A long-term intelligent prediction method for wind-induced response of super-high-rise building structures during typhoon passage

Abstract

A long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoons, characterized in that it includes the following steps: calculating the structural wind vibration response corresponding to the target super high-rise building; using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model; using the AI-driven Transformer meteorological prediction model and the linear typhoon model to predict the incoming wind field information of the target super high-rise building; inputting the incoming wind field prediction information and the corresponding structural modal parameters into the trained cascade forward BP neural network model, and quickly predicting the structural wind vibration response of the super high-rise building at different positions under different conditions at different times in the future. The present invention can quickly, flexibly and reliably realize the long-term prediction of the wind vibration response of the super high-rise building structure during the typhoon, and provide a reference for the prediction and early warning of the structural vibration state of the super high-rise building during the typhoon, thereby improving the city's relevant emergency response capabilities and realizing the transformation from passive emergency to forward decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of building structure response analysis, and in particular to a long-term intelligent prediction method for wind-induced response of a super-high-rise building structure during a typhoon. Background Art

[0002] Super high-rise buildings have low natural frequency and low damping, and their fundamental natural frequency is close to the dominant frequency band of typhoon pulsation loads. Under the action of typhoons, the wind effect of super high-rise building structures is significant, and the wind-induced response is usually large. When the structural vibration response reaches a certain limit, the occupants will feel uncomfortable. Therefore, in response to typhoon events, it is necessary to further improve the wind-resistant design level of high-rise structures, accurately predict the wind-induced response of super high-rise building structures during strong winds, and then evaluate the safety and comfort of the structure, and give corresponding event warnings.

[0003] Because the response of super high-rise building structures under typhoons is random, non-Gaussian and non-stationary, it is a challenging task to predict the response of building structures during typhoons, especially long-term predictions. Therefore, existing research mainly focuses on predictions within a specific period or a short period of time. For example, using structural dynamics equations combined with wind tunnel tests / numerical simulations to analyze and calculate the wind-induced response of structures is limited to only referring to individual cases such as wind pressure with a specific return period (such as 10 years, 1 year) in relevant specifications, and it is impossible to analyze the structural response during the entire typhoon process.

[0004] At the same time, the wind-induced vibration response of structures can also be predicted and analyzed using traditional linear statistical analysis models (such as autoregressive models, autoregressive moving average models, etc.) or machine learning models (such as BP neural networks, convolutional neural networks, long short-term memory neural networks, etc.). The above two model-based methods often use the measured time-history data of structural response as the training set to train the model. The prediction effect is closely related to the selection of the response training set. The acquisition of structural strong wind-induced vibration response data is limited by actual meteorological conditions and is difficult to obtain. The research on the above two model-based methods mainly focuses on the recovery of missing signals and the filling of abnormal data. The prediction time scale range is generally in the seconds and minutes. However, the long-term prediction of the vibration response of super-high-rise building structures during typhoon passage (time scale range: hours to days) is not ideal. Summary of the invention

[0005] Based on this, the purpose of the present invention is to provide a long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage, so as to solve the problem of inaccurate long-term prediction results of wind-induced response of super high-rise building structures during typhoon passage in the prior art.

[0006] A long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage comprises the following steps:

[0007] Step S11, establishing a building model and a finite element model of a target super high-rise building, obtaining aerodynamic loads of the structure by wind tunnel tests, and obtaining structural dynamic characteristics by finite element analysis methods, and considering the influence of changes in the incoming wind field and structural modal characteristics, so as to calculate the structural wind vibration response corresponding to the target super high-rise building;

[0008] Step S12, taking the wind-induced vibration response of the structure as a sample set, and taking wind speed, wind direction, structural frequency and damping ratio as input characteristic variables, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best;

[0009] Step S13, using the AI-driven Transformer weather forecast model and the linear typhoon model to predict the incoming wind field information of the target super high-rise building, where the incoming wind field information includes the center position, center pressure, wind speed and wind direction of the typhoon at different times in the future;

[0010] Step S14, inputting the incoming wind field prediction information and the corresponding structural modal parameters into the trained cascade forward BP neural network model, and quickly predicting the structural wind vibration response of the target super high-rise building at different locations under different conditions at different times in the future.

[0011] Preferably, the step S11 includes:

[0012] Step S111, establish a scaled model of the target super high-rise building, integrate the wind pressure distribution obtained by the wind tunnel test to obtain the concentrated force f of the unit height section along the x-axis x , the concentrated force f along the y-axis y and the torque m about the simplified center O of the resultant force z , and focus on x , concentration, f y and torque m z Linear interpolation is performed along the height to obtain the wind load on each floor of the target super high-rise building under different wind directions, where:

[0013] f x =-w r ∫ L cosθ·C p ds

[0014] f y =-w r ∫ L sinθ·Cp ds

[0015] m z =-w r ∫ L (xsinθ-ycosθ)·C p ds

[0016] In the formula, w r is the reference wind pressure at the reference point height, θ is the angle between the normal of any point on the edge of the measuring point layer section and the x-axis, L is the line integral area, C p is the wind pressure coefficient measured according to the wind pressure test, s is the unit integral length; x and y are the orthogonal axes of the Cartesian rectangular coordinate system plane;

[0017] Step S112, establishing a finite element model of the target super high-rise building, and obtaining the modal vibration shape and natural frequency of the target super high-rise building by using a finite element analysis method;

[0018] Step S113, considering the influence of the incoming wind field and the change of the structural modal characteristics, combining the structural dynamics equation, and using the modal analysis method to obtain the power spectrum density matrix [S aa (ω)], and integrate the power spectrum density to obtain the root mean square value σ of the wind-induced vibration response of the structure a ,in,

[0019] [M]{a}+[C]{V}+[K]{y}={p(t)}

[0020] [S aa (ω)]=ω 4 [Φ][H p (ω)] * [M p ] -1 [Φ] T [S PP (ω)][Φ][M p ] -1 [H p (ω)] T [Φ] T

[0021]

[0022] Where [M] is the mass matrix of the n×n structure, [C] is the damping matrix of the n×n structure, [K] is the stiffness matrix of the n×n structure, a is the acceleration of the structure, V is the velocity of the structure, y is the displacement response of the structure, p(t) is the n-dimensional load vector, ω is the circular frequency, [Φ] is the structural vibration mode, [H p (ω)] is the frequency response function matrix in modal coordinates, the superscript “*” indicates conjugation, [M p] is the modal mass matrix, the superscript “-1” indicates the inverse, [S PP (ω)] is the power spectral density matrix of the physical excitation, and the superscript “T” indicates the transpose.

[0023] Preferably, the step S12 includes:

[0024] Step S121, taking the wind-induced vibration response of the structure as a sample set, and randomly dividing it into a training set and a test set in a ratio of 8:2;

[0025] Step S122, taking the wind speed, wind direction, structural frequency and structural damping ratio as input characteristic variables for normalization processing, and taking the root mean square response in the x direction and the root mean square response in the y direction of the specific height plane of the target super high-rise building as output variables;

[0026] Step S123, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best.

[0027] Preferably, when designing the activation functions of each layer of the cascade forward BP neural network model, the activation function of the hidden layer is the logsig function, and the activation function of the output layer is the purelin identity mapping function.

[0028] Preferably, the step S13 includes:

[0029] Step S131, using the current global three-dimensional meteorological data as input parameters, and using the AI-driven Transformer meteorological prediction model to obtain the average air pressure at different times in the future at the sea level;

[0030] Step S132, extracting the center position and center pressure of the typhoon at different times in the future according to the average air pressure;

[0031] Step S133, according to the linear typhoon model, the typhoon motion equation is simplified by using the scale analysis method to obtain the radial u′ wind speed and the tangential v′ wind speed of the typhoon wind field, and the wind direction of the typhoon wind field is obtained according to the vector sum of the radial u′ wind speed and the tangential v′ wind speed, wherein,

[0032] Radial u′ and tangential v′ wind speed control equations:

[0033]

[0034] In the formula, v gis the gradient wind speed, r is the distance between the target position and the typhoon center, λ is the azimuth, f is the Coriolis parameter, K is the turbulent diffusion coefficient, z is the altitude, ρ is the air density, p is the air pressure, Δp is the pressure difference between the typhoon center and the standard atmospheric pressure, B is the radial pressure distribution coefficient, R max is the maximum wind speed radius.

[0035] Preferably, a statistical analysis model is used to determine the key parameters in the control equation in the linear typhoon model, wherein:

[0036] B=1.2858+8.6396×10 -3 Δp-8.7745×10 -3 φ

[0037] LqCy max =4.0441-1.2090×10 -2 Δp+7.2694×10 -3 φ

[0038] K = δ × exp(Ψ × v g )

[0039] Where φ is the latitude of the typhoon center, and δ and Ψ are fitting coefficients.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] First, the present invention inputs the obtained incoming wind field prediction information and corresponding structural modal parameters into a trained cascade forward BP neural network model, thereby realizing a rapid prediction of the structural wind vibration response of the target super-high-rise building at different locations under different conditions at different times in the future.

[0042] Second, the method proposed in the present invention can provide guidance for the prediction and evaluation of the structural vibration state during the typhoon period, realize typhoon process early warning, improve the rapid response capability of relevant emergency departments, and better realize the transformation from passive emergency response to forward decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a method for long-term intelligent prediction of wind-induced response of super-high-rise building structures during typhoon passage is provided for this application;

[0044] Figure 2 The calculation flow chart of the structural wind-induced vibration response data set of the target super-high-rise building in this application;

[0045] Figure 3a The model prediction of the training set and test set in the x-direction of the top plane of the neural network model structure constructed in the embodiment of the present application;

[0046] Figure 3bThe model prediction of the training set and test set in the y direction of the top plane of the neural network model structure constructed in the embodiment of the present application;

[0047] Figure 4 Comparison between the predicted results of the typhoon center pressure at different times and the measured results in the embodiment of the present application;

[0048] Figure 5a Comparison between the long-term prediction result and the measured result of the root mean square response of the acceleration in the x-direction of the target building plane in the embodiment of the present application;

[0049] Figure 5b This is a comparison between the long-term prediction results and the measured results of the root mean square response of the acceleration in the y-direction of the target building plane in the embodiment of the present application.

[0050] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0051] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0052] See also Figure 1 and Figure 2 An embodiment of the present invention provides a method for long-term intelligent prediction of wind-induced response of super high-rise building structures during typhoon passage, comprising the following steps:

[0053] Step S11, establishing a scaled model and a finite element model of the target super high-rise building, obtaining the structural aerodynamic load by wind tunnel test and the structural dynamic characteristics by finite element analysis method, and considering the influence of the incoming wind field and the change of the structural modal characteristics, so as to calculate the structural wind vibration response corresponding to the target super high-rise building;

[0054] In a preferred embodiment of the present application, the step S11 specifically includes:

[0055] Step S111, establish a building model of the target super high-rise building, integrate the wind pressure distribution obtained from the wind tunnel test to obtain the concentrated force f of the unit height section along the x-axis x , the concentrated force f along the y-axis y and the torque m about the simplified center O of the resultant force z , and focus on x , concentration y and torque m z Linear interpolation is performed along the height to obtain the wind load on each floor of the target super high-rise building under different wind directions, where:

[0056] f x =-wr ∫ L cosθ·C p ds

[0057] f y =-w r ∫ L sinθ·C p ds

[0058] m z =-w r ∫ L (xsinθ-ycosθ)·C p ds

[0059] In the formula, w r is the reference wind pressure at the reference point height, θ is the angle between the normal of any point on the edge of the measuring point layer section and the x-axis, L is the line integral area, C p is the wind pressure coefficient measured according to the wind pressure test, s is the unit integral length; x and y are the orthogonal axes of the Cartesian rectangular coordinate system plane;

[0060] Step S112, establishing a finite element model of the target super high-rise building, and obtaining the modal vibration shape and natural frequency of the target super high-rise building according to the finite element model;

[0061] Step S113, considering the influence of the incoming wind field and the change of the structural modal characteristics, combining the structural dynamics equation, and using the modal analysis method to obtain the power spectrum density matrix [S aa (ω)], and integrate the power spectrum density to obtain the root mean square value σ of the wind-induced vibration response of the structure a ,in,

[0062] [M]{a}+[C]{V}+[K]{y}={p(t)}

[0063] [S aa (ω)]=ω 4 [Φ][H p (ω)] * [M p ] -1 [Φ] T [S PP (ω)][Φ][M p ] -1 [H p (ω)] T [Φ] T

[0064]

[0065] Where [M] is the mass matrix of the n×n structure, [C] is the damping matrix of the n×n structure, [K] is the stiffness matrix of the n×n structure, a is the acceleration of the structure, V is the velocity of the structure, y is the displacement response of the structure, p(t) is the n-dimensional load vector, ω is the circular frequency, [Φ] is the structural vibration mode, [H p (ω)] is the frequency response function matrix in modal coordinates, the superscript “*” indicates conjugation, [M p ] is the modal mass matrix, the superscript “-1” indicates the inverse, [S PP (ω)] is the power spectral density matrix of the physical excitation, and the superscript “T” indicates the transpose.

[0066] The present invention considers the influence of changes in wind speed v, wind direction α, structural frequency f, and damping ratio ζ on structural response.

[0067] Step S12, taking the wind-induced vibration response of the structure as a sample set, and taking wind speed, wind direction, structural frequency and damping ratio as input characteristic variables, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best;

[0068] In a preferred embodiment of the present invention, step S12 specifically includes:

[0069] Step S121, taking the wind-induced vibration response of the structure as a sample set, and randomly dividing it into a training set and a test set in a ratio of 8:2;

[0070] Step S122, taking the wind speed, wind direction, structural frequency and structural damping ratio as input characteristic variables for normalization processing, and taking the root mean square response in the x direction and the root mean square response in the y direction of the specific height plane of the target super high-rise building as output variables;

[0071] Step S123, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best.

[0072] It should be noted that, considering the non-negativity of the root mean square of the structural response, when designing the activation functions of each layer of the cascaded forward BP neural network model, the activation function of the hidden layer uses the logsig function, and the activation function of the output layer uses the purelin identity mapping function.

[0073] Step S13, using the AI-driven Transformer weather forecast model and the linear typhoon model to predict the incoming wind field information of the target super high-rise building, where the incoming wind field information includes the center position, center pressure, wind speed and wind direction of the typhoon at different times in the future;

[0074] In a preferred implementation of the present invention, step S13 specifically includes:

[0075] Step S131, using the current global three-dimensional meteorological data as input parameters, and using the AI-driven Transformer meteorological prediction model to obtain the average air pressure at different times in the future at the sea level;

[0076] Step S132, extracting the center position and center pressure of the typhoon at different times in the future according to the average air pressure;

[0077] Step S133, according to the linear typhoon model, the typhoon motion equation is simplified by using the scale analysis method to obtain the radial u′ wind speed and the tangential v′ wind speed of the typhoon wind field, and the wind direction of the typhoon wind field is obtained according to the vector sum of the radial u′ wind speed and the tangential v′ wind speed, wherein,

[0078] Radial u′ and tangential v′ wind speed control equations:

[0079]

[0080]

[0081] In the formula, v g is the gradient wind speed, r is the distance between the target position and the typhoon center, λ is the azimuth, f is the Coriolis parameter, K is the turbulent diffusion coefficient, z is the altitude, ρ is the air density, p is the air pressure, Δp is the pressure difference between the typhoon center and the standard atmospheric pressure, B is the radial pressure distribution coefficient, R max is the maximum wind speed radius.

[0082] Preferably, a statistical analysis model is used to determine the key parameters in the control equation in the linear typhoon model, wherein:

[0083] B=1.2858+8.6396×10 -3 Δp-8.7745×10 -3 φ

[0084] LqCy max =4.0441-1.2090×10 -2 Δp+7.2694×10 -3 φ

[0085] K = δ × exp(ψ × v g )

[0086] Where φ is the latitude of the typhoon center, and δ and Ψ are fitting coefficients.

[0087] Step S14, inputting the incoming wind field prediction information and the corresponding structural modal parameters into the trained CFBPN neural network model, and quickly predicting the structural wind vibration response of the target super high-rise building at different locations under different conditions at different times in the future.

[0088] The present invention is described in detail below with specific embodiments:

[0089] First, taking the structure of a super high-rise building as an example, the tower is 441.8m high, with 100 floors above ground, a structural aspect ratio of 10.2, and a core tube aspect ratio of 19.1. The total number of measurement points of the wind tunnel test model is 376, with 15 layers of measurement points distributed along the building height. The inflow landform type of the target super high-rise building is determined as Class C according to the Chinese standard (GB50009-2012), that is, the wind profile index value is 0.22; the reference wind speed height corresponds to the prototype reference height of 400m.

[0090] The specific working conditions are selected as follows:

[0091] The root mean square of the response in two directions at the selected position (generally the top of the structure) is calculated by considering the influence of wind speed, wind direction, natural frequency (period) and damping ratio. This example takes the structural acceleration response of a super high-rise building as an example; the wind speed at the reference height ranges from 5 to 49 m / s, with an interval of 1 m / s; the wind direction ranges from 0 to 360°, with an interval of 10°, for a total of 36 wind direction angles; the structural frequency (period) variation is calculated using the natural period ratio (T), and the natural period ratio ranges from 0.85 to 1.15; combined with the measured data of existing super high-rise buildings, the damping ratio range is considered to be 0.3 to 3.5%, for a total of 302,940 working conditions.

[0092] Second, using the structural response samples calculated by wind tunnel tests, the total samples were first randomly divided into training sets and test sets with a ratio of 8:2; the input parameter variables (wind speed, wind direction, period ratio and damping ratio) were normalized; and the network was trained using the Levenberg-Marquardt algorithm. The neural network model adopts a cascade forward BP neural network model, with 3 hidden layers, and the first hidden layer unit is set to 36. As the number of hidden layers increases, the number of corresponding neural units gradually decreases, which is (36×18×9). The following indicators are used to evaluate the model prediction performance, including: correlation coefficient (R), root mean square error (RMSE) and mean absolute error (MAE), as shown below.

[0093]

[0094] In the formula, yi is the predicted value, is the mean of the predicted values, is the actual value, is the mean of the actual values, are the standard deviations of the predicted value and the actual value, respectively, and m is the number of predictions.

[0095] See also Figure 3a and Figure 3b The prediction results of the training set and test set model in two directions of the top plane of the constructed neural network model structure are given. The root mean square errors of the response predictions in the x and y directions of the test set under this model are 0.0094 and 0.0280 respectively, and the mean absolute errors are 0.0365 and 0.0567 respectively. The correlation coefficients R of the training set and the test set can reach above 0.99.

[0096] Third, taking the passage of Super Typhoon Mangkhut (201822) as an example, the closest distance between the target super-high-rise building and the typhoon center was about 130 km. In addition, the influence range of the typhoon circulation was relatively large. On September 16, 2018, the structure of the super-high-rise building produced a significant wind-induced vibration response.

[0097] A low-frequency and high-precision dual-axis wireless acceleration sensor measurement system is used to measure the structural wind-induced response of the target super-high-rise building to obtain the true response of the structure.

[0098] The Transformer neural network-based weather forecast model was used to make long-term predictions of the typhoon pressure field, wind field and other related parameters on September 16, 2018. The initial field data source was based on the fifth generation reanalysis data of the European Meteorological Center (ERA5), with a forecast time of 24 hours and an interval of 1 hour.

[0099] See also Figure 4 According to the changes in the pressure field at sea level at different times, the typhoon center pressure P can be obtained. i Changes over time, i represents the moment, such as Figure 4 As shown in the figure (0 time represents 2018-09-1600:00). At the same time, according to the tropical cyclone best path data set provided by the China Meteorological Administration (CMA), the typhoon center pressure data at different times can be obtained. By comparing it with the predicted data, it can be found that the changes in the typhoon center pressure over time are quite consistent.

[0100] By identifying the location of the lowest air pressure in the typhoon center, the typhoon movement path is extracted. Figure 4It can be seen that due to the limitation of longitude and latitude resolution, there are certain differences between the typhoon movement path prediction results (indicated by the red line segment of AI) and the measured results (indicated by the blue line segment of CMA), but the overall path change trend is consistent.

[0101] Fourth, based on the incoming wind field information and structural dynamic characteristics of the target super-high-rise building, the neural network response prediction model can be used to make long-term predictions of the root mean square acceleration response of the structure. The long-term structural response prediction method combining the meteorological prediction model and the response prediction model can more flexibly realize the prediction of structural response under strong wind conditions, thus effectively overcoming the shortcomings of previous studies that only used strong vibration measured data under limited conditions (such as wind speed restrictions and wind direction restrictions) as the neural network training set.

[0102] See also Figure 5a and Figure 5b , the predicted results and measured results of the acceleration root mean square response of the target super high-rise building in two aspects of the plane where the acceleration sensor is located are given, where the period ratio is taken as 1.0. The structural damping ratio has a great influence on the predicted response amplitude; Figure 5a The predicted results of the structural response in the x direction when the structural damping ratio ζ is 0.5%, 2.0% and 3.5% are given respectively. Figure 5b The prediction results of the structural response in the y direction are given when the structural damping ratio ζ is 0.5%, 0.8% and 2.0% respectively.

[0103] Depend on Figure 5a and Figure 5b It can be seen that the change trend of the response prediction results is basically consistent with the measured results. The acceleration root mean square response increases first and then decreases with time, reaching the maximum response value around 14:00. The damping ratio value has a great influence on the response prediction results, especially near the maximum response moment; taking the y direction as an example, when the damping ratio ζ is 0.5%, the maximum acceleration root mean square prediction value is 11.38cm / s 2 When the damping ratio ζ is 2.0%, the predicted maximum acceleration root mean square value is 4.20 cm / s 2 , a difference of 2.71 times.

[0104] In summary, the acceleration response value intervals predicted by different damping ratios are basically consistent with the actual measurements and are all within a reasonable range; and, when the damping ratio ζ is 2.0%, the correlation coefficient R between the predicted results and the measured results in the x direction is 0.67; when the damping ratio ζ is 0.8%, the correlation coefficient R between the predicted results and the measured results in the y direction is 0.89, which further illustrates the effectiveness of the long-term structural response prediction method during typhoon influence proposed in the present invention.

[0105] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage, characterized in that: The following steps are involved: Step S11, establishing a scaled model and a finite element model of the target super high-rise building, obtaining the structural aerodynamic load by wind tunnel test and the structural dynamic characteristics by finite element analysis method, and considering the influence of the incoming wind field and the change of the structural modal characteristics, so as to calculate the structural wind vibration response corresponding to the target super high-rise building; Step S12, taking the wind-induced vibration response of the structure as a sample set, and taking wind speed, wind direction, structural frequency and damping ratio as input characteristic variables, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best; Step S13, using the current global three-dimensional meteorological data as input parameters, and using the AI-driven Transformer meteorological prediction model to obtain the average air pressure at different times in the future at the sea level; The center position and center pressure of the typhoon at different times in the future are extracted based on the average air pressure; According to the linear typhoon model, the typhoon motion equation is simplified by using the scale analysis method to obtain the radial u′ wind speed and tangential v′ wind speed of the typhoon wind field, and the wind direction of the typhoon wind field is obtained according to the vector sum of the radial u′ wind speed and the tangential v′ wind speed. Step S14, inputting the incoming wind field prediction information and the corresponding structural modal parameters into the trained cascade forward BP neural network model, and quickly predicting the structural wind vibration response of the target super high-rise building at different locations under different conditions at different times in the future.

2. The long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage according to claim 1 is characterized in that: The step S11 comprises: Step S111, establish a scaled model of the target super high-rise building, integrate the wind pressure distribution obtained by the wind tunnel test to obtain the concentrated force f of the unit height section along the x-axis x , the concentrated force f along the y-axis y and the torque m about the simplified center O of the resultant force z , and focus on x , concentration y and torque m z Linear interpolation is performed along the height to obtain the wind load on each floor of the target super high-rise building under different wind directions, where: f x =-w r ∫ L cosθ·C p ds f y =-w r ∫ L sinθ·C p ds m z =-w r ∫ L (xsinθ-ycosθ)·C p ds In the formula, w r is the reference wind pressure at the reference point height, θ is the angle between the normal of any point on the edge of the measuring point layer section and the x-axis, L is the line integral area, C p is the wind pressure coefficient measured according to the wind pressure test, s is the unit integral length; x and y are the orthogonal axes of the Cartesian rectangular coordinate system plane; Step S112, establishing a finite element model of the target super high-rise building, and obtaining the modal vibration shape and natural frequency of the target super high-rise building by using a finite element analysis method; Step S113, considering the influence of the incoming wind field and the change of the structural modal characteristics, combining the structural dynamics equation, and using the modal analysis method to obtain the power spectrum density matrix [S aa (ω)], and integrate the power spectrum density to obtain the root mean square value σ of the wind-induced vibration response of the structure a ,in, [M]{a}+[C]{V}+[K]{y}={p(t)} [S aa (ω)]=ω 4 [Φ][H] p (oh)] * [M p ] -1 [F] T [S PP (ω)][Φ][M p ] -1 [H p (oh)] T [F] T Where [M] is the mass matrix of the n×n structure, [C] is the damping matrix of the n×n structure, [K] is the stiffness matrix of the n×n structure, a is the acceleration of the structure, V is the velocity of the structure, y is the displacement response of the structure, p(t) is the n-dimensional load vector, ω is the circular frequency, [Φ] is the structural vibration mode, [H p (ω)] is the frequency response function matrix in modal coordinates, the superscript "*" indicates conjugation, [M p ] is the modal mass matrix, the superscript "-1" indicates the inverse, [S PP (ω)] is the power spectral density matrix of the physical excitation, and the superscript “T” indicates the transpose.

3. The long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage according to claim 1 is characterized in that: The step S12 comprises: Step S121, taking the wind-induced vibration response of the structure as a sample set, and randomly dividing it into a training set and a test set in a ratio of 8:2; Step S122, taking the wind speed, wind direction, structural frequency and structural damping ratio as input characteristic variables for normalization processing, and taking the root mean square response in the x direction and the root mean square response in the y direction of the specific height plane of the target super high-rise building as output variables; Step S123, selecting a cascade forward BP neural network model, and using the Levenberg-Marquardt algorithm to train the cascade forward BP neural network model, so that the prediction performance of the cascade forward BP neural network model reaches the best.

4. The long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage according to claim 3 is characterized in that: When designing the activation functions of each layer of the cascade forward BP neural network model, the activation function of the hidden layer uses the logsig function, and the activation function of the output layer uses the purelin identity mapping function.

5. The long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage according to claim 1 is characterized in that: In step S13, the radial u′ and tangential v′ wind speed control equations are: In the formula, v g is the gradient wind speed, r is the distance between the target position and the typhoon center, λ is the azimuth, f is the Coriolis parameter, K is the turbulent diffusion coefficient, z is the altitude, ρ is the air density, p is the air pressure, Δp is the pressure difference between the typhoon center and the standard atmospheric pressure, B is the radial pressure distribution coefficient, R max is the maximum wind speed radius.

6. The long-term intelligent prediction method for wind-induced response of super high-rise building structures during typhoon passage according to claim 5 is characterized in that: A statistical analysis model is used to determine the key parameters in the control equations of the linear typhoon model, where: B=1.2858+8.6396×10 -3 Δp-8.7745×10 -3 f lnR max =4.0441-1.2090×10 -2 Δp+7.2694×10 -3 φ K=δ×exp(Ψ×v g ) Where φ is the latitude of the typhoon center, and δ and Ψ are fitting coefficients.

Images