A wind-induced vibration response early warning system and method for a long-span bridge
Through the two-stage GRU hybrid prediction model combined with GRU neural network and sensor technology, the problems of low prediction efficiency of wind vibration response and insufficient model reliability of large-span bridges in the existing technology are solved, and the rapid prediction of wind vibration response under typhoon effects and achievement of rapid prediction and Early warning ensures bridge structure and driving safety.
Patent Information
- Application Number
- CN202111038664.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-06
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-09-06
AI Technical Summary
The existing technology cannot effectively predict the wind vibration response of large-span bridges under the action of typhoons. The finite element method has low computational efficiency and insufficient model reliability. The pure data-driven machine learning algorithm lacks physical significance, and the generalization performance is poor, so it is impossible to achieve large-span. Quick prediction and early warning of bridge wind vibration response.
The sensor technology is used to collect wind environment and structural response data, establish a two-stage GRU hybrid prediction model, combine it with the GRU neural network to predict wind speed and wind vibration response, and improve the generalization ability and calculation efficiency of the model through data analysis and the weighted loss function of the motion equation to achieve the Quick prediction and early warning of wind vibration response of large span bridges.
It has achieved advanced prediction and timely warning of wind vibration response of large-span bridges, improved the safety of bridge structure and bridge decks, significantly improved the computing efficiency and model reliability, and can ensure the safety of bridge operations in extreme climates.
Smart Images

Figure CN114169036B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a wind-induced vibration response early warning system and method for a long-span bridge, and belongs to the field of wind resistance of long-span bridges in civil engineering. Background Art
[0002] As global climate change intensifies, natural disasters such as typhoons occur frequently. my country is located on the northwest coast of the Pacific Ocean and has long been attacked by Pacific typhoons. Cases of high-rise and large-span structures being damaged by wind vibration occur frequently. my country has built many large bridge projects across rivers and seas, such as the Runyang Yangtze River Bridge, the Hutong Yangtze River Bridge, and the Sutong Bridge. For such large-span structures, due to their greater flexibility and sensitivity to wind loads, they will produce violent dynamic responses under the action of extreme wind environments such as typhoons, and are prone to damage or even collapse. Therefore, the problem of wind-induced vibration has become one of the key factors threatening the safety of large-span bridge structures and the safety of bridge deck driving.
[0003] The bridge health monitoring system evaluates and diagnoses the service status of the bridge in real time by monitoring the environmental parameters of the bridge structure, operating loads, bridge response and other data over a long period of time. However, the system mainly provides data support for subsequent analysis work. When faced with emergencies such as typhoons, it is unable to make advance predictions, and the safety of bridge operation cannot be guaranteed. At present, the finite element method is often used to simulate the wind-induced vibration response of large-span bridges. This method requires the establishment of a bridge finite element model based on the design plan. The process is cumbersome and complicated, and the reliability of the model cannot be guaranteed. In addition, this method involves solving a large number of equations, and the calculation efficiency is very low, which basically cannot be used for rapid prediction and analysis of the wind-induced vibration response of bridges.
[0004] Nowadays, the rapid development of machine learning theory has led to its widespread use in various fields. In the field of structural wind engineering, it provides new ideas for complex and tedious numerical calculations, making it possible to predict the wind-induced vibration response of large-span bridges under typhoons in advance. However, purely data-driven machine learning algorithms have problems such as dependence on data and lack of physical meaning. Their generalization performance cannot be guaranteed, and changes in working conditions will make the trained model no longer reliable. How to effectively improve the accuracy of the neural network model so that it can replace the finite element method to simulate and predict the wind-induced vibration response of structures has been a key problem that the engineering community has long wanted to solve.
[0005] Therefore, in response to the above problems, it is urgent to develop a neural network-based rapid prediction system and method for the wind-induced vibration response of large-span bridges to ensure the structural and operational safety of bridge structures under sudden extreme climate events. Summary of the invention
[0006] Purpose of the invention: In view of the above status quo and existing problems, the present invention proposes a wind-induced vibration response early warning system and method for large-span bridges. The wind environment and structural wind-induced vibration response data of large-span bridges are collected through sensor technology, the characteristic parameters of the wind environment and the structural modal parameters of the bridge site are extracted, and a two-stage hybrid prediction model is established in combination with GRU neural network calculation to achieve rapid prediction of the wind-induced vibration response of large-span bridges, and to evaluate and warn the safe operation of large-span bridges based on the response development trend and assignment.
[0007] Technical solution: The technical solution provided by the present invention is:
[0008] A wind-induced vibration response warning system for a long-span bridge includes a monitoring front-end module, a data acquisition and transmission module, a data processing and calculation module, and an evaluation and warning module; wherein:
[0009] The monitoring front-end module is used to monitor the wind environment of the bridge site of the long-span bridge and the wind-induced vibration response of the bridge structure, including an anemometer, a displacement sensor, and an acceleration sensor;
[0010] The data acquisition and transmission module performs standardization processing on the data collected by the monitoring front-end module to establish a standardized database, and transmits the data of the standardized database to the data processing and calculation module; the data of the standardized database includes two parts, corresponding to typhoon monitoring data and bridge wind vibration response monitoring data;
[0011] The data processing and calculation module includes a data analysis module and a prediction module;
[0012] The data analysis module can analyze and process the typhoon monitoring data and bridge wind vibration response monitoring data transmitted by the data acquisition and transmission module, so as to obtain wind characteristic parameters and bridge structure modal parameters accordingly.
[0013] The prediction module is constructed based on the two-stage GRU hybrid prediction model, and can train, update, modify and apply the two-stage GRU hybrid prediction model in combination with GRU neural network calculation;
[0014] The two-stage GRU hybrid prediction model includes a first-stage typhoon wind speed prediction model and a second-stage bridge wind-induced vibration response prediction model; wherein:
[0015] The input of the first stage typhoon wind speed prediction model is wind characteristic parameters, and the output is the predicted wind speed when the typhoon passes through the bridge;
[0016] The input of the second-stage bridge wind-induced vibration response prediction model is the wind speed prediction result of the first stage and the modal parameters of the bridge structure, and the output is the predicted value of the bridge wind-induced vibration response x p , predicted value of bridge wind-induced vibration response x pIncluding displacement, velocity, and acceleration response predictions of bridge structures;
[0017] The evaluation and early warning module is based on the bridge wind vibration response prediction value x output by the data processing and calculation module. p , evaluate and warn the operating status of long-span bridges and assist in decision-making.
[0018] Preferably, when analyzing and processing the typhoon monitoring data, the data analysis module first uses the vector decomposition method to decompose the original wind speed collected by the anemometer into downwind, crosswind and vertical wind speeds, and then uses 10 minutes as the basic time interval to extract the average wind speed and pulsating wind speed by using the sliding average method, and calculates the downwind and vertical turbulence intensity, turbulence integral scale, and power spectral density.
[0019] Preferably, when analyzing and processing the wind-induced vibration response monitoring data of the bridge, the data analysis module uses a low-pass filter based on a Blackman window to extract the principal component information of the wind-induced vibration response to eliminate noise interference, uses a random reduction method to extract the free vibration component of the structural response, obtains the free vibration response signal attenuation curve, and uses a function fitting method to calculate the modal frequency, modal vibration shape and modal frequency damping ratio.
[0020] Preferably, the loss function of the first stage wind speed prediction model training is an empirical risk loss function in the form of mean square error; the loss function of the second stage wind vibration response prediction model training is a weighted sum of the empirical risk loss function in the form of mean square error and the loss function based on the motion equation.
[0021] Preferably, the loss function of the first stage wind speed prediction model training is expressed as:
[0022]
[0023] The loss function of the second stage wind vibration response prediction model training is:
[0024]
[0025]
[0026]
[0027]
[0028]
[0029] Where: N is the number of training samples; U p is the predicted wind speed; U m is the measured value of wind speed; k i Refers to the weighted coefficient of the loss function; L i(θ)(i=1,2,3,4) refers to the loss function based on displacement, velocity, acceleration, and motion equation; X p is the displacement prediction value; X m is the measured displacement value; M, C, K are the mass, damping and stiffness matrices of the bridge structure respectively; F is the sum of the buffeting force vector and the aerodynamic self-excited force vector.
[0030] Preferably, the evaluation and early warning module presets the bridge wind vibration response limit threshold x according to the long-span bridge design specification. l , and by comparing the bridge wind vibration response limit threshold x l And the predicted value of bridge wind vibration response x p , evaluate and warn the operation status of long-span bridges and assist in decision-making:
[0031] When x p >1.2x l When the red warning is issued, it means that the structural safety of the long-span bridge and the traffic on the bridge deck will be seriously threatened;
[0032] When 1.2x l >x p >x l When the warning is issued, an orange warning is issued, indicating that traffic on the bridge deck of a large-span bridge will be affected;
[0033] When x p <x l When the earthquake occurs, the long-span bridge can maintain normal operation without issuing an early warning.
[0034] Another technical purpose of the present invention is to provide an early warning method for wind-induced vibration response of a long-span bridge, comprising the following steps:
[0035] S1. Analyze and process the measured data of typhoon wind environment and wind-induced vibration response of large-span bridges
[0036] S1-1. Analyze and process the received measured data of typhoon wind environment to obtain wind characteristic parameters: first, use the vector decomposition method to decompose the received original wind speed into downwind, crosswind and vertical wind speeds, then use the sliding average method with 10 minutes as the basic time interval to extract the average wind speed and pulsating wind speed, and calculate the downwind and vertical turbulence intensity, turbulence integral scale, and power spectrum density;
[0037] S1-2, analyzing and processing the received wind-induced vibration response measured data of the large-span bridge to obtain the modal parameters of the bridge structure: firstly, using a low-pass filter based on the Blackman window to extract the principal component information of the wind-induced vibration response to eliminate noise interference, and then using the random reduction method to extract the free vibration component of the structural response, thereby obtaining the free vibration response signal attenuation curve, and using the function fitting method to calculate the modal frequency, modal vibration shape and modal frequency damping ratio;
[0038] S2, establish a two-stage GRU hybrid prediction model based on measured data
[0039] The two-stage GRU hybrid prediction model includes a first-stage typhoon wind speed prediction model and a second-stage bridge wind-induced vibration response prediction model. The first-stage typhoon wind speed prediction model takes wind environment data and wind characteristic parameters as input, and outputs the predicted wind speed when the typhoon passes over the bridge. The second-stage bridge wind-induced vibration response prediction model takes the predicted wind speed when the typhoon passes over the bridge and the modal parameters of the bridge structure output by the first-stage typhoon wind speed prediction model as input, and the predicted bridge wind-induced vibration response value x p Output: predicted value of bridge wind vibration response x p Including the predicted values of bridge structure displacement, velocity and acceleration response;
[0040] S3, according to the design specification of long-span bridges, set the bridge wind-induced vibration response limit threshold, denoted as: x l , if the predicted value of bridge wind-induced vibration response x p >1.2x l , a red warning is issued, indicating that the structural safety of long-span bridges and traffic on the bridge deck will be seriously threatened, prompting relevant departments to formulate and implement emergency plans in a timely manner to avoid major traffic accidents; if 1.2x l >x p >x l , an orange warning is issued, indicating that the traffic on the bridge deck of the large-span bridge will be affected, and the relevant departments should continue to pay attention to the changing trend of the wind-induced vibration response, and inform the vehicles in advance to pay attention to driving safety and formulate prevention plans; if x p <x l , indicating that the long-span bridge was able to maintain normal operation without the need to sound an alarm.
[0041] Preferably, in step S2, the parameters of the two-stage GRU hybrid prediction model have time-varying characteristics. Each time the latest measured data at the basic time interval is collected, the data processing and calculation module will retrain the two-stage GRU hybrid prediction model and update the model parameters.
[0042] Preferably, in step S2, the loss function of the first-stage wind speed prediction model training is an empirical risk loss function in the form of mean square error; the loss function of the second-stage wind vibration response prediction model training is a weighted sum of the empirical risk loss function in the form of mean square error and a loss function based on the motion equation.
[0043] Preferably, the loss function of the first stage wind speed prediction model training is expressed as:
[0044]
[0045] The loss function of the second stage wind vibration response prediction model training is:
[0046]
[0047]
[0048]
[0049]
[0050]
[0051] Where: N is the number of training samples; U p is the predicted wind speed; U m is the measured value of wind speed; k i Refers to the weighted coefficient of the loss function; L i (θ)(i=1,2,3,4) refers to the loss function based on displacement, velocity, acceleration, and motion equation; X p is the displacement prediction value; X m is the measured displacement value; M, C, K are the mass, damping and stiffness matrices of the bridge structure respectively; F is the sum of the buffeting force vector and the aerodynamic self-excited force vector.
[0052] Beneficial effects: With the increase of bridge span, the problem of wind-induced vibration response of bridges has become increasingly prominent. The theoretical means of traditional wind-induced vibration response analysis is the finite element method. However, the finite element analysis method has problems such as cumbersome modeling process and low computational efficiency, and its performance in rapid prediction of wind-induced vibration response of bridges is obviously insufficient. The invention proposes a neural network-based wind-induced vibration response early warning system and method for large-span bridges. The wind environment data and bridge service status of the bridge site are monitored in real time through sensor technology. A two-stage GRU hybrid prediction model is established on the basis of comprehensive consideration of the influence of wind environment parameters and the bridge's own structural modal on wind-induced vibration response. On the one hand, adding motion equations to the network can reduce the model's dependence on data and improve the model's generalization ability; on the other hand, since a large number of equations do not need to be solved, the invention significantly improves its computational efficiency. Therefore, the invention can realize the advance prediction and timely early warning of wind-induced vibration response of large-span bridges under typhoons, assist relevant departments in the formulation and implementation of emergency plans, and thus ensure the safety of large-span bridges. It has significant social benefits in improving the safety of bridge structures and bridge deck traffic. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 It is the overall framework of the system of the present invention;
[0054] Figure 2 It is a schematic diagram of a GRU unit;
[0055] Figure 3 This is a warning diagram;
[0056] Figure 4 This is a schematic diagram of the first stage GRU wind speed prediction model training;
[0057] Figure 5 This is a schematic diagram of the second-stage GRU bridge wind-induced vibration response prediction model training; DETAILED DESCRIPTION
[0058] The present invention is further described below in conjunction with the accompanying drawings. However, it should not be understood that the scope of the above subject matter of the present invention is limited to the following embodiments, and all technologies realized based on the content of the present invention belong to the scope of the present invention.
[0059] like Figure 1 As shown, a wind-induced vibration response warning system for a long-span bridge includes a monitoring front-end module, a data acquisition and transmission module, a data processing and calculation module, and an evaluation and warning module. The monitoring front-end module includes an anemometer, a displacement sensor, and an acceleration sensor, which are used to monitor the wind environment and structural wind-induced vibration response of the bridge site of the long-span bridge. The data acquisition and transmission module establishes a standardized database based on the data collected by the sensor, and transmits the data to the data processing and calculation module, which completes the data analysis and processing, prediction model training, updating, correction and application. The evaluation and warning module evaluates and warns the operating status of the long-span bridge and assists in decision-making based on the prediction results of the data processing and calculation module.
[0060] A wind-induced vibration response early warning method for a long-span bridge comprises the following steps:
[0061] S1, analyzes and processes the measured data of typhoon wind environment collected by the monitoring front-end module, uses vector decomposition method to decompose the original wind speed into downwind, crosswind and vertical wind speed, takes 10 minutes as the basic time interval, uses sliding average method to extract average wind speed and pulsating wind speed, and calculates the downwind and vertical turbulence intensity, turbulence integral scale and power spectrum density.
[0062] S2, analyze and process the measured data of wind-induced vibration response of the large-span bridge under the action of typhoon collected by the monitoring front-end module. Preferably, a low-pass filter based on the Blackman window is used to extract the principal component information of the wind-induced vibration response to eliminate noise interference, and the random reduction method is used to extract the free vibration component of the structural response, and the free vibration response signal attenuation curve is obtained. The structural damping ratio, modal frequency and vibration shape characteristics are calculated using the function fitting method.
[0063] S3, based on the measured data, establish a two-stage GRU hybrid prediction model, such as Figure 2 The figure shows a schematic diagram of a GRU unit, and its calculation method is shown in equations (1) to (4):
[0064] z t =σ g (W z xt +U z h t-1 +b z ) (1)
[0065] r t =σ g (W r x t +U r h t-1 +b r ) (2)
[0066] h′ t =φ h (W h x t +U h (r t ⊙h t-1 )+b h ) (3)
[0067] h t =(1-z t )⊙h t-1 +z t ⊙h′ t (4)
[0068] In the formula, x t is the input vector; h t is the output vector; h′ t is a candidate hidden state; z t is the update gate; r t is the reset gate; W, U, b are parameter matrices.
[0069] The two-stage GRU hybrid prediction model includes a first-stage typhoon wind speed prediction model and a second-stage bridge wind-induced vibration response prediction model; the first-stage wind speed prediction model inputs wind field data and wind characteristic parameters, and outputs the predicted wind speed when a typhoon passes over the bridge; the second-stage bridge wind-induced vibration response prediction model takes the first-stage wind speed prediction results and bridge structure modal parameters as input, and outputs the bridge structure displacement, velocity, and acceleration response.
[0070] S4, according to the design specification of long-span bridges, set the bridge wind vibration response limit threshold, denoted as: x l , if the predicted response x p >1.2x l , the structural safety of long-span bridges and traffic on the bridge deck will be seriously threatened, a red warning will be issued, and relevant departments should take emergency measures in time to avoid major traffic accidents; if 1.2x l >x p >x l, traffic on the bridge deck will be affected and an orange warning will be issued. Relevant departments should continue to pay attention to the changing trend of wind-induced vibration response and formulate prevention plans in advance; if x p <x l , long-span bridges can operate normally. Early warning methods such as Figure 3 shown.
[0071] In step S3, due to the continuous accumulation of measured data, the time correlation of wind speed and bridge wind-induced vibration response and the computational efficiency are considered, and the measured data within 100 times the prediction time interval is used as the training set to continuously update the network parameters. That is, each time new data is obtained, the network starts a new round of training and continues to optimize, such as Figure 4 and Figure 5 shown.
[0072] In step S3, the loss function of the first phase network training is the empirical risk loss function in the form of mean square error. The training process is as follows: Figure 4 The loss function of the second stage network training is the weighted sum of the empirical risk loss function in the form of mean square error and the loss function based on the motion equation. The training process is as follows: Figure 5 Considering the buffeting force and aerodynamic self-excitation force, the motion equation of the long-span bridge is shown in formula (5):
[0073]
[0074] Where M, C, and K are the mass, damping, and stiffness matrices of the bridge structure, respectively; F b is the buffeting force vector; F se represents the aerodynamic self-excited force vector; X is the displacement response vector of the bridge structure; t is time; is the velocity response vector of the bridge structure; is the acceleration response vector of the bridge structure.
[0075] Based on the quasi-steady aerodynamic theory, the buffeting force is expressed as shown in equations (6a) to (6c):
[0076]
[0077]
[0078]
[0079] Among them, L b (t), D b (t), M b (t) represent the lift, drag, and torque of the buffeting force, respectively. The buffeting force vector is represented by F b ={L b (t)D b (t)M b(t)} T ; ρ is air density; U is average wind speed; u(t), w(t) are the downwind and vertical pulsating wind speeds respectively; C L , C D , C M are lift, drag and lift moment coefficients respectively, measured by wind tunnel test; C′ L , C′ D , C′ M are the derivatives of lift, drag and lift moment coefficient with respect to the angle of attack α, respectively, L , χ′ L , χ D , χ′ D , χ M , χ′ M is the aerodynamic admittance coefficient, which is used to consider the correlation of the buffeting force along the width of the main beam.
[0080] According to Scanlan's aerodynamic self-excitation theory, the aerodynamic lift L on the main beam of a long-span bridge is se , aerodynamic resistance D se and aerodynamic torque M se The expressions of are shown in equations (7a) to (7c):
[0081]
[0082]
[0083]
[0084] Where, L se (t), D se (t), M se (t) represent the lift, drag, and torque of the aerodynamic self-excitation force, respectively. The aerodynamic self-excitation force vector is expressed as F se ={L se (t)D se (t)M se (t)} T ; B is the width of the main beam; K = ωB / U is the conversion frequency; H i *, P i *, A i *(i=1,2,…,6) are the flutter derivatives of the bridge section, measured by wind tunnel test; h, p, α are the vertical displacement, lateral displacement and torsional displacement of the main beam, respectively; are the time derivatives of the vertical, lateral and torsional displacements of the main beam respectively; other parameters are the same as those in equations (6a) to (6c).
[0085] Substitute the prediction results into the motion equation, move the load term in the motion equation to the right, take its 2-norm, and the loss function based on the motion equation can be defined as formula (8):
[0086]
[0087] The loss function of the second-stage network is defined as where k i In this embodiment, 0.25 is used.
Claims
1. A wind-induced vibration response early warning system for long-span bridges. Features: It includes monitoring front-end module, data acquisition and transmission module, data processing and calculation module and evaluation and early warning module; among which: The monitoring front-end module is used to monitor the wind environment of the bridge site of the long-span bridge and the wind-induced vibration response of the bridge structure, including an anemometer, a displacement sensor, and an acceleration sensor; The data acquisition and transmission module performs standardization processing on the data collected by the monitoring front-end module to establish a standardized database, and transmits the data of the standardized database to the data processing and calculation module; the data of the standardized database includes two parts, corresponding to typhoon monitoring data and bridge wind vibration response monitoring data; The data processing and calculation module includes a data analysis module and a prediction module; The data analysis module can analyze and process the typhoon monitoring data and bridge wind vibration response monitoring data transmitted by the data acquisition and transmission module, so as to obtain wind characteristic parameters and bridge structure modal parameters accordingly. The prediction module is constructed based on the two-stage GRU hybrid prediction model, and can train, update, modify and apply the two-stage GRU hybrid prediction model in combination with GRU neural network calculation; The two-stage GRU hybrid prediction model includes a first-stage typhoon wind speed prediction model and a second-stage bridge wind-induced vibration response prediction model; wherein: The input of the first stage typhoon wind speed prediction model is wind characteristic parameters, and the output is the predicted wind speed when the typhoon passes through the bridge; The input of the second-stage bridge wind-induced vibration response prediction model is the wind speed prediction result of the first stage and the modal parameters of the bridge structure, and the output is the predicted value of the bridge wind-induced vibration response x p , predicted value of bridge wind-induced vibration response x p Including displacement, velocity, and acceleration response predictions of bridge structures; The evaluation and early warning module is based on the bridge wind vibration response prediction value x output by the data processing and calculation module. p , evaluate and warn the operating status of long-span bridges and assist in decision-making; The loss function of the typhoon wind speed prediction model training in the first stage is the empirical risk loss function in the form of mean square error; the loss function of the bridge wind vibration response prediction model training in the second stage is the weighted sum of the empirical risk loss function in the form of mean square error and the loss function based on the motion equation; The expression of the loss function of the first stage typhoon wind speed prediction model training is: The loss function of the bridge wind-induced vibration response prediction model training in the second stage is: Where: N is the number of training samples; U p is the predicted wind speed; U m is the measured value of wind speed; k i Refers to the weighted coefficient of the loss function; L i (θ) refers to the loss function based on displacement, velocity, acceleration, and motion equation in order according to the value of i, i = 1, 2, 3, 4; X p is the displacement prediction value; X m is the measured displacement value; M, C, K are the mass, damping and stiffness matrices of the bridge structure respectively; F is the sum of the buffeting force vector and the aerodynamic self-excited force vector.
2. A wind-induced vibration response early warning system for a long-span bridge according to claim 1, Features: When analyzing and processing typhoon monitoring data, the data analysis module first uses the vector decomposition method to decompose the original wind speed collected by the anemometer into downwind, crosswind and vertical wind speeds, and then uses 10 minutes as the basic time interval to extract the average wind speed and pulsating wind speed using the sliding average method, and calculates the downwind and vertical turbulence intensity, turbulence integral scale, and power spectral density.
3. According to claim 1, a long-span bridge wind-induced vibration response early warning system, Features: The data analysis module, when analyzing and processing the wind-induced vibration response monitoring data of the bridge, uses a low-pass filter based on a Blackman window to extract the principal component information of the wind-induced vibration response to eliminate noise interference, uses a random reduction method to extract the free vibration component of the structural response, obtains the free vibration response signal attenuation curve, and uses a function fitting method to calculate the modal frequency, modal vibration shape and modal frequency damping ratio.
4. According to claim 1, a long-span bridge wind-induced vibration response early warning system, Features: The evaluation and early warning module presets the bridge wind vibration response limit threshold x according to the design specification of long-span bridges. l , and by comparing the bridge wind vibration response limit threshold x l And the predicted value of bridge wind vibration response x p , evaluate and warn the operation status of long-span bridges and assist in decision-making: When x p >1.2x l When the red warning is issued, it means that the structural safety of the long-span bridge and the traffic on the bridge deck will be seriously threatened; When 1.2x l >x p >x l When the warning is issued, an orange warning is issued, indicating that traffic on the bridge deck of a large-span bridge will be affected; When x p <x l When the earthquake occurs, the long-span bridge can maintain normal operation without issuing an early warning.
5. An early warning method for wind-induced vibration response of a long-span bridge. It is characterized in that The following steps are involved: S1, analyze and process the measured data of typhoon wind environment and the measured data of wind-induced vibration response of large-span bridges S1-1, analyze and process the received measured data of typhoon wind environment to obtain wind characteristic parameters: firstly, the received original wind speed is decomposed into downwind, crosswind and vertical wind speeds by using the vector decomposition method, and then, with 10 minutes as the basic time interval, the sliding average method is used to extract the average wind speed and the pulsating wind speed, and the downwind and vertical turbulence intensity, turbulence integral scale and power spectrum density are calculated; S1-2, analyzing and processing the received wind-induced vibration response measured data of the large-span bridge to obtain the modal parameters of the bridge structure: firstly, using a low-pass filter based on the Blackman window to extract the principal component information of the wind-induced vibration response to eliminate noise interference, and then using the random reduction method to extract the free vibration component of the structural response, thereby obtaining the free vibration response signal attenuation curve, and using the function fitting method to calculate the modal frequency, modal vibration shape and modal frequency damping ratio; S2, establish a two-stage GRU hybrid prediction model based on measured data The two-stage GRU hybrid prediction model includes a first-stage typhoon wind speed prediction model and a second-stage bridge wind-induced vibration response prediction model. The first-stage typhoon wind speed prediction model takes wind environment data and wind characteristic parameters as input, and outputs the predicted wind speed when the typhoon passes over the bridge. The second-stage bridge wind-induced vibration response prediction model takes the predicted wind speed when the typhoon passes over the bridge and the modal parameters of the bridge structure output by the first-stage typhoon wind speed prediction model as input, and the predicted bridge wind-induced vibration response value x p Output: predicted value of bridge wind vibration response x p Including the predicted values of bridge structure displacement, velocity and acceleration response; The loss function of the typhoon wind speed prediction model training in the first stage is the empirical risk loss function in the form of mean square error; the loss function of the bridge wind vibration response prediction model training in the second stage is the weighted sum of the empirical risk loss function in the form of mean square error and the loss function based on the motion equation; The expression of the loss function of the first stage typhoon wind speed prediction model training is: The loss function of the bridge wind-induced vibration response prediction model training in the second stage is: Where: N is the number of training samples; U p is the predicted wind speed; U m is the measured value of wind speed; k i Refers to the weighted coefficient of the loss function; L i (θ) refers to the loss function based on displacement, velocity, acceleration, and motion equation in order according to the value of i, i = 1, 2, 3, 4; X p is the displacement prediction value; X m is the measured displacement value; M, C, K are the mass, damping and stiffness matrices of the bridge structure respectively; F is the sum of the buffeting force vector and the aerodynamic self-excited force vector; S3, according to the design specification of long-span bridges, set the bridge wind-induced vibration response limit threshold, denoted as: x l , if the predicted value of bridge wind-induced vibration response x p >1.2x l , a red warning is issued, indicating that the structural safety of long-span bridges and traffic on the bridge deck will be seriously threatened, prompting relevant departments to formulate and implement emergency plans in a timely manner to avoid major traffic accidents; if 1.2x l >x p >x l , an orange warning is issued, indicating that the traffic on the bridge deck of the large-span bridge will be affected, and the relevant departments should continue to pay attention to the changing trend of the wind-induced vibration response, and inform the vehicles in advance to pay attention to driving safety and formulate prevention plans; if x p <x l , indicating that the long-span bridge was able to maintain normal operation without the need to sound an alarm.
6. The early warning method for wind-induced vibration response of a long-span bridge according to claim 5, Features: In step S2, the parameters of the two-stage GRU hybrid prediction model have time-varying characteristics. Each time the latest measured data at the basic time interval is collected, the data processing and calculation module will retrain the two-stage GRU hybrid prediction model and update the model parameters.
