Self-adaptive bridge section aerodynamic configuration adjusting device and control method thereof

Abstract

The invention discloses a self-adaptive bridge section pneumatic shape adjusting device and a control method thereof, and belongs to the technical field of large-span bridge pneumatic control. The device comprises a wind regime monitoring module, a wind regime control module, a deformable pneumatic adjustment module and a power module, and dynamic self-adaptive adjustment of the pneumatic profile is achieved through cooperative work. The wind regime monitoring module collects wind field and bridge vibration data by using a three-dimensional ultrasonic anemometer and an accelerometer; the control module integrates FFT, TDNN algorithms and a pneumatic optimization agent model based on an ARM architecture, extracts characteristic parameters, identifies a vibration state and outputs an optimal control instruction; the deformable pneumatic adjusting module optimizes the streaming form through a guide plate (angle adjustment from-30 degrees to + 30 degrees) and a stabilizing plate (0-0.20 D lifting adjustment); the power supply module adopts solar energy and storage battery hybrid power supply to guarantee reliability. According to the control method, closed-loop control is achieved through data collection, processing analysis, adjustment execution and feedback iteration, wind-induced vibration is effectively restrained, and the wind-resistant safety and economical efficiency of the bridge are improved.

Description

An adaptive bridge cross-section aerodynamic shape adjustment device and its control method Technical Field

[0001] This invention belongs to the field of aerodynamic control technology for long-span bridges, specifically relating to an adaptive bridge cross-section aerodynamic shape adjustment device and its control method. Background Technology

[0002] In bridge engineering, especially in the construction of long-span bridges, wind resistance is a key factor affecting bridge safety and stability. Currently, most bridges adopt a fixed aerodynamic shape design, with their geometric parameters (such as guide vane angles and stabilizing plate heights) locked after construction, making it impossible to dynamically adapt to changes in wind conditions during actual operation. This fixed design mode has significant limitations: when strong winds strike, different wind attack angles and wind speeds can cause complex wind-induced responses in the bridge structure, such as vortex-induced vibration, flutter, and galloping, leading to structural fatigue damage or even failure risks; while in light or calm wind environments, an overly conservative fixed shape may result in material redundancy and cost waste. Furthermore, existing wind-resistant technologies for long-span bridges mostly focus on structural reinforcement or passive protection. While these measures can locally improve aerodynamic performance, they are fundamentally unable to respond to real-time dynamic changes in the wind field during operation, making it difficult to achieve a balance between wind resistance and economy, and limiting optimization potential. Therefore, there is an urgent need for an intelligent system that can actively deform with wind conditions and adjust the aerodynamic shape of the bridge cross section in real time, so as to break through the bottleneck of the wind resistance performance of traditional fixed cross section and improve the wind-induced safety and service performance of the bridge throughout its entire life cycle. Summary of the Invention

[0003] Based on the above shortcomings, the purpose of this invention is to propose an adaptive bridge cross-section aerodynamic shape adjustment device and its control method, which aims to effectively suppress wind-induced responses such as vortex-induced vibration, flutter and gallop by dynamically adjusting the aerodynamic shape of the bridge cross-section, thereby improving the safety and economy of long-span bridges throughout their entire life cycle.

[0004] The technical solution adopted in this invention is as follows: An adaptive bridge cross-section aerodynamic shape adjustment device, comprising a wind condition monitoring module, a control module, a deformable aerodynamic adjustment module, and a power supply module, wherein each module works together to achieve dynamic adaptive adjustment of the bridge cross-section aerodynamic shape.

[0005] The wind condition monitoring module includes a wind field monitoring unit and a structural response monitoring unit. The wind field monitoring unit contains six three-dimensional ultrasonic anemometers, respectively arranged on both sides of the 1 / 4, 1 / 2, and 3 / 4 sections of the main span of the bridge. The wind speed measurement resolution is 0.01 m / s, accuracy is ±1%, and range is 0~40 m / s. The wind direction resolution is 0.1°, and accuracy is ±2%. The structural response monitoring unit contains twelve force-balanced single-axis accelerometers, arranged on the 1 / 4, 1 / 2, and 3 / 4 sections of the main span and the 1 / 2 section of the side span, with three measuring points on each section. Each sensor transmits real-time data to the control module wirelessly.

[0006] The control module employs an embedded microcontroller based on the ARM architecture, integrating a fast Fourier transform algorithm, a time-delay neural network algorithm, and an aerodynamic optimization surrogate model. This module receives data from the wind field monitoring unit and the structural response monitoring unit in real time and executes the following processes sequentially: First, feature parameter extraction is performed: the average wind speed and triaxial turbulence intensity within a set time window are calculated. Simultaneously, the fast Fourier transform algorithm is used to process the bridge acceleration time-domain signal to obtain the power spectral density spectrum, and the vibration state is determined based on the spectral characteristics: if the power spectral density spectrum exhibits a single main peak and the deviation between the peak frequency and the structure's natural frequency is within a set threshold, it is identified as vortex-induced resonance; if it exhibits a wide-band continuous spectrum or broadband resonance peak, it is identified as buffeting, and the characteristic frequencies of the dominant vibration mode are extracted. Subsequently, intelligent decision-making is performed: the time-delay neural network algorithm takes the average wind speed, triaxial turbulence intensity, structural vibration response statistics, and vibration frequency as inputs to calculate and output the deflection / expansion of the guide vane. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The aerodynamic optimization proxy model combines the current wind-induced vibration state and wind field characteristics to quickly match the optimal aerodynamic shape from the CFD simulation database and output the final control command to the deformable aerodynamic adjustment module.

[0007] The deformable pneumatic adjustment module includes a guide plate and a stabilizing plate. The guide plate is located on the side edge of the main beam and next to the maintenance track. It can extend and retract and adjust its angle from -30° to +30° through an electric push rod and a hinge mechanism. An angle encoder is integrated at the hinge. The stabilizing plate is located at the bottom and upper center of the bridge section. It can adjust its height from 0 to 0.20D through a hydraulic lifting mechanism, where D is the height of the bridge main beam. The lifting mechanism integrates a linear displacement sensor. Each actuator feeds back its actual status parameters to the control module via CAN bus or wirelessly.

[0008] The power module adopts a hybrid power supply system of solar panels and batteries. The solar panel power is not less than 500W, and the battery supports continuous operation for more than 72 hours. In extreme weather conditions, it switches to the main power interface of the bridge.

[0009] Furthermore, the control module is configured to execute the following prediction process based on the TDNN algorithm, with the specific steps as follows:

[0010] Data preprocessing and feature extraction: With an update step of 10 seconds, the downwind direction is calculated based on high-frequency sampled data within a sliding window over the past 10 minutes. Crosswind direction Vertical The turbulence intensity, root mean square vertical and torsional accelerations, and mean wind speed Vibration frequency f.

[0011] Input vector construction: Flatten the values ​​of the feature parameters obtained above at each time delay step to form a 420-dimensional input vector;

[0012] Network structure definition: The input layer of the TDNN algorithm contains 420 neurons, the first hidden layer contains 256 neurons and uses the ReLU activation function, and the output layer contains 4 neurons and uses the linear activation function;

[0013] Training process: 8500 sets of CFD simulation data were used as the training set. The Adam optimizer was used with an initial learning rate of 0.001, a batch size of 64, and 30,000 training rounds. An early stopping strategy was adopted to prevent overfitting. The training stopped when the validation set loss did not decrease for 20 consecutive rounds.

[0014] Inference and Fine-tuning: After the TDNN outputs the initial control parameters, a maximum of 5 gradient descent iterations are performed using a surrogate model based on multinomial regression, with a step size of 0.01. In real-time operation, the TDNN first outputs the initial control parameters: the extension / retraction amount of the deflector. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The aerodynamic optimization surrogate model is then input for verification and fine-tuning; the expression of the aerodynamic optimization surrogate model is:

[0015]

[0016] Among them, c ijklm The regression coefficients are used to minimize the root mean square amplitude of the predicted vertical and torsional vibrations, and the average aerodynamic drag coefficient does not exceed 1.05 times that of the uncontrolled state. When there are multiple combinations that meet the conditions, the scheme with the smallest adjustment amplitude is selected first.

[0017] Furthermore, the CFD simulation database of the control module covers Reynolds numbers. Strauha Number The wind attack angle is -10° to +10°, and includes 8,500 sets of full-scale CFD simulation results and corresponding wind tunnel test verification data, which can be incrementally updated every six months via OTA.

[0018] Furthermore, the control module also integrates closed-loop adjustment logic: within 5 seconds after issuing the adjustment command, it compares the actual state values ​​fed back by each actuator with the target values ​​set by the command. When any of the following deviations occur, a secondary fine-tuning is automatically triggered: the absolute value of the deviation between the actual value of the rotation angle of the guide plate or the maintenance track guide plate and the target value exceeds 1°; the absolute value of the deviation between the actual value of the lifting height of the stabilizing plate and the target value exceeds 5mm; the fine-tuning process is repeated a maximum of 3 times, and the total closed-loop adjustment time does not exceed 15 seconds.

[0019] The present invention also provides an adaptive bridge cross-section aerodynamic shape control method based on the adaptive bridge cross-section aerodynamic shape adjustment device described above, comprising the following steps:

[0020] S1. Data Acquisition: The sensor array of the wind condition monitoring module collects wind speed, wind direction data and bridge vibration response data in real time to form a wind condition and structural status dataset.

[0021] S2. Data Processing and Analysis: The control module first processes and analyzes the collected data, calculates the average wind speed, triaxial turbulence intensity, and root mean square of vertical and torsional acceleration, and performs spectral analysis on the 10-minute acceleration data using the Fast Fourier Transform algorithm to identify the vibration state; it then uses a time-delay neural network algorithm to process the time series features, directly predicts the optimal aerodynamic shape parameters using the trained time-delay neural network model, and obtains the control parameters after fine-tuning by the aerodynamic optimization surrogate model;

[0022] S3, Execution Adjustment: The control module converts the optimized control parameters into instructions, driving the electric push rod, hinge mechanism and hydraulic lifting mechanism of the deformable pneumatic adjustment module to adjust the state of the guide plate and the stabilizing plate;

[0023] S4. Feedback and Iteration: The sensor continuously monitors the bridge response, and the control module performs closed-loop adjustments based on the feedback data. With a 10-minute monitoring and statistical cycle, if the vibration amplitude decreases by less than 3% or the parameter changes twice consecutively: within two consecutive 10-minute monitoring cycles, the guide plate is less than 2° and the stabilizing plate is less than 0.01D, where D is the height of the main beam of the bridge, then the system enters the hold mode. After a maximum of 5 iterations, the current shape is locked and a maintenance reminder is issued.

[0024] Furthermore, in the data processing and analysis steps, the turbulence intensity is calculated using the following formula:

[0025]

[0026] In the formula, , , These represent turbulence intensity in the downwind, crosswind, and vertical directions, respectively. , , These represent the standard deviations of the wind speed fluctuation components in the corresponding directions; It is the average wind speed in the downwind direction, and the current 10-minute sliding window is used as a calculation period.

[0027] Advantages and beneficial effects of the present invention: The present invention achieves adaptive control of the aerodynamic shape of the bridge cross section through the coordinated adjustment of the guide plate and the stabilizing plate. The design concept is advanced, the installation and maintenance are convenient, and the engineering feasibility is high. The control effect is significant and has good controllability. The optimal aerodynamic performance can be obtained by adjusting the shape parameters in real time. Attached Figure Description

[0028] Figure 1 is a schematic diagram of the wind condition monitoring module arrangement according to the present invention, wherein (a) is a schematic diagram of the arrangement of the three-dimensional ultrasonic anemometer; (b) is a schematic diagram of the arrangement of the accelerometer; and (c) is a top view of the wind condition monitoring module arrangement.

[0029] Figure 2 is a schematic diagram of the control module process according to the present invention;

[0030] Figure 3 is a schematic diagram of the fixed position of the guide plate and the stabilizing plate on the bridge according to the present invention, wherein (a) is a schematic diagram of the original bridge cross section; and (b) is a schematic diagram of the installation position of the guide plate and the stabilizing plate.

[0031] Figure 4 shows the spanwise vorticity diagrams of the flow field around the box girder under uncontrolled and controlled conditions according to the present invention, wherein (a) is under uncontrolled conditions and (b) is under controlled conditions.

[0032] Figure 5 shows the time history curves of unsteady aerodynamic forces on the box girder under uncontrolled and controlled conditions according to the present invention, wherein (a) is under uncontrolled conditions and (b) is under controlled conditions.

[0033] Figure 6 shows the time-averaged streamlines and turbulent kinetic energy diagrams of the flow field around the box girder under uncontrolled and controlled conditions according to the present invention, wherein (a) is under uncontrolled conditions and (b) is under controlled conditions. Detailed Implementation

[0034] The invention will be further described below with reference to the accompanying drawings.

[0035] Example 1

[0036] An adaptive bridge cross-section aerodynamic shape adjustment device includes a wind condition monitoring module, a control module, a deformable aerodynamic adjustment module, and a power supply module.

[0037] The wind condition monitoring module comprises two parts: a wind field monitoring unit and a structural response monitoring unit. The wind field monitoring unit is equipped with six three-dimensional ultrasonic anemometers, located on both sides of the main span's 1 / 4, 1 / 2, and 3 / 4 sections. Its wind speed measurement resolution is 0.01 m / s, accuracy is ±1%, and effective range is 0–40 m / s; wind direction resolution is 0.1°, accuracy is ±2%. The structural response monitoring unit is equipped with twelve force-balanced single-axis accelerometers, located at the main span's 1 / 4, 1 / 2, and 3 / 4 sections and the side span's 1 / 2 section, with three measuring points at each monitoring section. All anemometers and accelerometers transmit real-time wind condition and structural response data to the control module wirelessly.

[0038] The control module employs an embedded microcontroller based on the ARM architecture, integrating a fast Fourier transform algorithm, a time-delay neural network algorithm, and an aerodynamic optimization surrogate model. This module receives data from the wind field monitoring unit and the structural response monitoring unit in real time and executes the following processes sequentially: First, feature parameter extraction is performed: the average wind speed and triaxial turbulence intensity within a set time window are calculated. Simultaneously, the fast Fourier transform algorithm is used to process the bridge acceleration time-domain signal to obtain the power spectral density spectrum, and the vibration state is determined based on the spectral characteristics: if the power spectral density spectrum exhibits a single main peak and the deviation between the peak frequency and the structure's natural frequency is within a set threshold, it is identified as vortex-induced resonance; if it exhibits a wide-band continuous spectrum or broadband resonance peak, it is identified as buffeting, and the characteristic frequencies of the dominant vibration mode are extracted. Subsequently, intelligent decision-making is performed: the time-delay neural network algorithm takes the average wind speed, triaxial turbulence intensity, structural vibration response statistics, and vibration frequency as inputs to calculate and output the deflection / expansion of the guide vane. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The aerodynamic optimization proxy model combines the current wind-induced vibration state and wind field characteristics to quickly match the optimal aerodynamic shape from the CFD simulation database and output the final control command to the deformable aerodynamic adjustment module. The decision response time is no more than 5 seconds.

[0039] The CFD simulation database contains data covering Reynolds numbers. The database contains approximately 8,500 full-scale CFD simulation results with Strauhaus numbers St = 0.08–0.35 and wind attack angles of -10° to +10°, along with corresponding wind tunnel test data for verification. The database supports incremental updates every six months via OTA (Over-The-Air) updates to continuously improve the model's adaptability to new operating conditions.

[0040] The specific implementation steps and parameters of the Time Delayed Neural Network (TDNN) algorithm are disclosed below:

[0041] (1) Input feature extraction: The control module collects wind field data and structural response data every 10 seconds and calculates turbulence intensity (downwind direction) in real time. Crosswind direction Vertical ), Root mean square of vertical and torsional acceleration, and average wind speed Vibration frequency Features and parameters are used to maintain the time series of the past 10 minutes (60 sampling points in total) to form a time delay window.

[0042] (2) Input vector construction: Flatten the values ​​of the above feature parameters at each time delay step to form an input vector with a dimension of 7×60=420.

[0043] (3) TDNN network structure: The input layer has 420 neurons, which are used to receive the flattened time series vector; the hidden layer has 1 layer with 256 neurons, which use the ReLU activation function. Each hidden layer neuron is connected to the time delay sequence of the input layer in a weighted manner (weights). Bias The output layer consists of four neurons, using a linear activation function, which output the expansion and contraction of the deflector plate, respectively. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height Control parameters, etc.

[0044] (4) Training process: The 8500 sets of CFD simulation data were used as the training set. Each set of data contained the input feature sequence and the corresponding optimal shape parameter label. The Adam optimizer was used with an initial learning rate of 0.001, a batch size of 64, and 30,000 training rounds. An early stopping strategy was used (the validation set loss stopped after 20 consecutive rounds if it did not decrease) to prevent overfitting.

[0045] (5) Reasoning and fine-tuning: During real-time operation, the TDNN first outputs preliminary control parameters (expander extension / retraction amount). Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The data was then input into a local aerodynamic optimization surrogate model for verification and fine-tuning. The surrogate model is based on multinomial regression, and its expression is:

[0046]

[0047] In the formula, The regression coefficients are used. The algorithm is fine-tuned with up to 5 gradient descent iterations (step size 0.01) to minimize the predicted root mean square amplitude of vertical and torsional vibrations.

[0048] (6) Optimization Objectives: The primary objective is to minimize the root mean square amplitude of the predicted vertical and torsional vibrations, and the secondary objective is to ensure that the average aerodynamic drag coefficient does not exceed 1.05 times that of the uncontrolled state. When multiple combinations satisfy the conditions exist, the scheme with the smallest adjustment amplitude is selected first. The entire decision response time does not exceed 5 seconds. At the same time, the control module also integrates a Fast Fourier Transform (FFT) algorithm to perform spectral analysis on 10 minutes of acceleration data to identify the vibration state.

[0049] The deformable pneumatic adjustment module is the actuation unit, comprising a deflector and a stabilizing plate, both made of lightweight, high-strength materials (such as carbon fiber composites) to reduce added weight. The deflectors are located on the side edges of the main beam and beside the maintenance track, and their extension, retraction, and angle adjustment are achieved via electric push rods and hinge mechanisms. The side deflector angle adjustment range is -30° to +30°. The stabilizing plate module is located at the bottom of the bridge cross-section and the upper center of the bridge cross-section, employing a hydraulic lifting mechanism with a lifting range of 0~0.20. ( (This refers to the height of the main bridge beam). Simultaneously, to achieve precise closed-loop control, a high-precision angle encoder (0.1° resolution) is integrated at the hinge of the side guide vane, and a linear displacement sensor (2mm resolution) is integrated into the hydraulic lifting mechanism of the stabilizing plate to measure the actual rotation angle of the guide vane in real time. and the actual height of the stabilizing plate The data is then transmitted back to the control module via CAN bus or wirelessly. The control module compares the actual value with the target value within 5 seconds of issuing the adjustment command. or If the error is not detected, a secondary fine-tuning command will be automatically triggered. The fine-tuning process will be repeated a maximum of three times until the error converges to within the threshold or the maximum number of iterations is reached. The typical total time for this closed-loop adjustment process is no more than 15 seconds, ensuring that the aerodynamic shape is always consistent with the optimal target.

[0050] The power module employs a hybrid power supply system combining solar panels and batteries, installed atop the bridge tower. The solar panels have a power output of at least 500W, and the battery capacity ensures continuous operation for over 72 hours. In extreme weather conditions, it can switch to the bridge's main power interface to ensure device reliability.

[0051] Through the coordinated actions of the aforementioned mechanisms, the flow pattern around the cross section can be effectively optimized, thereby improving the aerodynamic stability of the bridge under wind conditions.

[0052] The control method for this device includes the following steps:

[0053] 1. Data Acquisition: The sensor array monitors wind speed, wind direction and bridge vibration response in real time, forming a dataset of wind conditions and structural status.

[0054] 2. Data Processing and Analysis: The control unit, combined with an aerodynamic shape optimization model, analyzes the aerodynamic performance of the bridge cross-section under current wind conditions and identifies the risk of wind-induced vibration. It also utilizes an adaptive algorithm to calculate the optimal parameter combination for the deflector rotation angle and the stabilizing plate lifting height.

[0055] 3. Execution adjustment: The control unit converts the optimized parameters into instructions, which drive the adjustment of each module through the actuator to dynamically optimize the aerodynamic shape.

[0056] 4. Feedback and Iteration: Sensors continuously monitor the bridge response, and the control unit makes closed-loop adjustments based on the feedback data to ensure optimal aerodynamic shape.

[0057] The formula for determining wind field turbulence intensity during wind condition monitoring is as follows:

[0058]

[0059] In the formula, , , These represent the turbulence intensity in the following directions (longitudinal), crosswind directions (lateral), and vertical directions, respectively. , , These represent the standard deviations of the wind speed fluctuation components in the corresponding directions. It is the average wind speed in the downwind direction (usually taken as a 10-minute calculation period).

[0060] The device in this embodiment also integrates remote status monitoring and fault diagnosis functions. By analyzing parameters such as the working current, stroke resistance and adjustment accuracy of the actuator in real time, it can intelligently identify fault states such as guide plate jamming and drive mechanism abnormality, thereby reducing the difficulty of system maintenance and operation and maintenance costs.

[0061] Example 2

[0062] As shown in Figure 1, the wind condition monitoring module employs a distributed sensor array deployed at key sections of the bridge to collect real-time wind field parameters and structural dynamic response data. The module comprises a wind field monitoring unit and a structural response monitoring unit. The wind field monitoring unit is equipped with six three-dimensional ultrasonic anemometers, as shown in Figure 1(a), arranged on both sides of sections 1 / 4, 1 / 2, and 3 / 4 of the main span, preferably installed at the top of lighting columns, approximately 6 meters above the bridge deck. The anemometers use the bridge axis direction as a 0° reference, with wind direction angles increasing clockwise. They possess the following measurement performance: wind speed measurement resolution of 0.01 m / s, accuracy of ±1%, and effective range of 0~40 m / s; wind direction measurement resolution of 0.1° and accuracy of ±2%. The structural response monitoring unit is equipped with twelve force-balanced single-axis accelerometers, as shown in Figure 1(b). These accelerometers are installed at the 1 / 4, 1 / 2, and 3 / 4 sections of the main span and the 1 / 2 section of the side span. Each monitoring section has three measuring points: two accelerometers are installed on one side to measure lateral and vertical acceleration respectively; and one accelerometer is installed on the other side specifically to measure vertical acceleration. Each sensor transmits the real-time wind field data and structural response data to the control module via wireless transmission technology.

[0063] As shown in Figure 2, the control module first receives wind field data from the wind condition monitoring unit and acceleration signals from the structural response monitoring unit in real time, and calculates key parameters such as average wind speed, turbulence intensity, and vibration frequency. The structural response monitoring unit integrates a real-time spectrum analysis module, which uses the Fast Fourier Transform (FFT) algorithm to perform spectrum analysis on 10 minutes of acceleration data, converting the time-domain signal into a power spectral density distribution in the frequency domain, thereby accurately identifying the vibration state of the bridge under wind load.

[0064] The control module employs a Time-Delayed Neural Network (TDNN) that takes the time series of the aforementioned parameters as input and captures the dynamic characteristics of wind field and structural response through its unique time delay processing mechanism. The TDNN's specific structure consists of an input layer (420 neurons), a hidden layer (256 neurons, ReLU activation function), and an output layer. The time delay window is the data from the past 10 minutes, and it is trained offline for 30,000 epochs using the Adam optimizer and mean squared error loss function. The training and validation datasets are derived from datasets covering Reynolds numbers. Strauha Number This database contains approximately 8500 sets of full-scale CFD simulation results with wind attack angles ranging from -10° to +10°, along with corresponding wind tunnel test data. It supports incremental updates every six months via OTA (Over-The-Air) to continuously improve the model's adaptability to new operating conditions. The TDNN outputs preliminary control parameters (deflector extension / retraction). Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height Then, a local aerodynamic optimization surrogate model based on computational fluid dynamics (CFD) and structural dynamics is used for rapid decision verification and up to 5 gradient fine-tuning. The response time of the entire decision-making process does not exceed 5 seconds.

[0065] The primary optimization objective is to minimize the predicted root-mean-square amplitude of vertical and torsional vibrations. A secondary objective is to ensure that the average aerodynamic drag coefficient, Cd, does not exceed 1.05 times that of the uncontrolled state. When multiple shape combinations result in predicted vibration amplitudes differing by less than 5%, the scheme with the smallest total adjustment amplitude is prioritized. Closed-loop iteration uses a 10-minute complete adjustment cycle. At the end of each cycle, the measured root-mean-square amplitude reduction is compared. If the reduction is less than 3% or the change in adjustment parameters is less than 2° (guide vane) / 0.01D (stabilizer) for two consecutive adjustments, the system enters hold mode; otherwise, optimization continues, with a maximum of 5 iterations before the current shape is forcibly locked and a maintenance reminder is issued.

[0066] As shown in Figure 3(a), the original bridge cross-section has a streamlined single-box girder appearance, with a highway windbreak on the upper surface and a maintenance track on the bottom surface. As shown in Figure 3(b), this arrangement includes a stabilizing plate located at the center of the top of the bridge deck and side guide plates symmetrically arranged on both sides of the bridge. Both the guide plates and the stabilizing plate are made of lightweight, high-strength carbon fiber composite material and are fixed to the main body of the bridge through corresponding connection structures, forming an adjustable aerodynamic shape system. The angle adjustment range of the side guide plates is -30° to +30°; the stabilizing plate is located at the bottom and upper center of the bridge cross-section, with a lifting stroke of 0-0.20D. Through the coordinated action of electric push rods, hinge mechanisms, and hydraulic lifting systems, the aerodynamic shape of the cross-section can be adjusted in real time, effectively improving the flow pattern and enhancing the aerodynamic stability of the bridge.

[0067] The device adopts a modular design, allowing the guide plate and stabilizing plate to be controlled independently. In addition, the system features remote status monitoring and fault diagnosis capabilities. By analyzing parameters such as operating current and stroke resistance, it intelligently identifies abnormal conditions such as guide plate jamming, significantly improving system reliability and reducing maintenance costs.

[0068] Through high-precision numerical simulation, the flow field structure of the original cross-section (uncontrolled) and the flow field structure after applying the aerodynamic shape adjustment measures described in this invention were reproduced and visualized, as shown in Figure 4(a). In the uncontrolled state, the shear layer on the rear edge surface of the box girder rolled up, forming large-scale vortices that alternately detached from the upper and lower surfaces in the near-wake region of the box girder. Under controlled conditions, as shown in Figure 4(b), the wake vortex street structure still exists, but the vortex street width is significantly narrowed, the size of the upper and lower vortices is significantly reduced, the vortex intensity is somewhat weakened, and most importantly, the distance at which the vortex detaches from the rear edge of the main girder is significantly increased, and the vortex core center moves further away from the girder. This indicates that the adjustment measures of this invention effectively delay the formation and detachment position of the vortex, weakening the direct effect of the vortex on the main girder.

[0069] Figure 5 shows the lift coefficient C under uncontrolled and controlled conditions, respectively. l (Figure 5(a)) and drag coefficient C d (Figure 5(b)) Comparison of time history curves. It was found that under controlled conditions, the pulsation amplitudes of both the lift coefficient and drag coefficient were significantly lower than those under uncontrolled conditions. The lift coefficient amplitude decreased by more than 40%, and the pulsation component of the drag coefficient was also significantly suppressed, resulting in a significant improvement in aerodynamic drag stability. In addition, within the same time period, the number of cycles of change of the aerodynamic coefficient under controlled conditions was significantly greater than that under uncontrolled conditions. This indicates that the vortex shedding frequency increased, the vortex shedding period shortened, and the vortex size decreased accordingly. This is completely consistent with the phenomenon of narrowing wake vortex street width and vortex core shifting backward observed in the vorticity cloud diagram in Figure 4, further confirming the control mechanism of this invention, which effectively weakens the large-scale coherent vortex structure and improves flow stability by optimizing the cross-sectional aerodynamic shape.

[0070] Figure 6 shows the time-averaged streamlines and time-averaged turbulent kinetic energy distributions of the flow field around the box girder under uncontrolled and controlled conditions, respectively. The results show that under uncontrolled conditions, the flow undergoes significant separation near the leading edge, forming a vortex structure between the crash barriers in the middle of the box girder. Two unidirectional vortices form in the rear half of the upper surface, and a vortex forms at the trailing edge of the upper surface, which then detaches. On the lower surface, vortices are observed forming between the maintenance tracks, and a broad region of high turbulent kinetic energy concentration is formed near the wake, with a peak value reaching [missing value]. The magnitude is not specified. Under controlled conditions, the vortices between the upper surface crash barriers are absent due to the presence of the central stabilizing plate. Furthermore, the latter half of the upper surface exhibits a single vortex structure, with the tail vortex forming at the edge of the tail guide plate, shifting backward compared to uncontrolled conditions. Due to the presence of the lower surface guide plate, the vortex size on the lower surface decreases, and small vortices are generated between the maintenance track and the lower guide plate. In addition, the peak turbulent kinetic energy in the wake region drops sharply to approximately [missing information]. (The reduction rate reached over 96%), the high turbulent kinetic energy region was significantly compressed and closely attached to the rear edge of the beam, the wake width was significantly narrowed, and the diffusion of turbulent kinetic energy to the far field was effectively suppressed.

[0071] Under uncontrolled conditions (Fig. 6(a)), the incoming flow undergoes significant separation near the leading edge. A strong separation vortex forms on the upper surface between the central crash barriers, followed by two rotating separation bubbles in the latter half of the upper surface. These bubbles then swirl up into a large-scale wake near the trailing edge before detaching downstream. On the lower surface, a distinct local vortex structure forms between the maintenance vehicle tracks on both sides, resulting in a broad and intense high-turbulence kinetic energy concentration zone near the wake, with a peak value reaching [missing value]. The turbulent kinetic energy diffuses widely and decays slowly in the far field, exhibiting typical aerodynamic instability characteristics. Under controlled conditions (Figure 6(b)), due to the rise of the central stabilizing plate, the separation vortex between the upper surface crash barriers is suppressed, the latter half of the separation zone is significantly simplified, retaining only a single vortex structure, and the location of the upper surface wake vortex formation is significantly shifted to the edge of the tail guide plate; the lower surface is affected by the combined effect of the side edge guide plates and the maintenance track guide plates, and the original large vortex between the maintenance tracks is dispersed. The peak turbulent kinetic energy in the wake region drops sharply to approximately (The reduction exceeds 96%), the high turbulent kinetic energy region is greatly compressed and closely attached to the rear edge of the beam, the wake width is significantly narrowed, the diffusion of turbulent kinetic energy to the far field is effectively suppressed, and the flow field stability is greatly improved.

[0072] The above comparison clearly shows that the present invention, through the coordinated adaptive adjustment of the guide plate and the stabilizing plate, fundamentally improves the flow separation characteristics on the surface of the main beam, suppresses the formation and development of local vortices, and delays and weakens the wake vortex shedding process, thereby achieving a significant reduction in wake turbulence pulsation and a comprehensive improvement in overall aerodynamic stability, providing a new and reliable technical path for intelligent wind resistance during the operation of long-span bridges.

Claims

1. An adaptive bridge cross-section aerodynamic shape adjustment device, characterized in that, The system includes a wind condition monitoring module, a control module, a deformable aerodynamic adjustment module, and a power supply module. These modules work together to achieve dynamic adaptive adjustment of the bridge's cross-sectional aerodynamic shape. The wind condition monitoring module comprises a wind field monitoring unit and a structural response monitoring unit. The wind field monitoring unit contains six three-dimensional ultrasonic anemometers, respectively positioned on both sides of the 1 / 4, 1 / 2, and 3 / 4 sections of the main span of the bridge. The wind speed measurement resolution is 0.01 m / s, accuracy is ±1%, and range is 0~40 m / s; the wind direction resolution is 0.1°, and accuracy is ±2%. The structural response monitoring unit contains 12 force-balanced single-axis accelerometers, arranged at the 1 / 4, 1 / 2, 3 / 4 sections of the main span and the 1 / 2 section of the side span, with 3 measuring points at each section. Each sensor transmits real-time data to the control module wirelessly. The control module uses an embedded microcontroller based on the ARM architecture, integrating a fast Fourier transform algorithm, a time-delay neural network algorithm, and an aerodynamic optimization surrogate model. This module receives data from the wind field monitoring unit and the structural response monitoring unit in real time and performs the following processing sequentially: First, feature parameter extraction is performed: The average wind speed and triaxial turbulence intensity within a set time window are calculated. Simultaneously, the bridge acceleration time-domain signal is processed using the Fast Fourier Transform algorithm to obtain the power spectral density spectrum. The vibration state is determined based on the spectral characteristics: if the power spectral density spectrum exhibits a single main peak and the deviation between the peak frequency and the structure's natural frequency is within a set threshold, it is identified as vortex-induced resonance; if it exhibits a wide-band continuous spectrum or broadband resonance peak, it is identified as buffeting. The characteristic frequencies of the dominant vibration mode are then extracted. Subsequently, intelligent decision-making is performed: the time-delay neural network algorithm takes the average wind speed, triaxial turbulence intensity, structural vibration response statistics, and vibration frequency as inputs to calculate and output the expansion and contraction of the guide vane. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The aerodynamic optimization proxy model, combined with the current wind-induced vibration state and wind field characteristics, quickly matches the optimal aerodynamic shape from the CFD simulation database and outputs the final control command to the deformable aerodynamic adjustment module. The deformable aerodynamic adjustment module includes a guide vane and a stabilizing plate. The guide vane is located on the side edge of the main beam and next to the maintenance track, and its extension and retraction, as well as its angle adjustment from -30° to +30°, are achieved through an electric push rod and a hinge mechanism. An angle encoder is integrated at the hinge. The stabilizing plate is located at the bottom and upper center of the bridge section, and its lifting adjustment from 0 to 0.20D is achieved through a hydraulic lifting mechanism, where D is the height of the bridge main beam. The lifting mechanism integrates a linear displacement sensor. Each actuator feeds back the actual status parameters to the control module via CAN bus or wirelessly. The power module adopts a hybrid power supply system of solar panels and batteries, switching to the bridge's main power interface in extreme weather conditions.

2. The adaptive bridge cross-section aerodynamic shape adjustment device according to claim 1, characterized in that, The control module is configured to execute the following prediction process based on the TDNN algorithm, with the following specific steps: Data preprocessing and feature extraction: Every 10 seconds is used as an update step, and based on the high-frequency sampling data within the sliding window of the past 10 minutes, the downwind direction is calculated. Crosswind direction Vertical The turbulence intensity, root mean square vertical and torsional accelerations, and mean wind speed Vibration frequency f; Input vector construction: Flatten the values ​​of the above-obtained feature parameters at each time delay step to form a 420-dimensional input vector; Network structure definition: The input layer of the TDNN algorithm contains 420 neurons, the first hidden layer contains 256 neurons and uses the ReLU activation function, and the output layer contains 4 neurons and uses the linear activation function; Training process: Using 8500 sets of CFD simulation data as the training set, the Adam optimizer is used, the initial learning rate is 0.001, the batch size is 64, and the training is conducted for 30000 rounds. An early stopping strategy is used to prevent overfitting. The validation set loss stops after 20 consecutive rounds; Inference and fine-tuning: After the TDNN outputs the initial control parameters, a maximum of 5 gradient descent fine-tuning operations are performed using a surrogate model based on multinomial regression, with a step size of 0.01; In real-time operation, the TDNN first outputs the initial control parameters: the extension and retraction of the deflector. Rotation angle 1. Inspect the rotation angle of the guide vane. stabilizing plate lifting height The aerodynamic optimization surrogate model is then input for verification and fine-tuning; the expression of the aerodynamic optimization surrogate model is: Where cijklm is the regression coefficient, the predicted root mean square amplitude of vertical and torsional vibration is minimized, and the average aerodynamic drag coefficient does not exceed 1.05 times that of the uncontrolled state. When there are multiple combinations that meet the conditions, the scheme with the smallest adjustment amplitude is selected first.

3. The adaptive bridge cross-section aerodynamic shape adjustment device according to claim 1, characterized in that, The CFD simulation database of the control module covers Reynolds numbers. Strauha Number The wind attack angle is -10° to +10°, and includes 8,500 sets of full-scale CFD simulation results and corresponding wind tunnel test verification data, which can be incrementally updated every six months via OTA.

4. The adaptive bridge cross-section aerodynamic shape adjustment device according to claim 1, characterized in that, The control module also integrates closed-loop adjustment logic: within 5 seconds after issuing the adjustment command, it compares the actual state values ​​fed back by each actuator with the target values ​​set by the command. When any of the following deviations occur, a secondary fine-tuning is automatically triggered: the absolute value of the deviation between the actual value of the rotation angle of the guide plate or the maintenance track guide plate and the target value exceeds 1°; the absolute value of the deviation between the actual value of the lifting height of the stabilizing plate and the target value exceeds 5mm; the fine-tuning process is repeated a maximum of 3 times, and the total closed-loop adjustment time does not exceed 15 seconds.

5. An adaptive bridge cross-section aerodynamic shape control method based on the adaptive bridge cross-section aerodynamic shape adjustment device according to any one of claims 1-4, characterized in that, Includes the following steps: S1. Data Acquisition: The sensor array of the wind condition monitoring module collects wind speed, wind direction data, and bridge vibration response data in real time, forming a wind condition and structural status dataset; S2. Data Processing and Analysis: The control module first processes and analyzes the collected data, calculating the average wind speed, triaxial turbulence intensity, and root mean square of vertical and torsional acceleration. It then performs spectral analysis on the 10-minute acceleration data using a fast Fourier transform algorithm to identify the vibration state. A time-series neural network algorithm is used to process the time series features, and a trained time-series neural network model is used to directly predict the optimal aerodynamic shape parameters. These parameters are then fine-tuned using an aerodynamic optimization surrogate model to obtain the control parameters; S3. Execution Line Adjustment: The control module converts the optimized control parameters into instructions, driving the electric push rod, hinge mechanism, and hydraulic lifting mechanism of the deformable pneumatic adjustment module to adjust the state of the guide vane and stabilizing plate; S4, Feedback and Iteration: The sensor continuously monitors the bridge response, and the control module performs closed-loop adjustment based on the feedback data; with a 10-minute monitoring and statistical cycle, if the vibration amplitude decrease is less than 3% or the change in adjustment parameters is less than 2° for the guide vane and less than 0.01D for the stabilizing plate within two consecutive 10-minute monitoring cycles, where D is the height of the main beam of the bridge, then it enters the hold mode, and after a maximum of 5 iterations, it locks the current shape and issues a maintenance reminder.

6. The adaptive bridge cross-section aerodynamic shape control method according to claim 5, characterized in that: In the data processing and analysis steps, the turbulence intensity is calculated using the following formula: In the formula, ， ， These represent turbulence intensity in the downwind, crosswind, and vertical directions, respectively. ， ， These represent the standard deviations of the wind speed fluctuation components in the corresponding directions; It is the average wind speed in the downwind direction, and the current 10-minute sliding window is used as a calculation period.

Images