Bridge structural health monitoring system based on industrial internet

Abstract

This invention relates to the field of bridge structural health monitoring systems based on the Industrial Internet, specifically disclosing a bridge structural health monitoring system based on the Industrial Internet. The system includes a bridge structural sensor network, a micro-meteorological sensing unit, a computational fluid dynamics simulation engine, a physical information neural network proxy model, a real-time boundary correction unit, and a structural health assessment center. By fusing micro-meteorological data with high-fidelity wind field simulation, a wind pressure-structural response mapping model with physical consistency is constructed, and real-time boundary correction is used to achieve closed-loop synchronization between the digital twin and the physical entity. The structural health assessment center identifies anomalies based on the expected response and measured deviations, and reduces false alarms by combining an adaptive threshold mechanism. Furthermore, the model has the ability to extrapolate and predict wind-induced vibrations, enabling early risk warnings. This invention can achieve high-precision damage identification and proactive safety warnings, improving the proactive safety assurance and intelligent operation and maintenance level of bridges in complex wind environments.

Description

Technical Field

[0001] This invention belongs to the field of bridge structural health monitoring and industrial internet integration application, specifically involving a bridge structural health monitoring system based on the industrial internet. Background Technology

[0002] With the rapid development of Industrial Internet technology, its application in the operation and maintenance of large-scale transportation infrastructure has become an important way to improve bridge safety and management efficiency. Bridge structural health monitoring integrates sensor perception, data transmission, and intelligent analysis technologies to achieve real-time assessment of the service status of bridges in complex environments. Especially for long-span bridges, establishing a comprehensive monitoring system plays a crucial role in identifying early damage, preventing structural failure, and optimizing maintenance decisions.

[0003] Digital twin technology, which integrates multiphysics simulation, is a key direction for achieving accurate monitoring. This technology aims to simulate the dynamic response of bridges under different climatic conditions by constructing a digital mirror image that includes meteorological environment, mechanical properties, and structural parameters. By combining the fluid dynamics principles of meteorology with structural mechanics models, the system can reconstruct the realistic load transfer mechanism in digital space, providing in-depth theoretical support for the accurate diagnosis of bridge conditions.

[0004] Existing technologies typically treat environmental wind loads as simplified boundary inputs, making it difficult to accurately characterize the complex unsteady flows generated by local microclimates on bridges, leading to discrepancies between simulation results and actual operating conditions. Furthermore, traditional monitoring schemes lack efficient decoupling capabilities between environmental excitations and structural responses, failing to accurately identify the essential differences between minute damage signals and normal environmental disturbances, resulting in high false alarm rates and insufficient robustness. In addition, due to the lack of deep learning analysis constrained by physical laws, the monitoring process relies excessively on discrete sensor sampling data, making it difficult to provide proactive early warnings for sudden vibration risks, resulting in monitoring accuracy and real-time performance that fail to meet the demands of modern bridge maintenance.

[0005] Therefore, a bridge structure health monitoring system based on the Industrial Internet is desired. Summary of the Invention

[0006] The purpose of this invention is to provide a bridge structural health monitoring system based on the Industrial Internet, which can solve the problems of difficulty in decoupling environmental excitation and structural response, high false alarm rate and lack of early warning capability in the above-mentioned background technology.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A bridge structural health monitoring system based on the Industrial Internet includes a bridge structural sensor network, a micro-meteorological sensing unit, a computational fluid dynamics simulation engine, a physical information neural network proxy model, a real-time boundary correction unit, and a structural health assessment center, wherein:

[0009] The bridge structure sensor network is deployed at key stress-bearing parts of the bridge and is configured to collect structural response data of the bridge in real time during its service life, including acceleration, displacement, strain and vibration frequency.

[0010] The micro-meteorological sensing unit consists of several wind speed sensors distributed in the main beam, tower and cable area of ​​the bridge, and is configured to acquire real-time wind speed and direction information of the local wind field.

[0011] The computational fluid dynamics simulation engine is based on the three-dimensional geometric model of the bridge and its surrounding terrain data to construct a high-precision unsteady flow simulation model, which is used to simulate the local vortices, separation flow and reattachment phenomena formed by wind on the surface of complex structures, and to generate a dynamic wind pressure distribution field throughout the entire bridge.

[0012] The physical information neural network proxy model takes the wind pressure distribution field output by the computational fluid dynamics simulation engine as input, the structural response data collected by the bridge structure sensor network as supervision signal, and embeds the physical residuals of the basic fluid dynamics equations into its loss function to train and form a wind pressure-structure response mapping relationship model with physical consistency.

[0013] The real-time boundary correction unit receives the measured wind speed data from the micro-meteorological sensing unit and dynamically adjusts the entry boundary conditions of the computational fluid dynamics simulation engine to keep the simulated wind field synchronized with the actual micro-meteorological state, thereby realizing closed-loop interaction between the digital twin and the physical entity.

[0014] The structural health assessment center integrates the output of the physical information neural network proxy model, and identifies abnormal signals exceeding a preset threshold by comparing the deviation between the current structural response and the expected response derived from the corrected wind field, thereby determining whether structural damage exists.

[0015] Preferably, the computational fluid dynamics simulation engine uses unstructured meshes to perform fine modeling of the complex geometric features of the bridge, which can capture the three-dimensional turbulent structure of the cable wake, the flow around the bridge deck, and the intersection of the tower and the beam. Its time step is set according to the lowest natural frequency of the bridge structure to ensure accurate capture of low-frequency responses such as vortex-induced vibration.

[0016] Furthermore, the physical information neural network proxy model adopts a multilayer perceptron architecture. Its input layer receives spatially discretized wind pressure field data, and its output layer corresponds to the structural response prediction values ​​at each sensor location. The hidden layer embeds the continuity and momentum conservation constraints of the Navier-Stokes equations, so that the model not only fits the data during the training process, but also satisfies the basic physical laws of fluid motion.

[0017] Furthermore, the real-time boundary correction unit employs a Kalman filter algorithm to fuse sparse observation data from the micro-meteorological sensing unit with the prior wind field from the computational fluid dynamics simulation engine, dynamically updating the velocity profile and turbulence intensity parameters at the wind field inlet, thus maintaining the accuracy of the wind field simulation even in areas with insufficient sensor coverage.

[0018] Preferably, the structural health assessment center is equipped with an adaptive threshold determination mechanism. This adaptive threshold determination mechanism dynamically adjusts the sensitivity of damage identification based on the current wind speed level, temperature changes, and traffic load level, so as to avoid false alarms triggered by fluctuations in normal structural response under strong wind or temperature change conditions.

[0019] Furthermore, the bridge structure sensor network and micro-meteorological sensing unit are connected to the edge computing node through the Industrial Internet Protocol. The edge computing node preprocesses and compresses the raw data before uploading it to the cloud platform. The computational fluid dynamics simulation engine and physical information neural network proxy model are deployed on the cloud high-performance computing cluster to realize large-scale parallel simulation and model inference.

[0020] Furthermore, the physical information neural network proxy model has extrapolation capabilities, enabling it to predict the wind-induced vibration response that will occur in a critical area based on upstream wind speed data and the aerodynamic shape of the bridge before the current wind field reaches that area, thereby issuing an early warning for structural safety.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. The bridge structural health monitoring system based on the Industrial Internet provided by this invention achieves high-fidelity simulation of complex local wind fields and accurate prediction of structural response of bridges by constructing a micro-meteorological-structural coupled digital twin that integrates meteorology, structural mechanics and deep learning.

[0023] 2. The system utilizes a physical information neural network to take the basic equations of fluid dynamics as intrinsic constraints, which improves the generalization ability and physical consistency of the surrogate model, distinguishes between normal responses caused by environmental wind loads and abnormal signals caused by structural damage, and reduces the false alarm rate.

[0024] 3. By achieving virtual-real interaction through real-time boundary correction units, the digital twin can dynamically track real micro-meteorological changes, ensuring the timeliness and reliability of simulation results. The system has the ability to extrapolate wind-induced effects in unobserved areas and can predict risk events such as vortex-induced vibration in advance, upgrading the traditional post-event alarm mode to a proactive pre-event warning mechanism, thereby improving the proactive safety assurance capabilities and intelligent operation and maintenance level of large bridges in extreme wind environments. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0026] Figure 2 This is a schematic diagram of the core principle framework of the physical information neural network proxy model in this invention;

[0027] Figure 3 This is a logical flowchart of computational fluid dynamics simulation and real-time boundary correction in this invention;

[0028] Figure 4 This is a schematic diagram illustrating the multi-level interaction relationship and data flow between the physical entity of the bridge and the cloud-based digital twin in this invention;

[0029] Figure 5 This is a flowchart illustrating the logical process of structural health assessment based on expected response comparison and adaptive threshold in this invention. Detailed Implementation

[0030] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0031] A bridge structural health monitoring system based on the Industrial Internet includes a bridge structural sensor network, a micro-meteorological sensing unit, a computational fluid dynamics simulation engine, a physical information neural network proxy model, a real-time boundary correction unit, and a structural health assessment center.

[0032] The bridge structure sensor network, deployed at key stress-bearing components of the bridge, is configured to collect structural response data in real time during service. This data includes at least acceleration, displacement, strain, and vibration frequency. The network comprises multiple independent and time-synchronized sensing nodes, each connected to an edge processing terminal via an industrial Ethernet or high-bandwidth wireless dedicated network to ensure high concurrency and millisecond-level latency in data acquisition. Physically, the network includes triaxial accelerometers distributed at the mid-span of the main girder to capture the vertical and lateral vibration characteristics under wind loads; high-precision satellite positioning displacement monitoring stations distributed at the top of the main towers to acquire real-time wind-induced displacement of the tower tops under strong winds; and fiber optic strain sensors embedded in key stress zones of the main girder cross-sections to monitor the stress state of structural components under complex environmental loads. The network also possesses self-diagnostic capabilities, monitoring the power supply status and communication link quality of each sensing node in real time, automatically marking and filtering channels with abnormal signals to output a high-fidelity original structural response sequence.

[0033] The micro-meteorological sensing unit, composed of several wind speed sensors, is distributed across the main girder, towers, and cable-stayed areas of the bridge. It is configured to acquire real-time wind speed and direction information of the local wind field. The wind speed sensors in the micro-meteorological sensing unit are three-dimensional ultrasonic anemometers, characterized by no starting wind speed limit, high linearity across the entire measurement range, and the ability to simultaneously measure three-dimensional wind speed vector components. The spatial layout of the micro-meteorological sensing unit follows the principle of wind field gradient coverage. Gradient observation towers are set upstream of the bridge, and dense measuring points are set at the bridge deck height to finely characterize the initial flow field distribution of wind entering the bridge structure. The real-time data acquired by the micro-meteorological sensing unit is preprocessed via an edge gateway, including removing outliers caused by airflow pulsations and calculating characteristic parameters such as average wind speed, turbulence intensity, and gust factor according to a preset time window, providing input benchmarks for subsequent flow field simulations.

[0034] The computational fluid dynamics (CFD) simulation engine, based on the bridge's three-dimensional geometric model and surrounding terrain data, constructs a high-precision unsteady flow simulation model to simulate local vortices, separation flows, and reattachment phenomena formed by wind on complex structural surfaces, generating a dynamic wind pressure distribution field across the entire bridge. The CFD simulation engine is configured to use unstructured meshes for refined modeling of the bridge's complex geometric features. These unstructured meshes include hexahedral boundary layer meshes distributed in the near-wall region and tetrahedral meshes distributed in the far-field region. This multi-scale meshing strategy captures the three-dimensional turbulent structure of cable wakes, flow around the bridge deck, and the tower-beam intersection area. The CFD simulation engine's computational logic is based on Navier's and Stokes' equations and employs high-fidelity turbulence numerical models such as large eddy simulation (LES) or separated eddy simulation (SES). Its implementation process includes discretizing the fluid momentum equations and using an implicit format in the time dimension. The time step of the computational fluid dynamics simulation engine is set according to the lowest natural frequency of the bridge structure, specifically, the time step is set to be less than 1 / 10 of the bridge's first-order natural period to ensure accurate capture of low-frequency responses such as vortex-induced vibration. The dynamic wind pressure distribution field includes the instantaneous pressure coefficients at each discrete grid point on the bridge surface, which can dynamically show the non-uniform distribution characteristics of wind-induced loads in space.

[0035] The physical information neural network proxy model takes the wind pressure distribution field output by the computational fluid dynamics simulation engine as input, and the structural response data collected by the bridge structure sensor network as the supervision signal. It embeds the physical residuals of the fundamental fluid dynamics equations into its loss function, training to form a physically consistent wind pressure-structural response mapping model. The physical information neural network proxy model employs a multilayer perceptron architecture, including an input layer, multiple hidden layers, and an output layer. The input layer receives spatially discretized wind pressure field data, and the output layer corresponds to the predicted structural response values ​​at each sensor location. During training, the physical information neural network proxy model introduces physical constraint terms into the loss function, ensuring that while fitting data points, its output satisfies the constraints of the continuity equation and the momentum conservation equation across the entire field. This physical constraint is achieved by substituting the partial derivatives of the network output with respect to spatial position and time into the fluid motion control equations and calculating the corresponding residual values. When the residual value approaches zero, it indicates that the model's predicted physical quantity distribution not only matches the known observations but also conforms to the fundamental physical logic of fluid dynamics. The physical information neural network proxy model constructs a high-dimensional mapping with strong generalization ability by deeply integrating massive simulation data and measured data, and achieves near real-time prediction of the response of complex nonlinear structures while reducing computational overhead.

[0036] The real-time boundary correction unit receives measured wind speed data from the micro-meteorological sensing unit and dynamically adjusts the inlet boundary conditions of the computational fluid dynamics simulation engine to keep the simulated wind field synchronized with the actual micro-meteorological state, achieving closed-loop interaction between the digital twin and the physical entity. The real-time boundary correction unit employs a Kalman filter algorithm to fuse the sparse observation data from the micro-meteorological sensing unit with the prior wind field from the computational fluid dynamics simulation engine. Within each correction cycle, the real-time boundary correction unit inversely calculates the velocity profile parameters and turbulence intensity parameters at the global flow field inlet based on the wind speed residuals at the measured points, and dynamically updates the input parameters of the simulation engine accordingly. This closed-loop correction mechanism ensures that the digital twin system can track the drift and fluctuations of the actual meteorological environment in real time, maintaining the accuracy of wind field prediction through simulation calculations even in areas with insufficient sensor coverage. The real-time boundary correction unit is also configured with error self-learning capabilities, continuously optimizing the filter gain coefficient by analyzing the deviation characteristics between historical observation sequences and simulation sequences, thereby improving the accuracy of virtual-real synchronization.

[0037] The structural health assessment center integrates the output of the physical information neural network surrogate model. By comparing the deviation between the current structural response and the expected response extrapolated from the corrected wind field, it identifies abnormal signals exceeding a preset threshold and determines whether structural damage exists. The structural health assessment center is equipped with an adaptive threshold determination mechanism, which dynamically adjusts the sensitivity of damage identification based on the current wind speed level, temperature changes, and traffic load levels. When the wind speed is low, the system automatically lowers the alarm threshold to improve the sensitivity to capturing minor damage signals; while in severe environments such as strong winds or extreme cold, the system raises the judgment benchmark accordingly based on a preset environmental response correction curve to avoid false alarms triggered by normal structural stress fluctuations. The structural health assessment center also includes an early warning extrapolation module. Utilizing the extrapolation capability of the physical information neural network surrogate model, it predicts the amplitude of wind-induced vibrations that will occur in the bridge's critical sensitive areas before the current wind field reaches them, based on sentinel data from upstream wind speed sensors and the bridge's aerodynamic shape characteristics. When the predicted response value reaches the fatigue damage critical point, the system automatically issues an early warning command, upgrading the traditional post-event alarm mode to a proactive pre-event safety defense mechanism.

[0038] In a further preferred embodiment, each sensing node in the bridge structure sensor network employs high-precision timestamp synchronization technology. By using a Global Positioning System (GPS) time signal, the time synchronization error of data acquisition is controlled within the microsecond range to ensure the accuracy of phase relationships during multi-point synchronous analysis. The sensor network is also equipped with an intelligent self-testing circuit, capable of periodically performing sensor self-calibration procedures. By applying a known physical excitation signal to the sensing element, it detects the response offset at the output end and automatically compensates for sensitivity attenuation in the data processing logic.

[0039] The micro-meteorological sensing unit is further configured to include a rainfall monitoring subunit and a visibility monitoring subunit. The rainfall monitoring subunit employs a piezoelectric rain gauge to sense rainfall intensity in real time and incorporates it as an additional physical parameter into the computational fluid dynamics simulation engine. When rainfall is detected, the simulation engine automatically adjusts the air density and dynamic viscosity parameters of the fluid to account for the damping effect of raindrops on airflow characteristics. The visibility monitoring subunit is used to assist in determining traffic flow distribution characteristics under foggy weather conditions, thereby helping the structural health assessment center to more accurately decouple the contributions of environmental loads and traffic loads to the structural response.

[0040] The computational fluid dynamics simulation engine employs a local mesh refinement strategy, specifically targeting the cable-stayed bridge region, during unstructured mesh generation. Due to their slender shape, cables are highly susceptible to vortex-induced vibrations under strong winds. The simulation engine arranges at least 15 boundary layer meshes along the normal direction of the cable surface. The height of the first mesh layer is rigorously calculated based on the dimensionless distance requirements of the wall function to ensure the resolution of the minute vortex shedding frequency on the leeward side of the cable. The simulation engine also supports parallel computing task scheduling, enabling the decomposition and distribution of complex full-bridge flow field calculations across different computing nodes in a cloud-based high-performance computing cluster. This leverages many-core parallelism technology to shorten the time required for a single unsteady computation.

[0041] In its construction, the physical information neural network surrogate model, in addition to embedding fluid dynamics equations, further incorporates d'Alembert's principle from solid mechanics, thereby establishing a fully coupled constraint between wind pressure distribution and structural dynamic response. An improved activation function is used in the hidden layer, possessing continuous second derivatives across the entire domain to ensure numerical stability and prevent gradient vanishing during partial differential equation operator calculations. Model training employs a combination of online and offline training. Specifically, long-term accumulated historical monitoring data is used periodically for deep optimization of the global network weights, while incremental learning techniques are used during real-time operation to fine-tune local weights based on recent micro-meteorological characteristics, ensuring the model consistently maintains optimal predictive performance.

[0042] The real-time boundary correction unit employs a nonlinear correction strategy during Kalman filtering to address the strong nonlinear characteristics exhibited by fluid motion. This unit not only corrects the wind speed inlet conditions but also adaptively adjusts the equivalent roughness parameters of the bridge surface. When the observed wind pressure distribution on the bridge deck deviates regularly from the simulated values, the system uses an optimization algorithm to search for the best-matching surface roughness correction coefficient to reflect the aerodynamic changes on the actual structural surface caused by corrosion, icing, or dust accumulation.

[0043] The structural health assessment center possesses a multi-dimensional damage identification algorithm library. Besides threshold alarms based on response deviation, it integrates a damage localization module based on modal curvature change rate and a non-stationary signal processing module based on Hilbert-Huang transform. The adaptive threshold determination mechanism comprehensively considers the periodic impact of seasonal temperature fluctuations on the main beam stiffness. By constructing a regression model of temperature and stiffness, it automatically removes temperature-induced frequency shifts during the assessment process, thereby significantly improving the robustness of identifying internal physical damage. When issuing an early warning, the early warning simulation module automatically generates a three-dimensional visualized dynamic map of wind field evolution, intuitively displaying the trajectory of the approaching strong wind mass and its predicted impact on specific bridge components on a digital twin platform.

[0044] Furthermore, to ensure the system's operational stability under high load, the industrial internet architecture employs a dual-route redundancy design. The core data switch is connected to the edge computing server via a 10 Gigabit fiber optic cable and has an independent backup wireless microwave communication link. In the event of an accidental fiber optic cable breakage, the system can automatically switch to the microwave link within milliseconds, ensuring uninterrupted structural monitoring data flow. The edge computing nodes are configured with data caching and local decision-making capabilities, enabling them to perform basic health assessments and emergency warnings based on locally stored simplified models even in extreme cases of cloud connection loss.

[0045] Furthermore, the bridge structural health monitoring system also integrates an autonomous drone inspection interface. When the structural health assessment center identifies a suspected damaged area, it automatically sends coordinate commands to the drone inspection system. Based on the received coordinates, the drone, carrying a high-resolution industrial camera and an infrared thermal imager, automatically takes off and flies to the target area to perform close-range visual inspection and temperature field scanning. The acquired image information is then fed back to the assessment center, achieving cross-verification of macroscopic monitoring data and microscopic visual features.

[0046] The physical information neural network proxy model is also equipped with a long-term performance prediction branch, which is specifically used to assess the fatigue life evolution of bridges during long-term service. By processing the historical wind pressure cumulative effect output by the physical information neural network with rainflow counting, the system can calculate the equivalent fatigue stress amplitude of each key weld and connector. Combined with the fatigue characteristic curve of the material, the remaining service life of the structure can be dynamically estimated, providing a quantitative basis for maintenance departments to formulate reasonable maintenance plans.

[0047] The structural health assessment center is also connected to a visual management terminal, which uses a digital twin large-screen display interface built on a 3D geographic information system. Maintenance personnel can click on virtual bridge components on the large screen to view real-time comparison curves of predicted physical quantities and actual sensor measurements for that location. They can also use the retrospective function to view the dynamic response process of the bridge during any historical storm event. The system also supports automatic tiered alarm notifications, automatically sending warning reports to different levels of management personnel via SMS, email, or instant messaging tools based on the severity of the damage.

[0048] Example 2: As a supplement and evolution of Example 1, this example describes an implementation of a bridge structure health monitoring system based on a distributed edge computing architecture and hardware redundancy mechanism to meet the high reliability application requirements in ultra-long span bridges or extreme geographical environments.

[0049] A bridge structural health monitoring system based on the Industrial Internet includes a multi-level distributed sensor cluster, a heterogeneous network transmission module, a distributed CFD acceleration platform, an enhanced physical information neural network unit, a collaborative boundary correction gateway, and a multi-criteria integrated health management system.

[0050] The multi-level distributed sensor cluster divides the entire bridge into multiple logical monitoring sections. Each monitoring section is equipped with an independent local data aggregation center, which uses an industrial-grade ARM processor as its core to directly manage the bridge structure sensor network within that section. Specifically, sensor data in each section is first synchronized, aligned, and feature-extracted at the local aggregation center, with only the extracted high-order modal features or anomaly statistics uploaded to the central station. This distributed architecture effectively solves the bandwidth pressure and computational bottlenecks faced by traditional centralized acquisition schemes when processing thousands of monitoring channels. The sensor selection in the multi-level distributed sensor cluster further emphasizes environmental adaptability, employing fully sealed stainless steel housings and surge protection modules to resist the corrosive effects of high salt spray and strong lightning strikes on electronic components in the cross-sea bridge environment.

[0051] The heterogeneous network transmission module is configured as a composite architecture integrating a 5G private network, satellite links, and an industrial-grade self-healing ring network. Under normal operating conditions, sensor data is transmitted via a bidirectional self-healing fiber optic ring network deployed within the bridge maintenance tunnel. This ring network supports millisecond-level automatic fault switching, ensuring that a break in any link does not affect the overall data flow. In the event of fiber optic damage or ground base station failure due to a major disaster, the heterogeneous network transmission module automatically activates a geostationary orbit satellite communication terminal, using the satellite link to upload critical structural alarm signals, achieving all-weather, all-region connectivity assurance.

[0052] The distributed CFD acceleration platform employs a multi-GPU (Graphics Processing Unit) parallel computing node cluster in its hardware architecture. Each node is equipped with large-capacity video memory and a high-speed interconnect bus, specifically designed for processing flow field calculations on large-scale unstructured meshes. In terms of algorithm implementation, the distributed CFD acceleration platform utilizes a multi-overlapping mesh technique, dividing the full-bridge flow field into multiple overlapping sub-regions. Each sub-region is solved in parallel by different GPU cores, and global consistency is achieved through boundary data exchange in the overlapping areas. This approach shortens the simulation time for unsteady flows, enabling the system to update the full-field wind pressure data at a real-time frequency, thus meeting the requirements for rapid response to sudden gusty wind loads.

[0053] The enhanced physical information neural network unit, based on Embodiment 1, introduces an attention mechanism and a long short-term memory (LSM) network structure. The input layer not only receives the instantaneous wind pressure field but also incorporates historical wind field evolution information from the previous time series, automatically identifying key flow field feature regions that contribute most to the structural response using the attention mechanism. For example, the system automatically enhances the extraction of detached vortex features at the leading edge of the main beam while suppressing background noise in the far-field uniform flow region. The LSM network structure is configured to capture time lag effects in the structural response, which is particularly important for bridge components with large damping characteristics. The enhanced unit adds an additional structural energy conservation constraint to the loss function, requiring the predicted system input power to remain balanced with the sum of damping dissipation and kinetic energy changes, further strengthening the physical realism of the model from the perspective of energy flow.

[0054] The collaborative boundary correction gateways are deployed at key nodes of the bridge, including the main span, tower base, and anchorages. Each gateway operates an independent edge-side Kalman filter, configured to correct not only the inlet wind speed but also the bridge structure's dynamic parameters, such as modal damping ratio, in real time. Since the damping ratio is complexly affected by environmental humidity, temperature, and vibration amplitude, it is difficult to pre-calibrate precisely. The collaborative boundary correction gateways utilize response data from local high-density measurement points to online invert the structure's dynamic damping parameters and synchronously feed the corrected parameters back to the physical information neural network. This bidirectional correction mechanism enables the digital twin to accurately locate environmental excitations and precisely describe the structure's own energy dissipation characteristics.

[0055] The multi-criteria integrated health management system employs a Bayesian inference framework in its evaluation logic. This system not only relies on a single criterion of response deviation but also incorporates prior knowledge from historical maintenance records, material aging models, and collective experience data from similar bridges. By weighted fusion of various damage indicators (such as stiffness reduction coefficient and coherence function attenuation rate), the posterior damage probability distribution of the structure is calculated. The management system also integrates a fault tracing module. When an abnormal response is detected, this module automatically analyzes the gradient contribution rate in the physical information neural network to pinpoint the most likely load location or component part causing the anomaly, providing precise navigation for on-site manual inspection.

[0056] In the hardware implementation of the distributed system, each sensing node employs energy harvesting technology, enabling it to generate its own power using micro-vibrations in the bridge environment or solar energy, thus reducing maintenance costs and system failure risks associated with large-scale cabling. Each edge computing node is equipped with a security chip based on a hardware encryption engine, ensuring the confidentiality and integrity of monitoring data during transmission through the Industrial Internet, and preventing malicious network attacks from tampering with or stealing bridge safety data.

[0057] The enhanced physical information neural network unit also includes an adaptive mesh mapping module, which dynamically interpolates the wind pressure values ​​at discrete mesh points generated by computational fluid dynamics simulations to the stress nodes of the structural finite element model. During the interpolation process, the system employs high-precision interpolation algorithms such as radial basis functions to ensure the conservation of the total spatial integral of the load during the mapping process, thus avoiding spurious numerical excitations caused by mesh mismatch.

[0058] The multi-criteria integrated health management system is also configured with macro-level decision support capabilities. It can automatically simulate the bridge's evolution over the next 48 hours, incorporating typhoon warnings issued by the local meteorological station. By running tens of thousands of Monte Carlo simulations, the system assesses the probability of bridge overturning, structural damage, or instability under the expected maximum gust of wind and provides traffic control recommendations (such as closing some lanes or prohibiting traffic on the entire bridge). This decision support function significantly improves the effectiveness of traffic management departments in responding to risks during extreme weather events.

[0059] Furthermore, this embodiment introduces a digital twin growth mechanism to address long-term structural performance degradation. As the bridge's service life increases, the system automatically records the actual evolution trajectory of the structure. By analyzing the trend of the slow drift of sensor reference values, the system dynamically corrects the reference stiffness and boundary constraints in the digital twin model. This growth mechanism ensures that the digital twin remains a true mapping of the physical entity throughout the bridge's entire lifespan, rather than a rigid initial state model.

[0060] Regarding the user interface, this embodiment provides a mobile maintenance terminal based on augmented reality technology. When maintenance engineers walk onto the bridge wearing AR glasses, the system directly overlays and displays real-time sensor data hidden inside the structure, structural stress cloud maps, and historical damage records in their real-world field of view. Through this interactive method, engineers can intuitively observe the virtual trajectories of wind flowing on the bridge surface and perceive the distribution of subtle vibrations invisible to the naked eye, facilitating on-site inspections and technical assessments.

[0061] Example 3: This example focuses on describing the extended application of the system in dealing with multiple disaster coupling and sudden load conditions, as well as the advanced analysis module of cloud collaboration, aiming to demonstrate the system's deep protection capabilities in complex and extreme environments.

[0062] A bridge structural health monitoring system based on the Industrial Internet, in addition to the aforementioned modules, further integrates an earthquake early warning triggering module, a traffic load precision sensing subsystem, a multi-hazard coupling analysis cloud platform, and an emergency response automated command center;

[0063] The earthquake early warning triggering module is interconnected in real time with the National Seismic Network to receive early warning signals of arriving seismic waves. Upon detecting the arrival of a P-wave, the module rapidly increases the sampling frequency of the bridge structure sensor network to the kilohertz level and triggers each module to enter high-speed operating mode. The physical information neural network surrogate model automatically switches to a dynamic-dominated mode at this time, prioritizing the calculation of the large displacement response of the structure caused by foundation vibration. By introducing the seismic wave input as a condition for base motion, the system can decouple the combined effects of seismic forces and wind loads in real time and assess the dynamic safety margin of the structure during strong earthquakes.

[0064] The traffic load precision sensing subsystem utilizes weighing sensors (WIMs) deployed beneath the bridge deck pavement and video AI recognition technology to acquire real-time data on vehicle type, axle load, speed, and precise location. This dynamic traffic information is input into the structural health assessment center in real time, serving as another major environmental excitation factor besides wind load. The system employs a physical information neural network to simultaneously process the superimposed effects of aerodynamic loads and moving vehicle loads, enabling a more accurate interpretation of local strain fluctuations in the structure under vehicle impact and eliminating interference from traffic noise in damage identification.

[0065] The multi-hazard coupled analysis cloud platform, deployed in a cloud center with supercomputing capabilities, can handle quadruple coupled simulations of wind, rain, earthquakes, and traffic loads. During typhoon and rainy weather, the cloud platform can utilize a refined raindrop impact model to simulate the minute changes in aerodynamic shape caused by the accumulation of rainwater on the bridge deck. This ability to simulate extreme detail allows the system to predict specific nonlinear vibration phenomena such as wind-induced vibration. The cloud platform also features a large-scale historical database, enabling the mining of monitoring data from similar bridges nationwide. Using transfer learning techniques, damage features learned from other damaged bridges are transferred to the system's discrimination model, achieving intelligent early warning sharing across bridges.

[0066] The automated emergency response command center is connected to the bridge's broadcasting system, traffic lights, and emergency escape signs. When the system detects an immediate risk of structural failure (such as main beam displacement exceeding the safety limit or cable breakage), the center can issue a bridge-wide closure order within seconds without manual intervention and automatically activate evacuation guidance for the flood discharge area. The center also has an interface with the surrounding intelligent transportation system, enabling vehicles to automatically avoid damaged bridge sections and minimize loss of life and property caused by secondary disasters.

[0067] In the simulation logic involving multiple coupled hazards, the system employs a layered processing strategy. First, a computational fluid dynamics simulation engine generates pure wind field data. Then, the traffic load subsystem superimposes moving point loads. Finally, a physical information neural network surrogate model uniformly calculates the nonlinear contribution to the structural response. During this process, the system can identify whether dangerous conditions arise due to resonance between wind-induced vibrations and vehicle driving frequencies, and accordingly provide specific speed limits or vibration reduction recommendations.

[0068] To further enhance the interpretability of the physical information neural network surrogate model, this embodiment also introduces symbolic regression analysis. While training the neural network, the system attempts to extract simplified explicit algebraic expressions from massive amounts of data. These expressions represent empirical formulas relating wind speed, vehicle type, and key bridge displacements, providing engineers with an intuitive tool based on first principles, thus making the "black box" model transparent to a certain extent.

[0069] The structural health assessment center also incorporates a dynamic evaluation system for structural vulnerability. This system uses real-time monitored conditions as input, combined with a structural resistance degradation model, to calculate the failure probability contour map of the bridge under different reliability indices. This evaluation system not only provides a "yes or no" alarm, but also a quantitative level of the overall structural safety. For example, the system might indicate that "the fatigue reserve of the main span section has been depleted to 75% under the current working conditions," providing guidance for precise preventative maintenance.

[0070] The system is also equipped with a virtual sensor interpolation module. When the actual number of sensors deployed is limited, the physical information neural network proxy model can act as a "virtual sensor," generating virtual response sequences in areas where no physical hardware is deployed. By comparing and verifying the virtual sensors with the actual measured sensors, the system can assess the stress level at any point within the entire bridge range, truly achieving full-area digital coverage of the physical entity.

[0071] In summary, the bridge structural health monitoring system based on the Industrial Internet provided by this invention constructs a micro-meteorological and structural coupled digital twin with high physical fidelity and logical evolution capabilities by deeply integrating meteorological CFD models, physical information neural network AI architecture, and real-time closed-loop correction technology. The system effectively solves the core technical pain points of traditional monitoring schemes, such as insufficient accuracy in depicting complex wind fields, difficulty in decoupling environmental excitations, and lagging early warning modes. By accurately predicting and identifying anomalies in structural responses, the system reduces the false alarm rate and achieves proactive early warning of risk events such as wind-induced vibrations, thereby improving the proactive safety management level of long-span bridges in harsh environments.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions of the present invention or equivalent substitutions can be made to some of the technical features, and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

[0073] All calculation processes, comparison logic, parameter relationships, and logical operations involved in this invention have been completely converted into pure Chinese text descriptions, without containing any mathematical formulas or algebraic expressions. The terms "first," "second," etc., used in this specification are only used to distinguish different components or modules and do not imply any order or importance. In practical applications, the modules of this invention can be integrated into a single hardware server or distributed across edge computing nodes and cloud clusters; such variations in deployment are all within the scope of protection of this invention.

[0074] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in the system claims may also be implemented by the same unit or device in software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any particular order.

Claims

1. An industrial internet-based bridge structure health monitoring system, characterized by, include: A bridge structure sensor network is deployed at key stress-bearing parts of the bridge and configured to collect structural response data of the bridge in real time during its service life. The structural response data includes acceleration, displacement, strain, and vibration frequency. Micro-meteorological sensing units are distributed in the main beams, towers, and cable areas of the bridge, and are configured to acquire real-time wind speed and direction information of the local wind field; The computational fluid dynamics simulation engine is configured to build an unsteady flow simulation model based on the three-dimensional geometric model of the bridge and surrounding terrain data, simulate the local vortices, separation flow and reattachment phenomena formed by wind on the structural surface, and generate a dynamic wind pressure distribution field throughout the entire bridge area. A physical information neural network proxy model is configured to take the dynamic wind pressure distribution field as input, the structural response data as a supervision signal, and embed the physical residuals of the basic equations of fluid dynamics into its loss function to construct a wind pressure and structural response mapping relationship model with physical consistency. The real-time boundary correction unit is configured to receive the measured wind speed data from the micro-meteorological sensing unit and dynamically adjust the inlet boundary conditions of the computational fluid dynamics simulation engine to keep the simulated wind field synchronized with the actual micro-meteorological state. The structural health assessment center is configured to integrate the output of the physical information neural network proxy model. By comparing the deviation between the current structural response and the expected response derived from the corrected wind field, it identifies abnormal signals that exceed a preset threshold and determines whether structural damage exists. 2.The bridge structure health monitoring system based on industrial internet of claim 1, wherein: The bridge structure sensor network includes multiple independent and time-synchronized sensing nodes, each of which is connected to an edge processing terminal via an industrial Ethernet or a high-bandwidth wireless dedicated network. The physical deployment of the sensing nodes includes triaxial accelerometers distributed at the mid-span of the main girder of the bridge, high-precision satellite positioning displacement monitoring stations distributed at the top of the main tower, and fiber optic strain sensors embedded in the key stress areas of the main girder section. The bridge structure sensor network has a self-diagnostic function and is configured to monitor the power supply status and communication link quality of each sensor node in real time, and automatically mark and filter channels with abnormal signals. Each of the aforementioned sensing nodes employs high-precision timestamp synchronization technology, using global positioning system timing signals to control the time synchronization error of data acquisition within the microsecond range; The sensor network is also equipped with an intelligent self-testing circuit configured to periodically perform a sensor self-calibration process. By applying a known physical excitation signal to the sensing element, it detects the response offset at the output end and automatically compensates for sensitivity decay in the data processing logic. 3.The bridge structure health monitoring system based on industrial internet of claim 2, wherein: The micro-meteorological sensing unit consists of several three-dimensional ultrasonic anemometers. The three-dimensional ultrasonic anemometers have full-range linearity characteristics and are configured to measure three-dimensional wind speed vector components. The micro-meteorological sensing unit follows the principle of wind field gradient coverage in its spatial layout. Gradient observation towers are set up in the upstream direction of the bridge, and dense measuring points are set up at the height of the bridge deck. The real-time data acquired by the micro-meteorological sensing unit is preprocessed by the edge gateway, including removing outliers caused by airflow pulsation and calculating the average wind speed, turbulence intensity and gust factor characteristic parameters according to a preset time window. The micro-meteorological sensing unit also includes a rainfall monitoring subunit and a visibility monitoring subunit. The rainfall monitoring subunit uses a piezoelectric rain gauge to sense the rainfall intensity in real time and introduces the rainfall intensity as an additional physical parameter into the computational fluid dynamics simulation engine, so that the computational fluid dynamics simulation engine adjusts the air density and dynamic viscosity parameters of the fluid according to the rainfall intensity. The visibility monitoring subunit is configured to assist in determining the traffic flow distribution characteristics under foggy weather conditions, and to help the structural health assessment center decouple the contribution of environmental loads and traffic loads to the structural response. 4.The bridge structure health monitoring system based on industrial internet of claim 3, wherein: The computational fluid dynamics simulation engine uses unstructured meshes to model the geometric features of the bridge. The unstructured meshes include hexahedral boundary layer meshes distributed in the near-wall region of the bridge and tetrahedral meshes distributed in the far-field region. The computational fluid dynamics simulation engine applies a local mesh refinement strategy to the cable area of ​​the bridge, arranging at least fifteen boundary layer meshes in the normal direction of the cable surface, and the height of the first layer mesh is calculated according to the dimensionless distance requirement of the wall function, so as to analyze the vortex shedding frequency on the leeward side of the cable. The computational fluid dynamics simulation engine is based on the Navier and Stokes equations in its computational logic, and uses large eddy simulation or separated eddy simulation turbulence numerical models. Its execution process includes discretizing the fluid momentum equation and advancing it in the time dimension using an implicit format. The time step of the computational fluid dynamics simulation engine is set to be less than one-tenth of the first-order natural period of the bridge. The computational fluid dynamics simulation engine supports parallel computing task scheduling and is configured to decompose and distribute full-bridge flow field calculation tasks to different computing nodes of the cloud-based high-performance computing cluster, thereby shortening the time consumption of a single unsteady calculation by utilizing many-core parallel technology. 5.The bridge structure health monitoring system based on industrial internet of claim 4, wherein: The physical information neural network proxy model adopts a multilayer perceptron architecture, which includes an input layer, multiple hidden layers, and an output layer. The input layer receives spatially discretized wind pressure field data, and the output layer corresponds to the structural response prediction values ​​at each sensor location. The physical information neural network proxy model introduces physical constraint terms into the loss function, so that the network output satisfies the constraints of the continuity equation and the momentum conservation equation throughout the entire field. The physical constraints are achieved by substituting the partial derivatives of the network output with respect to spatial position and time into the fluid motion control equations to calculate the residual values. The hidden layer employs an activation function with continuous second derivatives across the entire domain to ensure numerical stability during partial differential equation operator computation. The physical information neural network proxy model is also configured with a long-term performance prediction branch, which is configured to evaluate the fatigue life evolution of the bridge during long-term service. By performing rainflow counting processing on the historical wind pressure accumulation effect output, the equivalent fatigue stress amplitude of key connecting parts is calculated, and the remaining service life of the structure is dynamically estimated by combining the fatigue characteristic curve of the material.

6. The bridge structural health monitoring system based on the Industrial Internet according to claim 5, characterized in that: The real-time boundary correction unit uses the Kalman filter algorithm to fuse the observation data of the micro-meteorological sensing unit with the prior wind field of the computational fluid dynamics simulation engine. Within each correction cycle, the real-time boundary correction unit inverts and calculates the velocity profile parameters and turbulence intensity parameters of the global flow field inlet based on the wind speed residuals at the measured points, and dynamically updates the input parameters of the computational fluid dynamics simulation engine accordingly. The real-time boundary correction unit has error self-learning capability and is configured to analyze the deviation characteristics between historical observation sequences and simulation sequences to optimize the gain coefficient of the filter. During the Kalman filtering process, the real-time boundary correction unit adopts a nonlinear correction strategy and simultaneously adaptively adjusts the equivalent roughness parameter of the bridge surface. The optimization algorithm searches for a surface roughness correction coefficient that matches the actual observed wind pressure distribution on the bridge surface to reflect the changes in aerodynamic characteristics of the structural surface caused by corrosion, icing, or dust accumulation.

7. A bridge structural health monitoring system based on the Industrial Internet according to claim 6, characterized in that: The structural health assessment center is equipped with an adaptive threshold determination mechanism, which dynamically adjusts the sensitivity of damage identification based on the current wind speed level, temperature changes, and traffic load level. When the wind speed is at a preset low level, the system automatically lowers the alarm threshold. In strong wind or extremely cold environments, the system raises the judgment benchmark according to the preset environmental response correction curve to avoid false alarms triggered by normal structural stress fluctuations. The structural health assessment center also includes an early warning and simulation module. Utilizing the extrapolation capability of the physical information neural network proxy model, it can predict the amplitude of wind-induced vibrations that will occur in the area before the current wind field reaches the critical and sensitive area of ​​the bridge, based on the sentinel data transmitted back from the upstream wind speed sensor and the aerodynamic shape characteristics of the bridge. The early warning simulation module is configured to automatically issue an early warning command when the forecast response value reaches the fatigue damage critical point, and generate a three-dimensional visualized dynamic map of wind field evolution, displaying the movement trajectory of the strong wind mass and its predicted impact on bridge components on the digital twin platform.

8. A bridge structural health monitoring system based on the Industrial Internet according to claim 7, characterized in that: The system adopts an industrial internet architecture, which includes edge computing nodes and cloud high-performance computing clusters. The edge computing node preprocesses and compresses the raw data collected by the bridge structure sensor network and the micro-meteorological sensing unit before uploading it to the cloud-based high-performance computing cluster. The industrial internet architecture adopts a dual-route redundancy design. The core data switch is connected to the edge computing server through optical fiber and has an independent wireless microwave communication link as backup. It is configured to automatically switch to the microwave link within milliseconds when the optical fiber breaks. The edge computing node is equipped with a data cache and a local decision-making module, which is used to perform basic health assessment and emergency warning operations based on a locally stored simplified model in the event of a loss of cloud connection. The edge computing node is equipped with a security chip based on a hardware encryption engine to ensure the confidentiality and integrity of monitoring data during transmission.

9. A bridge structural health monitoring system based on the Industrial Internet according to claim 8, characterized in that: The system also integrates an autonomous inspection interface for unmanned aerial vehicles, a traffic load precision sensing subsystem, and an earthquake early warning triggering module. When the structural health assessment center identifies a suspected damaged area, it automatically sends coordinate instructions to the drone inspection system to guide the drone carrying a high-resolution industrial camera and an infrared thermal imager to perform close-range visual inspection of the target area. The traffic load precision sensing subsystem utilizes weighing sensors deployed under the bridge deck pavement and video recognition technology to obtain the axle load and position of vehicles on the bridge, and inputs them as environmental excitation terms into the physical information neural network proxy model to decouple the contribution of traffic load to the structural response. The earthquake early warning triggering module is interconnected with the seismic network. When a seismic wave arrival signal is detected, it rapidly increases the sampling frequency of the bridge structure sensor network to the kilohertz level, causing the physical information neural network proxy model to switch to the dynamic-dominated mode to assess the dynamic safety margin of the structure in real time during strong earthquakes.

10. A bridge structural health monitoring system based on the Industrial Internet according to claim 9, characterized in that: The system is also equipped with a virtual sensor interpolation module and a visual management terminal; The virtual sensor interpolation module uses the physical information neural network proxy model to generate a virtual response sequence in areas where no physical hardware sensors are deployed. By comparing and verifying the virtual response sequence with the measured response sequence, the module can calculate and digitally cover the stress level at any location within the entire bridge range. The visualization management terminal is based on a three-dimensional geographic information system to build a digital twin display interface, configured to display in real time the comparison curve between the predicted physical quantities of bridge components and the measured values ​​of sensors, and supports automatic hierarchical push of alarm levels. The system also introduces a digital twin growth mechanism, configured to record the bridge's actual evolution trajectory. By analyzing the trend of sensor reference values, the system dynamically corrects the reference stiffness and boundary constraints in the digital twin model of the digital twin, ensuring that the digital twin maintains consistent characteristics with the physical entity throughout the bridge's entire life cycle.

Images