Large-span asymmetric steel-UHPC concrete superposed beam cable-stayed bridge sunlight temperature effect real-time monitoring and early warning system

Abstract

The invention discloses a large-span asymmetric steel-UHPC concrete superposed beam cable-stayed bridge sunlight temperature effect real-time monitoring and early warning system, which relates to the technical field of bridges and comprises a multi-scale sensing unit, a simulation prediction unit, an intelligent early warning unit and an autonomous iteration unit. Cooperative monitoring of a microcosmic interface and a macrostructure is achieved through the multi-scale sensing unit, the problems that capture of steel-UHPC interface thermal stress is insufficient and an asymmetric cantilever end blind area exists are solved, the simulation prediction unit fuses a knowledge graph and dynamic modeling, the temperature field simulation precision and timeliness are improved, and the temperature field simulation precision and timeliness are improved. According to the method, the interface stripping risk is accurately predicted in combination with a fatigue accumulated damage model, the intelligent early warning unit adopts dynamic threshold adjustment and prediction early warning, the false alarm and missing alarm rate is reduced, the comprehensiveness, accuracy and response timeliness of monitoring of the sunlight temperature effect of the large-span asymmetric steel-UHPC superposed beam cable-stayed bridge are greatly improved, and safe operation of the bridge is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bridges, in particular to a real-time monitoring and early warning system for solar temperature effect of a large-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge. BACKGROUND

[0002] The large-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge refers to a bridge with a cable-stayed bridge structure, a main span reaching the large-span standard, an overall structure (span, stress, etc.) showing asymmetric characteristics, and a beam body composed of a steel material and an ultra-high performance concrete (UHPC) composite structure, and specific responses of the bridge under the action of solar temperature.

[0003] For example, a bridge structure temperature field monitoring method disclosed in CN111723509A includes the following steps: S1: calculating convective heat transfer and radiative heat transfer using measured air temperature and real-time wind speed; S2: calculating solar radiation using a day number and solar radiation on a horizontal plane; S3: calculating the thermal boundary conditions of the bridge according to the convective heat transfer, the radiative heat transfer, and the solar radiation; S4: determining the heat exchange model of the bridge according to the thermal boundary conditions of the bridge; and S5: solving the heat exchange model of the bridge using finite element numerical simulation to obtain the temperature field of the bridge, thereby completing the monitoring of the bridge structure temperature field.

[0004] In the prior art, due to the large difference in thermal physical properties between steel and UHPC materials, the characteristics of superimposed asymmetric structures, and the complex temperature gradient formed under the action of solar radiation, insufficient attention is paid to the temperature gradient and interlayer interaction at the interface of the steel-UHPC composite layer during solar temperature monitoring, and the traditional monitoring method only focuses on the surface temperature, which increases the calculation error of the interfacial shear stress and cannot accurately assess the impact of temperature effect on the safety of the structure, resulting in insufficient precision of the current temperature effect analysis and early warning strategy, which may endanger the safety of the bridge due to missed reports, and cannot meet the needs of bridge life cycle maintenance and management. SUMMARY

[0005] The present application aims to provide a real-time monitoring and early warning system for solar temperature effect of a large-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge to solve the problem of insufficient attention to the temperature gradient and interlayer interaction at the interface of the steel-UHPC composite layer during solar temperature monitoring, and the traditional monitoring method only focuses on the surface temperature, which increases the calculation error of the interfacial shear stress and exists the problem of missed reports.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solution: a real-time monitoring and early warning system for solar temperature effect of a large-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge, including a multi-scale perception unit, a simulation prediction unit, an intelligent early warning unit, and an autonomous iteration unit.

[0007] The multiscale perception unit is used for constructing a multidimensional perception network and acquiring multi-source data; the simulation and prediction unit mines deep correlations between temperature effects and structural responses by integrating the multi-source heterogeneous data acquired by the multiscale perception unit, and performs simulation and prediction of a temperature field and structural responses based on coupling of a physical model and real-time data;

[0008] The simulation and prediction unit comprises a data preprocessing module, a feature fusion module, a dynamic modeling module and a response analysis module; the data preprocessing module repairs missing sensor data based on a generative adversarial network, introduces a time attention layer in an LSTM, allocates weights of temperature data at different times, and identifies dynamic time lag relationships between temperature and deformation; the feature fusion module acquires monitoring data of the multiscale perception unit, abstracts the monitoring data into environmental entities, temperature entities and structural response entities, mines relationships between the defined entities, constructs a knowledge graph comprising triplets, converts the knowledge graph into a weighted graph, calculates feature correlation weights by using a graph sampling aggregation algorithm, identifies key influence factors, and updates node weights; the dynamic modeling module constructs a dynamic temperature field model coupled with a physical model and real-time data; and the response analysis module converts node temperature values of the temperature field output by the dynamic modeling module into equivalent thermal loads, introduces a fatigue cumulative damage model to calculate interfacial shear stress, extracts a typical temperature cycle spectrum based on historical monitoring data, and predicts interfacial peeling risks after long-term temperature cycles.

[0009] The intelligent early warning unit performs early warning and decision-making for dynamic adjustment of working conditions based on the prediction results of the simulation and prediction unit; the autonomous iteration unit develops a user feedback closed-loop mechanism through continuous learning and case accumulation, reversely corrects LSTM prediction model parameters, and performs autonomous iteration of system analysis capability.

[0010] Preferably, the multiscale perception unit comprises a micro monitoring module, a macro monitoring module and an environmental perception module.

[0011] The micro monitoring module implants a micro fiber grating array in the steel-UHPC composite layer, performs high-frequency sampling through a distributed demodulator, and collects interface fine data by arranging self-calibrating vibrating wire strain gauges at the joint between the steel box girder and the UHPC layer; the macro monitoring module collects bridge macro data by combining a BOTDA distributed optical fiber and an unmanned aerial vehicle infrared thermal imager; and the environmental perception module outputs a solar elevation angle, an azimuth angle and radiation intensity components by measuring a six-component solar radiation tracker, and acquires environmental information data in combination with an ultrasonic anemometer.

[0012] Preferably, in the dynamic modeling module, constructing the dynamic temperature field model comprises the following steps:

[0013] S1, physical model foundation construction: for the layered structure of the steel-UHPC composite beam, i.e. the steel beam, the UHPC layer and the interface layer, a three-dimensional unsteady heat conduction equation is established, the calculation domain is divided according to the structure level, which is divided into a steel beam layer, a UHPC layer and a steel-UHPC interface layer, the historical temperature field data of the monitoring system are used as the initial value, and the boundary conditions are set;

[0014] S2, coupling of physical model and real-time data: based on the real-time temperature data, the target parameters are established, the Bayesian inversion algorithm is used for iterative solution, the parameters with high uncertainty in the physical model are inverted, the parameters are updated in real time, the dynamic adjustment is performed, and the macro and local monitoring data are fused;

[0015] S3, dynamic prediction of temperature field model: the finite element method is used to discretely solve the coupled heat conduction equation, and continuous iteration is performed with real-time data input, and the dynamic prediction of the temperature field is performed. Every time a batch of new real-time data is obtained, the model is updated once.

[0016] Preferably, in step S1, the formula of the three-dimensional unsteady heat conduction equation is as follows:

[0017]

[0018] Wherein, T(x, y, z, t) represents the temperature of the spatial coordinates (x, y, z) at time t, p represents the density of the material, c represents the specific heat capacity of the material, k represents the thermal conductivity of the material, and Q(x, y, z, t) represents the internal heat source term.

[0019] Preferably, in step S1, the thickness direction of the steel beam layer is refined when the grid is divided, the thermal physical parameters of the steel material are used, the equivalent thermal conductivity and the specific heat capacity of the UHPC layer are determined considering the characteristics of the multiphase composite material, and the interface thermal resistance coefficient reflecting the contact heat conduction resistance of the two materials is introduced in the steel-UHPC interface layer.

[0020] Preferably, in step S1, the boundary conditions include the solar radiation boundary, the convection heat transfer boundary and the long-wave radiation boundary, the solar radiation model is used to calculate the direct and scattered radiation of the sun, and the heat flux density is converted into heat flux density combined with the structure surface absorption rate, the convection heat transfer boundary is based on real-time wind speed and air temperature, and the long-wave radiation boundary considers the long-wave radiation exchange between the structure surface and the atmosphere and the sky.

[0021] Preferably, in step S2, the real-time meteorological data and the temperature monitoring data are used to dynamically adjust the solar radiation correction and the convection heat transfer coefficient correction, the solar radiation correction is used to correct the calculation results of the dynamic temperature field model with the real-time data of the solar radiation sensor, the prediction deviation of the radiation amount caused by cloud cover is avoided, and the convection heat transfer coefficient correction is based on the real-time wind speed and the structure surface temperature monitoring value. The calculation formula of the convection heat transfer coefficient is updated in real time by the least square method.

[0022] Preferably, in the response analysis module, the fatigue accumulation damage model is the Miner rule, and a specific formula is as follows:

[0023]

[0024] Wherein, D represents a fatigue damage degree, when D is 1, theoretically, it is considered that the material or structure is damaged by fatigue, m j represents actual cycle times of the material or structure under the jth stress level, M j represents a fatigue life of the material or structure under the jth stress level, that is, cycle times when fatigue damage is reached, and is obtained through material fatigue test.

[0025] Preferably, the intelligent early warning unit comprises a dynamic adjustment module, a prediction and early warning module and a decision support module.

[0026] The dynamic adjustment module is based on reinforcement learning training, takes the current temperature field distribution, stress state, environmental parameter and time information of the bridge as state variables, forms a state space, selects a suitable early warning threshold according to the real-time monitored state information, and dynamically updates the early warning threshold; the prediction and early warning module calculates attention weights on different time steps by constructing a time attention network, predicts future hour stress states, and timely triggers early warning when the prediction value exceeds a safety threshold; and the decision support module constructs a plan knowledge graph, integrates historical cases and real-time data, and generates disposal suggestions.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] The present application realizes collaborative monitoring of micro interfaces and macro structures through the multi-scale perception unit, solves the problems of insufficient capture of thermal stress of the steel-UHPC interface and blind area of the asymmetric cantilever end in traditional monitoring, the simulation prediction unit fuses a knowledge graph and dynamic modeling, improves the temperature field simulation precision and timeliness, combines the fatigue accumulation damage model to accurately predict the interface peeling risk, the intelligent early warning unit adopts dynamic threshold adjustment and prediction and early warning, reduces the false alarm and missed alarm rate, the autonomous iteration unit realizes continuous optimization of the model, and the overall forms a closed loop, greatly improves the comprehensiveness, accuracy and response timeliness of the sunshine temperature effect monitoring of the long-span asymmetric steel-UHPC composite beam cable-stayed bridge, and guarantees the safe operation of the bridge. BRIEF DESCRIPTION OF DRAWINGS

[0029] Fig. 1 It is a flow chart of the long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge sunshine temperature effect real-time monitoring and early warning system of the present application.

[0030] Fig. 2 It is a system block diagram of the long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge sunshine temperature effect real-time monitoring and early warning system of the present application.

[0031] In the figure: 1, multi-scale perception unit; 11, micro monitoring module; 12, macro monitoring module; 13, environment perception module; 2, simulation prediction unit; 21, data preprocessing module; 22, feature fusion module; 23, dynamic modeling module; 24, response analysis module; 3, intelligent early warning unit; 31, dynamic adjustment module; 32, prediction and early warning module; 33, decision support module; 4, autonomous iteration unit. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0033] Embodiment one: refer to Figs. 1-2 The real-time monitoring and early warning system for the solar temperature effect of the long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge is shown in the figure. Through the collaborative work of the four core units of "perception-simulation-early warning-iteration", the whole process monitoring and early warning of the solar temperature effect of the bridge is realized. The specific composition and functions are as follows:

[0034] I. Technical solutions

[0035] 1. The multi-scale perception unit 1 is responsible for building an integrated perception network of "micro-macro-environment", obtaining accurate and comprehensive raw data, and solving the problem of insufficient coverage of traditional monitoring on interface details and blind areas of asymmetric structures, including the micro monitoring module 11, the macro monitoring module 12 and the environment perception module 13.

[0036] 2. The simulation prediction unit 2 uses a generative adversarial network to repair missing sensor data; introduces a time attention layer in the LSTM (Long Short Term Memory Network) to dynamically allocate the weights of temperature data at different times and automatically identify the time lag relationship between temperature and structural deformation.

[0037] Among them, the feature fusion module 21 abstracts the monitoring data into three types of entities, constructs a knowledge graph containing entity relationship triples, converts the knowledge graph into a weighted graph, calculates the feature correlation weight through a graph sampling aggregation algorithm, identifies the key influence factor, and provides priority basis for modeling; the dynamic modeling module 22 is the core of building a dynamic temperature field model coupled with "physical model + real-time data"; the response analysis module 23 converts the temperature value output by the dynamic temperature field model into an equivalent thermal load, introduces a fatigue cumulative damage model (Miner's rule) to calculate the interface shear stress, extracts a typical temperature cycle spectrum from historical data, and predicts the interface peeling risk after long-term cycling.

[0038] 3. Intelligent early warning unit 3 realizes accurate early warning and efficient decision-making based on simulation prediction results, and solves the false alarm / misreporting problem of traditional static threshold early warning.

[0039] Among them, the dynamic adjustment module 31 adjusts the early warning threshold based on reinforcement learning, taking the bridge temperature field, stress state, environmental parameters, etc. as state variables, dynamically adjusts the early warning threshold (such as increasing the stress threshold sensitivity according to high-temperature weather), and improves the early warning adaptability; the prediction and early warning module 32 constructs a time attention network, weights the historical temperature and stress data (emphasizes the influence of key time steps), and predicts the stress state in the next 3 hours; when the predicted value exceeds the safety threshold, the early warning is triggered, and the response time is obtained for maintenance; the decision support module 33 constructs a plan knowledge graph (integrates historical cases and real-time data), generates disposal suggestions, and superimposes operation instructions (such as operation steps of local cooling measures) in the field through AR technology, improving disposal efficiency.

[0040] 4. Autonomous iteration unit 4 realizes self-evolution through a closed-loop feedback mechanism. After the engineer labels the early warning effectiveness, the system reversely corrects the LSTM model parameters, iterates once a month, and continuously improves the data processing, prediction and early warning accuracy.

[0041] II. Core principle

[0042] 1. Multi-scale perception principle, aiming at the characteristics of large difference in thermal physical properties of steel and UHPC materials and complex temperature gradient of asymmetric structure, through the cooperation of "microscopic interface monitoring + macroscopic whole bridge monitoring + environmental parameter acquisition", realize the data full coverage from local to whole, from structure to environment, lay the foundation for accurate modeling.

[0043] 2. Physical-data fusion modeling principle, based on three-dimensional unsteady heat conduction equation (physical model), combined with real-time monitoring data (data driven) through Bayesian inversion dynamic correction parameters, solve the problem that traditional pure physical model is difficult to adapt to environmental dynamic change, pure data model lacks theoretical support, realize the "high precision + high timeliness" of temperature field simulation.

[0044] 3. Dynamic early warning and iteration principle, based on reinforcement learning dynamic threshold adjustment, time attention network forward-looking prediction, improve the accuracy of early warning; through user feedback closed-loop correction model parameters, realize the autonomous evolution of system ability, meet the bridge whole life cycle maintenance demand

[0045] The application realizes multi-dimensional perception, physical-data fusion simulation, dynamic early warning and autonomous iteration, accurately captures the temperature effect of steel-UHPC interface and structural response, and provides full-process technical support for the safe operation of long-span asymmetric composite beam cable-stayed bridge.

[0046] Example two: reference Figs. 1-2The real-time monitoring and early warning system for the temperature effect of the long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge in sunlight includes a multi-scale sensing unit 1, a simulation prediction unit 2, an intelligent early warning unit 3, and an autonomous iteration unit 4.

[0047] The multi-scale sensing unit 1 is used to build a multi-dimensional sensing network, obtain multi-source data, and realize fine temperature field and response monitoring of the steel-UHPC interface and the whole bridge structure.

[0048] The multi-scale sensing unit 1 includes a microscopic monitoring module 11, a macroscopic monitoring module 12, and an environmental sensing module 13. The microscopic monitoring module 11 implants a micro FBG array at the interface of the steel-UHPC composite layer, synchronously collects the interface temperature and shear strain, realizes high-frequency sampling through a distributed demodulator, captures the instantaneous thermal stress fluctuation between the layers, and arranges self-calibrating vibrating wire strain gauges at the joint of the steel box girder and the UHPC layer to control the long-term drift and solve the failure problem caused by temperature fatigue. The macroscopic monitoring module 12 combines BOTDA distributed optical fiber and an unmanned aerial vehicle infrared thermal imager. The BOTDA distributed optical fiber covers the longitudinal temperature gradient of the main girder, and the unmanned aerial vehicle infrared thermal imager scans the blind area at the asymmetric cantilever end twice a week. The environmental sensing module 13 measures the solar elevation angle, azimuth angle, and radiation intensity components through a six-component solar radiation tracker, and combines an ultrasonic anemometer to provide fine boundary conditions for temperature field modeling.

[0049] The simulation prediction unit 2 is used to integrate multi-source heterogeneous data, mine the deep correlation between temperature effect and structural response, and realize fine simulation and prediction of the temperature field and structural response based on the coupling of physical models and real-time data.

[0050] The simulation prediction unit 2 includes a data preprocessing module 21, a feature fusion module 22, a dynamic modeling module 23, and a response analysis module 24. The data preprocessing module 21 uses a generative adversarial network to repair missing sensor data, introduces a time attention layer in the LSTM, realizes weight distribution of temperature data at different times, and automatically identifies the dynamic time lag relationship between temperature and deformation.

[0051] The feature fusion module 22 abstracts the monitoring data into environmental entities, temperature entities, and structural response entities by constructing a knowledge graph. The environmental entities include solar radiation intensity, wind speed, and solar azimuth angle. The temperature entities include cross-section maximum temperature, gradient temperature difference, and interface temperature. The structural response entities include main girder strain, cable force change, and beam end displacement. The module defines the relationship between entities through expert experience and data mining, forms triples, constructs a knowledge graph containing triples, converts the knowledge graph into a weighted graph, calculates the feature correlation weight using a graph sampling aggregation algorithm, identifies key influence factors, updates node weights, generates a "feature-contribution degree" ranking table, and provides priority basis for subsequent modeling.

[0052] The dynamic modeling module 23 is used to build a dynamic temperature field model coupled with a physical model and real-time data, so as to realize high-precision and high-time-effect simulation of the full-bridge temperature distribution.

[0053] In the dynamic modeling module 23, building the dynamic temperature field model includes the following steps:

[0054] S1, physical model foundation construction: for the layered structure of the steel-UHPC composite beam, i.e., the steel beam, the UHPC layer and the interface layer, a three-dimensional unsteady heat conduction equation is established, the calculation domain is divided according to the structure level according to the asymmetry and material difference of the steel-UHPC composite beam, and is divided into the steel beam layer, the UHPC layer and the steel-UHPC interface layer, the historical temperature field data of the monitoring system is used as the initial value, and the boundary condition is set;

[0055] In step S1, the formula of the three-dimensional unsteady heat conduction equation is as follows:

[0056]

[0057] Wherein, T(x, y, z, t) represents the temperature of the spatial coordinates (x, y, z) at time t, p represents the density of the material, c represents the specific heat capacity of the material, k represents the thermal conductivity of the material, and Q(x, y, z, t) represents the internal heat source term.

[0058] In step S1, the thickness direction of the steel beam layer is refined when the grid is divided, the thermal physical parameters of the steel material are adopted, the equivalent thermal conductivity and the specific heat capacity of the UHPC layer are determined through the test or the mixing rule considering the characteristics of the multiphase composite material, and the interface thermal resistance coefficient reflecting the contact heat conduction impedance of the two materials is introduced in the steel-UHPC interface layer, so as to avoid the error caused by directly assuming perfect contact;

[0059] The boundary conditions include the solar radiation boundary, the convection heat transfer boundary and the long-wave radiation boundary, the solar radiation model is used to calculate the direct and scattered radiation of the sun, the structure surface absorption rate (difference between steel and UHPC coating) is combined to convert into a heat flux density, the convection heat transfer boundary is calculated through the Newton cooling formula based on the real-time wind speed and air temperature, and the long-wave radiation boundary considers the long-wave radiation exchange between the structure surface and the atmosphere and the sky, and is calculated by using the Stefan-Boltzmann law.

[0060] S2, coupling of physical model and real-time data: based on real-time monitoring temperature data, the error sum of squares of the predicted temperature and the monitoring temperature is minimized as the target of the physical model, the target parameters are established, the Bayesian inversion algorithm is used for iterative solution, the parameters with high uncertainty in the physical model are inverted, the parameters are updated in real time to adapt to the changes of thermal parameters caused by material degradation, the real-time meteorological data and temperature monitoring data are used to dynamically adjust the time-varying boundary conditions, and the macro and local monitoring data are fused to adapt to the spatial heterogeneity of the temperature field of the asymmetric structure;

[0061] In step S2, the dynamic adjustment of real-time meteorological data and temperature monitoring data includes solar radiation correction and convective heat transfer coefficient correction. The solar radiation correction uses real-time data of the solar radiation sensor to correct the calculation results of the dynamic temperature field model, avoiding the prediction deviation of radiation caused by cloud cover. The convective heat transfer coefficient correction is based on real-time wind speed and structure surface temperature monitoring value, and the calculation formula of the convective heat transfer coefficient is updated in real time by the least square method.

[0062] S3, dynamic prediction of temperature field model: the coupled heat conduction equation is discretely solved by using the finite element method, the grid is encrypted at the steel-UHPC interface and the main beam section mutation, the calculation accuracy is improved, the implicit time difference format is adopted, the stability and calculation efficiency are considered, the long-period temperature field simulation is adapted, and the dynamic prediction of the temperature field is realized. The model is updated once for every batch of new real-time data, forming a closed loop of monitoring, correction and prediction.

[0063] The response analysis module 24 converts the temperature field node temperature value output by the dynamic modeling module 23 into an equivalent thermal load, introduces a fatigue cumulative damage model for interface shear stress calculation, extracts a typical temperature cycle spectrum based on historical monitoring data, and predicts the interface peeling risk after long-term temperature cycles.

[0064] In the response analysis module 24, the fatigue cumulative damage model is the Miner rule, and its specific formula is as follows:

[0065]

[0066] Where D represents the fatigue damage degree, when D is 1, theoretically, the material or structure is considered to have fatigue failure, m j represents the actual cycle number of the material or structure at the jth stress level, M j represents the fatigue life of the material or structure at the jth stress level, i.e. the cycle number when fatigue failure is reached, which is obtained through material fatigue test.

[0067] The intelligent early warning unit 3 breaks through the limitation of static threshold, realizes early warning and intelligent decision based on working condition dynamic adjustment.

[0068] The intelligent early warning unit 3 comprises a dynamic adjustment module 31, a prediction and early warning module 32 and a decision support module 33; the dynamic adjustment module 31 is based on reinforcement learning training, and takes the current temperature field distribution, stress state, environmental parameter and time information of the bridge as state variables to form a state space, and the action space is a set of adjustable early warning threshold values, and different values of the temperature difference threshold value and the stress threshold value can be set, according to the state information monitored in real time, the trained strategy network is selected to select a suitable early warning threshold value, and the dynamic update of the early warning threshold value is realized; the prediction and early warning module 32 constructs a time attention network, inputs the preprocessed time sequence data such as temperature and stress into the network according to a time step of 10 minutes, calculates the attention weight on different time steps, highlights the time step which has a greater influence on the future stress state, predicts the stress state in the next 3 hours, and timely triggers the early warning when the prediction value exceeds the safety threshold value, so as to provide a response time for the bridge management and maintenance personnel in advance, so that measures can be taken to ensure the safety of the bridge; the decision support module 33 constructs a plan knowledge graph, integrates historical cases and real-time data, generates a disposal suggestion, and superimposes operation instructions in the field through AR labeling technology, thereby improving the disposal efficiency and accuracy.

[0069] The autonomous iteration unit 4 develops a user feedback closed loop mechanism through continuous learning and case accumulation, engineers label the early warning effectiveness through the platform, and reversely correct the LSTM prediction model parameters, so that the system analysis capability is autonomously iterated once a month.

[0070] Firstly, the micro monitoring module 11 of the multi-scale perception unit 1 collects the temperature and shear strain of the steel-UHPC interface, the macro monitoring module 12 obtains the temperature gradient of the main beam and the data of the asymmetric cantilever end, and the environmental perception module 13 collects environmental parameters such as solar radiation and wind speed, and the three cooperate to provide multi-source original data, which enters the simulation prediction unit 2, the data preprocessing module 21 repairs missing data and identifies the temperature and deformation time lag relationship, the feature fusion module 22 constructs a knowledge graph, calculates the feature correlation weight to generate a ranking table, and the dynamic modeling module 23 combines the layered structure to establish a heat conduction equation, iteratively optimizes the model with real-time data, realizes dynamic simulation of the temperature field, and the response analysis module 24 converts the temperature field into a thermal load and predicts the interface peeling risk by the Miner rule, the simulation results drive the intelligent early warning unit 3, the dynamic adjustment module 31 dynamically updates the early warning threshold value based on reinforcement learning, the prediction and early warning module 32 predicts the stress in the next 3 hours by using a time attention network, and the decision support module 33 generates a disposal suggestion in combination with the knowledge graph, and superimposes operation instructions through AR technology, thereby forming an early warning decision closed loop, finally, the autonomous iteration unit 4 corrects the model parameters according to the labeling of the engineers on the early warning effectiveness, reversely optimizes the simulation prediction and intelligent early warning unit, realizes the continuous iteration of the system, and the units form a complete closed loop through data circulation and feedback.

[0071] Although the present application has been described in detail with reference to the foregoing embodiments, the technical solutions recorded in the foregoing embodiments can be modified, or some of the technical features can be replaced by equivalent features, by those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A real-time monitoring and early warning system for the temperature effects of sunlight on long-span asymmetric steel-UHPC concrete composite girder cable-stayed bridges, characterized in that: The system comprises a multi-scale perception unit (1), a simulation prediction unit (2), an intelligent early warning unit (3) and an autonomous iteration unit (4); The multi-scale perception unit (1) is used for constructing a multi-dimensional perception network and acquiring multi-source data; the simulation prediction unit (2) is used for integrating the multi-source heterogeneous data acquired by the multi-scale perception unit (1), mining the deep correlation between the temperature effect and the structural response, coupling the physical model and the real-time data, simulating and predicting the temperature field and the structural response; The simulation prediction unit (2) comprises a data preprocessing module (21), a feature fusion module (22), a dynamic modeling module (23) and a response analysis module (24); the data preprocessing module (21) is used for repairing the missing data of sensors based on a generative adversarial network, introducing a time attention layer in an LSTM, assigning the weight of temperature data at different times, identifying the dynamic time lag relationship between temperature and deformation, and providing clean data for the feature fusion module (22); The feature fusion module (22) is used for abstracting the clean data into environmental entities, temperature entities and structural response entities, mining the relationship between the defined entities, constructing a knowledge graph comprising triplets, converting the knowledge graph into a weighted graph, calculating the feature correlation weight by using a graph sampling aggregation algorithm, identifying key influence factors, updating the node weight, and guiding the parameter optimization of the dynamic modeling module (23); the dynamic modeling module (23) is used for constructing a dynamic temperature field model coupled with the physical model and the real-time data; the response analysis module (24) is used for converting the temperature value of the temperature field node output by the dynamic modeling module (23) into an equivalent thermal load, introducing a fatigue cumulative damage model to calculate the interfacial shear stress, extracting a typical temperature cycle spectrum based on the historical monitoring data, and predicting the interfacial peeling risk after long-term temperature cycles; The intelligent early warning unit (3) is used for early warning and decision-making of dynamic adjustment of working conditions based on the prediction result of the simulation prediction unit (2); the autonomous iteration unit (4) is used for developing a user feedback closed-loop mechanism through continuous learning and case accumulation, reversely correcting the LSTM prediction model parameters, and performing autonomous iteration of the system analysis capability. 2.The real-time monitoring and early warning system for the sunshine temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 1, characterized in that: The multi-scale perception unit (1) comprises a micro monitoring module (11), a macro monitoring module (12) and an environmental perception module (13); The micro monitoring module (11) implants a micro optical fiber grating array in the steel-UHPC composite layer, performs high-frequency sampling through a distributed demodulator, and arranges a self-calibrating vibrating wire strain gauge at the joint of the steel box girder and the UHPC layer to collect fine interface data; the macro monitoring module (12) adopts a combination of a BOTDA distributed optical fiber and an unmanned aerial vehicle infrared thermal imager to collect macro bridge data; the environmental perception module (13) measures the solar elevation angle, azimuth angle and radiation intensity component through a six-component solar radiation tracker, and acquires environmental information data in combination with an ultrasonic anemometer. 3.The real-time monitoring and early warning system for the sunshine temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 1, characterized in that: In the dynamic modeling module (23), the construction of the dynamic temperature field model comprises the following steps: S1, physical model foundation construction: for the layered structure of steel-UHPC composite beam, i.e. steel beam, UHPC layer and interface layer, a three-dimensional unsteady heat conduction equation is established, the calculation domain is divided into steel beam layer, UHPC layer and steel-UHPC interface layer according to the structure level, the historical temperature field data of the monitoring system are taken as the initial value, and the boundary conditions are set; S2, coupling of physical model and real-time data: based on the real-time monitored temperature data, the target parameters are established, the Bayesian inversion algorithm is used for iterative solution, the parameters with high uncertainty in the physical model are inverted, the parameters are updated in real time, the dynamic adjustment is carried out, and the macro and local monitoring data are fused; S3, dynamic prediction of temperature field model: the finite element method is used to discretely solve the coupled heat conduction equation, and continuously iterates with real-time data input to dynamically predict the temperature field. Every time a batch of new real-time data is obtained, the model is updated once. 4.The real-time monitoring and early warning system for the sunshine temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 3, characterized in that: In step S1, the formula of the three-dimensional unsteady heat conduction equation is as follows: Wherein, T(x, y, z, t) represents the temperature of spatial coordinates (x, y, z) at time t, p represents the density of the material, c represents the specific heat capacity of the material, k represents the thermal conductivity of the material, and Q(x, y, z, t) represents the internal heat source term. 5.The real-time monitoring and early warning system for the sunshine temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 3, characterized in that: In step S1, the thickness direction is refined when the steel beam layer is divided into grids, the thermal physical parameters of steel are used, the equivalent thermal conductivity and specific heat capacity of the UHPC layer are determined considering its multi-phase composite material characteristics, and the interface thermal resistance coefficient reflecting the contact heat conduction resistance of the two materials is introduced in the steel-UHPC interface layer. 6.The real-time monitoring and early warning system for the sunshine temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 3, characterized in that: In step S1, the boundary conditions include the solar radiation boundary, the convection heat transfer boundary and the long wave radiation boundary. The solar radiation boundary uses the solar radiation model to calculate the direct and scattered radiation of the sun, and converts it into heat flux density combined with the surface absorption rate of the structure. The convection heat transfer boundary is based on real-time wind speed and air temperature, and the long wave radiation boundary considers the long wave radiation exchange between the structure surface and the atmosphere and the sky. 7.The real-time monitoring and early warning system for the sunshine temperature effect of the long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 3, characterized in that: In step S2, the real-time meteorological data and temperature monitoring data are used to dynamically adjust the solar radiation correction and the convection heat transfer coefficient correction. The solar radiation correction uses the real-time data of the solar radiation sensor to correct the calculation results of the dynamic temperature field model, avoiding the prediction deviation of the radiation amount caused by cloud cover. The convection heat transfer coefficient correction is based on real-time wind speed and structure surface temperature monitoring value, and the calculation formula of the convection heat transfer coefficient is updated in real time by least square method. 8.The real-time monitoring and early warning system for the solar temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 1, characterized in that: In the response analysis module (24), the fatigue accumulation damage model is the Miner rule, and its specific formula is as follows: where D represents the fatigue damage degree, and when D is 1, it is theoretically considered that the material or structure is damaged by fatigue, m j Nj represents the actual cycle number of the material or structure under the jth stress level, M j Nj represents the fatigue life of the material or structure under the jth stress level, i.e., the cycle number when fatigue failure is reached, which is obtained through material fatigue tests. 9.The real-time monitoring and early warning system for the solar temperature effect of a long-span asymmetric steel-UHPC concrete composite beam cable-stayed bridge according to claim 1, characterized in that: The intelligent early warning unit (3) includes a dynamic adjustment module (31), a prediction and early warning module (32), and a decision support module (33); The dynamic adjustment module (31) is based on reinforcement learning training, the current temperature field distribution of the bridge, the stress state, the environmental parameters and the time information are taken as the state variables to form a state space, and the appropriate early warning threshold is selected according to the real-time monitored state information, the early warning threshold is dynamically updated, and the dynamic standard is provided for the prediction and early warning module (32); the prediction and early warning module (32) predicts the stress state in the future 3 hours by constructing a time attention network and calculating the attention weight at different time steps, and timely triggers the early warning when the predicted value exceeds the safety threshold; The decision support module (33) constructs a plan knowledge graph, integrates historical cases and real-time data, and generates a treatment suggestion.

Images