Load balance control method and device in lifting process of synchronous jacking bridge

Abstract

The invention relates to the technical field of bridge jacking, and discloses a load balancing control method and device in a synchronous jacking bridge lifting process. According to the method, multiple items of data of each jacking point of a bridge are collected in real time through a multi-source sensor array and transmitted to an edge computing node. The edge computing node uses an improved Transform architecture to extract cross-sensor spatial-temporal characteristics, generates real-time virtual mapping of a bridge structure based on a digital twinning technology, and calculates dynamic control parameters of each jacking point through a reinforcement learning strategy. A self-adaptive PID controller and a hydraulic executing mechanism are adopted for cooperative operation, multi-jacking-point load balance control is achieved, and meanwhile lifting track deviation is corrected in real time through a vibration feedback loop. According to the method, by integrating advanced algorithms and technologies, the safety and efficiency in the bridge jacking process are improved, balanced distribution of loads is ensured, and the lifting track is corrected in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge jacking, and in particular to a load balancing control method and device during the synchronous jacking bridge raising process. Background Art

[0002] Synchronous jacking technology is a key process in bridge repair, reinforcement or lifting. However, due to the complexity and unevenness of bridge structures, as well as the influence of various uncertain factors during the jacking process, such as uneven load distribution, differences in hydraulic cylinder output, structural deformation, etc., it is difficult to maintain a balanced load at each jacking point. Traditional control methods often rely on empirical formulas or simple feedback mechanisms, which make it difficult to adjust the output of each jacking point in real time and accurately, and can easily cause problems such as structural stress concentration, local damage, or deviation in the lifting trajectory. Therefore, a more intelligent and efficient load balancing control method is needed to ensure the safety and stability of the bridge lifting process. Summary of the Invention

[0003] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a load balancing control method and device during the synchronous jacking bridge lifting process, so as to solve the problem in the prior art that the loads at each jacking point are uneven and difficult to balance in real time.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a load balancing control method during a synchronous jacking bridge raising process, the method comprising: The multi-source sensor array collects real-time pressure data, displacement data, tilt angle and vibration frequency data of each jacking point of the bridge, and transmits them synchronously to the edge computing node; Based on the real-time data collected by the multi-source sensor array, the edge computing node uses an improved Transformer architecture to extract cross-sensor spatiotemporal features. Based on digital twin technology, it generates a real-time virtual map of the bridge structure. It then iterates through a reinforcement learning strategy to calculate the dynamic control parameters of each jacking point in real time. These dynamic control parameters include the optimal pressure adjustment parameters and displacement compensation. An adaptive PID controller is used in coordination with the hydraulic actuator according to real-time dynamic control parameters to achieve load balancing control at multiple lifting points. At the same time, the lifting trajectory deviation is corrected in real time through the vibration feedback loop.

[0005] Preferably, in a possible implementation manner of the first aspect, the multi-source sensor array consists of a pressure sensor, a displacement sensor, a three-axis tilt sensor and an acceleration sensor, and data alignment is achieved through a timestamp synchronization mechanism.

[0006] Preferably, in a possible implementation of the first aspect, the improved Transformer architecture includes: The encoder layer is used to encode the spatiotemporal position of multi-source sensor data and generate high-dimensional feature vectors; The decoder layer generates initial predictions of dynamic control parameters based on the high-dimensional feature vector; The multi-head spatiotemporal attention mechanism captures the dynamic correlation of load distribution by allocating joint attention weights across sensor channels and time dimensions; The residual connection layer is used to fuse multi-source sensor data with attention-weighted features to obtain cross-sensor spatiotemporal features.

[0007] Preferably, in a possible implementation manner of the first aspect, the spatiotemporal position encoding of the encoder layer is implemented by the following formula:

[0008] in Indicates the Lifting points at time The encoding vector of is a single-dimensional convolution operation, For the Lifting points at time The raw sensor input data, is the total number of sine functions, is the index, and is a learnable parameter, is the time scale parameter is the dimension of the encoding vector, It is the encoding function of the spatial position of the lifting point.

[0009] Preferably, in a possible implementation of the first aspect, the multi-head spatiotemporal attention mechanism includes: The high-dimensional feature vector output by the encoder layer is expanded into a spatiotemporal sequence according to the sensor channel and time dimensions. The query matrix Q, key matrix K, and value matrix V are generated through linear projection. Each matrix is divided into H independent attention heads. Each attention head performs the following operations through the spatiotemporal joint weight calculation module:

[0010] in, For the The output of an attention head, 、 、 Respectively The query, key, and value matrices of the attention heads, 、 、 For the The learnable projection matrix of the attention heads, For transpose calculation, is the subspace dimension, represents the time decay function, represents the spatial distance penalty term; The spatiotemporal joint weight calculation module generates an attention weight distribution with physical constraints by coupling the spatial topological relationship and time series correlation between sensors; The outputs of H attention heads are concatenated in the feature dimension, feature fusion is performed through the feedforward neural network layer, and layer normalization technology is used to stabilize the training process; The multi-head spatiotemporal attention mechanism captures the load transfer patterns at different spatiotemporal scales in parallel, dynamically establishes a mechanical correlation model across jacking points, and achieves robust perception of asymmetric load distribution.

[0011] Preferably, in a possible implementation of the first aspect, the digital twin technology generates a real-time virtual mapping of the bridge structure, including: Establish the bridge dynamic equations based on the three-dimensional finite element model:

[0012] in, is the mass matrix, is the damping matrix, is the stiffness matrix, is the displacement vector, s is the time variable, External load vector driven by real-time sensor data; The stress distribution and deformation field in the virtual map are updated using the bridge dynamic equations.

[0013] Preferably, in a possible implementation of the first aspect, the state space in the reinforcement learning strategy is the virtually mapped stress gradient distribution, the real-time pressure difference of each lifting point, and the historical displacement sequence; the action space is the pressure adjustment amount and displacement compensation amount of each lifting point; and the reward function in the reinforcement learning strategy is designed as follows:

[0014] in, is the reward value, For the Real-time pressure value of each lifting point, is the total number of jacking points, is the average pressure value, For the lifting point and The height difference, 、 is the index of the lifting point, and is the weight coefficient.

[0015] Preferably, in a possible implementation of the first aspect, the multi-lifting point load balancing control includes: dynamically allocating the hydraulic cylinder output of each lifting point according to the real-time pressure difference; when it is detected that the pressure difference between any two points exceeds a set threshold, triggering the gradient descent algorithm to recalculate the target displacement sequence, and applying priority compensation control to the lagging lifting point.

[0016] Preferably, in a possible implementation of the first aspect, the vibration feedback loop includes: extracting the main frequency component through real-time vibration spectrum analysis, and when it is detected that the main frequency energy exceeds the safety threshold, generating a reverse compensation signal to inject into the integral term of the PID controller, and triggering the lifting speed reduction mode until the spectrum characteristics are restored to the permitted range.

[0017] In a second aspect, the present invention provides a load balancing control device during the synchronous jacking bridge raising process, the device comprising: The multi-source sensor array module, consisting of pressure sensors, displacement sensors, three-axis tilt sensors, and acceleration sensors, is deployed at each lifting point and data alignment is achieved through a timestamp synchronization mechanism; The edge computing node module, which includes an improved Transformer processing unit, a digital twin modeling unit, and a reinforcement learning strategy unit, is used to extract cross-sensor spatiotemporal features, generate a real-time virtual map of the bridge, and calculate dynamic control parameters; The adaptive PID controller module receives dynamic control parameters and coordinates with the hydraulic actuator module to adjust the output of each lifting point; Hydraulic actuator module, including multiple sets of hydraulic cylinders and servo valves, performs load-balancing jacking operations; The vibration feedback loop module uses spectrum analysis to correct PID control parameters and lifting speed in real time.

[0018] The beneficial effects of the present invention are as follows: the load balancing control method of the present invention in the process of synchronous jacking bridge lifting collects data of each jacking point of the bridge in real time through a multi-source sensor array, and combines the processing power of the edge computing node to realize real-time monitoring and analysis of the bridge structure. The improved Transformer architecture is used to extract cross-sensor spatiotemporal features, and a real-time virtual mapping of the bridge structure is generated based on the digital twin technology, thereby improving the accuracy and real-time perception of the bridge state. The dynamic control parameters of each jacking point are iteratively calculated through a reinforcement learning strategy, and an adaptive PID controller is used to coordinate with the hydraulic actuator to realize load balancing control of multiple jacking points. At the same time, the lifting trajectory deviation is corrected in real time through a vibration feedback loop to ensure the safety and stability of the bridge lifting process. This method has the advantages of intelligence, efficiency, and precision, and provides a new solution for load balancing control in the process of synchronous bridge lifting. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 A flow chart of a load balancing control method during the synchronous jacking bridge raising process is provided for this application.

[0021] Figure 2 A structural diagram of a load balancing control device during the synchronous jacking bridge raising process is provided for this application. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] Example 1: Figure 1 As shown, the present invention provides a load balancing control method during the synchronous jacking bridge raising process, comprising: The pressure data, displacement data, tilt angle and vibration frequency data of each jacking point of the bridge are collected in real time through a multi-source sensor array and transmitted synchronously to the edge computing node.

[0024] In this embodiment, a multi-source sensor array is configured at each jacking point of the bridge. The sensor group at each jacking point consists of a pressure sensor, a displacement sensor, a three-axis inclination sensor and an acceleration sensor. The pressure sensor is embedded in the bottom of the jacking support seat with a spoke-type structure to monitor the output pressure value of the hydraulic system in real time; the displacement sensor is configured with a composite architecture of a laser ranging module and a contact displacement meter to measure the relative displacement of the contact surface between the jacking bracket and the bridge body with millimeter-level accuracy; the three-axis inclination sensor uses an inertial measurement unit that integrates a MEMS gyroscope and an accelerometer, and is mounted on the bottom surface of the bridge body to monitor three-dimensional attitude angle changes; the acceleration sensor is arranged at key positions of the bridge body around the jacking point to collect structural vibration spectrum characteristics through the piezoelectric effect.

[0025] The multi-source sensor array uses hardware trigger signals to synchronize data acquisition across sensor channels. Each sensor module has a built-in high-precision clock module, and the edge computing node periodically broadcasts the NTP network time protocol for clock calibration. During data acquisition, the pressure sensor captures dynamic pressure fluctuations at a 10kHz sampling rate, the displacement sensor records displacement increments at a frequency of 200Hz, and the triaxial inclinometer and accelerometer synchronously output attitude angle parameters and vibration acceleration data at rates of 1kHz and 5kHz, respectively.

[0026] After receiving multi-source sensor data streams via Ethernet, the edge computing node first performs time window alignment. Using a sliding window mechanism, the heterogeneous sensor data is resampled using a 20ms time unit. High-frequency acceleration data is downsampled, while cubic spline interpolation is used to compensate for the temporal resolution of low-frequency sensors. After timestamp synchronization, the multimodal data forms a space-time matrix with a unified time base. This matrix is dimensionally arranged according to the spatial position index of the jacking point, constructing a space-time tensor structure containing four characteristic channels: pressure, displacement, inclination, and vibration frequency.

[0027] Based on the real-time data collected by the multi-source sensor array, the edge computing node uses an improved Transformer architecture to extract cross-sensor spatiotemporal features, and generates a real-time virtual mapping of the bridge structure based on digital twin technology. The dynamic control parameters of each jacking point are iteratively calculated in real time through reinforcement learning strategies. The dynamic control parameters include the optimal pressure adjustment parameters and displacement compensation.

[0028] In this embodiment, after receiving the synchronized data from the multi-source sensor array, the edge computing node first extracts cross-sensor spatiotemporal features through the improved Transformer architecture. In the encoder layer, the multi-source sensor data of each lifting point is processed through spatiotemporal position encoding, and a one-dimensional convolution operation is used to perform preliminary feature abstraction on the original input. At the same time, a sinusoidal time encoding term with learnable parameters and a spatial position encoding function are injected to form a coding vector containing spatiotemporal information. Specifically, for the first Lifting points at time Sensor data , through the formula Realize the fusion of time and space dimensions, Indicates the Lifting points at time The encoding vector of is a single-dimensional convolution operation, For the Lifting points at time The raw sensor input data, is the total number of sine functions, is the index, and is a learnable parameter, is the current time point, is the time scale parameter is the dimension of the encoding vector, This is the spatial position encoding function for the jacking point. The convolution kernel dynamically captures local time series patterns, while the sine function term establishes multi-scale temporal associations. The spatial position encoding is generated based on the preset geometric topology of the bridge, ensuring that the encoding vector simultaneously incorporates physical spatial constraints and dynamic temporal evolution characteristics.

[0029] The high-dimensional feature vector output by the encoder then enters the multi-head spatiotemporal attention mechanism. This mechanism expands the feature sequence by sensor channel and time dimensions, generates a query matrix Q, a key matrix K, and a value matrix V through linear projection, and divides them into H independent attention heads. When calculating the weights of each attention head, a spatiotemporal joint weight calculation module is introduced to perform the following operations: ,in For the The output of an attention head, 、 、 Respectively The query, key, and value matrices of the attention heads, 、 、 For the The learnable projection matrix of the attention heads, For transpose calculation, is the subspace dimension, Make exponential decay correction to the timeliness of historical data. is the time attenuation coefficient, the spatial distance penalty term Apply mechanical constraints based on the actual coordinate distance between the lifting points to suppress abnormal correlation between non-adjacent nodes. is the spatial distance penalty coefficient. The attention weights, normalized by Softmax, are weighted and summed with the value matrix to generate a physically interpretable spatiotemporal feature representation. After concatenation and layer normalization, the outputs of each attention head are combined with multi-source sensor data through a residual connection layer for feature fusion, ultimately forming a dynamic load correlation model across jacking points.

[0030] After the decoder generates the initial predicted value of the dynamic control parameter based on the fused spatiotemporal features, the digital twin engine starts the virtual mapping update of the bridge structure synchronously. The engine is based on the three-dimensional finite element model and solves the dynamic equations. Real-time calculation of structural response, including is the mass matrix, is the damping matrix, is the stiffness matrix, is the displacement vector, s is the time variable, and the external load vector The update is driven by the actual data measured by the sensor. The numerical solution process uses implicit The time discretization is performed using the Jacobian matrix, and the nonlinear deformation effect is corrected iteratively through the Jacobian matrix to ensure that the stress distribution and deformation field in the virtual mapping can accurately reflect the mechanical state of the actual bridge.

[0031] In this stage, the reinforcement learning strategy constructs the state space with the virtual mapped stress gradient distribution, real-time pressure difference matrix and historical displacement sequence, and uses the deep Q network framework to optimize the dynamic control parameters. The network input layer receives the spatiotemporal state vector after feature extraction and maps it to the action space through a three-layer fully connected network, that is, the pressure adjustment amount and displacement compensation amount of each lifting point. The reward function in the reinforcement learning strategy is designed as ,in, is the reward value, For the Real-time pressure value of each lifting point, is the total number of jacking points, is the average pressure value, For the lifting point and The height difference, 、 is the index of the lifting point, and is the weight coefficient. During training, a double-delayed deep deterministic policy gradient algorithm is used for policy optimization. An experience replay mechanism balances exploration and exploitation, allowing the control strategy to gradually converge to the optimal solution that minimizes both load deviation and height difference. The control parameters generated in each iteration are verified by the digital twin model and then output to the adaptive PID controller as feedforward input, forming a data-driven closed-loop optimization system.

[0032] An adaptive PID controller is used in coordination with the hydraulic actuator according to real-time dynamic control parameters to achieve load balancing control at multiple lifting points. At the same time, the lifting trajectory deviation is corrected in real time through the vibration feedback loop.

[0033] In this embodiment, an adaptive PID controller receives dynamic control parameters from the edge computing node and generates hydraulic cylinder output commands for each lifting point with a 50ms control cycle. A real-time pressure difference monitoring module continuously calculates the pressure deviation between adjacent lifting points. When the pressure difference between any two points exceeds 15% of the average pressure (the threshold is set at 300 kN in this embodiment), a gradient descent algorithm is triggered to replan the target displacement sequence. This algorithm constructs an objective function based on a weighted combination of the sum of squared pressure differences and the displacement synchronization error. The target displacement for each lifting point is updated through gradient iteration at a learning rate of 0.01. Displacement compensation control is prioritized for lifting points with a lag exceeding 3mm. The compensation amount is 1.2 times the lag amount and is superimposed with a feedforward acceleration component. In pressure-equalizing mode, the hydraulic actuator employs an output limiting strategy, limiting the maximum output of a single point to 120% of the theoretical calculated value. A pressure interlock mechanism across lifting points is implemented via a ring bus protocol.

[0034] The vibration feedback loop collects accelerometer data at a 200Hz sampling rate and extracts the vibration energy distribution in the 2-150Hz frequency band using a fast Fourier transform. When the amplitude of the monitored main frequency (8-12Hz) exceeds the safety threshold of 5m / s², the vibration suppression module generates a reverse compensation signal with an opposite phase and an amplitude attenuation factor of 0.8. This signal is injected into the PID controller via an integral gain correction. This triggers a step-by-step lifting speed reduction strategy, reducing the current lifting rate to 50% of the original rate and activating a second-order Butterworth low-pass filter to smooth the control signal. The compensation process continues until the main frequency energy drops below the threshold. The system then automatically executes a gradient recovery algorithm, gradually increasing the lifting speed to the target at a rate of 5% per cycle until the preset operating parameters are restored. During the vibration suppression phase, the digital twin engine simultaneously updates the modal parameters of the virtual bridge map. The finite element model verifies in real time the impact of the compensation strategy on the structural stress distribution, ensuring that the dynamic adjustment process complies with the bridge's mechanical integrity constraints.

[0035] Example 2: Figure 2As shown, the present invention provides a load balancing control device during the synchronous jacking bridge raising process, comprising: The multi-source sensor array module, consisting of pressure sensors, displacement sensors, three-axis tilt sensors, and acceleration sensors, is deployed at each lifting point and data alignment is achieved through a timestamp synchronization mechanism; The edge computing node module, which includes an improved Transformer processing unit, a digital twin modeling unit, and a reinforcement learning strategy unit, is used to extract cross-sensor spatiotemporal features, generate a real-time virtual map of the bridge, and calculate dynamic control parameters; The adaptive PID controller module receives dynamic control parameters and coordinates with the hydraulic actuator module to adjust the output of each lifting point; Hydraulic actuator module, including multiple sets of hydraulic cylinders and servo valves, performs load-balancing jacking operations; The vibration feedback loop module uses spectrum analysis to correct PID control parameters and lifting speed in real time.

[0036] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A load balancing control method during synchronous jacking bridge raising process, characterized in that: The method comprises: The multi-source sensor array collects real-time pressure data, displacement data, tilt angle and vibration frequency data of each jacking point of the bridge, and transmits them synchronously to the edge computing node; Based on the real-time data collected by the multi-source sensor array, the edge computing node uses an improved Transformer architecture to extract cross-sensor spatiotemporal features. Based on digital twin technology, it generates a real-time virtual map of the bridge structure. It then iterates through a reinforcement learning strategy to calculate the dynamic control parameters of each jacking point in real time. These dynamic control parameters include the optimal pressure adjustment parameters and displacement compensation. An adaptive PID controller is used in coordination with the hydraulic actuator according to real-time dynamic control parameters to achieve load balancing control at multiple lifting points. At the same time, the lifting trajectory deviation is corrected in real time through the vibration feedback loop.

2. The load balancing control method during the synchronous jacking bridge raising process according to claim 1 is characterized in that: The multi-source sensor array consists of a pressure sensor, a displacement sensor, a three-axis tilt sensor and an acceleration sensor, and data alignment is achieved through a timestamp synchronization mechanism.

3. The load balancing control method during the synchronous jacking bridge raising process according to claim 1 is characterized in that: The improved Transformer architecture includes: The encoder layer is used to encode the spatiotemporal position of multi-source sensor data and generate high-dimensional feature vectors; The decoder layer generates initial predictions of dynamic control parameters based on the high-dimensional feature vector; The multi-head spatiotemporal attention mechanism captures the dynamic correlation of load distribution by allocating joint attention weights across sensor channels and time dimensions; The residual connection layer is used to fuse multi-source sensor data with attention-weighted features to obtain cross-sensor spatiotemporal features.

4. The load balancing control method during the synchronous jacking bridge raising process as claimed in claim 3 is characterized in that: The spatiotemporal position encoding of the encoder layer is achieved by the following formula: in Indicates the Lifting points at time The encoding vector of is a single-dimensional convolution operation, For the Lifting points at time The raw sensor input data, is the total number of sine functions, is the index, and is a learnable parameter, is the time scale parameter is the dimension of the encoding vector, It is the encoding function of the spatial position of the lifting point.

5. The load balancing control method during the synchronous jacking bridge raising process as claimed in claim 3 is characterized in that: The multi-head spatiotemporal attention mechanism includes: The high-dimensional feature vector output by the encoder layer is expanded into a spatiotemporal sequence according to the sensor channel and time dimensions. The query matrix Q, key matrix K, and value matrix V are generated through linear projection. Each matrix is divided into H independent attention heads. Each attention head performs the following operations through the spatiotemporal joint weight calculation module: in, For the The output of an attention head, 、 、 Respectively The query, key, and value matrices of the attention heads, 、 、 For the The learnable projection matrix of the attention heads, For transpose calculation, is the subspace dimension, represents the time decay function, represents the spatial distance penalty term; The spatiotemporal joint weight calculation module generates an attention weight distribution with physical constraints by coupling the spatial topological relationship and time series correlation between sensors; The outputs of H attention heads are concatenated in the feature dimension, feature fusion is performed through the feedforward neural network layer, and layer normalization technology is used to stabilize the training process; The multi-head spatiotemporal attention mechanism captures the load transfer patterns at different spatiotemporal scales in parallel, dynamically establishes a mechanical correlation model across jacking points, and achieves robust perception of asymmetric load distribution.

6. The load balancing control method during synchronous jacking bridge elevation as claimed in claim 1, characterized in that: The digital twin technology generates a real-time virtual mapping of the bridge structure including: Establish the bridge dynamic equations based on the three-dimensional finite element model: in, is the mass matrix, is the damping matrix, is the stiffness matrix, is the displacement vector, s is the time variable, External load vector driven by real-time sensor data; The stress distribution and deformation field in the virtual map are updated using the bridge dynamic equations.

7. The load balancing control method during the synchronous jacking bridge raising process according to claim 1 is characterized in that: The state space in the reinforcement learning strategy is the virtual mapped stress gradient distribution, the real-time pressure difference of each lifting point, and the historical displacement sequence. The action space is the pressure adjustment amount and displacement compensation amount of each lifting point. The reward function in the reinforcement learning strategy is designed as: in, is the reward value, For the Real-time pressure value of each lifting point, is the total number of jacking points, is the average pressure value, For the lifting point and The height difference, 、 is the index of the lifting point, and is the weight coefficient.

8. The load balancing control method during the synchronous jacking bridge raising process as claimed in claim 1 is characterized in that: The multi-lifting point load balancing control includes: dynamically allocating the hydraulic cylinder output of each lifting point according to the real-time pressure difference; triggering the gradient descent algorithm to recalculate the target displacement sequence when the pressure difference between any two points exceeds a set threshold; and applying priority compensation control to the lagging lifting points.

9. The load balancing control method during the synchronous jacking bridge raising process as claimed in claim 8, characterized in that: The vibration feedback loop includes: extracting the main frequency component through real-time vibration spectrum analysis. When it is detected that the main frequency energy exceeds the safety threshold, a reverse compensation signal is generated and injected into the integral term of the PID controller, and the lifting speed reduction mode is triggered until the spectrum characteristics return to the allowable range.

10. The load balancing control device during the synchronous jacking bridge raising process is characterized in that: The device comprises: The multi-source sensor array module, consisting of pressure sensors, displacement sensors, three-axis tilt sensors, and acceleration sensors, is deployed at each lifting point and data alignment is achieved through a timestamp synchronization mechanism; The edge computing node module, which includes an improved Transformer processing unit, a digital twin modeling unit, and a reinforcement learning strategy unit, is used to extract cross-sensor spatiotemporal features, generate a real-time virtual map of the bridge, and calculate dynamic control parameters; The adaptive PID controller module receives dynamic control parameters and coordinates with the hydraulic actuator module to adjust the output of each lifting point; Hydraulic actuator module, including multiple sets of hydraulic cylinders and servo valves, performs load-balancing jacking operations; The vibration feedback loop module uses spectrum analysis to correct PID control parameters and lifting speed in real time.

Images