Digital twin long-span bridge earthquake wind vibration and vibration reduction and isolation system and method

Abstract

The invention discloses a digital twinning long-span bridge earthquake wind vibration and vibration reduction and isolation system and method, the system takes a high-fidelity digital model for constructing a long-span bridge as a basis, comprises full life cycle monitoring and management of the bridge under earthquake, wind vibration and vibration reduction and isolation conditions, and is based on a sky-air-ground-water integrated ubiquitous sensing module. In combination with artificial intelligence algorithms such as deep learning and a pulse neural network, real-time state evaluation of a bridge structure, material nonlinear analysis, geometric nonlinear analysis, multi-point excitation and pile-soil-structure interaction simulation, wind-induced fatigue damage prediction, earthquake vulnerability evaluation and closed-loop control scientific decision based on an artificial intelligence body are realized. According to the invention, system support and method guidance can be provided for large-span bridge anti-seismic and anti-wind design, intelligent monitoring, operation and maintenance management and vibration reduction and isolation system optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent protection technology in civil engineering, particularly to the field of seismic and wind protection in bridge engineering, and especially to the application tools and implementation methods for intelligent seismic and wind monitoring, risk warning and active protection of long-span bridges. Specifically, it is a digital twin system and method for seismic and wind vibration reduction and isolation of long-span bridges. Background Technology

[0002] The level of informatization and intelligence in the operation and maintenance of bridge engineering seismic and wind-induced vibration monitoring and vibration reduction and isolation systems is still under continuous improvement. Although existing systems have relatively rich data acquisition and processing capabilities, and can support a certain degree of structural assessment and analysis prediction, they still lack high-precision modeling and simulation mechanisms to cope with the complex seismic and wind-induced vibration conditions of long-span bridges, especially in multi-physics coupling analysis. This makes it difficult to achieve real-time prediction and intelligent closed-loop control of nonlinear responses. In recent years, with the continuous development and integration of key technologies such as BIM, GIS, FEM, IoT, AI, deep learning, and big data, bridge seismic and wind-induced vibration monitoring and digital twin technology have been gradually applied to the operation and maintenance management of bridge vibration reduction and isolation systems. By constructing high-fidelity digital models, mapping and dynamic interaction of physical entities are achieved, promoting the transformation of bridges from traditional periodic detection to real-time, continuous sensing and intelligent decision-making. Relying on new surveying and mapping technologies, intelligent sensing networks, collaborative computing platforms, multi-dimensional information expression and deep learning algorithms, the bridge digital twin system has initially formed a system framework covering core functions such as perception, modeling, analysis and decision-making. It has effectively improved the data perception accuracy, model expression capability, interaction level and response speed of bridge health monitoring, provided a technical foundation for intelligent bridge management, and is becoming an important support for the digital transformation of bridge operation and maintenance management (Zheng Gangkang, He Yuelei, Wan Leshan. Research on data visualization and early warning of health monitoring of extra-large span railway bridges based on digital twin [J]. Information Technology of Civil Engineering, 2024, 16(4): 65-69).

[0003] However, current digital twin systems in bridge engineering both domestically and internationally still have certain limitations. On the one hand, existing bridge digital twin systems mainly focus on monitoring the static mechanical parameters of prefabricated or ordinary bridges. By constructing a basic digital model of the bridge and combining it with sensing modules to collect pressure and displacement data, they achieve simple assessments of the bridge's operational status and provide decision support for maintenance. However, these systems are significantly lacking in their ability to perform refined modeling and real-time prediction of the nonlinear response of bridge structures under extreme environmental loads such as earthquakes and strong winds. For example, when faced with the complex material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges, existing systems often cannot accurately simulate various damage forms such as wind-induced fatigue and seismic vulnerability, let alone achieve high-precision prediction of bridge responses under these complex conditions. Furthermore, they lack a deep fusion mechanism for multi-source heterogeneous data and a closed-loop control strategy based on prediction results, making it difficult to meet the vibration reduction and isolation safety requirements of long-span bridges under seismic and wind-induced vibration conditions. On the other hand, existing bridge digital twin systems have shortcomings in their modeling methods. They are mainly designed for scenario modeling during the bridge construction phase, using a knowledge graph of the construction scenario combined with traditional machine learning models to predict changes in the bridge construction process. However, these systems have limited capabilities in analyzing the seismic and wind-induced vibration characteristics of bridges during the operational phase, especially in real-time prediction and active control of multiphysics coupling analysis. Summary of the Invention

[0004] To address at least one of the problems existing in current technologies, this invention provides a digital twin system and method for earthquake and wind-induced vibration reduction and isolation of long-span bridges. Based on multiphysics coupling analysis, real-time data acquisition and processing, intelligent prediction models, and closed-loop control mechanisms, it provides comprehensive support for the safety and reliability of long-span bridges under extreme environmental loads. This system integrates multi-source sensing data, constructs a high-fidelity digital model, achieves real-time prediction and assessment, and optimizes the design of the vibration reduction and isolation system, thereby comprehensively improving the performance of bridges under earthquakes and strong winds.

[0005] The present invention is achieved by at least one of the following technical solutions.

[0006] A digital twin system for earthquake and wind vibration reduction and isolation of long-span bridges includes a benchmark modeling module, a ubiquitous sensing module, a virtual-real interaction module, a real-time prediction module, and a closed-loop control module. The benchmark modeling module is used to construct a high-fidelity digital model of a long-span bridge. The digital model can characterize the multi-dimensional features of the bridge from aspects such as geometry, physics, behavior, rules, evolution, safety, and failure. It can simulate the material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges under earthquake, wind vibration, and vibration reduction and isolation conditions. It can also achieve accurate simulation of structural mechanics, wind environment, and seismic response models through multiphysics coupling analysis. The ubiquitous sensing module is used to collect multi-source data from the sky, air, ground, and water, and to sense the interaction between the structural performance of the bridge structure during construction and operation and the external environment of the project; the multi-source data includes real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data; The virtual-real interaction module is used to realize real-time digital mapping and bidirectional dynamic updates between the physical entity and digital model of long-span bridges, ensuring the synchronization between the physical entity and the digital model. The real-time prediction module receives multi-source data from the ubiquitous sensing module and digital models from the benchmark modeling module. It then uses artificial intelligence algorithms, including deep learning and spiking neural networks (SNNs), to train and obtain a prediction model. This model monitors, predicts, and evaluates the seismic, wind-induced vibration, and vibration isolation nonlinear performance of long-span bridges, as well as the interaction between bridge structural performance and the external engineering environment. The nonlinear performance includes material nonlinearity and geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction. The module also predicts potential performance degradation and failures. The closed-loop control module receives the evaluation and prediction results from the real-time prediction module and performs predictive maintenance decisions, risk management, and maintenance optimization through the artificial intelligence agent AIAgent. This plays a core role in the full life cycle management of long-span bridges, which spans all stages of design, construction, operation, and maintenance to achieve closed-loop control of the digital twin.

[0007] Furthermore, the benchmark modeling module acquires data through finite element simulation and a bridge health monitoring system. The finite element simulation covers refined analysis of material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges.

[0008] Furthermore, the ubiquitous sensing module includes an advanced sensor network, a data acquisition and transmission unit, and a data processing unit; The advanced sensor network includes a satellite InSAR system, an unmanned aerial vehicle (UAV) inspection system, fixed surface sensors, an intelligent mobile monitoring platform, and a water monitoring system. The satellite InSAR system is used for large-scale surface deformation monitoring. The UAV inspection system, equipped with visible light cameras, infrared thermal imagers, and lidar, is used for visual inspection and temperature field monitoring. The fixed surface sensors include fiber optic grating sensors, accelerometers, displacement sensors, GPS / GNSS receivers, anemometers, and temperature sensors, used to monitor strain, vibration, displacement, three-dimensional position, wind environment, and temperature changes, respectively. The intelligent mobile monitoring platform includes unmanned vehicles and robots for refined inspection of key components. The water monitoring system includes unmanned vessels and unmanned underwater vehicles for hydrological monitoring and underwater structure inspection, comprehensively acquiring multi-dimensional data required for the interaction between bridge structural performance and the external environment. The data acquisition and transmission unit is used to acquire data from the advanced sensor network and to achieve real-time and accurate transmission of monitoring data through optical fiber and wireless transmission technologies, while ensuring data security and integrity. The data processing unit is used to perform multi-scale data fusion on the collected data, clean and reduce noise by using filtering or wavelet analysis, extract key feature parameters such as frequency, damping, mode shape, stress distribution, environmental load characteristics, and structural deformation mode from massive monitoring data, and use machine learning and deep learning technologies for data pattern recognition, anomaly detection, early warning of damage, and intelligent identification of the interaction between bridge structural performance and external environment.

[0009] Furthermore, the real-time prediction module integrates structural mechanics models, wind environment models, and seismic response models through multiphysics coupling analysis to achieve understanding of the behavior of long-span bridges in complex environments and accurate response prediction; the multiphysics coupling analysis adopts a two-way coupling mechanism to achieve deep feedback between structural dynamic response and environmental factors. The wind environment model combines computational fluid dynamics (CFD) technology and wind tunnel test data to finely simulate the wind field around the bridge, obtain the aerodynamic coefficient of the bridge section, and couple it with the structural dynamics model to accurately predict wind-induced vibration response, vortex-induced vibration, flutter, and buffeting and other wind-induced damage modes. The earthquake response model combines real-time ground motion data from earthquake monitoring stations with a bridge structure model to simulate the dynamic response of the bridge under earthquake action. It also considers material nonlinearity, geometric nonlinearity, and the characteristics of energy-consuming devices to perform multi-scale simulation and nonlinear behavior simulation, so as to evaluate the damage state, vulnerability, and probability of failure of the bridge under different earthquake intensities.

[0010] Furthermore, the real-time prediction module adopts an intelligent prediction model based on artificial intelligence algorithms and data-driven methods. The intelligent prediction model includes prediction models based on recurrent neural networks, graph neural networks, and reinforcement learning. It learns the health evolution pattern of the bridge from historical monitoring data and data on the interaction between structural performance and the external environment, predicts future performance degradation and potential failures, and assesses the long-term impact of the interaction between the bridge structural performance and the external environment. The intelligent prediction model is also used for data-driven damage identification. By comparing the real-time monitoring data with the response of the digital twin model under normal conditions, it identifies changes in structural stiffness and damping parameters, thereby determining the occurrence and extent of damage and predicting the damage evolution path and remaining life.

[0011] Furthermore, the virtual-real interaction module is responsible for mapping the massive multi-source data collected by the ubiquitous sensing module to the corresponding nodes or units of the digital model in real time through data fusion technology, and dynamically correcting and updating the topology, physical parameters and behavioral patterns of the digital model based on real-time data, thereby driving the dynamic evolution of the digital model to achieve digital management and precise control of the entire life cycle of the long-span bridge; the mapping process adopts weighted least squares method, Bayesian estimation or state-space model method.

[0012] Furthermore, the closed-loop control module uses an AI Agent to perform automated analysis based on time-series data provided by the ubiquitous sensing module, the virtual-real interaction module, and the real-time prediction module. The AI ​​Agent can perform in-depth processing of the evaluation results and output a comprehensive evaluation report on the seismic, wind-induced vibration, and vibration reduction and isolation performance of long-span bridges, including damage classification, risk level, performance degradation trend, and the evolution of the interaction state between structural performance and the external environment.

[0013] Furthermore, the artificial intelligence entity is constructed based on the time-series data provided by the ubiquitous perception module, the virtual-real interaction module, and the real-time prediction module. It utilizes the Bridge Engineering Large Language Model LLM-BE, the Bridge Advanced Construction and Operation Knowledge Graph KGTADC, the Bridge Engineering Finite Element Simulation Module FEM-BE, and the Machine Learning Algorithm Intelligent Identification of Construction Drawings / BIM / GIS Technical Indicators and Verification Module MLTE as its core components. It completes intelligent decision-making and control by automatically generating APG through prompts.

[0014] The method for implementing the aforementioned digital twin long-span bridge seismic and wind-induced vibration and vibration reduction / isolation system includes the following steps: S01, Benchmark Modeling: Construct a high-fidelity digital model of a long-span bridge. The digital model is characterized from multiple dimensions such as geometry, physics, behavior, rules, evolution, safety, and failure. It simulates the material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges under earthquake, wind vibration, and vibration reduction and isolation conditions. It also achieves accurate simulation of structural mechanics, wind environment, and seismic response through multiphysics coupling analysis. S02, Ubiquitous Sensing: Conduct integrated multi-source data acquisition from the sky, air, ground, and water to obtain real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data of long-span bridges, and perceive the interaction between the structural performance of bridge construction and operation and the external environment of the project. S03, Virtual-Real Interaction: Realizes real-time digital mapping and bidirectional dynamic updates between the physical entity of a long-span bridge and the digital model, ensuring the synchronization between the physical entity and the digital model; S04. Real-time prediction: Receive multi-source data from ubiquitous sensing and digital models from benchmark modeling, use artificial intelligence algorithms such as deep learning and spiking neural networks (SNN) to train and obtain prediction models, monitor, predict and evaluate the seismic, wind vibration and vibration reduction and isolation nonlinear performance of long-span bridges, as well as the interaction between bridge structural performance and the external environment of the project. S05. Closed-loop control: Receives real-time prediction assessment results and performs predictive maintenance decisions, risk management, and maintenance optimization through an artificial intelligence agent, thereby playing a core role in the full life cycle management of long-span bridges. The full life cycle management runs through all stages such as design, construction, operation, and maintenance to achieve closed-loop control of the digital twin.

[0015] Furthermore, the decision-making process of the AI ​​Agent in the closed-loop control step includes: It receives time-series data from ubiquitous sensing, virtual-real interaction, and real-time prediction, and generates a situation description using APG technology through automated prompts. The LLM-BE (Large Language Model for Bridge Engineering) combined with the KGTADC (Knowledge Graph for Advanced Bridge Construction and Operation) is used for multi-step reasoning to diagnose the causes of anomalies. The bridge engineering finite element simulation module FEM-BE was used to verify the diagnostic results and assess the risk level. Based on machine learning algorithms, the intelligent identification module MLTE for construction drawings / BIM / GIS technical indicators and verification calculations analyzes construction drawings and BIM / GIS models to generate optimized maintenance plans that include work steps, material lists, and time-series planning. After maintenance, continuously monitor the effects and store the decision-making process and results in a knowledge graph for continuous model optimization.

[0016] Compared with the prior art, the beneficial effects of the present invention are at least as follows: This invention provides a digital twin system for earthquake and wind-induced vibration and vibration reduction / isolation of long-span bridges, comprising five core modules: a baseline modeling module, a ubiquitous sensing module, a virtual-real interaction module, a real-time prediction module, and a closed-loop control module. It enables intelligent monitoring and management of long-span bridges throughout their entire lifecycle under earthquake, wind-induced vibration, and vibration reduction / isolation conditions. This system achieves the perception of the interaction between structural performance and the external engineering environment, structural health monitoring, earthquake and wind-induced vibration response prediction, wind-induced fatigue damage assessment, earthquake vulnerability analysis, and intelligent decision-making based on artificial intelligence agents (AIAgent) throughout the design, construction, operation, and maintenance processes of long-span bridges. This improves the accuracy, reliability, and timeliness of earthquake and wind-induced vibration monitoring of long-span bridges, significantly enhances the safety of bridges in extreme environments, and saves on lifecycle maintenance costs.

[0017] In past engineering projects, existing bridge digital twin systems primarily focused on monitoring the static mechanical parameters of prefabricated or conventional bridges. Their ability to perform refined modeling and real-time prediction of the nonlinear responses of bridge structures under extreme environmental loads such as earthquakes and strong winds was significantly insufficient. When faced with the complex material nonlinearities, geometric nonlinearities, multi-point excitations, and pile-soil-structure interactions of long-span bridges, they could not accurately simulate various damage forms such as wind-induced fatigue and seismic vulnerability. Furthermore, they lacked a deep fusion mechanism for multi-source heterogeneous data and a closed-loop control strategy based on prediction results. In this invention, the baseline modeling module constructs a high-fidelity digital model through deep fusion of the finite element model with building information models and geographic information models. This model encompasses complex characteristics such as material nonlinearities, geometric nonlinearities, multi-point excitations, and pile-soil-structure interactions. It also achieves integrated modeling of structural mechanics, wind environment, and seismic response through multiphysics coupling analysis. The real-time prediction module constructs a complex nonlinear mapping relationship between input and output data using artificial intelligence algorithms such as deep learning and spiking neural networks. This enables dynamic tracking and trend prediction of the evolution of the interaction between structural performance and the external environment, significantly accelerating the model prediction rate compared to traditional finite element analysis models.

[0018] Because various specialized software programs (such as finite element analysis software, building information modeling software, and geological information modeling software) often lack interfaces that support each other's data formats, data output from different software programs is difficult to transfer between each other, thus reducing the efficiency of data analysis. On the other hand, data and materials in various formats (such as design drawings, monitoring data, and digital models) typically lack a comprehensive management system with high visualization and ease of use to facilitate data retrieval and collaborative work among engineering and technical personnel from different departments. The digital twin bridge seismic and wind-induced vibration and vibration reduction / isolation system constructed in this invention uses a virtual-real interaction module to facilitate the mutual transmission and conversion of various models and data, thereby reducing the workload of pre-processing and post-processing during data analysis and significantly increasing work efficiency. The front end uses a closed-loop control module based on an artificial intelligence agent (AI Agent) and its core components, including the Bridge Engineering Large Language Model (LLM-BE), the Bridge Advanced Construction and Operation Knowledge Graph (KGTADC), the Bridge Engineering Finite Element Simulation Module (FEM-BE), and the Machine Learning Algorithm Intelligent Recognition and Verification Module (MLTE) for technical indicators of construction drawings / Building Information Models / Geological Information Models. It also uses Automated Prompt Generation (APG) technology to design a highly visualized digital twin platform, which facilitates engineers to monitor the seismic and wind vibration response status of bridge structures in real time during the design, construction, and operation phases. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the seismic and wind-induced vibration and vibration reduction and isolation system for long-span bridges in an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the structural module of the seismic and wind-induced vibration reduction and isolation system for long-span bridges in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the AI ​​Agent decision-making process in the seismic and wind-induced vibration and vibration reduction closed-loop control module for long-span bridges in an embodiment of the present invention. Detailed Implementation

[0022] Exemplary embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. However, it should be understood that the invention can be implemented in various forms and should not be limited to the exemplary embodiments set forth herein; rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0023] like Figure 1 As shown in the figure, an embodiment of the present invention provides a digital twin method for seismic and wind-induced vibration reduction and isolation of long-span bridges, comprising the following steps: S01. Benchmark Modeling: Construct a high-fidelity digital model of a long-span bridge. This digital model characterizes the bridge from multiple dimensions, including geometry, physics, behavior, rules, evolution, safety, and failure. It covers complex characteristics such as material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction. Through multi-physics coupling analysis, it achieves accurate simulation of structural mechanics, wind environment, and seismic response.

[0024] In one embodiment of the present invention, a high-fidelity digital model covering the main beam, supports, pile caps, pile foundations, and surrounding soil medium is formed by combining Building Information Modeling (BIM), Geological Information System (GIS), and Finite Element Model (FEM). This digital model encompasses multiphysics coupling simulations of material nonlinearity, geometric nonlinearity, multi-point excitation effects, and pile-soil-structure interactions, supporting dynamic boundary condition loading and real-time parameter updates. Each structural element is assigned an independent constitutive relation and damage evolution equation, where the material constitutive relation includes linear elastic and elastoplastic constitutive relations, and the boundary condition loading includes the seismic motion input spectrum and wind pressure time history curve.

[0025] The high-fidelity digital model needs to refer to bridge design and construction drawings, construction records, material performance reports, and historical monitoring data to complete the initial state assignment and boundary condition setting. During model operation, it can receive real-time feedback data from the ubiquitous sensing module to dynamically correct local element parameters, such as elastic modulus, Poisson's ratio, and yield strength, thereby realizing the online evolution and adaptive updating of the model.

[0026] S02. Ubiquitous Sensing: Deploy an advanced ubiquitous sensing network to conduct integrated multi-source data acquisition across space, air, land, and water. Specifically, "integrated space, air, land, and water" refers to a three-dimensional sensing system encompassing four dimensions: space (satellite InSAR remote sensing monitoring), air (unmanned aerial vehicle inspection), land (surface monitoring equipment such as mobile monitoring platforms like unmanned vehicles, robots, and robotic dogs), and water (surface monitoring using unmanned vessels and underwater vehicles for water area and underwater infrastructure monitoring). Through this integrated sensing network, real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data of the bridge are acquired, and the interaction between the bridge's structural performance and the external environment is perceived.

[0027] The types of data collected include, but are not limited to: satellite InSAR providing large-scale surface deformation monitoring data for monitoring bridge foundation settlement and regional geological activity; UAVs equipped with high-resolution cameras, infrared thermal imagers, lidar, and other equipment for bridge appearance inspection, crack identification, and temperature field distribution monitoring; fiber optic grating sensors deployed on the ground to monitor strain, temperature, and vibration; accelerometers to capture vibration response; displacement sensors to record relative displacement and settlement; GPS / GNSS receivers to provide high-precision three-dimensional displacement information; anemometers to acquire wind environment data; temperature sensors to monitor structural temperature changes; and portable detection equipment carried by mobile monitoring platforms such as unmanned vehicles, robots, and robotic dogs to conduct close-range, detailed inspections of key bridge components; for bridges spanning rivers, seas, and lakes, unmanned surface vessels are deployed for hydrological monitoring, pier scour monitoring, and water quality monitoring, while unmanned underwater vehicles are deployed underwater to inspect underwater structures and measure scour depth of bridge foundations, piers, and abutments.

[0028] In one embodiment of the present invention, the ubiquitous sensing module is used not only to collect response data of the bridge structure itself, but also to comprehensively perceive the interaction between the structural performance and the external environment of the bridge structure during construction and operation. Specifically, the interaction includes, but is not limited to: the influence of temperature field changes on structural stress distribution, the coupling relationship between wind load and structural vibration response, the interaction between the bridge-foundation system under seismic excitation, the accumulation of structural fatigue caused by traffic loads, and the long-term deterioration effect of environmental corrosive media (such as chloride ions and carbon dioxide) on material properties.

[0029] The perception of the interactive state is achieved through spatiotemporal correlation analysis of multi-source data. For example, by correlating ambient temperature data monitored by temperature sensors with structural strain data collected by strain sensors, strain caused by temperature effects and strain caused by load effects can be separated; by coupling wind field data collected by anemometers with structural vibration data captured by accelerometers, the characteristic frequencies and vibration modes of wind-induced vibrations can be identified; by comparing and analyzing seismic ground motion data from earthquake monitoring stations with displacement and acceleration data at bridge foundations, the degree of influence of pile-soil-structure interaction on seismic response can be assessed.

[0030] Furthermore, the perception of the interactive state also includes monitoring the interaction between structural performance and the construction environment. For example, during cantilever construction, the relationship between concrete pouring temperature, hydration heat rise, ambient temperature and humidity, and structural stress and strain is monitored to assess the safety and quality control effectiveness of the structure during the construction phase. During the hoisting of steel box girders, the interaction between wind load, hoisting equipment load, and structural deformation and stress is monitored to ensure the safety and controllability of the construction process. During the operation and maintenance phase, the impact of solar radiation intensity, diurnal temperature difference, and seasonal temperature cycles on the temperature field distribution of the bridge structure, as well as the resulting superposition effect of temperature stress and traffic load stress, is continuously monitored to comprehensively understand the dynamic interaction between structural performance and the external environment.

[0031] In one embodiment of the invention, integrated multi-source data acquisition from the air, ground, and water is employed. This multi-source data includes real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data. Real-time environmental data covers instantaneous wind speed, wind direction, ambient temperature, humidity, air pressure, and key ground motion parameters (such as peak ground acceleration, peak velocity, and dominant frequency) for the area where the bridge is located. Operational load data includes detailed traffic flow information, such as real-time traffic volume, vehicle type identification, lane distribution, and vehicle axle load and total weight obtained through weighing sensors. Structural static and dynamic response data includes key parameters such as strain, vibration acceleration, tilt angle, and settlement at critical structural components. Historical monitoring data comprises a vast amount of data accumulated during the long-term operation of long-span bridges, containing records of structural response under different load conditions, environmental factors, and service durations.

[0032] In a preferred embodiment of the present invention, the ubiquitous sensing module further includes an advanced sensor network, a data acquisition and transmission unit, and a data processing unit.

[0033] 1. An advanced sensor network is configured with a variety of high-precision sensors to build a four-dimensional three-dimensional perception system covering the sky, air, ground, and water.

[0034] At the space-based level, deploy satellite InSAR (Interferometric Synthetic Aperture Radar) systems. Utilize multi-temporal images acquired by synthetic aperture radar satellites and employ differential interferometry to achieve large-scale surface deformation monitoring with millimeter-level precision. This will be used to identify bridge foundation settlement, regional geological activity, and long-term deformation trends.

[0035] At the airborne level, multi-rotor and fixed-wing UAV inspection systems are equipped with high-resolution visible light cameras, infrared thermal imagers, and lidar (LiDAR) sensors to perform bridge appearance inspections regularly or as needed. These systems enable automatic identification of surface cracks, detection of coating peeling, monitoring of temperature field distribution, and 3D point cloud modeling, achieving rapid and safe inspection of areas that are difficult to reach manually.

[0036] At the foundation level, fiber optic grating sensors, accelerometers, displacement sensors, GPS / GNSS receivers, anemometers, and temperature sensors are fixedly deployed at key components of the bridge structure. The fiber optic grating sensors, due to their distributed measurement capabilities, electromagnetic interference resistance, and long-term stability, are widely deployed on critical load-bearing components of the bridge (such as main cables, steel box girder welds, and tension zones of concrete bridge decks) to monitor strain, temperature, and high-frequency vibration response. Accelerometers are installed on bridge towers, main girders, and suspension cables to capture the vibration response of the bridge structure under various loads; for example, low-frequency accelerometers are used for overall modal analysis and damage identification, while high-frequency accelerometers are used for local damage identification and fatigue analysis. Displacement sensors monitor the relative displacement and settlement of the bridge. GPS / GNSS receivers are deployed at key control points of the bridge (such as the top of the main tower and the mid-span of the main girder) to provide high-precision three-dimensional displacement information and long-term settlement trends. Anemometers are installed at different heights on the bridge deck and towers to acquire real-time wind environment data for the area where the bridge is located, including instantaneous wind speed, average wind speed, wind direction, and turbulence intensity. The temperature sensors are used to monitor temperature changes inside the bridge structure (such as inside the concrete and on the surface of the steel) and in the external environment to assess structural deformation and stress caused by temperature effects. Simultaneously, an intelligent mobile monitoring platform is deployed, including unmanned vehicles, wall-climbing robots, and robotic dogs, equipped with portable ultrasonic detectors, eddy current flaw detectors, crack width measuring instruments, and other equipment. This allows for close-range, detailed inspection and regular checks of key bridge components (such as bearings, expansion joints, and the bottom of the main beam), compensating for blind spots covered by fixed sensors.

[0037] At the water-based level, a three-dimensional water monitoring system is deployed for bridges spanning rivers, seas, and lakes. Unmanned surface vessels (USVs) equipped with multibeam echo sounders, side-scan sonar, and water quality sensors are deployed to monitor the hydrological environment around bridge piers, measure scour depth in real time, and collect water quality parameters (such as pH, dissolved oxygen, and chloride ion concentration). Underwater unmanned submersibles (UUVs) equipped with high-definition cameras, sonar imaging systems, and underwater laser scanners are deployed to inspect the structural integrity of the bridge's underwater foundations, piers, abutments, and pile foundations, identifying underwater cracks, spalling, scour pits, marine organism attachment, and other defects, and acquiring three-dimensional morphological data of the underwater structure.

[0038] 2. The data acquisition and transmission unit is used to acquire massive amounts of data generated by the advanced sensor network at high speed and with high reliability. It adopts a distributed data acquisition system, with each acquisition node having multiple synchronous acquisition channels. These acquisition nodes use fiber optic communication technology or wireless transmission technology (such as 5G cellular network, LoRaWAN, satellite communication) to achieve real-time and accurate transmission of monitoring data and ensure data security during transmission.

[0039] 3. The data processing unit is configured to perform data fusion, data cleaning and noise reduction, feature parameter extraction, and pattern recognition and early warning on the collected data. The data fusion employs advanced algorithms, such as the Extended Kalman Filter (EKF), to integrate data from different types of sensors and different sampling frequencies, thereby improving data integrity, consistency, and accuracy. The data cleaning and noise reduction utilizes signal processing techniques, such as the Butterworth filter, to remove environmental noise and sensor drift, and wavelet analysis (such as the Daubechies wavelet) to perform multi-scale decomposition of the data, removing transient interference and high-frequency noise, thus ensuring data quality. The feature parameter extraction extracts key dynamic parameters of the bridge structure, such as frequency, damping ratio, and mode shape, from the massive monitoring data. For example, the Stochastic Subspace Identification (SSI) algorithm is used to identify structural modal parameters from the vibration response under environmental excitation. The data processing unit further utilizes machine learning and deep learning technologies for data pattern recognition, anomaly detection, and early warning of damage. For example, it uses a support vector machine (SVM) to identify abnormal patterns and an autoencoder to perform unsupervised anomaly detection. The autoencoder learns the distribution characteristics of normal data and identifies data points that deviate significantly from the normal pattern.

[0040] S03, Virtual-Real Interaction: Realizes real-time digital mapping and bidirectional dynamic updates between the physical entity and digital model of long-span bridges. Utilizing 5G mobile communication technology, massive amounts of multi-source data collected by ubiquitous sensing modules are transmitted to the digital model in real time, dynamically correcting and updating the parameters of the digital model to ensure synchronization between the physical entity and the digital model.

[0041] In one embodiment of the present invention, the massive amount of multi-source data collected by the ubiquitous sensing module is mapped in real time to the corresponding nodes or units of the digital model using data fusion technology. This data mapping process employs methods such as weighted least squares, Bayesian estimation, or state-space models, optimizing weight allocation based on the accuracy, reliability, and data latency of different sensors. This accurately associates monitoring data such as displacement, velocity, acceleration, strain, temperature, and wind load with the corresponding physical locations and time steps in the digital model. For example, GPS-measured bridge deck displacement can directly update the degrees of freedom of the corresponding nodes in the model, while strain sensor data is used to calibrate the stress state of components in the model.

[0042] Furthermore, the parameters of the digital model are dynamically corrected and updated in real time using model calibration algorithms, such as those based on genetic algorithms (GA), particle swarm optimization (PSO), or Bayesian model updating (BMU). These algorithms drive the dynamic evolution of the digital model by iteratively adjusting its structural stiffness, mass, damping parameters, material constitutive parameters, and boundary conditions, minimizing the error between the digital model's predicted response and the actual monitored response. This dynamic evolution refers to the digital model's ability to continuously adjust its intrinsic parameters and response behavior in response to changes in the physical entity's service condition (e.g., concrete creep, steel fatigue damage), environmental changes (e.g., stress changes caused by temperature gradients, foundation settlement), and damage accumulation (e.g., crack propagation, local stiffness degradation), ensuring that the digital model always accurately reflects the physical entity's current condition, damage accumulation state, and future trends.

[0043] S04. Real-time prediction: Based on the digital model of the self-ubiquitous sensing multi-source data and the benchmark modeling module, the prediction model is trained and obtained using artificial intelligence algorithms such as deep learning and spiking neural network (SNN) to monitor, predict and evaluate the seismic, wind vibration and vibration reduction and isolation nonlinear performance of long-span bridges. The nonlinear performance includes material nonlinearity, geometric nonlinearity, multi-point excitation and pile-soil-structure interaction, and predicts potential performance degradation and failure.

[0044] In one embodiment of the present invention, a multiphysics coupling analysis is used to integrate a structural mechanics model, a wind environment model, and a seismic response model to achieve an understanding of the behavior of long-span bridges in complex environments and accurate response prediction. The wind environment model combines computational fluid dynamics (CFD) technology and wind tunnel test data to finely simulate the wind field distribution and turbulence characteristics around the bridge, obtaining aerodynamic coefficients (such as lift coefficient CL, drag coefficient CD, and moment coefficient CM) of the bridge cross-sections (such as box girder cross-sections and tower cross-sections). The seismic response model takes real-time ground motion data (such as acceleration time history data) from seismic monitoring stations as input and combines it with a high-fidelity digital model provided by the baseline modeling module to simulate the dynamic response of the bridge under seismic loading.

[0045] The real-time prediction module, when conducting performance evaluation, pays particular attention to the evolution of the interaction between bridge structural performance and the external engineering environment. Specifically, it establishes a multivariate nonlinear mapping relationship between environmental factors and structural response to achieve dynamic tracking and trend prediction of the interaction state. For example, a coupled temperature field-stress field prediction model is established. This model uses historical temperature monitoring data and stress monitoring data as training samples, employing a Long Short-Term Memory (LSTM) network to learn the time-delay relationship and nonlinear mapping law between temperature change patterns and stress response. This allows it to predict the temperature stress distribution and potential temperature crack risk of the bridge based on future temperature forecasts. Another example is a coupled wind load-vibration response prediction model. This model integrates wind field monitoring data, bridge vibration monitoring data, and CFD simulation results, using a Graph Neural Network (GNN) to capture the topological correlation between the bridge's spatial structure and wind field distribution. This enables rapid prediction of the bridge's vibration response under different wind speed and direction combinations and identifies dangerous wind conditions that may trigger vortex-induced vibration and flutter.

[0046] Furthermore, the real-time prediction module employs a multi-timescale prediction strategy, using different prediction time windows for different types of interaction effects. For rapidly changing interaction processes (such as wind-induced vibration and seismic response), short-term predictions at the second to minute level are used to provide decision-making time for emergency response; for slowly evolving interaction processes (such as temperature effects, fatigue accumulation, and material degradation), medium- to long-term predictions at the day to year level are used to provide a basis for the formulation of preventive maintenance plans. The multi-timescale prediction strategy is implemented through a cascaded prediction model: the short-term prediction model is based on high-frequency monitoring data from the most recent period, using a recurrent neural network (RNN) or a temporal convolutional network (TCN) to capture short-term dynamic features; the medium- to long-term prediction model is based on the statistical characteristics of long-term historical data and physical degradation models, using combined prediction methods (such as the integration of physical models and data-driven models) to improve the accuracy and reliability of predictions.

[0047] In a preferred embodiment of the present invention, the real-time prediction module is used to train and obtain an intelligent prediction model using artificial intelligence algorithms such as deep learning and spiking neural networks (SNNs) on a large-scale dataset. This intelligent prediction model is specifically designed for real-time monitoring, high-precision prediction, and comprehensive analysis of the nonlinear performance of long-span bridges under seismic, wind-induced vibration, and vibration reduction / isolation conditions. It can effectively predict potential performance degradation and early failures of bridge structures, thereby providing a time window for subsequent decision-making.

[0048] S05. Closed-Loop Control: Based on the evaluation and prediction results of the real-time prediction module, and combined with a pre-trained AI agent, predictive maintenance decisions, risk management, and maintenance optimization are performed. The AI ​​agent combines time-series data provided by the ubiquitous sensing module, the virtual-real interaction module, and the real-time prediction module. Through automated analysis and evaluation, it outputs a comprehensive evaluation report including damage classification, risk level, and performance degradation trend. It also dynamically assesses the risk level of the bridge under different loads and environmental conditions, identifies potential weaknesses, and formulates optimal maintenance plans, including maintenance timing, maintenance content, and resource allocation, to avoid sudden failures and reduce maintenance costs.

[0049] In one embodiment of the present invention, the closed-loop control module includes an intelligent decision-making and control system based on an artificial intelligence agent (AIAgent). This system integrates multiple specialized modules to work together to achieve intelligent management of the entire life cycle of long-span bridges.

[0050] In specific implementation, the AI ​​Agent is constructed using the Large Language Model for Bridge Engineering (LMM-BE) as the core inference engine, combined with four key modules: the Knowledge Graph of Technologies for Advanced Design and Construction (KGTADC), the Finite Element Method for Bridge Engineering (FEM-BE), and the Machine Learning for Technical Evaluation (MLTE) module, which uses machine learning algorithms to intelligently identify and verify technical indicators in construction drawings, BIM, and GIS.

[0051] The Bridge Engineering Large Language Model (LLM-BE) is based on the Transformer architecture and adopts a two-stage training strategy of "pre-training-fine-tuning". In specific implementation, a professional corpus in the field of bridge engineering is first constructed, including but not limited to: domestic and foreign bridge design specifications (such as the "Code for Seismic Design of Highway Bridges" JTG / T 2231-01-2020), academic papers, technical manuals, construction cases, monitoring reports, damage analysis reports and maintenance records, etc., covering multi-dimensional knowledge such as bridge structural design, construction technology, material properties, monitoring technology, damage mechanisms and maintenance strategies.

[0052] During the model pre-training phase, a general-purpose large language model (such as a mainstream domestic large-scale model) is selected as the pre-training base. Domain-adaptive pre-training is performed using the aforementioned bridge engineering-specific corpus, enabling the model to learn the professional terminology, conceptual relationships, and knowledge structure of the bridge engineering field. Subsequently, for specific tasks such as bridge health monitoring, damage diagnosis, risk assessment, and maintenance decision-making, targeted optimization is performed using supervised fine-tuning and reinforcement learning from human feedback (RLHF) techniques, enabling the model to achieve expert-level reasoning capabilities in bridge engineering tasks. The LLM-BE model can be selected at different scales based on the computational resources and response speed requirements of the actual application scenario. Model deployment can utilize a combination of cloud servers and edge computing devices to achieve efficient inference and rapid response.

[0053] The bridge advanced construction and operation knowledge graph (KGTADC) is constructed using graph database technology, specifically Neo4j, ArangoDB, or the domestic graph database TuGraph as the underlying storage engine. The knowledge graph construction process includes: 1. Knowledge Acquisition and Extraction. Structured knowledge is extracted from bridge engineering design specifications, technical standards, academic literature, engineering cases, monitoring data, and expert experience. Natural language processing technology combined with manual annotation is used to identify bridge types (suspension bridges, cable-stayed bridges, arch bridges, etc.), structural components (main beams, main cables, bridge towers, bearings, pile foundations, etc.), material types (concrete, steel, rubber, etc.), types of defects (cracks, corrosion, fatigue, erosion, etc.), environmental factors (temperature, wind load, earthquake, traffic load, etc.), and their relationships.

[0054] 2. Knowledge Representation and Modeling. An ontology modeling approach is used to define formal schemas for entity types, attributes, and relation types. Entity types include bridge structure classes, material classes, load classes, defect classes, sensor classes, and maintenance measure classes; relation types include semantic associations such as "belongs to," "contains," "causes," "affects," "applies to," and "monitoring." Based on the definition of entity and relation types, attribute information (such as bridge span, material strength, defect level, maintenance cost, etc.) is assigned to each entity node, and weights or confidence levels are assigned to each relation edge.

[0055] 3. Knowledge Fusion and Reasoning. Entity alignment, conflict resolution, and consistency verification are performed on multi-source heterogeneous knowledge to construct a unified knowledge graph. Multi-hop reasoning, path querying, and knowledge completion are achieved based on the graph's topology. For example, the possible causes of fatigue damage in the main cable can be diagnosed through the reasoning path "suspension bridge - main cable - fatigue damage - stress amplitude - wind-induced vibration".

[0056] 4. Knowledge Updates and Maintenance. The knowledge graph has a dynamic update mechanism. After each anomaly diagnosis and maintenance decision is completed, new case data, diagnostic results, and maintenance effects will be automatically added to the database, continuously expanding the coverage and reasoning capabilities of the knowledge graph.

[0057] The implementation method of the bridge engineering finite element simulation module (FEM-BE) is as follows: Based on the computational kernel of mature finite element analysis software (such as ANSYS, ABAQUS, SAP2000, Midas Civil, etc.), an automated simulation interface for AI agents is constructed through secondary development or script-based encapsulation. Specific implementations include: utilizing Python scripts to call the batch processing function of the finite element software to automatically complete model parameter modification, mesh generation, boundary condition application, solver invocation, and result extraction; establishing standardized input and output data formats so that the AI ​​agent can transmit structural parameters, load cases, and analysis types through the API interface, and receive calculation results such as displacement, stress, and modal parameters; pre-setting analysis templates for common analysis scenarios (such as modal analysis, time history analysis, nonlinear analysis, fluid-structure interaction analysis, etc.) to improve simulation efficiency; and deploying high-performance computing clusters to support parallel computing of large-scale models and batch analysis of multiple load cases, ensuring rapid acquisition of verification simulation results during the AI ​​agent's decision-making process.

[0058] The machine learning algorithm-based intelligent identification and verification module for construction drawings / BIM / GIS technical indicators (MLTE) comprises three parts: a front-end visual interface, an intelligent drawing identification submodule, and a specification verification submodule. 1. Front-end visual interface. A desktop application developed using Qt and Python enables the uploading, display, and annotation of construction drawings, as well as the 3D visualization and browsing of BIM / GIS models.

[0059] 2. Intelligent Drawing Recognition Submodule. Employing deep learning object detection algorithms (such as Faster R-CNN and YOLO series) and optical character recognition (OCR) technology, this submodule automatically extracts key information from construction drawings, including component dimensions, material specifications, reinforcement details, and technical requirements, and stores the recognition results in a structured format. For BIM models, it parses model files using the IFC (Industry Foundation Classes) standard interface to extract component geometric parameters, material properties, spatial relationships, and other information. For GIS models, it uses geographic data format parsing tools such as Shapefile and GeoJSON to extract information such as bridge locations, topography, geological conditions, and environmental parameters.

[0060] 3. Standardized Verification and Calculation Submodule. Based on bridge engineering design standards (such as the "Standard for Technical Condition Assessment of Highway Bridges" JTG / T H21-2011 and the "Code for Seismic Design of Highway Bridges" JTG / T 2231-01-2020), a verification program is written. It takes as input technical parameters extracted from drawings and models, automatically performs bearing capacity verification, seismic performance verification, wind resistance performance verification, technical condition assessment, durability evaluation, etc., and outputs the verification results and standard compliance judgment. The verification module adopts a modular design, encapsulating different standard clauses into independent verification functions, facilitating expansion and version updates.

[0061] Through the coordinated work of the above four core modules, the AI ​​Agent can achieve a complete closed-loop control process from data perception, knowledge reasoning, simulation verification to decision generation.

[0062] The specific implementation process of the closed-loop control includes the following steps: 1. Data Reception and Situational Awareness. The Artificial Intelligence Agent (AIAgent) receives real-time multi-source monitoring data streams from the ubiquitous sensing module, digital model updates from the virtual-real interaction module, and performance evaluation results from the real-time prediction module. Through Automated Prompt Generation (APG) technology, the AI ​​Agent automatically integrates key information such as timestamps, measurement point locations, monitoring parameters and their values, exceedance conditions, environmental conditions, and historical trends into a structured situational description text. For example, when the accelerometer detects that the main beam vibration amplitude exceeds the design threshold, APG technology automatically generates a situational report containing complete information such as "Time: October 6, 2025, 14:35; Location: Main beam L / 4 span section; Parameter: Lateral vibration acceleration; Value: 0.35g; Threshold: 0.25g; Exceedance rate: 40%; Environment: Southwest wind, level 8; Historical comparison: 120% increase compared to the average of the last 30 days."

[0063] 2. Intelligent Diagnosis and Cause Analysis. After receiving the situational description, the Large Language Model for Bridge Engineering (LLM-BE) leverages its expertise in bridge engineering and combines it with the Knowledge Graph of Technologies for Advanced Design and Construction (KGTADC) to perform multi-step reasoning. This knowledge graph stores structured knowledge such as the correlation between bridge types and typical defects, the mapping between defect characteristics and causal mechanisms, and the causal relationship between environmental factors and structural responses. For example, for the aforementioned vibration anomaly, LLM-BE first searches the knowledge graph for the path "suspension bridge - main girder - lateral vibration - wind-induced vibration," identifying possible causes including vortex-induced vibration, flutter precursors, or abnormal support constraints. Subsequently, LLM-BE calls the Finite Element Model for Bridge Engineering (FEM-BE) module, inputting the current wind speed, wind direction, and bridge structural parameters to perform rapid CFD-structural coupling analysis and calculate the theoretical vibration response under the current wind conditions. By comparing the measured values ​​with the theoretical values, LLM-BE determined that the measured vibration amplitude was significantly higher than the theoretical prediction, further inferring that there might be bearing constraint degradation or damper performance deterioration. Next, LLM-BE retrieved the bridge bearing's historical inspection records, maintenance records, and real-time monitoring data, and found that the friction coefficient monitoring value of a certain bearing had been continuously decreasing over the past 6 months, confirming that insufficient bearing lubrication leading to reduced constraint stiffness was the main cause of the abnormal vibration.

[0064] 3. Risk Assessment and Graded Early Warning. Based on the diagnostic results, the AI ​​Agent automatically assesses the risk level. The risk level assessment strictly follows the assessment method and grading standards of the "Technical Condition Assessment Standard for Highway Bridges" JTG / T H21-2011.

[0065] In response to the abnormal bearing constraint situation in this embodiment of the invention, the AI ​​Agent invokes the Bridge Engineering Large Language Model-BE (LLM-BE) to parse the clauses of the JTG / T H21-2011 standard, and performs inference by combining the mapping relationship between bearing defect types, damage characteristics and assessment levels stored in the Bridge Advanced Construction and Operation Knowledge Graph (KGTADC). According to Clause 3.2.1 of the JTG / TH21-2011 standard, LLM-BE identifies bridge components as primary components and secondary components. The assessment scale for primary components is divided into Class 1, Class 2, Class 3, Class 4 and Class 5, and the assessment scale for secondary components is divided into Class 1, Class 2, Class 3 and Class 4.

[0066] Furthermore, the AI ​​Agent, based on monitoring data, identified defects in the supports such as "continuously decreasing friction coefficient and insufficient lubrication," and the measured value of the lateral vibration acceleration at the L / 4 span of the main beam was 0.35g, exceeding the design threshold of 0.25g. It then used machine learning algorithms to intelligently identify technical indicators from construction drawings / BIM / GIS and the verification and calculation module (MLTE) to assess the technical condition. According to Clause 3.2.2 and Table 3.2.2 of the standard, the supports are considered major components. The AI ​​Agent, referring to the assessment scale for major component supports in Table 3.2.4, classified them into three categories: moderate material defects; or mild functional defects that are developing slowly and still maintain normal functionality.

[0067] Furthermore, the AI ​​Agent invoked FEM-BE to perform nonlinear time history analysis, simulating the bridge's response under design wind speed and seismic loads in the current support condition, calculating the maximum displacement of the main girder, stress level, and support forces. The results show that the bridge still meets safety requirements under common wind loads in the current condition, but the main girder displacement exceeds the limit by 15% under design wind speed, and there is a risk of support slippage under rare earthquake loads.

[0068] Based on the assessment results and risk analysis, the AI ​​Agent outputs maintenance recommendations: the support components are assessed as Category 3, and it is recommended that professional personnel be arranged to conduct on-site inspection and confirmation within 72 hours, and that maintenance work such as support cleaning, lubrication replenishment, and friction coefficient testing be completed within 10 working days. During the maintenance period, an encrypted monitoring plan should be activated (monitoring frequency increased to once per hour), and a traffic control plan should be activated when the wind speed exceeds 20m / s.

[0069] 4. Maintenance Decision and Solution Generation. The AI ​​Agent automatically generates optimized maintenance solutions based on multiple objectives, including risk level, current bridge condition, available resources, maintenance costs, and traffic impact. For bearing lubrication maintenance, the AI ​​Agent searches the knowledge graph for the association between "bearing-insufficient lubrication-maintenance solution," extracting recommended maintenance measures such as cleaning the bearing surface, replenishing grease, checking sealing devices, testing the coefficient of friction, and replacing worn parts if necessary. It further utilizes machine learning algorithms to intelligently identify technical indicators from construction drawings / BIM / GIS and uses a verification module (Machine Learning for Technical Evaluation, MLTE) to automatically parse the bridge bearing design drawings and BIM model, extracting technical parameters such as bearing model, specifications, grease type, and lubrication cycle, and checking for matching spare parts and materials in the current inventory. The AI ​​Agent also uses historical bridge traffic flow data and a time series prediction model to predict traffic flow distribution for the next 7 days, identifying 11:00 PM to 5:00 AM the following day as a low-traffic period and recommending maintenance work during this time to minimize traffic impact. The final maintenance plan includes detailed work steps, a list of required materials, personnel allocation recommendations, estimated work duration, traffic management plan, and emergency response plan.

[0070] 5. Full Lifecycle Performance Evolution Prediction. The closed-loop control module not only responds to current abnormal states but also makes long-term predictions of the bridge's future performance evolution based on historical data and physical degradation models. The closed-loop control module plays a core role in the full lifecycle management of long-span bridges, spanning all stages including design, construction, operation, and maintenance. The AI ​​Agent integrates short-term predictions from the real-time prediction module with long-term degradation predictions based on physical models, employing a digital twin model for accelerated evolution simulation. For example, regarding carbonation and chloride ion corrosion of concrete main beams, the AI ​​Agent first uses machine learning algorithms to intelligently identify technical indicators in construction drawings / BIM / GIS and automatically parses the concrete structure design drawings and BIM model using the MLTE (Made for Detailed) module. This extracts key technical indicators such as cement type, water-cement ratio, protective layer thickness, and concrete strength grade. Then, combining the Bridge Engineering Large Language Model-BE (LLM-BE) and the Bridge Advanced Construction and Operation Knowledge Graph (KGTADC), the deterioration mechanism is analyzed, and the "Standard for Durability Assessment of Existing Concrete Structures" (GB / T 51355-2019) is used for evaluation. Subsequently, the built-in GB / T standard in MLTE is invoked. The 51355-2019 verification procedure integrates extracted technical indicators, environmental monitoring data (temperature, humidity, chloride ion concentration), and historical monitoring data (carbonation depth, chloride ion concentration distribution). Based on the carbonation depth and chloride ion diffusion calculation methods specified in Chapters 5 and 6 and Appendices B, C, and D of this standard, it comprehensively considers factors such as concrete location, curing conditions, working stress, strength grade, and environmental temperature and humidity to predict the concrete durability degradation process over the next 10, 20, and 50 years, and assess the risk of steel corrosion and the trend of load-bearing capacity degradation. An AI agent presents the prediction results in a timeline format, marking key expected time points, such as "It is predicted that the carbonation depth will reach the protective layer thickness in 15 years, and it is recommended to repair the protective layer in the 12th year; it is predicted that the load-bearing capacity will decrease by 10% due to steel corrosion in 25 years, and it is recommended to reinforce in the 20th year." This forward-looking prediction provides a scientific basis for developing preventative maintenance plans, optimizing maintenance timing, and allocating resources.

[0071] 6. Closed-Loop Feedback and Continuous Optimization. After the maintenance plan is implemented, the ubiquitous sensing module continuously collects response data of the bridge structure, the virtual-real interaction module updates the digital model parameters based on the actual state after maintenance, the real-time prediction module reassesses the bridge performance, and the AI ​​Agent compares the monitoring data and performance indicators before and after maintenance to evaluate the maintenance effect. For example, after the bearing lubrication maintenance is completed, the AI ​​Agent compares the vibration response of the main beam under the same wind conditions before and after maintenance to verify whether the vibration amplitude has been reduced to the normal range; it calls FEM-BE to recalculate the response of the bridge under the design load after maintenance to confirm whether the safety margin has been restored. If the maintenance effect is not good or new anomalies occur, the secondary diagnostic process is automatically initiated to adjust the maintenance plan or upgrade the warning level. At the same time, the complete closed-loop process from anomaly detection, diagnostic analysis, risk assessment, maintenance decision to effect verification is recorded, including the decision basis, reasoning path, measures taken and final effect, forming case data stored in the knowledge graph for continuous training and optimization of the LLM-BE model and decision algorithm, continuously improving the intelligence level and decision accuracy of the AI ​​Agent. Through this closed-loop feedback mechanism, the digital twin system has achieved a transformation from passive response to proactive prevention, and from experience-based decision-making to intelligent optimization, significantly improving the safety management level and life-cycle operation and maintenance efficiency of long-span bridges under complex conditions such as earthquakes and wind vibrations.

[0072] like Figure 2 As shown, the digital twin long-span bridge seismic and wind vibration reduction and isolation system includes a benchmark modeling module, a ubiquitous sensing module, a virtual-real interaction module, a real-time prediction module, and a closed-loop control module.

[0073] In specific implementation, the data flow and collaborative control process between modules is as follows: The baseline modeling module constructs a high-fidelity digital model of the long-span bridge, providing initial parameters and theoretical basis for the sensor deployment scheme of the ubiquitous sensing module and the digital mapping of the virtual-real interaction module; the ubiquitous sensing module acquires multi-source data through its data acquisition and transmission unit, while deeply sensing the interaction state between the bridge structural performance and the external environment of the project, and transmits it to the virtual-real interaction module in real time; the virtual-real interaction module realizes low-latency, high-bandwidth data interaction between the physical entity and the digital model based on 5G communication technology, completes the dynamic calibration and parameter update of the digital model, and transmits the updated digital model and real-time monitoring data to the real-time prediction module; the real-time prediction module uses artificial intelligence algorithms such as deep learning and spiking neural networks (SNN) to train the input data, outputs a comprehensive evaluation result including structural performance assessment, damage identification, and risk warning, and transmits it to the closed-loop control module; the closed-loop control module is based on artificial intelligence agents (AI) The Agent and its core components (including the Bridge Engineering Large Language Model LLM-BE, the Bridge Advanced Construction and Operation Knowledge Graph KGTADC, the Bridge Engineering Finite Element Simulation Module FEM-BE, and the Machine Learning Algorithm Intelligent Recognition of Construction Drawings / BIM / GIS Technical Indicators and Verification Module (MLTE)) perform in-depth analysis and decision-making on the evaluation results. Through Automated Prompt Generation (APG) technology, intelligent collaboration between modules is achieved to generate optimal maintenance strategies, including maintenance sequence planning, resource scheduling schemes, and emergency response measures. Predictive maintenance, risk management, and optimized allocation of maintenance resources are implemented through a feedback mechanism. This collaborative working mechanism among the modules ensures the system's stable operation and intelligent decision-making capabilities under multi-field coupling, complex loads, and uncertain environments.

[0074] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A digital twin system for seismic and wind-induced vibration reduction and isolation of long-span bridges, characterized in that, It includes a baseline modeling module, a ubiquitous sensing module, a virtual-real interaction module, a real-time prediction module, and a closed-loop control module; The benchmark modeling module is used to construct a high-fidelity digital model of a long-span bridge. The digital model can characterize the multi-dimensional features of the bridge from aspects such as geometry, physics, behavior, rules, evolution, safety, and failure. It can simulate the material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges under earthquake, wind vibration, and vibration reduction and isolation conditions. It can also achieve accurate simulation of structural mechanics, wind environment, and seismic response models through multiphysics coupling analysis. The ubiquitous sensing module is used to collect multi-source data from the sky, air, ground, and water, and to sense the interaction between the structural performance of the bridge structure during construction and operation and the external environment of the project; the multi-source data includes real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data; The virtual-real interaction module is used to realize real-time digital mapping and bidirectional dynamic updates between the physical entity and digital model of long-span bridges, ensuring the synchronization between the physical entity and the digital model. The real-time prediction module receives multi-source data from the ubiquitous sensing module and digital models from the benchmark modeling module. It then uses artificial intelligence algorithms, including deep learning and spiking neural networks (SNNs), to train and obtain a prediction model. This model monitors, predicts, and evaluates the seismic, wind-induced vibration, and vibration isolation nonlinear performance of long-span bridges, as well as the interaction between bridge structural performance and the external engineering environment. The nonlinear performance includes material nonlinearity and geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction. The module also predicts potential performance degradation and failures. The closed-loop control module receives the evaluation and prediction results from the real-time prediction module and performs predictive maintenance decisions, risk management, and maintenance optimization through the artificial intelligence agent AIAgent. This plays a core role in the full life cycle management of long-span bridges, which spans all stages of design, construction, operation, and maintenance to achieve closed-loop control of the digital twin.

2. The digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system according to claim 1, characterized in that, The benchmark modeling module acquires data through finite element simulation and a bridge health monitoring system. The finite element simulation covers refined analysis of material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges.

3. The digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system according to claim 1, characterized in that, The ubiquitous sensing module includes an advanced sensor network, a data acquisition and transmission unit, and a data processing unit. The advanced sensor network includes satellite InSAR systems, unmanned aerial vehicle (UAV) inspection systems, fixed surface sensors, intelligent mobile monitoring platforms, and water monitoring systems. The satellite InSAR system is used for large-scale surface deformation monitoring; the UAV inspection system is equipped with visible light cameras, infrared thermal imagers, and lidar for appearance inspection and temperature field monitoring; the fixed surface sensors include fiber optic grating sensors, accelerometers, displacement sensors, GPS / GNSS receivers, anemometers, and temperature sensors for monitoring strain, vibration, displacement, three-dimensional position, wind environment, and temperature changes, respectively; the intelligent mobile monitoring platform includes unmanned vehicles and robots for refined inspection of key components; and the water monitoring system includes unmanned vessels and unmanned underwater vehicles for hydrological monitoring and underwater structure inspection, in order to comprehensively acquire multi-dimensional data required for the interaction between bridge structural performance and the external environment of the project. The data acquisition and transmission unit is used to acquire data from the advanced sensor network and to achieve real-time and accurate transmission of monitoring data through optical fiber and wireless transmission technologies, while ensuring data security and integrity. The data processing unit is used to perform multi-scale data fusion on the collected data, clean and reduce noise by using filtering or wavelet analysis, extract key feature parameters such as frequency, damping, mode shape, stress distribution, environmental load characteristics, and structural deformation mode from massive monitoring data, and use machine learning and deep learning technologies for data pattern recognition, anomaly detection, early warning of damage, and intelligent identification of the interaction between bridge structural performance and external environment.

4. The digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system according to claim 1, characterized in that, The real-time prediction module integrates structural mechanics models, wind environment models, and seismic response models through multiphysics coupling analysis to achieve understanding of the behavior of long-span bridges in complex environments and accurate response prediction. The multiphysics coupling analysis adopts a two-way coupling mechanism to achieve deep feedback between structural dynamic response and environmental factors. The wind environment model combines computational fluid dynamics (CFD) technology and wind tunnel test data to finely simulate the wind field around the bridge, obtain the aerodynamic coefficient of the bridge section, and couple it with the structural dynamics model to accurately predict the wind-induced vibration response, vortex-induced vibration, flutter, and buffeting failure modes. The earthquake response model combines real-time ground motion data from earthquake monitoring stations with a bridge structure model to simulate the dynamic response of the bridge under earthquake action. It also considers material nonlinearity, geometric nonlinearity, and the characteristics of energy-consuming devices to perform multi-scale simulation and nonlinear behavior simulation, so as to evaluate the damage state, vulnerability, and probability of failure of the bridge under different earthquake intensities.

5. The digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system according to claim 1, characterized in that, The real-time prediction module adopts an intelligent prediction model based on artificial intelligence algorithms and data-driven methods. The intelligent prediction model includes prediction models based on recurrent neural networks, graph neural networks and reinforcement learning. It learns the health evolution law of the bridge from historical monitoring data and data on the interaction between structural performance and the external environment, predicts future performance degradation and potential failures, and assesses the long-term impact of the interaction between the bridge structural performance and the external environment. The intelligent prediction model is also used for data-driven damage identification. By comparing the real-time monitoring data with the response of the digital twin model under normal conditions, it identifies changes in structural stiffness and damping parameters, thereby determining the occurrence and extent of damage and predicting the damage evolution path and remaining life.

6. The digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system according to claim 1, characterized in that, The virtual-real interaction module is responsible for mapping the massive multi-source data collected by the ubiquitous sensing module to the corresponding nodes or units of the digital model in real time through data fusion technology, and dynamically correcting and updating the topology, physical parameters and behavior patterns of the digital model based on real-time data, thereby driving the dynamic evolution of the digital model to achieve digital management and precise control of the entire life cycle of the long-span bridge. The mapping process employs weighted least squares, Bayesian estimation, or state-space modeling methods.

7. A digital twin system for seismic and wind-induced vibration reduction and isolation of long-span bridges according to claim 1, characterized in that, The closed-loop control module uses an AI Agent to automatically analyze time-series data provided by the ubiquitous sensing module, virtual-real interaction module, and real-time prediction module. The AI ​​Agent can process the evaluation results in depth and output a comprehensive evaluation report on the seismic, wind-induced vibration, and vibration reduction and isolation performance of long-span bridges, including damage classification, risk level, performance degradation trend, and evolution of the interaction between structural performance and the external environment.

8. A digital twin system for earthquake and wind-induced vibration reduction and isolation of long-span bridges according to claim 1, characterized in that, The AI ​​entity is built around time-series data provided by the ubiquitous perception module, virtual-real interaction module, and real-time prediction module. It utilizes the Bridge Engineering Large Language Model LLM-BE, the Bridge Advanced Construction and Operation Knowledge Graph KGTADC, the Bridge Engineering Finite Element Simulation Module FEM-BE, and the Machine Learning Algorithm Intelligent Identification of Construction Drawings / BIM / GIS Technical Indicators and Verification Module MLTE as its core components. It achieves intelligent decision-making and control by automatically generating APGs through prompts.

9. A method for implementing a digital twin long-span bridge seismic and wind-induced vibration reduction and isolation system as described in claim 1, characterized in that, Includes the following steps: S01, Benchmark Modeling: Construct a high-fidelity digital model of a long-span bridge. The digital model is characterized from multiple dimensions, including geometry, physics, behavior, rules, evolution, safety, and failure. It simulates the material nonlinearity, geometric nonlinearity, multi-point excitation, and pile-soil-structure interaction of long-span bridges under seismic, wind-induced vibration, and vibration reduction and isolation conditions. It also achieves accurate simulation of structural mechanics, wind environment, and seismic response through multiphysics coupling analysis. S02, Ubiquitous Sensing: Conduct integrated multi-source data acquisition from the sky, air, ground, and water to obtain real-time environmental data, operational load data, structural static and dynamic response data, and historical monitoring data of long-span bridges, and perceive the interaction between the structural performance of bridge construction and operation and the external environment of the project. S03, Virtual-Real Interaction: Realizes real-time digital mapping and bidirectional dynamic updates between the physical entity of a long-span bridge and the digital model, ensuring the synchronization between the physical entity and the digital model; S04. Real-time prediction: Receive multi-source data from ubiquitous sensing and digital models from benchmark modeling, use artificial intelligence algorithms to train and obtain prediction models, monitor, predict and evaluate the seismic, wind-induced vibration and vibration reduction and isolation nonlinear performance of long-span bridges, as well as the interaction between bridge structural performance and the external environment of the project. S05. Closed-loop control: Receives real-time prediction assessment results and performs predictive maintenance decisions, risk management, and maintenance optimization through an artificial intelligence agent (AI Agent), thereby playing a core role in the full life cycle management of long-span bridges. The full life cycle management runs through all stages of design, construction, operation, and maintenance to achieve closed-loop control of the digital twin.

10. A digital twin method for seismic and wind-induced vibration reduction and isolation of long-span bridges according to claim 9, characterized in that, The decision-making process of the AI ​​Agent in the closed-loop control steps includes: It receives time-series data from ubiquitous sensing, virtual-real interaction, and real-time prediction, and generates a situation description using APG technology through automated prompts. The LLM-BE (Large Language Model for Bridge Engineering) combined with the KGTADC (Knowledge Graph for Advanced Bridge Construction and Operation) is used for multi-step reasoning to diagnose the causes of anomalies. The bridge engineering finite element simulation module FEM-BE was used to verify the diagnostic results and assess the risk level. Based on machine learning algorithms, the intelligent identification module MLTE for construction drawings / BIM / GIS technical indicators and verification calculations analyzes construction drawings and BIM / GIS models to generate optimized maintenance plans that include work steps, material lists, and time-series planning. After maintenance, continuously monitor the effects and store the decision-making process and results in a knowledge graph for continuous model optimization.

Images