Bridge floor global driving wind environment early warning method, system and equipment based on limited measuring points and storage medium

Abstract

The invention discloses a bridge floor global driving wind environment early warning method, system and device based on limited measuring points and a storage medium, and relates to the technical field of wind speed early warning and artificial intelligence. Future short-time wind speed is predicted based on wind speed time history data, and a two-dimensional high-temporal-spatial-resolution prediction flow field of multiple planes is obtained according to the future short-time wind speed; main beam global wind-induced vibration response data are obtained based on the wind-induced vibration response data; obtaining three-dimensional high-temporal-spatial-resolution prediction flow field data according to the two-dimensional high-temporal-spatial-resolution prediction flow field and the girder global wind vibration response data; analyzing the three-dimensional high-temporal-spatial-resolution prediction flow field data to obtain a safe driving critical wind speed, and setting a wind speed-vehicle speed limiting criterion; and when the predicted wind environment exceeds the wind speed limit criterion, outputting the predicted data and the early warning level of the corresponding driving wind environment, and carrying out advanced early warning. According to the invention, the problem of low accuracy of existing driving safety early warning is solved, and wind speed multi-stage early warning for different vehicles in different driving environments is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind speed early warning and artificial intelligence, more particularly to a bridge deck global driving wind environment early warning method, system, device and storage medium based on limited measuring points. BACKGROUND

[0002] The existing driving wind speed early warning method is based on local wind environment data obtained from limited wind speed measuring points, without considering the difference of wind environment at different heights and positions of the bridge deck. The wind speed early warning based on this cannot reflect the actual wind environment distribution of the bridge deck, thereby affecting the driving safety on the bridge.

[0003] The wind speed threshold in the current safe driving early warning method is fixed, however, the fixed threshold is prone to cause false early warning information, resulting in missed reports and false reports. Bridge operation and maintenance units need to use different thresholds to cope with the complex and variable bridge service environment and the types of driving vehicles.

[0004] And the current safe driving early warning method only considers the influence of wind speed on driving, without considering the dynamic effect of wind load. The wind-induced vibration state of the bridge not only affects the driving wind environment of the bridge, but also affects the wind environment to judge the driving safety of the bridge. However, there is currently a lack of safe driving early warning method considering the wind-induced vibration state of the bridge.

[0005] Therefore, how to provide a bridge deck global driving wind environment early warning method, system, device and storage medium based on limited measuring points to improve the accuracy of the existing driving safety early warning is a problem that needs to be solved by those skilled in the art. SUMMARY

[0006] Therefore, the present application provides a bridge deck global driving wind environment early warning method, system, device and storage medium based on limited measuring points to solve the problem of low accuracy of the existing driving safety early warning. The wind speed multi-level early warning for different vehicles under different driving environments is realized. The health monitoring cloud platform is used to realize the advanced early warning of the future short-time driving wind environment prediction data and early warning level as the output with the current measured wind speed and wind vibration state as the input.

[0007] In order to achieve the above purpose, the present application adopts the following technical scheme: a bridge deck global driving wind environment early warning method based on limited measuring points, comprising:

[0008] Respectively collect wind speed time history data and wind vibration response data at different positions of the bridge site;

[0009] Predict the future short-time wind speed based on the wind speed time history data, and obtain a plurality of two-dimensional high-temporal and spatial resolution prediction flow fields of planes according to the future short-time wind speed;

[0010] Obtain global wind vibration response data of the main girder based on the wind vibration response data;

[0011] obtaining three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment according to the two-dimensional high-temporal and spatial resolution prediction flow field of the multiple planes and the global wind vibration response data of the main girder;

[0012] performing wind-vehicle-bridge coupled vibration analysis on the three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment to obtain a safe driving critical wind speed, and setting a wind speed-vehicle speed limit criterion;

[0013] based on the three-dimensional high-temporal and spatial resolution prediction flow field of the bridge deck wind environment and the wind speed-vehicle speed limit criterion, when the predicted wind environment exceeds the wind speed limit criterion, outputting corresponding prediction data and warning level of the driving wind environment for early warning.

[0014] Preferably, the future short-time wind speed is predicted based on the wind speed time history data, and the two-dimensional high-temporal and spatial resolution prediction flow field of the multiple planes is obtained according to the future short-time wind speed, comprising:

[0015] using a time convolution neural network to predict the wind speed time history data at the bridge site to obtain future short-time wind speeds at different measurement points of the bridge site;

[0016] constructing a bridge deck two-dimensional driving wind environment reconstruction model based on a spatio-temporal deep neural network, training the bridge deck two-dimensional driving wind environment reconstruction model through wind speed time history data and CFD numerical simulation wind speed data, and obtaining a trained bridge deck two-dimensional driving wind environment reconstruction model;

[0017] inputting the future short-time wind speeds at different measurement points of the bridge site into the trained bridge deck two-dimensional driving wind environment reconstruction model to obtain the two-dimensional high-temporal and spatial resolution prediction flow field of the multiple planes.

[0018] Preferably, the global wind vibration response data of the main girder is obtained based on the wind vibration response data, comprising:

[0019] extending the wind vibration response data using a modal expansion method to obtain the global wind vibration response data of the main girder.

[0020] Preferably, the global wind vibration response data of the main girder is obtained by extending the wind vibration response data using a modal expansion method, comprising:

[0021] establishing a relationship between the wind vibration displacement of the finite measurement points and the global wind vibration displacement of the main girder by a modal expansion method, obtaining the wind vibration displacement at any position of the main girder, and obtaining the wind vibration response at any position of the main girder based on the theory of material mechanics;

[0022] the global wind vibration displacement is calculated according to the bending theory to obtain the global wind vibration strain, the global stress is obtained according to the stress-strain relationship, and the global wind vibration response data of the main girder is obtained by taking the second derivative of the global displacement.

[0023] Preferably, the spatio-temporal deep neural network comprises a convolutional autoencoder and a bidirectional recurrent neural network with gated recurrent units;

[0024] The convolutional autoencoder maps the input low temporal resolution velocity field to a latent vector through an encoding function, the latent vector being a low frequency high spatial resolution CAE modal coefficient;

[0025] The bidirectional recurrent neural network with gated recurrent units takes a high frequency discrete velocity time course as input, learns its mapping relationship to the low frequency high spatial resolution CAE modal coefficient, and calculates the CAE modal coefficient of high temporal resolution after training and convergence;

[0026] The decoding function in the convolutional autoencoder is used to inversely map the CAE modal coefficient of high temporal resolution to reconstruct a high spatio-temporal resolution two-dimensional flow field, realizing the conversion from the original low temporal resolution velocity field to the high spatio-temporal resolution two-dimensional flow field.

[0027] Preferably, the three-dimensional high spatio-temporal resolution prediction flow field data of the bridge deck wind environment is obtained according to the two-dimensional high spatio-temporal resolution prediction flow field of the multiple planes and the global wind vibration response data of the main beam, comprising:

[0028] A physical-enhanced multi-scale deep learning neural network is used to construct a bridge deck three-dimensional driving wind environment reconstruction model, the bridge deck three-dimensional driving wind environment reconstruction model is trained by the two-dimensional high spatio-temporal resolution prediction flow field of the multiple planes and the global wind vibration response data of the main beam, and the three-dimensional high spatio-temporal resolution prediction flow field data of the bridge deck wind environment is obtained through the trained bridge deck three-dimensional driving wind environment reconstruction model.

[0029] Preferably, the physical-enhanced multi-scale deep learning neural network is composed of n parallel sub-networks, the input of the mth sub-network is a scaled spatio-temporal coordinate (mt, mx), and the output is a flow field velocity vector V m at a spatio-temporal position (t, x).

[0030] The output of the intermediate layer of the multi-scale network is:

[0031]

[0032] Through network automatic differentiation, the residual e M corresponding to the momentum equation is output.

[0033] The two-dimensional high spatio-temporal resolution flow field prediction data of the bridge deck of the multiple planes, the Navier-Stokes control equation, the incoming flow wind characteristic parameters, the wind vibration state, and the two-dimensional flow field unsupervised feature item extracted by the convolutional discriminator are embedded into the multi-scale deep network through a composite loss function.

[0034] Finally, the gradient of the back propagation algorithm is derived, and the multi-scale deep learning neural network is trained using the stochastic gradient descent algorithm to obtain the three-dimensional high spatio-temporal resolution prediction flow field of the bridge deck wind environment.

[0035] Preferably, a bridge deck global driving wind environment early warning system based on limited measurement points comprises:

[0036] A data acquisition module is configured to respectively collect wind speed time history data and wind vibration response data at different positions of the bridge site.

[0037] A wind environment prediction module is configured to predict future short-term wind speed based on the wind speed time history data, obtain a plurality of two-dimensional high spatio-temporal resolution prediction flow fields of multiple planes according to the future short-term wind speed, obtain global wind vibration response data of the main girder based on the wind vibration response data, obtain three-dimensional high spatio-temporal resolution prediction flow field data of the bridge deck wind environment according to the plurality of two-dimensional high spatio-temporal resolution prediction flow fields of the multiple planes and the global wind vibration response data of the main girder, perform wind-vehicle-bridge coupled vibration analysis on the three-dimensional high spatio-temporal resolution prediction flow field data of the bridge deck wind environment to obtain a safe driving critical wind speed, and set a wind speed-vehicle speed limit criterion.

[0038] An early warning module is configured to, based on the three-dimensional high spatio-temporal resolution prediction flow field of the bridge deck wind environment and the wind speed-vehicle speed limit criterion, when the predicted wind environment exceeds the wind speed limit criterion, output corresponding prediction data and warning level of the driving wind environment and perform early warning.

[0039] Preferably, an apparatus comprises a memory, a processor, and a bridge deck global driving wind environment early warning program based on limited measurement points stored on the memory and running on the processor, and the bridge deck global driving wind environment early warning program based on limited measurement points implements the steps of the above-mentioned bridge deck global driving wind environment early warning method based on limited measurement points when executed by the processor.

[0040] Preferably, a storage medium has a computer program stored thereon, and the computer program implements the steps of the above-mentioned bridge deck global driving wind environment early warning method based on limited measurement points when executed by a processor.

[0041] Compared with the prior art, the bridge deck global driving wind environment early warning method, system, device and storage medium based on limited measuring points provided by the technical solution disclosed in the present application comprise: predicting future short-time wind speed based on wind speed time history data, and obtaining a plurality of two-dimensional high-temporal and spatial resolution prediction flow fields according to the future short-time wind speed; obtaining global wind vibration response data of the main girder based on wind vibration response data; obtaining three-dimensional high-temporal and spatial resolution prediction flow field data according to the two-dimensional high-temporal and spatial resolution prediction flow fields and the global wind vibration response data of the main girder; analyzing the three-dimensional high-temporal and spatial resolution prediction flow field data to obtain a safe driving critical wind speed, and setting a wind speed-vehicle speed limit criterion; when the predicted wind environment exceeds the wind speed limit criterion, outputting corresponding prediction data and early warning level of the driving wind environment for early warning. The present application significantly improves the prediction accuracy by predicting the future wind speed at the bridge site through the time convolution neural network; the modal expansion technology is used to deduce the wind vibration response of the limited measuring points to the whole bridge, and the wind vibration response at any position of the main girder is obtained; the spatio-temporal deep neural network is introduced to realize the high-resolution reconstruction of the multiple two-dimensional flow fields; the global wind vibration response of the main girder and the multiple two-dimensional flow fields are input into the physically enhanced multi-scale deep learning neural network to realize the high-temporal and spatial resolution reconstruction of the global three-dimensional flow field of the bridge deck; based on the results of the wind-vehicle-bridge coupled vibration analysis, the wind speed-vehicle speed limit criterion for safe driving is established, and the wind speed multi-level warning for different vehicles in different driving environments is realized; by using the health monitoring cloud platform, the early warning is finally realized with the current measured wind speed and wind vibration state as the input and the prediction data and early warning level of the driving wind environment as the output. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0043] Figure 1 A bridge deck global driving wind environment early warning method based on limited measuring points provided by the present application is shown in the flowchart. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0045] The embodiment of the application discloses a bridge deck global driving wind environment early warning method based on limited measuring points, as shown in the figure, comprising: Figure 1

[0046] Respectively collect wind speed time history data and wind vibration response data at different positions of the bridge site;

[0047] Based on the wind speed time history data, predict the future short-time wind speed at different measuring points, and obtain a plurality of two-dimensional high-temporal and spatial resolution prediction flow fields of multiple planes according to the future short-time wind speed at different measuring points;

[0048] Based on the wind vibration response data, obtain the global wind vibration response data of the main girder;

[0049] According to the two-dimensional high-temporal and spatial resolution prediction flow fields of multiple planes and the global wind vibration response data of the main girder, obtain the three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment;

[0050] Perform wind-vehicle-bridge coupled vibration analysis on the three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment, obtain the safe driving critical wind speed, and set the wind speed-vehicle speed limit criterion;

[0051] Based on the three-dimensional high-temporal and spatial resolution prediction flow field of the bridge deck wind environment and the wind speed-vehicle speed limit criterion, when the predicted wind environment exceeds the wind speed limit criterion, output the corresponding driving wind environment prediction data and warning level for early warning.

[0052] The embodiment of the application takes the current measured wind speed and wind vibration state as input, and the prediction data and warning level of the future short-time driving wind environment as output, performs early warning, solves the problem of low accuracy of existing driving safety warning, and realizes multi-level wind speed warning for different vehicles in different driving environments.

[0053] Specifically, in the bridge site area, a plurality of monitoring positions are arranged according to the cross distribution of the main girder of the bridge deck, the transverse partition (such as the left side, the middle part and the right side of the bridge deck) and the wind environment sensitive area (such as the position where the main girder flutter and vortex vibration are prone to occur), wind speed time history data (including wind speed, wind direction and time sequence information) of the corresponding position are collected by installing wind speed sensors (such as ultrasonic anemometers) at each monitoring position; at the same time, vibration sensors (such as acceleration sensors) are arranged at key structure positions of the bridge deck (such as the midspan of the main girder, the vicinity of the support and other positions with significant wind vibration response), and wind vibration response data (including vibration acceleration, displacement and time sequence information) are collected.

[0054] Specifically, the collected wind speed time history data and wind vibration response data are collected in continuous time, and the collected wind speed time history data and wind vibration response data are preprocessed, and the preprocessing includes data cleaning. Data cleaning mainly includes rejecting outliers and filling missing data, wherein the missing data is supplemented by using a cubic spline interpolation method.​

[0055] Specifically, the future short-time wind speed at different measuring points is predicted based on the wind speed time history data, and a two-dimensional high-temporal and spatial resolution predicted flow field of multiple planes is obtained according to the future short-time wind speed at different measuring points, including:

[0056] The wind speed time history data at the bridge site is predicted by using a time convolution neural network to obtain the future short-time wind speed at different measuring points at the bridge site;

[0057] A two-dimensional driving wind environment reconstruction model of the bridge deck based on a spatio-temporal deep neural network is constructed, the two-dimensional driving wind environment reconstruction model of the bridge deck is trained by the wind speed time history data in the wind tunnel test and the wind speed data in the CFD numerical simulation, a mapping relationship between the high-temporal and low-spatial resolution wind speed time history of different positions of the bridge deck in the wind tunnel test and the low-temporal and high-spatial resolution flow field of different planes in the CFD numerical simulation is established, and a trained two-dimensional driving wind environment reconstruction model of the bridge deck is obtained.

[0058] The future short-time wind speed at different measuring points at the bridge site is input into the trained two-dimensional driving wind environment reconstruction model of the bridge deck, and a two-dimensional high-temporal and spatial resolution predicted flow field of multiple planes is obtained.

[0059] Specifically, the global wind vibration response data of the main girder is obtained based on the wind vibration response data, including:

[0060] The wind vibration response data is expanded by using a modal expansion method to obtain the global wind vibration response data of the main girder.

[0061] Preferably, the wind vibration response data is expanded by using a modal expansion method to obtain the global wind vibration response data of the main girder, including:

[0062] The relationship between the wind vibration displacement of the finite measuring points and the global wind vibration displacement of the main girder is established by using a modal expansion method to obtain the wind vibration displacement at any position of the main girder, and the wind vibration response at any position of the main girder is obtained based on the theory of material mechanics.

[0063] The global wind vibration displacement is used to calculate the global wind vibration strain according to the bending theory, the global stress is obtained according to the stress-strain relationship, and the global wind vibration response data of the main girder is obtained by taking the second derivative of the global displacement.

[0064] Wherein, the relationship between the displacement of the finite measuring points and the global wind vibration displacement of the main girder is established according to the expansion technique as follows:

[0065]

[0066] In the formula, [U a ], U d and [U nand U represents the expansion matrix. Since the structural modal shape is a continuous function, the expansion matrix can be calculated by using the modal theory.

[0067] where, according to the superposition theory of linear system, the response of the system and each modal can be expressed as follows:

[0068] [U n ]=[Φ n ][q];

[0069] where, [Φ n ] represents the modal shape of the whole field of the system, and [q] represents the contribution of each modal to the response, also known as modal coordinate. The displacement of the whole field can be obtained by combining the above two formulas as follows:

[0070]

[0071] where, Φ a and Φ d represent the modal shape of the measured point and the unmeasured point respectively, so the response at the measured point can be expressed as follows:

[0072] [U a ]=[Φ a ][q];

[0073] Since [Φ a ] is not necessarily a square matrix, the modal coordinate [q] cannot be directly calculated by using the inverse method. In order to calculate the modal coordinate, the above formula is multiplied by [Φ a ] T as follows:

[0074] [Φ a ] T [U a ]=[Φ a ] T [Φ a ][q];

[0075] where, [Φ a ] T [Φ a ] is a square matrix, and the modal coordinate [q] can be expressed as follows:

[0076] [q]=([Φ a ] T [Φ a ]) -1 [Φ a ] T [U a ];

[0077] The vibration displacement [U n ] of the whole field can be obtained by combining the above formulas as follows:

[0078] [U n ] = [Φ n ]([Φ a ] T [Φ a ]) -1 [Φ a ] T [U a ];

[0079] The extended matrix T can be obtained as follows:

[0080] T = [Φ n ]([Φ a ] T [Φ a ]) -1 [Φ a ] T ;

[0081] The wind vibration displacement of the finite measuring points can be extended to the global wind vibration displacement by the above modal extension method.

[0082] Specifically, the space-time deep neural network comprises two parts of a convolutional autoencoder (CAE) and a bidirectional recurrent neural network (bi-RNN) with a gated recurrent unit (GRU).

[0083] The convolutional autoencoder maps the input low-time-resolution velocity field to a hidden vector through an encoding function, and the hidden vector is a low-frequency high-spatial-resolution CAE modal coefficient.

[0084] The bidirectional recurrent neural network with a gated recurrent unit takes a high-frequency discrete velocity time course as input, learns the mapping relationship from the high-frequency discrete velocity time course to the low-frequency high-spatial-resolution CAE modal coefficient, and calculates the high-time-resolution CAE modal coefficient after training and convergence.

[0085] The decoding function in the convolutional autoencoder is used to inversely map the high-time-resolution CAE modal coefficient to reconstruct a high-spatial-temporal-resolution two-dimensional flow field, thereby realizing conversion from the original low-time-resolution velocity field to the high-spatial-temporal-resolution two-dimensional flow field.

[0086] The future short-time wind speed of different measuring points at the bridge site is input into the trained space-time deep neural network to obtain a two-dimensional high-spatial-temporal-resolution predicted flow field of multiple planes of the bridge.

[0087] Specifically, the wind tunnel test discrete measuring point wind speed time history data and the corresponding CFD numerical simulation results are used to train a bridge deck two-dimensional driving wind environment reconstruction model to obtain a two-dimensional high-spatial-temporal-resolution flow field of multiple planes of the bridge, including:

[0088] The mapping relationship between the high-time low-space resolution wind speed time history of the bridge deck at different positions and the low-time high-space resolution flow field at different planes in the CFD numerical simulation is established by using a space-time deep neural network to obtain the two-dimensional high-time and space resolution flow field of multiple planes of the bridge.

[0089] The high-time low-space resolution wind speed time history is the instantaneous fluctuating wind speed time history measured by the three-dimensional cobra probe installed at different positions on the bridge deck in the wind tunnel test, and the high-time low-space resolution flow field on multiple two-dimensional planes distributed in the spanwise direction of the bridge can be obtained through the wind tunnel test.

[0090] The low-time high-space resolution flow field is the 3D LES numerical analysis carried out by a large general fluid calculation software Ansys The key points of the calculation scheme are as follows:

[0091] 1. Three-dimensional large eddy simulation (3D LES);

[0092] 2. Use the non-stationary wind field CFD inlet turbulence numerical simulation method suitable for 3D LES to generate a non-stationary wind field consistent with the wind tunnel test;

[0093] 3. Compare the CFD calculated values and test values of each aerodynamic force coefficient to verify the accuracy of the numerical simulation.

[0094] Based on the above CFD simulation, the low-time high-space resolution three-dimensional flow field distribution on the bridge deck can be obtained, and multiple low-time high-space resolution two-dimensional flow fields along the representative planes in the spanwise direction of the bridge can be obtained.

[0095] Specifically, the space-time deep neural network is composed of a convolutional autoencoder (CAE) and a bidirectional recurrent neural network (bi-RNN) with a gated recurrent unit (GRU).

[0096] The convolutional autoencoder maps the input low-time resolution velocity field u to the low-time resolution hidden vector A through the encoding function f θ : u→A, and then inversely maps the reconstructed velocity field θ' through the decoding function f n : A→u. As follows:

[0097]

[0098] In the formula, represents the decoding process, and Φ=[φ1(x),φ2(x),...φ n (x)] TA = [α1(t), α2(t), ... α n (t)] T , where φ i (x)=f θ' ([α1=0,α2=0,...,α i =1,...,α n =0] T ), where Φ is the mode extracted by CAE, and A represents the CAE mode coefficient.

[0099] Among them, a bidirectional recurrent neural network with gated recurrent units is used to learn high temporal resolution mode coefficients. A many-to-one RNN is used to learn the mapping relationship from high-frequency discrete velocity time history to low-frequency high spatial resolution CAE mode coefficients, as shown in the following formula:

[0100]

[0101] In the formula, The high-frequency discrete velocity time history (constitutes the input of the RNN), k is the number of time steps before and after time t, and τ k -τ k-1 It is the time sampling interval of the wind speed sensor. These are the coefficients corresponding to the i-th CAE mode (forming the output of the RNN), f i It is the mapping function determined by the bidirectional RNN, θ i These are the trainable parameters of the RNN. When the RNN training converges, the CAE modal coefficients with high temporal resolution can be calculated.

[0102] First, the low temporal resolution velocity field is mapped to low-frequency, high spatial resolution CAE mode coefficients using the encoding function in the convolutional autoencoder. Then, the corresponding high temporal resolution CAE mode coefficients are calculated using the bidirectional recurrent neural network with gated recurrent units. Finally, the high temporal resolution CAE mode coefficients are decoded using the decoding function in the convolutional autoencoder, thereby reconstructing the high spatiotemporal resolution two-dimensional flow field, as shown in the following equation:

[0103]

[0104] in, This represents the CAE mode coefficients learned by the RNN with high temporal resolution.

[0105] Specifically, based on the two-dimensional high spatiotemporal resolution predicted flow field of the multiple planes and the wind vibration response data of the main girder, three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment are obtained, including:

[0106] The physical enhanced multi-scale deep learning neural network is used to construct a bridge deck three-dimensional driving wind environment reconstruction model, the bridge deck three-dimensional driving wind environment reconstruction model is trained by using a plurality of two-dimensional high space-time resolution prediction flow fields and global wind vibration response data of the main beam, and the three-dimensional high space-time resolution prediction flow field data of the bridge deck wind environment is obtained by using the trained bridge deck three-dimensional driving wind environment reconstruction model.

[0107] Specifically, the input of the physical enhanced multi-scale deep learning neural network is the inflow wind characteristic parameter, the wind vibration response and the three-dimensional coordinate, the output is the three-dimensional wind speed field, the multi-plane two-dimensional high space-time resolution prediction flow field data of the bridge and the global wind vibration state are embedded into the multi-scale deep network by using a composite loss function, and the three-dimensional high space-time resolution prediction flow field of the bridge deck wind environment considering the wind vibration state of the main beam is obtained.

[0108] Specifically, the physical enhanced multi-scale deep learning neural network is composed of n parallel sub-networks, the input of the mth sub-network is the scaled space-time coordinate (mt, mx), and the output is the flow field velocity vector V m ;

[0109] The output of the intermediate layer of the multi-scale network is:

[0110]

[0111] The residual error e corresponding to the momentum equation is output by network automatic differentiation M ;

[0112] The physical knowledge such as the two-dimensional high space-time resolution flow field prediction data of the bridge deck of the bridge, the Navier-Stokes control equation, the inflow wind characteristic parameter, the wind vibration state, the two-dimensional flow field unsupervised feature item extracted by the convolution discriminator is embedded into the multi-scale deep network by using a composite loss function; the expression of the composite loss function is:

[0113] L=λ Dat L Dat +λ NS L NS +λ Dis L Dis +λ Vib L Vib +λ Cha L Cha ;

[0114] Wherein, L Dat is the two-dimensional plane flow field prediction data loss function, L NS is the equation residual loss function, L Dis is the unsupervised feature loss function, L Vib is the wind vibration state loss function, L ChaThe incoming flow wind characteristic loss function is λ i The weight coefficient of each loss function is λ i

[0115] Finally, the gradient of the back propagation algorithm is derived, and the multi-scale deep learning neural network is trained using the stochastic gradient descent algorithm to obtain a three-dimensional high spatio-temporal resolution prediction flow field of the bridge deck wind environment.

[0116] Specifically, the three-dimensional coordinates of different positions on the bridge deck are input into the physically enhanced multi-scale deep learning neural network, and the two-dimensional high spatio-temporal resolution flow field prediction data of multiple planes of the bridge deck are embedded into the composite loss function in the multi-scale deep learning neural network. When the neural network training is completed, a three-dimensional high spatio-temporal resolution prediction flow field of the bridge deck wind environment can be reconstructed.

[0117] Specifically, wind-vehicle-bridge coupled vibration analysis is performed on the three-dimensional high spatio-temporal resolution prediction flow field data of the bridge deck wind environment to obtain a safe driving critical wind speed, and a wind speed-vehicle speed limit criterion is set, including:

[0118] Through wind-vehicle-bridge coupled vibration analysis, the safe driving critical wind speed of different vehicles in different driving environments is obtained, and the safe driving wind speed multi-level warning threshold of each vehicle is established according to the size of the critical wind speed.

[0119] Among them, the wind-vehicle-bridge coupled vibration analysis selects small cars, small buses, medium buses, large buses, van trucks, and container trucks as research objects, and evaluates the driving safety of the vehicles under the action of different wind speeds.

[0120] Three safety accident models of rollover, sideslip, and sideslip are established, and the sideslip safety model includes dry, wet, snow, and icy bridge deck conditions.

[0121] Among them, the calculation process of the lowest critical wind speed under the occurrence of rollover safety accidents is as follows:

[0122] Under the most unfavorable combination of horizontal and lateral buffeting forces, the resultant force of the rollover-related forces in the vertical bridge deck (downward) and parallel to the side wind action direction can be represented as follows:

[0123]

[0124] Among them, the lateral aerodynamic force is The aerodynamic lift is The centrifugal force on the curve is The gravity G = mg, the horizontal buffeting inertia force F bH = ma bH The horizontal buffeting force produces an acceleration a bH The vertical buffeting inertia force F bV = mabV , the acceleration of the vertical buffeting force is a bV .

[0125] When the lateral wind blows into the curve, the resultant of the vertical bridge deck (downward) and the parallel side wind action direction under the most unfavorable combination of horizontal and lateral buffeting forces is expressed by the following formula:

[0126]

[0127] The vehicle height and width are H and B respectively, and the force acting on the vehicle is assumed to act on the center of height and width, which is conservative. The moment of the above force to the downstream side of the wind direction of the vehicle axle, i.e. the rollover moment, is as follows:

[0128] M over = 0.5HF para - 0.5BF perp ;

[0129] With the rollover moment, the judgment criterion for the vehicle to overturn accident can be given, i.e. when the total overturning moment M over ≥ 0, the vehicle will overturn accident.

[0130] The minimum critical wind speed for the occurrence of the side slip safety accident is calculated as follows:

[0131] When the vehicle is under the action of the lateral wind load, the wheel will generate a force F para parallel to the wind direction acting on the road surface. When the action force is greater than the static friction force between the wheel and the road surface, the wheel will slide in the direction of the action force, and the driving safety accident occurring in this case is the side slip safety accident. The friction force between the vehicle and the road surface can be expressed as:

[0132] F f = μ s F perp ;

[0133] In the formula, μ s is the friction coefficient, and F perp is the vertical force of the wheel on the road surface.

[0134] The judgment criterion for the vehicle to overturn accident:

[0135] F para -F f > 0;

[0136] The minimum critical wind speed for the occurrence of the side slip safety accident is calculated as follows:

[0137] The deflection of the side-slip safety accident is not the deflection of the vehicle caused by the steering of the driver, but the lateral displacement of the vehicle under the joint action of all lateral forces. When the displacement reaches a certain level, the vehicle enters another lane, which may cause a traffic safety accident. The frequency of the driver controlling the vehicle is related to the lateral wind speed. If the wind speed is too high, the driver will intentionally adjust the vehicle speed, which will increase the reaction time. In the driver's reaction time, the side slip of the wheel is not taken into account. The process of the lateral displacement of the vehicle under the action of the lateral resultant force F pera The lateral displacement of the vehicle in 0.5s can be calculated by the following formula:

[0138]

[0139] According to the relevant experience, the side-slip safety accident criterion can be expressed as: D s > 0.5m, it is generally considered that the vehicle will have a side-slip safety accident.

[0140] Based on the above three safety accident models and the induction of several basic vehicle types, the critical state wind speed of the basic vehicle type under the conditions of rollover, side-slip and side-slip on different types of bridge can be obtained. The lowest critical wind speed for the occurrence of safety accidents is used as the limiting wind speed, and the corresponding limiting speed can be obtained. The wind speed between different lanes should be reduced accordingly;

[0141] The control speed obtained in this way needs to be targeted for various different situations, which can be summarized as three types of bridge (including dry, wet and snow bridge), two lanes (inside and outside lanes), six types of vehicles (small car, small passenger car, medium passenger car, large passenger car, box type freight car and container car), and five levels of wind speed (blue, yellow, orange, red and black) as shown in Tables 1-3.

[0142] Table 1 Snow bridge driving safety speed control (km / h)

[0143]

[0144] Table 2 Wet bridge driving safety speed control (km / h)

[0145]

[0146]

[0147] Table 3 Dry bridge driving safety speed control (km / h)

[0148]

[0149]

[0150] In one embodiment of the present application, a bridge deck global driving wind environment early warning system based on limited measuring points comprises:

[0151] A data acquisition module is configured to collect wind speed time history data and wind vibration response data at different positions of the bridge site.

[0152] A wind environment prediction module is configured to predict future short-term wind speed at different measuring points based on the wind speed time history data, obtain a plurality of two-dimensional high-temporal and spatial resolution prediction flow fields of multiple planes according to the future short-term wind speed at different measuring points, obtain global wind vibration response data of the main girder based on the wind vibration response data, obtain three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment according to the two-dimensional high-temporal and spatial resolution prediction flow fields of the multiple planes and the global wind vibration response data of the main girder, perform wind-vehicle-bridge coupling vibration analysis on the three-dimensional high-temporal and spatial resolution prediction flow field data of the bridge deck wind environment to obtain a safe driving critical wind speed, and set a wind speed-vehicle speed limit criterion.

[0153] An early warning module is configured to, based on the three-dimensional high-temporal and spatial resolution prediction flow field of the bridge deck wind environment and the wind speed-vehicle speed limit criterion, output corresponding prediction data and warning level of the driving wind environment and perform early warning when the predicted wind environment exceeds the wind speed limit criterion.

[0154] In one embodiment of the present application, an apparatus comprises a memory, a processor, and a bridge deck global driving wind environment early warning program based on limited measuring points stored on the memory and running on the processor, and the bridge deck global driving wind environment early warning program based on limited measuring points implements the steps of the above-mentioned bridge deck global driving wind environment early warning method based on limited measuring points when executed by the processor.

[0155] In one embodiment of the present application, a storage medium stores a computer program, and the computer program implements the steps of the above-mentioned bridge deck global driving wind environment early warning method based on limited measuring points when executed by a processor.

[0156] In one embodiment of the present application, in order to solve the problem of low accuracy of the current bridge driving wind speed early warning method, as shown in Figure 1 the flow of a bridge deck global driving wind environment early warning method based on limited measuring points, which starts from data acquisition, goes through model training and analysis calculation, and finally realizes the prediction and early warning of the driving wind environment, and the specific steps are as follows:

[0157] 1. Data collection and preliminary processing: wind speed time history data is collected from discrete measuring points at the bridge site, and wind vibration displacement data of key measuring points at the bridge site is obtained.

[0158] 2. Wind speed prediction and wind-induced response expansion: Time convolutional neural network is used to predict the time history data of wind speed at the bridge site to obtain the future short-term predicted wind speed at different measuring points. At the same time, the global wind-induced response of the main girder is obtained: first, the relationship between the bridge system response and each mode is established, then the relationship between the global displacement of the main girder and the modal coordinates is derived, and the expansion matrix from the wind-induced displacement of the finite measuring points to the global wind-induced displacement is obtained. With the help of this matrix, the wind-induced displacement data of the key measuring points are expanded using the modal expansion technique to obtain the global wind-induced displacement of the main girder, and further based on the theory of material mechanics, the global wind-induced response of the main girder is obtained.

[0159] 3. Two-dimensional predicted flow field construction: Discrete wind speed data is obtained through wind tunnel test, and low-time high-space resolution flow field is obtained by combining 3D LES numerical simulation; the time history data of wind speed at discrete measuring points in wind tunnel test and the corresponding CFD numerical simulation results are used as training data, which are input into spatio-temporal deep neural network for training; after training, the future short-term predicted wind speed at different measuring points is input into the trained spatio-temporal deep neural network, which can output multi-plane two-dimensional high spatio-temporal resolution predicted flow field.

[0160] 4. Three-dimensional predicted flow field construction: the multi-plane two-dimensional high spatio-temporal resolution predicted flow field obtained above and the global wind-induced response of the main girder are input into the physically enhanced multi-scale deep learning network. The network takes the flow wind characteristic parameters, wind-induced response and three-dimensional coordinates as input, embeds the relevant physical knowledge into the network through the composite loss function, and finally outputs the global three-dimensional high spatio-temporal resolution predicted flow field of the bridge.

[0161] 5. Safety wind speed threshold setting: Taking various vehicle types such as small cars and buses as the research object, and aiming at different bridge deck conditions (dry, wet, snow accumulation and icing, etc.), wind-vehicle-bridge coupled vibration analysis is carried out; by establishing safety accident models such as rollover, sideslip and sidesway, the critical wind speed for safe driving of different vehicles in different driving environments is determined, and the wind speed-vehicle speed limit criterion is set according to the critical wind speed, and the multi-level warning threshold of safe driving wind speed is established.

[0162] 6. Advanced warning: Use the health monitoring (SHMS) cloud platform to build the data interface between the "short-term wind speed prediction module", "global wind-induced response analysis module" and "bridge deck wind environment multi-scale deep learning network". The "short-term wind speed prediction module" predicts the wind speed in the future period according to the measured wind speed data at the bridge site, and the "global wind-induced response analysis module" analyzes the global wind-induced state of the main girder online according to the wind-induced response data of the key measuring points. The predicted wind speed data and global wind-induced state data are input into the "bridge deck wind environment multi-scale deep learning network" to obtain the driving wind environment prediction data of different bridge decks and different lanes. Finally, the prediction data are compared with the wind speed multi-level warning threshold set above, and the corresponding multi-level warning level is output, realizing the advanced warning of driving wind environment.

[0163] Specifically, by using the health monitoring (SHMS) cloud platform, a data interface of the "short-time wind speed prediction module" and the "global wind vibration response analysis module" to the "bridge deck wind environment multi-scale deep learning network" is constructed, and an early warning system taking the current measured wind speed and wind vibration state as input and the prediction data and warning level of the future short-time driving wind environment as output is developed.

[0164] Specifically, the data interface of the "short-time wind speed prediction module" and the "global wind vibration response analysis module" to the "bridge deck wind environment multi-scale deep network module" is established through the cloud platform center server;

[0165] The current instantaneous fluctuating wind speed is obtained through the wind characteristic monitoring subsystem to determine the extreme gale condition; when the extreme gale is determined to occur, the "short-time wind speed prediction model" module calls the original wind monitoring data of the discrete measuring point to perform wind speed prediction in the future period;

[0166] The "global wind vibration response analysis module" calls the wind vibration response data of the key measuring point to perform online analysis of the global wind vibration state of the main girder;

[0167] The "driving safety evaluation module" evaluates the driving comfort and driving safety under the gale to obtain the quantitative relationship between the wind speed, vehicle speed and vehicle traffic volume and the driving safety;

[0168] The prediction wind speed data and the global wind vibration state database are transmitted to the "bridge deck wind environment multi-scale deep network module" through the center server to predict the wind environment space-time characteristics of different lanes on the bridge deck in the future period, and the safe driving wind speed obtained based on the "driving safety evaluation module" is used to perform multi-level early warning on the future wind environment.

[0169] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0170] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for early warning of bridge deck wind environment based on a limited number of measuring points, characterized in that, include: Wind speed time history data and wind vibration response data were collected at different locations at the bridge site. Predict future short-term wind speeds based on wind speed time history data, and obtain two-dimensional high spatiotemporal resolution predicted flow fields on multiple planes based on future short-term wind speeds. Obtain the full-domain wind vibration response data of the main beam based on wind vibration response data; Based on the two-dimensional high spatiotemporal resolution predicted flow field of the multiple planes and the wind vibration response data of the main beam, the three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment are obtained. A wind-vehicle-bridge coupled vibration analysis was performed on the three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment to obtain the critical wind speed for safe driving, and a wind speed-vehicle speed limit criterion was set. Based on the three-dimensional high spatiotemporal resolution predicted flow field of the bridge deck wind environment and the wind speed-vehicle speed limit criterion, when the predicted wind environment exceeds the wind speed limit criterion, the corresponding predicted data and warning level of the driving wind environment are output to provide an early warning.

2. The bridge deck full-area traffic wind environment early warning method based on finite measuring points according to claim 1, characterized in that, Predicting future short-term wind speeds based on wind speed time-history data, and obtaining two-dimensional high spatiotemporal resolution predicted flow fields in multiple planes based on the future short-term wind speeds, including: The wind speed time history data at the bridge site is predicted using a temporal convolutional neural network to obtain the future short-term wind speed at different measuring points at the bridge site; A two-dimensional traffic wind environment reconstruction model for bridge deck based on spatiotemporal deep neural network is constructed. The two-dimensional traffic wind environment reconstruction model for bridge deck is trained by wind speed time history data in wind tunnel test and wind speed data in CFD numerical simulation to obtain the trained two-dimensional traffic wind environment reconstruction model for bridge deck. The future short-term wind speeds at different measuring points at the bridge site are input into the trained two-dimensional traffic wind environment reconstruction model on the bridge deck to obtain two-dimensional high spatiotemporal resolution predicted flow fields on multiple planes.

3. The bridge deck global driving wind environment early warning method based on finite measuring points according to claim 1, characterized in that, The wind-induced vibration response data of the main beam was obtained based on the wind-induced vibration response data, including: The wind-induced vibration response data were extended using the modal extension method to obtain the full-domain wind-induced vibration response data of the main beam.

4. The bridge deck full-area traffic wind environment early warning method based on limited measuring points according to claim 3, characterized in that, The wind-induced vibration response data were extended using the modal extension method to obtain the full-domain wind-induced vibration response data of the main girder, including: The relationship between the wind-induced vibration displacement at finite measurement points and the wind-induced vibration displacement over the entire main beam is established by modal extension method. The wind-induced vibration displacement at any position of the main beam is obtained. Based on the theory of mechanics of materials, the wind-induced vibration response at any position of the main beam is obtained. The wind-induced vibration displacement across the entire field is calculated based on bending theory to obtain the wind-induced vibration strain across the entire field. The stress across the entire field is obtained based on the stress-strain relationship. The second derivative of the displacement across the entire field is used to obtain the wind-induced vibration response data of the main beam across the entire field.

5. The bridge deck full-area traffic wind environment early warning method based on finite measuring points according to claim 2, characterized in that, The spatiotemporal deep neural network includes: a convolutional autoencoder and a bidirectional recurrent neural network with gated recurrent units; The convolutional autoencoder maps the low temporal resolution velocity field of the input into a latent vector through an encoding function. The latent vector is the low-frequency, high spatial resolution CAE mode coefficient. A bidirectional recurrent neural network with gated recurrent units takes high-frequency discrete velocity time history as input, learns its mapping relationship to low-frequency high spatial resolution CAE mode coefficients, and calculates high temporal resolution CAE mode coefficients after training convergence. The high temporal resolution CAE mode coefficients are inversely mapped using the decoding function in the convolutional autoencoder to reconstruct a high spatiotemporal resolution two-dimensional flow field.

6. The bridge deck global driving wind environment early warning method based on finite measuring points according to claim 1, characterized in that, Based on the two-dimensional high spatiotemporal resolution predicted flow field of the multiple planes and the wind vibration response data of the main girder, three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment are obtained, including: A three-dimensional traffic wind environment reconstruction model for the bridge deck is constructed using a physically enhanced multi-scale deep learning neural network. The model is trained using two-dimensional high spatiotemporal resolution predicted flow field data from multiple planes and wind vibration data from the entire main beam. The trained model is then used to obtain three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment.

7. The bridge deck full-area traffic wind environment early warning method based on finite measuring points according to claim 6, characterized in that, The physically enhanced multi-scale deep learning neural network consists of n parallel sub-networks. The input of the m-th sub-network is the scaled spatiotemporal coordinates (mt, mx), and the output is the flow field velocity vector V at the spatiotemporal position (t, x). m ; The intermediate layer output of a multi-scale network is: The residual e of the corresponding momentum equation is output through automatic differentiation via the network. M ; By using a composite loss function, the two-dimensional high spatiotemporal resolution flow field prediction data of multiple planes on the bridge deck, the Navier-Stokes governing equations, the incoming wind characteristic parameters, the wind vibration state, and the two-dimensional flow field unsupervised feature terms extracted by the convolution discriminator are embedded into a multi-scale deep learning neural network. Finally, the gradient of the backpropagation algorithm is derived, and the stochastic gradient descent algorithm is used to train the three-dimensional traffic wind environment reconstruction model of the bridge deck.

8. A bridge deck full-area driving wind environment early warning system based on limited measuring points, using the bridge deck full-area driving wind environment early warning method based on limited measuring points as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect wind speed time history data and wind vibration response data at different locations at the bridge site. The wind environment prediction module is used to predict the future short-term wind speed based on wind speed time history data, and to obtain a two-dimensional high spatiotemporal resolution predicted flow field in multiple planes based on the future short-term wind speed. The wind vibration response data of the main beam is obtained based on the wind vibration response data; the three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment is obtained based on the two-dimensional high spatiotemporal resolution predicted flow field of the multiple planes and the wind vibration response data of the main beam; the wind-vehicle-bridge coupled vibration analysis is performed on the three-dimensional high spatiotemporal resolution predicted flow field data of the bridge deck wind environment to obtain the critical wind speed for safe driving, and the wind speed-vehicle speed limit criterion is set. The advanced warning module is used to predict the flow field based on the three-dimensional high spatiotemporal resolution of the bridge surface wind environment and the wind speed-vehicle speed limit criterion. When the predicted wind environment exceeds the wind speed limit criterion, it outputs the corresponding predicted data of the driving wind environment and the warning level to provide an advanced warning.

9. A device, wherein, The device includes: a memory, a processor, and a bridge deck omnidirectional wind environment early warning program based on finite measurement points stored in the memory and running on the processor. When the bridge deck omnidirectional wind environment early warning program based on finite measurement points is executed by the processor, it implements the steps of a bridge deck omnidirectional wind environment early warning method based on finite measurement points as described in any one of claims 1 to 7.

10. A storage medium, wherein, The storage medium stores a computer program, which, when executed by a processor, implements the steps of a bridge deck global driving wind environment early warning method based on finite measurement points as described in any one of claims 1 to 7.

Images