud, which also puts higher requirements on the application of technology.

[0007] In short, the existing bridge construction monitoring technology still has many deficiencies in data processing capabilities, system usability, real-time performance, security, etc. With the continuous development of civil engineering, information technology, artificial intelligence and BIM technology, the development of a more intelligent, real-time, convenient and safe monitoring platform has become an urgent problem to be solved in the field of bridge construction. Summary of the invention

[0008] The purpose of the present invention is to overcome the problems of data integration difficulties, lack of real-time performance, complex operation and low safety in the prior art, and to provide a construction monitoring method and system for a cross-rail swivel bridge based on digital twin technology. The present invention can realize real-time monitoring and dynamic analysis by integrating data from different types of sensors, and timely discover and predict potential risks, thereby improving the safety and efficiency of the construction process. The present invention also introduces artificial intelligence technology to improve decision-making support capabilities and optimize construction management through data modeling and prediction functions. Combined with Web development technology and cloud computing, the present invention realizes cross-platform data visualization and sharing, and improves the convenience and coordination of user operations. At the same time, through the in-depth application of building information modeling (BIM), the present invention realizes dynamic information management and model updating throughout the construction process. Finally, the present invention provides an intelligent, efficient and safe monitoring solution for bridge construction.

[0009] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0010] A construction monitoring method for a cross-railway swivel bridge based on digital twins comprises the following steps:

[0011] S1. Obtaining real-time monitoring data of the cross-rail swivel bridge; the monitoring data includes stress, displacement, temperature and cable force;

[0012] S2. Preprocessing the monitoring data, including data cleaning, standardization and denoising;

[0013] S3. Input the preprocessed monitoring data into the monitoring model of the digital twin platform, the monitoring model processes the monitoring data in combination with the time series algorithm, generates the change trend and current status data of the real-time monitoring data, and outputs the data to the early warning module;

[0014] S4. Compare the current status data and change trend data generated by the monitoring model with the preset warning thresholds to generate a warning notification.

[0015] Preferably, the monitoring data is obtained through sensors deployed at the construction site of the cross-rail swivel bridge, and the sensors include strain gauges, displacement sensors, temperature sensors, and cable force monitoring equipment; the step of acquiring the monitoring data includes real-time monitoring of key interfaces, and using the principle of symmetry to predict monitoring data at other symmetrical positions. By deploying multiple types of sensors at the construction site and collecting stress, displacement, temperature, and cable force data in real time, the comprehensiveness and accuracy of monitoring can be ensured; using the principle of symmetry to predict monitoring data at symmetrical positions not only reduces the risk of data loss caused by sensor damage, but also improves the system's fault tolerance and monitoring efficiency.

[0016] Preferably, the preprocessing step in S2 includes normalizing the monitoring data and removing invalid data through an outlier detection algorithm. Normalization and outlier removal in the preprocessing stage can effectively improve data quality, avoid interference of invalid or abnormal data on monitoring results, provide a reliable data basis for subsequent prediction and analysis, and enhance the stability and accuracy of the system.

[0017] Preferably, the digital twin platform includes:

[0018] BIM model, used to describe the geometric structure and component information of the cross-railway swivel bridge;

[0019] Finite element model for analyzing the mechanical behavior of a cross-rail swivel bridge;

[0020] The monitoring model is used to generate real-time status data and change trend data based on the monitoring data, and output them to the early warning module.

[0021] By integrating the BIM model, finite element model and monitoring model through the digital twin platform, a comprehensive linkage between the bridge's geometric structure, mechanical behavior and real-time monitoring data is achieved; the monitoring model can intuitively display the deviation between the monitoring data and the theoretical model, help quickly identify structural anomalies, and improve the efficiency of construction management and the scientific nature of decision-making.

[0022] Preferably, the monitoring model realizes the processing and prediction of monitoring data by combining time series algorithm and multi-feature input data; the multi-feature data includes historical data, real-time monitoring data and theoretical model calculation results. Combining time series algorithm with multi-feature input data (including historical data, real-time data and theoretical model calculation results) can realize accurate prediction of monitoring data under complex working conditions; this multi-source data input mode improves the robustness and accuracy of prediction and enhances the intelligent analysis ability of the system.

[0023] Preferably, the time series algorithm uses long short-term memory network, random forest, recursive neural network, back propagation algorithm. Using a variety of time series algorithms, the most suitable algorithm can be flexibly selected to process the monitoring data, improve the system's ability to analyze complex nonlinear data, and ensure efficient and accurate prediction in different scenarios.

[0024] Preferably, the generation of the warning notification is based on the comparison results of the current state data and the change trend data output by the monitoring model with the warning thresholds, and is divided into four warning levels, specifically including:

[0025] A Warning: indicates that the construction status is stable and safe;

[0026] B Warning: Triggered when the stress reaches 80% of the warning threshold;

[0027] C Warning: Triggered when the stress reaches 90% of the warning threshold or the cable force monitoring value exceeds 5% of the cable force threshold;

[0028] D Warning: Triggered when the stress reaches 95% of the warning threshold or the difference between the linear monitoring data and the design value reaches 15mm, or exceeds 5% of the cable tension threshold.

[0029] By dividing the warning into four levels and generating warning notifications based on the comparison results of real-time monitoring data and warning thresholds, the safety level of the construction status can be more accurately reflected. The response measures at different warning levels can effectively guide the construction party to take targeted actions in a timely manner, thereby reducing construction risks and ensuring the safety and stability of the bridge structure.

[0030] Preferably, the digital twin platform uses WebGL technology to achieve three-dimensional visualization of the cross-railway swivel bridge, including dynamic interaction of BIM models, construction animations, and GIS geographic information. The three-dimensional visualization achieved through WebGL technology can present the BIM model, construction animation, and GIS geographic information of the cross-railway swivel bridge in a dynamic interactive manner, improving the visibility of monitoring data and ease of operation; the technology supports multi-terminal access, allowing the construction party to understand the construction status of the bridge anytime and anywhere, and improve management efficiency.

[0031] Preferably, the digital twin platform uses WebGL technology to achieve three-dimensional visualization of the cross-railway swivel bridge, including dynamic interaction of BIM models, construction animations, and GIS geographic information.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The present invention proposes a construction monitoring method for a cross-railway swivel bridge based on digital twins, which can monitor and analyze key parameters such as stress, displacement, temperature, and cable force in real time during bridge construction. By improving data quality through preprocessing, combining monitoring data with multiple models of the digital twin platform, and introducing a time series algorithm to predict the changing trend of monitoring data, it is possible to achieve early warning of potential construction risks and improve the intelligence level and safety of construction management. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of a method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to Embodiment 1 of the present invention;

[0035] Figure 2 This is a structural diagram of a digital twin model according to Embodiment 2 of the present invention;

[0036] Figure 3 This is a flow chart of the algorithm for LSTM model training according to Embodiment 3 of the present invention;

[0037] Figure 4 A main technical roadmap for platform development of Example 2 of the present invention;

[0038] Figure 5 This is a structural diagram of a safe construction monitoring and early warning system according to Embodiment 2 of the present invention;

[0039] Figure 6 This is a strain measurement point diagram of the cross-railway swivel bridge according to Example 3 of the present invention;

[0040] Figure 7 The measured stress data of sections 2, 5, 7 and 10 of Example 3 of the present invention;

[0041] Figure 8 This is a prediction result diagram of Example 3 of the present invention;

[0042] Fig. 9 This is the pycharm output value of Example 3 of the present invention. DETAILED DESCRIPTION

[0043] The following is a further detailed description of a digital twin-based cross-railway swivel bridge construction monitoring method and system provided by the present invention in conjunction with the accompanying drawings and specific embodiments. However, this should not be understood as the scope of the above-mentioned subject matter of the present invention being limited to the following embodiments, and all technologies implemented based on the content of the present invention belong to the scope of the present invention. Combined with the following description, the advantages and features of the present invention will be clearer. It should be noted that the drawings are all in a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0044] This embodiment provides a specific implementation method of a cross-railway swivel bridge construction monitoring method based on digital twin, including the following steps:

[0045] S1. Obtaining real-time monitoring data of the cross-rail swivel bridge; the monitoring data includes stress, displacement, temperature and cable force;

[0046] S2. Preprocessing the monitoring data, including data cleaning, standardization and denoising;

[0047] S3. Input the preprocessed monitoring data into the monitoring model of the digital twin platform, the monitoring model processes the monitoring data in combination with the time series algorithm, generates the change trend and current status data of the real-time monitoring data, and outputs the data to the early warning module;

[0048] S4. Compare the current status data and change trend data generated by the monitoring model with the preset warning thresholds to generate a warning notification.

[0049] Preferably, the monitoring data is obtained through sensors deployed at the construction site of the cross-rail swivel bridge, and the sensors include strain gauges, displacement sensors, temperature sensors, and cable force monitoring equipment; the step of acquiring the monitoring data includes real-time monitoring of key interfaces, and using the principle of symmetry to predict monitoring data at other symmetrical positions. By deploying multiple types of sensors at the construction site and collecting stress, displacement, temperature, and cable force data in real time, the comprehensiveness and accuracy of monitoring can be ensured; using the principle of symmetry to predict monitoring data at symmetrical positions not only reduces the risk of data loss caused by sensor damage, but also improves the system's fault tolerance and monitoring efficiency.

[0050] Preferably, the preprocessing step in S2 includes normalizing the monitoring data and removing invalid data through an outlier detection algorithm. Normalization and outlier removal in the preprocessing stage can effectively improve data quality, avoid interference of invalid or abnormal data on monitoring results, provide a reliable data basis for subsequent prediction and analysis, and enhance the stability and accuracy of the system.

[0051] Preferably, the digital twin platform includes:

[0052] BIM model, used to describe the geometric structure and component information of the cross-railway swivel bridge;

[0053] Finite element model for analyzing the mechanical behavior of a cross-rail swivel bridge;

[0054] The monitoring model is used to generate real-time status data and change trend data based on the monitoring data, and output them to the early warning module.

[0055] By integrating the BIM model, finite element model and monitoring model through the digital twin platform, a comprehensive linkage between the bridge's geometric structure, mechanical behavior and real-time monitoring data is achieved; the monitoring model can intuitively display the deviation between the monitoring data and the theoretical model, help quickly identify structural anomalies, and improve the efficiency of construction management and the scientific nature of decision-making.

[0056] Preferably, the monitoring model realizes the processing and prediction of monitoring data by combining time series algorithm and multi-feature input data; the multi-feature data includes historical data, real-time monitoring data and theoretical model calculation results. Combining time series algorithm with multi-feature input data (including historical data, real-time data and theoretical model calculation results) can realize accurate prediction of monitoring data under complex working conditions; this multi-source data input mode improves the robustness and accuracy of prediction and enhances the intelligent analysis ability of the system.

[0057] Preferably, the time series algorithm uses long short-term memory network, random forest, recursive neural network, back propagation algorithm. Using a variety of time series algorithms, the most suitable algorithm can be flexibly selected to process the monitoring data, improve the system's ability to analyze complex nonlinear data, and ensure efficient and accurate prediction in different scenarios.

[0058] Preferably, the generation of the warning notification is based on the comparison results of the current state data and the change trend data output by the monitoring model with the warning thresholds, and is divided into four warning levels, specifically including:

[0059] A Warning: indicates that the construction status is stable and safe;

[0060] B Warning: Triggered when the stress reaches 80% of the warning threshold;

[0061] C Warning: Triggered when the stress reaches 90% of the warning threshold or the cable force monitoring value exceeds 5% of the cable force threshold;

[0062] D Warning: Triggered when the stress reaches 95% of the warning threshold or the difference between the linear monitoring data and the design value reaches 15mm, or exceeds 5% of the cable tension threshold.

[0063] By dividing the warning into four levels and generating warning notifications based on the comparison results of real-time monitoring data and warning thresholds, the safety level of the construction status can be more accurately reflected. The response measures at different warning levels can effectively guide the construction party to take targeted actions in a timely manner, thereby reducing construction risks and ensuring the safety and stability of the bridge structure.

[0064] Preferably, the digital twin platform uses WebGL technology to achieve three-dimensional visualization of the cross-railway swivel bridge, including dynamic interaction of BIM models, construction animations, and GIS geographic information. The three-dimensional visualization achieved through WebGL technology can present the BIM model, construction animation, and GIS geographic information of the cross-railway swivel bridge in a dynamic interactive manner, improving the visibility of monitoring data and ease of operation; the technology supports multi-terminal access, allowing the construction party to understand the construction status of the bridge anytime and anywhere, and improve management efficiency.

[0065] Preferably, the digital twin platform uses WebGL technology to achieve three-dimensional visualization of the cross-railway swivel bridge, including dynamic interaction of BIM models, construction animations, and GIS geographic information.

[0066] Example 2

[0067] As a further optimization of Example 1, this embodiment provides a specific implementation method of a cross-railway swivel bridge construction monitoring method based on digital twins. Through the data feedback of the monitoring instruments on the construction site, a large amount of data collected by the sensors is analyzed and processed to achieve real-time monitoring, prediction and early warning, timely discover potential risks and take effective measures. The system uses the digital twin platform to combine model information interaction and feedback to ensure that all relevant parties can receive alerts in a timely manner and respond to problems quickly.

[0068] Based on WEB development technology, the present invention constructs a monitoring and early warning digital twin platform and method for cross-railway swivel bridges. The platform uses WEB technology to dynamically interact with bridge stress, deformation and cable force data, and includes the following functional modules:

[0069] (1) Display the detailed component information of the BIM model of the cross-railway swivel bridge, combine the swivel construction animation and GIS geographic information fusion model, and display it on the platform through WebGL technology;

[0070] (2) Use digital twin models (theoretical models, finite element models, and actual field models) for visual monitoring, prediction, and early warning;

[0071] (3) Data fusion visualization method: Collect monitoring data from various sensors (such as strain gauges, displacement sensors, and temperature sensors) in real time. Remove noise and outliers from the data to ensure data accuracy and consistency.

[0072] Integrate data from different sensors and monitoring devices to form a comprehensive bridge health model. Standardize data of different types and sources so that they can be analyzed and visualized on the same platform. View historical monitoring data compared to current data to analyze trends and changes.

[0073] Utilize multiple sensors and other measurement data acquisition equipment to extract bridge features for data prediction to achieve monitoring and early warning of the construction process; establish BIM models, finite element models, monitoring models, etc. to achieve twin data collection.

[0074] Digital twin model architecture Figure 2 As shown in the figure, specifically, three time-dimensional digital twin models are established, which are divided into three stages according to the time span: the initial design stage of the bridge, the on-site construction stage of the bridge, and the data feedback and correction stage.

[0075] In the design stage, the BIM initial model and finite element model are established through the provided engineering information, their theoretical values ​​are calculated, and the calculation results and model characteristics are transmitted to the platform for visualization;

[0076] During the construction phase, on-site sensors are used to collect real-time data on the physical information of the construction object, and the force and deformation of each structural component under each construction state are monitored in real time. The bridge construction monitoring data is preprocessed, feature engineered, and data sets are constructed to perform data normalization, feature dimension processing, and create a time series relationship for the data;

[0077] In the feedback correction stage, the processed data model is used for time series prediction. By selecting the eigenvalues ​​that have an impact on the bridge physical structure model and analyzing the distribution and correlation of the eigenvalues, the model is tuned and improved, and the appropriate algorithm and parameter settings are selected for data prediction. Multiple training rounds (epochs) are carried out. In each round: the model is trained and the training loss is calculated. The model is verified and the verification loss is calculated. The best model weight is saved to correct the time series algorithm. The actual model data on site is corrected by combining the predicted value, actual value, and theoretical value, and then the actual construction process is adjusted. At the same time, the feedback verification design model is fed back and updated to correct the BIM model and finite element model.

[0078] Preferably, the time series algorithm can use long short-term memory network (LSTM), random forest (RandomForest), recursive neural network (RNN) and back propagation algorithm (BP) for data analysis and prediction. Among them, due to the particularity of railway bridge construction, on-site monitoring is difficult to carry out frequently, and it only relies on the monitoring equipment buried in the early stage. Once these devices are damaged or cannot be maintained, it may lead to interruption of monitoring data. Therefore, the LSTM algorithm is preferred for prediction and early warning due to its advantages in processing time series data, so as to minimize the vacancy period caused by the failure of monitoring equipment.

[0079] The algorithm flow of LSTM model training is as follows: Figure 3As shown in Figure 1, specifically, the algorithm flowchart describes a bridge monitoring data prediction and early warning method based on the LSTM model. First, real-time monitoring data is collected from the left and right bridges with symmetrical working conditions, including key parameters such as stress, cable force and deformation, and the data is denoised and normalized to improve data quality. Subsequently, position monitoring data with the same characteristics are extracted according to the symmetrical working conditions and used as input features to train the LSTM model. The monitoring data of the target position is predicted by the LSTM model, and the predicted value is compared with the theoretical value. If the predicted value exceeds the theoretical value, it is considered that the construction status may be abnormal, and the system will send an early warning message to the web platform and notify relevant personnel to handle it; otherwise, the system continues to collect data and conduct the next round of prediction and monitoring to achieve real-time monitoring and intelligent early warning of the bridge construction status.

[0080] The main technical routes for platform development are as follows: Figure 4 As shown, the following steps are included:

[0081] 1. First, parameterize the established BIM model using Dynamo to make the information of the structural parts concrete;

[0082] 2. Combine drone oblique photography to generate 3D cache slices and combine them with GIS information in SuperMap software and BIM models;

[0083] 3. Secondary development of Revit Exporter export interface to convert Revit files into gltf files, and use three.js technology to load the model to the web page; specifically, use Revit software to build a bridge information model, use the family library's lofting fusion and hollow stretching to build the bridge beam ends and box beams, splice the beam ends into the main frame of the bridge, stretch to build the inclined cables, main towers, piers, and rotating device bridge components, create a new Revit project through the above content according to the drawing elevation axis network splice the complete bridge model, and export the model file from the Revit export interface as a gltf file;

[0084] 4. Export FBX files from Revit, import models, materials, and geometries from 3DMAX, set scene parameters, add keyframes, and customize construction animations by moving and rotating;

[0085] 5. Visualize the data transmitted by the sensors at the construction site on the platform webpage for real-time monitoring; process the on-site data, extract the physical structure characteristic values ​​of the model, establish data set division, and construct a time series algorithm to predict the monitoring data of key parts;

[0086] 6. Classify the above data information and store it in the MySQL database through the python-flask framework and create corresponding data tables; access, view, delete, modify, and update data through the URL. Complete the update and correction of the digital twin model and the actual model on site. The specific steps are as follows:

[0087] Use Revit software to build a bridge information model, use the family library's lofting fusion and hollow stretching to build the bridge beam ends and box beams, splice the beam ends into the main frame of the bridge, stretch to build the bridge components of the inclined cable, main tower, pier body, and rotating device, and create a new Revit project to splice the complete bridge model according to the elevation axis network of the drawing through the above content, and export the model file from the Revit export interface as a gltf file;

[0088] Use DJI Phantom 4 drone to take oblique photography of the surrounding environment near the bridge on site to obtain the terrain 3D cache, import the 3D cache terrain file into Supermap software, add the scene generation environment, add Revit's BIM model, determine the real position of the model through the pile position coordinate table x, y of the bridge drawing, set the position x, y of the imported model to merge the model with the oblique photography model.

[0089] Set the height H of the Revit model and the earth distance H1. Use SuperMap iserver to publish and get the url network address. Use the intersection surface of each beam end as the main beam elevation detection surface and use a total station to monitor the displacement change. Use the acquisition module (JMZX-8A) to collect data from different sensors on site through its 8 channels. The DTU module (JMZX-2007) transmits this data to the remote monitoring system or cloud platform in real time through CMNET (mobile network). The three modules of the fully sealed chassis (JMBV-1116) form a wireless automatic monitoring bridge stress change system to monitor the stress changes at the key beam ends.

[0090] The cable force dynamic measuring instrument was used to monitor the cable force of the 44 cables in the whole bridge. The frequency spectrum method was used to measure the cable force during the construction stage.

[0091] The principle of the spectrum method is as follows:

[0092]

[0093] Where: T——cable force (N);

[0094] W——weight per unit length of cable (kg / m);

[0095] f n ——nth order natural frequency of the cable (Hz);

[0096] L——Calculated length of cable (m);

[0097] n——cable natural frequency order;

[0098] g——acceleration due to gravity (9.81m / s 2 );

[0099] F——cable’s fundamental frequency.

[0100] Next, use the python-flask framework and mysql database to build a web platform; create app.py as the main program, set the mysql database link string parameters, host as root, password root, host address localhost, port number 3306 to connect to the local database. The app.py file creates routing functions and corresponding html files. Use the three.js underlying technology to set the model's click function and component information window. Use the iframe tag to integrate the aforementioned url link into the web platform. Create a data receiving route to receive the data transmitted by the sensor, and use the jsonify library to set the incoming data to js format. Create a new data route in the database, add data in js format and save it into the mysql database, set the section and node id, and create a section and node warning threshold data table. Query and save the data in app.py as a new parameter data and pass it into the html file. Use data.id in the web page to query the data corresponding to the id in the mysql database. Create a neural network route, and use a query statement to save the section and node data as new data for neural network route training simulation data.

[0101] Set the device to the GPU or CPU available on the computer system, set the model training data to 80% of the sample data, and the remaining 20% ​​is divided into 10% of the data for the validation set and 10% of the data for the test set. Design the training batch size to be 50, the time step number to be input=4, and the output=1. Specify the loss function as the mean square error MSE, set the learning rate to 0.01, and the weight decay to 0.001 to prevent overfitting. The number of LSTM hyperparameter stacking times is 2, and the hidden layer dim is 128. Use the normalized scalar to normalize the imported data. The training cycle epochs is 50 times. Use the torch.optim.AdamW optimizer to handle regularization, define the training model as LSTM Main_model, and input the above parameters into the model for prediction.

[0102] Use TensorDataset and DataLoader to build batch loaders for training, validation, and test sets. Use matplotlib.pyplot to visualize the prediction results and input values. Save the generated data to mysql data through mysql add data statement. Extract monitoring data, prediction data, and thresholds in mysql for comparative analysis. When the warning value is reached, a mysql add statement is generated to add data to the warning table. Set the warning table to send a warning to the web platform and send a text message to relevant staff if it is not null.

[0103] Safety warning system such as Figure 5 As shown in the figure, the safety construction monitoring and early warning system consists of a data acquisition module, a digital twin platform module, and an early warning threshold setting module. The data acquisition module integrates BIM model information, finite element analysis results, theoretical data generated by time series algorithm, and field monitoring data. Field monitoring data is input into the digital twin platform in real time through directly connected monitoring equipment. On this platform, early warning thresholds are set for different monitoring types. The early warning disposal mechanism supports all parties to conduct analysis, remote collaboration, emergency response, and process approval when an early warning occurs.

[0104] The digital twin-based cross-railway swivel bridge construction monitoring system has the following notable features: comprehensive coverage of key node monitoring types, real-time monitoring point maps, stress monitoring data, linear displacement monitoring, cable force monitoring of low-tower cable stays, and information on buildings and surrounding GIS terrain. In the digital twin model of monitoring information, the actual buried location and type are annotated and visualized, and it is associated with the monitoring data of the safety monitoring and early warning platform, and different early warning states are displayed based on whether the real-time monitoring data exceeds the early warning threshold. There are four early warning levels:

[0105] Green warning: The monitoring point displays green, indicating that the construction status is stable and safe. At this time, regular inspections can be carried out to check the changes in equipment status and time series algorithms.

[0106] Yellow warning: When the stress reaches 80% of the warning threshold, it is triggered and the construction site is immediately notified to check the on-site equipment and record relevant warning information, including location, type, warning threshold, predicted value and normal value.

[0107] Orange warning: If the stress reaches 90% of the warning threshold or the cable tension monitoring value exceeds 5% of the cable tension threshold, the construction party must immediately check the on-site sensor equipment and record the warning information. At the same time, each unit must take corresponding measures to reinforce the bridge.

[0108] Red warning: When the stress reaches 95% of the warning threshold or the difference between the linear monitoring and the design value reaches 15mm or exceeds the cable tension threshold by 5%, a red warning is immediately issued. The relevant construction units must quickly stop construction, check the sensors, and record the warning information. They must arrange measurements on site for re-measurement. Construction can continue after meeting the design requirements.

[0109] The system automatically activates the warning mechanism according to the set warning level and sends warning notifications to relevant personnel. The security monitoring and early warning platform displays the location and value of the warning through the digital twin platform to facilitate the setting of the early warning mechanism, send reminders and collaboratively handle warning events online.

[0110] The present invention provides a bridge construction monitoring and management system based on digital twin technology. By combining the BIM three-dimensional model with the finite element model and time series algorithm, the theoretical model and field monitoring data are corrected to achieve long-term monitoring, early warning and WEB display of creep deformation of concrete bridges, significantly reducing the complexity and risk of bridge operation and maintenance. The system uses sensors to collect stress, linearity, cable force and other data at the bridge construction site in real time, and completes data integration and accurate prediction through an intelligent platform, providing comprehensive data support and scientific decision-making basis for bridge construction and operation and maintenance.

[0111] The system integrates the BIM 3D model into the cloud platform, supports real-time viewing on PC, mobile phone and large screen, without installing professional software, thus overcoming the problems of data transmission delay and lack of real-time performance in traditional monitoring methods, greatly improving the response capability to potential risks and the efficiency of bridge status assessment. At the same time, the time series algorithm is trained and optimized in combination with real environment data, and the accuracy and reliability of system prediction are improved through simulation and dynamic adjustment.

[0112] The monitoring platform based on WEB technology extracts the structural characteristics of the bridge through digital twin technology, and combines time series algorithms to predict and analyze real-time data, which can visually present the health status of the bridge. The system provides data analysis and simulation functions to support engineers in making scientific decisions, and can predict the maintenance needs of the bridge, reduce unnecessary inspections, optimize maintenance processes, and improve management efficiency.

[0113] The present invention significantly improves the real-time, accuracy and reliability of monitoring by continuously monitoring and analyzing the construction and operation status of bridges. At the same time, the system reduces the risk of accidents, ensures the construction and operation safety of bridges, and promotes the development of bridge construction monitoring towards intelligence, precision and efficiency.

[0114] Example 3

[0115] Based on the symmetrical structural characteristics of the cross-railway swivel bridge and the optimization of sensor layout, this embodiment proposes a specific implementation method of a cross-railway swivel bridge construction monitoring method based on digital twins. Since the left and right bridges of the cross-railway swivel bridge are symmetrical and have the same working conditions, and the left and right bridges themselves have consistent working condition characteristics in their centrally symmetrical positions, the actual monitoring data at symmetrical positions can be used to predict the monitoring data at other positions. Even if some sensors are damaged and data is missing, the system can still provide real-time monitoring by analyzing and predicting symmetrical or homogeneous cross-sectional data.

[0116] like Figure 6 As shown, sections 2, 5, 7, and 8 are sections with the same characteristics and can be used as data with the same characteristics for prediction. Similarly, sections 1 and 11 are sections with the same characteristics, 12 and 14 are sections with the same characteristics, 3, 4, 8, and 9 are sections with the same characteristics, and 13 and 15 are sections with the same characteristics. For these sections with the same characteristics, this method uses multi-sensor input data to train the long short-term memory network (LSTM) algorithm to achieve the prediction of monitoring data. Specifically, it includes:

[0117] 1. Receive strain data from the automated sensor to the platform;

[0118] 2. The platform uses MySQL's new statement to upload data with the same characteristic section data to the MySQL table

[0119] 3. Query the strain values ​​in the MySQL table and assign new variables for training

[0120] 4. Normalize the data and divide the data to facilitate training, verification, and testing.

[0121] 5. Use the LSTM algorithm to input variable values, set the training feature value object to features_num = 2, 4 for multiple features, and features_num = 1 for single feature. Set the time step dimension to features_num and the output dimension to a linear layer.

[0122] 6. Map the LSTM output at each time step to the prediction space through a linear layer.

[0123] 7. Restore the output layer tensor shape output, summarize the predictions for the entire sequence, and return the predicted value with the shape of (training batch size, number of time steps, output length).

[0124] 8. Visualize and compare using matplotlib.pyplot.

[0125] Through the above steps, we can get Figure 7The measured stress data of cross-sections 2, 5, 7, and 10 shown are imported into the above neural network algorithm. The feature dimension is set to 4, the training set is 80% of the data, and the test set and validation set are each 10%. The dim hidden layer is 128. Set the time step, input length, and output length, and call the encapsulated LSTM algorithm above to obtain the prediction results as Figure 8 , it can be seen that the prediction results obtained by the LSTM algorithm are basically consistent with the measured data.

[0126] Calculate and evaluate the model metrics MSE, MAE, and MAPE. R2 is used to evaluate the reliability of the training:

[0127]

[0128] Among them, MSE is the mean squared error, obtained by calculating the square of the difference between the predicted value and the true value and taking the average; MAE is the mean absolute error, obtained by calculating the absolute difference between the predicted value and the true value and taking the average; is the predicted value of the i-th sample, is the true value of the i-th sample, and n is the number of samples.

[0129] In the code, mean_absolute_error calculates the MAE and MSE of the test_labels actual data and the pre_array predicted data.

[0130]

[0131] Among them, MAPE is the mean absolute percentage error, obtained by calculating the proportion of the absolute error between the predicted value and the true value to the true value and taking the average, and is usually used to measure the relative size of the prediction error. A smaller MAPE indicates a smaller relative error of the model. Generally speaking, a MAPE value less than 10% is considered a very good result. The reasonable range of MAPE: excellent model: MAPE < 10%; acceptable model: 10% < MAPE < 20%; less ideal model: MAPE > 20%.

[0132] Adopt the coefficient of determination R 2 to measure the degree to which the model explains the variance of the independent variable. The formula is:

[0133]

[0134] Among them, is the mean of the true values; R 2 measures the explanatory power of the model, that is, the degree of fitting of the model to the data. The value range of R 2 is [0, 1]. The closer it is to 1, the better the model fits; R 2Negative values ​​indicate a model that is worse than a simple mean model.

[0135] R 2 Value and model evaluation: R 2 ≥0.9: The model is very good and fits the data well; R 2 ≥0.8: A good model that performs very well and can capture the changes in the data well; R 2 ≥0.6: A reasonable model that is applicable to many practical problems, but there is still room for further optimization; R 2 ≥0.5: Average performance, further data processing or selection of a more appropriate model may be required; R 2 <0.5: The model fit is poor and you may need to adjust the model or reconsider the characteristics of the data.

[0136] like Fig. 9 Output the values ​​of the above four indicators. It can be seen that MAPE loss less than 10% is an excellent model, R 2 Greater than or equal to 0.9, it can fit the data well.

[0137] In summary, using multi-feature inputs from multiple sensors to train the LSTM algorithm has the following main advantages over traditional LSTM predictions using single-position sensors: improved prediction accuracy, because multi-source data provides more comprehensive system status information, which helps to improve the prediction accuracy of target sensor data; enhanced robustness and fault tolerance of the model, even if a single sensor fails or data is missing, the system can still predict the state of the target sensor through data at the same feature position; richer contextual information is provided, and multi-sensor input provides rich contextual information, which helps to model complex dependencies in space and time; improved generalization ability, avoiding model overfitting through multi-feature input, and improving adaptability to different construction conditions and environmental changes; optimized data utilization efficiency, making full use of the same feature section data of the cross-rail swivel bridge to improve fault detection and diagnosis capabilities.

[0138] Example 4

[0139] This embodiment provides a cross-railway swivel bridge construction monitoring system based on digital twin, which is used to implement the monitoring method described in the above claims. The system consists of the following modules:

[0140] 1. Data acquisition module

[0141] The data acquisition module is deployed at the construction site of the cross-railway swivel bridge and includes the following types of sensors:

[0142] Strain gauges, used to collect stress changes in key parts of the bridge;

[0143] Displacement sensors are used to monitor the displacement of bridge components;

[0144] Temperature sensors are used to monitor the temperature changes of bridge construction environment and components;

[0145] Cable tension monitoring equipment is used to measure the cable tension status of bridges in real time.

[0146] These sensors transmit the monitoring data to the data preprocessing module in real time via wireless or wired means.

[0147] 2. Data preprocessing module

[0148] The data preprocessing module cleans and normalizes the collected monitoring data, and removes invalid data through an outlier detection algorithm to ensure the accuracy and consistency of the input data. The module also integrates multiple types of data to provide a unified data basis for subsequent analysis and prediction.

[0149] 3. Digital Twin Platform

[0150] The digital twin platform is the core module of this system, which includes the following sub-modules:

[0151] BIM model module: used to construct and manage the three-dimensional geometric structure and component details of the cross-rail swivel bridge;

[0152] Finite element model module: used to analyze the mechanical behavior of bridge structures;

[0153] Monitoring model module: used to display the deviation between monitoring data and theoretical model in real time and update it dynamically.

[0154] The platform uses WebGL technology to achieve three-dimensional visualization display, including dynamic interaction of BIM models, construction animations and GIS geographic information.

[0155] 4. Forecast and warning module

[0156] Use long short-term memory network (LSTM), random forest (Random Forest), recursive neural network (RNN) or back propagation algorithm (BP) to make trend predictions based on historical data and real-time monitoring data.

[0157] According to the comparison between monitoring data and theoretical values, the warning is divided into four levels:

[0158] A Warning: indicates that the construction status is stable and safe;

[0159] B Warning: When the stress reaches 80% of the warning threshold, an on-site inspection is prompted;

[0160] C Warning: When the stress reaches 90% of the warning threshold or the cable force exceeds the threshold by 5%, it prompts to strengthen monitoring and disposal;

[0161] D Warning: When the stress reaches 95% of the warning threshold or the difference between the linear monitoring data and the design value reaches 15mm, a high-risk alarm is issued and construction is stopped immediately.

[0162] Early warning notifications are sent to relevant personnel in real time via SMS, email or platform.

[0163] 5. Data Visualization Module

[0164] The data visualization module displays monitoring data and prediction results through the digital twin platform, supports interactive operations in three-dimensional scenes, and enables users to intuitively understand the status of bridge construction. The platform is compatible with PC, mobile and large-screen terminals, enabling multi-terminal access and operation.

[0165] 6. System integration and operation

[0166] By integrating various modules, the system can monitor, predict and warn the bridge construction status in real time, and guide the construction party to take corresponding measures according to the warning level.

[0167] The system is based on cloud computing technology to achieve unified storage, processing and sharing of data, support multi-user collaboration, and reduce the complexity and risks of construction management.

[0168] The system described in this embodiment can efficiently integrate monitoring data, predict trends and issue timely warnings, overcoming the problems of difficult data integration and lack of real-time performance in traditional methods. At the same time, it improves the intuitiveness and scientificity of construction management through a visualization platform, effectively improving the intelligence level and safety assurance capabilities of bridge construction monitoring.

[0169] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0170] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0171] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A construction monitoring method for a cross-railway swivel bridge based on digital twin, characterized in that: The following steps are involved: S1. Obtaining real-time monitoring data of the cross-rail swivel bridge; the monitoring data includes stress, displacement, temperature and cable force; S2. Preprocessing the monitoring data, including data cleaning, standardization and denoising; S3. Input the preprocessed monitoring data into the monitoring model of the digital twin platform, the monitoring model processes the monitoring data in combination with the time series algorithm, generates the change trend and current status data of the real-time monitoring data, and outputs the data to the early warning module; S4. Compare the current status data and change trend data generated by the monitoring model with the preset warning thresholds to generate a warning notification.

2. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 1, characterized in that: The monitoring data is obtained through sensors deployed at the construction site of the cross-railway swivel bridge, and the sensors include strain gauges, displacement sensors, temperature sensors and cable tension monitoring equipment; the step of obtaining the monitoring data includes real-time monitoring of key interfaces and using the principle of symmetry to predict monitoring data at other symmetrical positions.

3. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 2, characterized in that: The preprocessing step in S2 includes normalizing the monitoring data and eliminating invalid data through an outlier detection algorithm.

4. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 3, characterized in that: The digital twin platform includes: BIM model, used to describe the geometric structure and component information of the cross-railway swivel bridge; Finite element model for analyzing the mechanical behavior of a cross-rail swivel bridge; The monitoring model is used to generate real-time status data and change trend data based on the monitoring data, and output them to the early warning module.

5. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 4, characterized in that: The monitoring model realizes the processing and prediction of monitoring data by combining time series algorithm and multi-feature input data; the multi-feature data includes historical data, real-time monitoring data and theoretical model calculation results.

6. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 5, characterized in that: The time series algorithm adopts long short-term memory network, random forest, recursive neural network and back propagation algorithm.

7. A method for monitoring the construction of a cross-railway swivel bridge based on digital twins according to claim 6, characterized in that: The generation of the warning notification is based on the comparison results of the current state data and the change trend data output by the monitoring model with the warning threshold, which are divided into four warning levels, including: A Warning: indicates that the construction status is stable and safe; B Warning: Triggered when the stress reaches 80% of the warning threshold; C Warning: Triggered when the stress reaches 90% of the warning threshold or the cable force monitoring value exceeds 5% of the cable force threshold; D Warning: Triggered when the stress reaches 95% of the warning threshold or the difference between the linear monitoring data and the design value reaches 15mm, or exceeds 5% of the cable tension threshold.

8. The method for monitoring the construction of a cross-railway swivel bridge based on digital twin according to claim 7, characterized in that: The digital twin platform uses WebGL technology to achieve three-dimensional visualization of the cross-railway swing bridge, including dynamic interaction of BIM models, construction animations, and GIS geographic information.

9. A construction monitoring system for a cross-railway rotating bridge based on digital twin, comprising: A data acquisition module, used to collect monitoring data of the cross-railway swivel bridge in real time, wherein the monitoring data includes stress, displacement, temperature and cable force; A data preprocessing module is used to clean, normalize and remove outliers from the monitoring data; A digital twin platform is used to integrate the monitoring data with the BIM model, finite element model and monitoring model of the bridge, and realize three-dimensional visualization through Web technology; The prediction and warning module is used to predict the changing trend of monitoring data based on time series algorithm, and generate warning notifications based on the comparison between monitoring data and preset thresholds.

Images