Intelligent hoisting control method for bridge tower component reinforcing steel bars

Abstract

The invention relates to a bridge tower component reinforcing steel bar intelligent hoisting control method, which belongs to the technical field of bridge construction, and comprises component reinforcing steel bar butt joint, and specifically comprises the following steps: S601, conical permanent magnet array modules are circumferentially embedded in the end surfaces of component reinforcing steel bars, and annular electromagnet array modules are correspondingly arranged on the end surfaces of embedded reinforcing steel bars; on the basis of the fusion parameter set in the step S2, obtaining the distance value between the embedded steel bar and the component steel bar, and then triggering an electromagnet to be powered on; s602, closed-loop feedback adjustment is conducted, the pose deviation and contact pressure of the component reinforcing steel bars and the embedded reinforcing steel bars are monitored in real time through a multi-mode sensing unit, and data are transmitted back to the edge computing nodes; and S603, electromagnetic parameter dynamic optimization is conducted, the current intensity of the annular electromagnet array is dynamically adjusted according to the pose deviation and the spacing value, and the magnetic attraction force is enhanced. Efficient and accurate butt joint of the component reinforcing steel bars and the embedded reinforcing steel bars is achieved through magnetic butt joint and dynamic current adjustment.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge construction, and in particular relates to an intelligent hoisting control method for componentized steel bars of a bridge tower. Background Art

[0002] Traditional bridge tower reinforcement construction mostly uses on-site binding or segmented hoisting methods, which requires a large number of workers to work at the bridge tower end. Not only is it highly dependent on manual operation, but the construction efficiency is also low. Therefore, a componentized construction method has emerged, in which the steel bars are uniformly bound at the bottom and then hoisted to the bridge tower for installation. This method saves 3-5 days compared to manual binding efficiency.

[0003] There is now a componentized lifting method on the market. Due to the heavy weight and large volume of the tied steel mesh, special crane equipment is required to lift it to the bridge tower. Therefore, BIM-based construction simulation has emerged to achieve intelligent monitoring and resolution of the steel lifting process, transforming the lifting judgment based on manual experience into relying on intelligent monitoring, which can improve the accuracy and safety of steel mesh lifting.

[0004] However, in existing construction, a large number of multi-dimensional reference values ​​are needed to judge the lifting height and posture of the steel mesh, but there is a lack of multi-sensor data fusion and intelligent decision-making.

[0005] In addition, the existing process of lifting modular steel bars requires lifting the modular steel bars from the bottom to the top of the tower to connect with the pre-buried steel bars. However, since multiple sensors are installed in the modular steel bar cage and the density of the modular steel bars is high, manual guidance of the steel bars is usually required during docking. In addition, the area of ​​the modular steel bars is large, and manual guidance of the docking leads to low efficiency and is prone to safety issues.

[0006] Therefore, there is an urgent need for an intelligent lifting control method for the modular steel bars of bridge towers to solve the problem that the existing modular steel bars and pre-embedded steel bars are not intelligent enough and have low efficiency. Summary of the Invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides an intelligent lifting control method for componentized steel bars of bridge towers, which solves the problem that the existing connection between componentized steel bars and embedded steel bars is not intelligent enough and has low efficiency.

[0008] The purpose of the present invention can be achieved through the following technical solutions:

[0009] A method for intelligent lifting and controlling componentized steel bars of a bridge tower, comprising the following contents:

[0010] S1: Obtaining the lifting data of the componentized steel bars through the multimodal sensing unit deployed on the componentized steel bars;

[0011] S2: Based on the edge computing node, the lifting data is temporally and spatially aligned and feature-level fused to obtain the fusion parameter set of componentized steel bars;

[0012] S3: Drives the dual models built on the cloud based on the fusion parameter set and outputs the model difference value:

[0013] S4: Based on the real-time comparison of model difference values ​​by the self-correction module, a model parameter correction value is generated;

[0014] S5: The adaptive path planner generates a dynamic obstacle avoidance path based on the model parameter corrections;

[0015] It also includes the butt joint of componentized steel bars, as follows:

[0016] S601: Embed a conical permanent magnet array module circumferentially on the end face of the componentized steel bar, and a corresponding annular electromagnet array module on the end face of the embedded steel bar. The electromagnet is energized after obtaining the spacing value between the embedded steel bar and the componentized steel bar based on the fusion parameter set in step S2.

[0017] S602: Closed-loop feedback adjustment: Using a multimodal sensing unit, the position deviation and contact pressure between the prefabricated and embedded steel bars are monitored in real time, and the data is transmitted back to the edge computing node.

[0018] S603: Dynamic optimization of electromagnetic parameters. The current intensity of the annular electromagnet array is dynamically adjusted according to the posture deviation and spacing value to enhance the magnetic attraction. The real-time dynamic magnetic attraction is calculated as follows:

[0019] F(t)=ke αd(t) I(t) 2

[0020] d(t) is the real-time distance, I(t) is the dynamic current, k is the magnetic coupling coefficient, and α is the distance sensitivity coefficient.

[0021] Preferably, the outer side of the annular electromagnet array module on the end face of the embedded steel bar is covered with a magnetic shielding layer, the inner diameter of which matches the outer diameter of the electromagnet array; the magnetic shielding layer is used to confine the magnetic field generated by the electromagnet to the docking area, reducing the magnetic induction docking interference of the surrounding steel bars.

[0022] Preferably, the electromagnet array generates a gradient magnetic field with a polarity complementary to that of the permanent magnet array according to a preset magnetic pole arrangement pattern, so that the componentized steel bars generate a rotational self-alignment torque.

[0023] Preferably, the fusion parameter set includes multi-dimensional space posture parameters, stress-strain characteristic parameters, environmental coupling parameters, dynamic response parameter set and time-space correlation parameters.

[0024] Preferably, step S3 further includes the following:

[0025] S301: The dual model consists of a digital twin benchmark model and a reinforcement learning dynamic model running collaboratively on the cloud, wherein:

[0026] The digital twin benchmark model is constructed based on the BIM model and historical hoisting data, and uses a finite element-multibody dynamics coupling algorithm to simulate the rigid-flexible coupled motion characteristics of componentized steel bars in real time.

[0027] The reinforcement learning dynamic model is trained through a deep deterministic policy gradient algorithm to establish an end-to-end mapping relationship between the fusion parameter set and the control quantity of the lifting machinery;

[0028] S302: Model difference value generation adopts a double closed-loop feedback mechanism:

[0029] The first closed loop injects the real-time fusion parameter set into the digital twin benchmark model through the online simulation engine, and outputs the theoretical motion trajectory and stress distribution cloud map;

[0030] The second closed loop sends the control instructions output by the reinforcement learning dynamic model to the lifting mechanical actuator to obtain feedback on the actual motion state.

[0031] Preferably, step S4 includes the following:

[0032] S401: Difference value calculation is based on the dynamic time warping algorithm, which performs multi-dimensional alignment matching on the theoretical motion trajectory and the actual motion trajectory to generate a difference vector including spatial posture deviation, stress excess rate, and energy loss coefficient;

[0033] S402: Establish an online correction channel for model parameters. When any dimension in the difference vector exceeds a preset threshold, an incremental learning mechanism based on a generative adversarial network is triggered to convert the difference vector into a correction value for the boundary conditions of the twin model and an adjustment value for the reinforcement learning reward function.

[0034] S403: The dual model adopts a cross-project transfer learning architecture, uses the knowledge distillation technology to use the trained model parameters as the initialization weights of the new project model, and dynamically updates the model parameter library during the lifting process.

[0035] Preferably, step S5 includes the following contents:

[0036] S501: Global path planning, using the A* algorithm, introduces the structural stress risk coefficient and wind load impact factor output by the digital twin benchmark model to generate the optimal path;

[0037] S502: Establish a dynamic cost map update mechanism, integrate environmental perception data in real time through edge computing nodes, and make progressive corrections to the planned path to ensure that the global path always matches the current working conditions.

[0038] Preferably, a laser scanner is used to construct a point cloud model of the steel bar end face, and ICP point cloud matching is performed with the BIM data of the digital twin benchmark model.

[0039] The beneficial effects of the present invention are:

[0040] This application uses a multimodal sensing unit to collect the position and posture information of steel bars in real time, combined with the prediction and optimization of the cloud-based dual model, to achieve intelligent lifting of steel bars, and further realize the docking of componentized steel bars and embedded steel bars through dynamic current regulation, reducing dependence on manual labor, and achieving accurate and efficient docking of componentized steel bars through magnetic docking. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0042] Figure 1 A flowchart of intelligent lifting of componentized steel bars provided in one embodiment of the present invention;

[0043] Figure 2 A flowchart of connecting modular steel bars and pre-buried steel bars provided in one embodiment of the present invention; DETAILED DESCRIPTION

[0044] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0045] like Figure 1-Figure 2 As shown, a method for intelligent lifting and control of componentized steel bars of a bridge tower includes the following contents:

[0046] S1: Deploy multimodal sensor units on modular steel bars to collect key data during the lifting process in real time, including the spatial position, posture, stress state, and environmental information of the steel bars, and obtain the lifting data of the modular steel bars; these data provide the basis for subsequent intelligent control.

[0047] S2: The collected data is transmitted to the edge computing node for spatiotemporal alignment and feature-level fusion. Through algorithmic processing, the data of different modalities, namely spatial position, posture, stress state, and environmental information, are correlated and integrated to generate a fusion parameter set for componentized steel bars, effectively improving data utilization efficiency and accuracy.

[0048] S3: Based on the fused parameter set, it drives the dual models built in the cloud to perform calculations. The dual models include a digital twin baseline model and a reinforcement learning dynamic model. Through the collaborative work of the dual models, the model difference values ​​are output, providing a basis for subsequent parameter correction:

[0049] S4: The self-correction module compares model difference values ​​in real time and generates model parameter correction values ​​according to a preset algorithm. It can dynamically adjust model parameters to ensure the accuracy and stability of the lifting process.

[0050] S5: The adaptive path planner generates a dynamic obstacle avoidance path based on model parameter corrections and real-time environmental information. This path can avoid obstacles during the lifting process and ensure that the componentized rebar reaches the designated location safely and efficiently.

[0051] Based on the above, traditional methods are prone to problems such as inaccurate docking and large deviations during the rebar docking process due to the lack of precise sensing and control methods. By using a multimodal sensing unit to collect real-time information on the position and posture of the rebar, combined with the prediction and optimization of the cloud-based dual model, intelligent lifting of rebar is achieved.

[0052] The existing jointing of modular rebars relies on manual visual guidance and requires the collaboration of multiple people. To address the visual obstruction caused by the high density and large area of ​​modular rebars in traditional manual jointing, magnetic pre-guiding is used to replace manual rough positioning, including the jointing of modular rebars. The details are as follows:

[0053] S601: electromagnetic coupling pre-docking;

[0054] Conical permanent magnet array design: A conical permanent magnet array module is embedded circumferentially on the end surface of the componentized steel bar. The conical structure is used to achieve radial focusing of the magnetic lines of force and enhance the concentration of magnetic attraction at the end.

[0055] Annular electromagnet array response: A circular electromagnet array module is set up corresponding to the end face of the embedded steel bar. After obtaining the spacing value between the two through the fusion parameter set in step S2, the electromagnet is triggered to energize. At this time, the conical permanent magnet and the annular electromagnet form a non-contact magnetic coupling field, which preliminarily achieves the posture pre-calibration of the componentized steel bar. By replacing manual coarse positioning with magnetic pre-guiding, it can solve the visual occlusion problem caused by the high density and large area of ​​the componentized steel bar in traditional manual guidance docking.

[0056] S602: Closed-loop feedback regulation;

[0057] Multimodal sensing unit: Integrates a pressure sensor array and a six-degree-of-freedom posture sensor on the end face of the componentized steel bar to monitor the contact pressure distribution and spatial posture deviation of the docking surface in real time. Edge computing node processing: Sensor data is transmitted back to the edge computing node, and multi-source information is integrated through the Kalman filter algorithm to generate posture correction instructions. This overcomes the local perception blind spots caused by the centralized deployment of sensors in traditional manual docking and realizes closed-loop control of the posture of the entire cross-section.

[0058] S603: Dynamic optimization of electromagnetic parameters. The current intensity of the annular electromagnet array is dynamically adjusted according to the posture deviation and spacing value to enhance the magnetic attraction. The real-time dynamic magnetic attraction is calculated as follows:

[0059] F(t)=ke αd(t) I(t) 2

[0060] d(t) is the real-time distance, I(t) is the dynamic current, k is the magnetic coupling coefficient, and α is the distance sensitivity coefficient;

[0061] The formula for calculating the real-time dynamic magnetic attraction describes the dynamic relationship between the magnetic attraction, current, and distance. Its core is to achieve precise control of the magnetic attraction by adjusting the current I(t) in real time. The specific adjustment logic is as follows:

[0062] Long-distance stage: When the distance between the prefabricated steel bars and the embedded steel bars is large, the system generates a strong magnetic attraction with a fixed high current to quickly shorten the distance;

[0063] Close-range stage: PID control is used to dynamically adjust the current according to the real-time posture deviation to ensure that the magnetic attraction force is sufficient to attract the component steel bars while avoiding posture loss control due to excessive magnetic force, until the component steel bars are connected to the embedded steel bars;

[0064] In summary, this application realizes the intelligent lifting of steel bars by collecting the position and posture information of steel bars in real time through multimodal sensing units, and combines the prediction and optimization of cloud-based dual models. It further realizes the docking of componentized steel bars and embedded steel bars through dynamic current regulation, reduces dependence on manual labor, and realizes accurate and efficient docking of componentized steel bars through magnetic docking.

[0065] Due to the characteristics of the steel bars themselves, the annular electromagnet array module will also generate a certain attraction to the steel bars. Therefore, in order to make the steel bars in the preset docking area generate suction force, while reducing the influence of suction force on the steel bars in the non-magnetic docking area;

[0066] In one embodiment, the magnetic shielding layer is made of high magnetic permeability material (such as silicon steel, Permalloy, etc.), and its magnetic permeability is much higher than that of the surrounding non-magnetic area. After the electromagnet is energized, the surrounding steel bars generate additional attraction due to induced magnetization, causing the componentized steel bars to shift. However, the magnetic shielding layer provides a low magnetic resistance path, so that the magnetic flux lines form a closed loop along the inside of the shielding layer. The magnetic layer forms a low magnetic resistance path, which concentrates the magnetic field lines to the end faces of the steel bars in the preset docking area, reducing the diffusion of the magnetic field to non-target areas. The inner diameter of the magnetic shielding layer is precisely matched with the outer diameter of the electromagnet array to form a ring-shaped closed magnetic circuit, further limiting the escape of the magnetic field. The magnetic shielding layer of this embodiment can reduce the influence of the steel bars in the non-magnetic docking area through magnetic field directional constraint and magnetic circuit optimization, significantly improve docking accuracy and system reliability, and reduce energy consumption and interference risks.

[0067] In one embodiment, the electromagnet array generates a gradient magnetic field with complementary polarity to that of the permanent magnet array according to a preset magnetic pole arrangement pattern. On the end face of the embedded rebar, the annular electromagnet array is arranged in a circular pattern with alternating polarity (NSNS), forming a complementary polarity relationship with the conical permanent magnet array (SNSN) alternating polarity on the end face of the component rebar. By controlling the current direction of each electromagnet coil, the magnetic field forms a gradient distribution in the radial direction (for example, the outer electromagnets generate a stronger magnetic field), thereby forming a non-uniform magnetic field around the component rebar. When the component rebar deviates from its posture, the magnetic torque drives it to rotate around its axis until the magnetic torque aligns with the magnetic field, achieving automatic alignment.

[0068] In summary, this embodiment greatly shortens the time required to connect a single steel bar by means of a rotational self-alignment torque; and realizes automatic rotational alignment of componentized steel bars by means of magnetic pole complementarity and gradient magnetic field control of the electromagnet array, fundamentally solving the efficiency, precision, and safety problems of traditional connection technology.

[0069] In one embodiment, the fusion parameter set includes multi-dimensional spatial posture parameters collected by inertial measurement units deployed at the four corners of the componentized steel bars, including three-dimensional coordinate deviation and dynamic change rate of pitch angle / roll angle, and the spatial posture solution is realized using a quaternion fusion algorithm;

[0070] The stress and strain characteristic parameters are acquired by an embedded fiber Bragg grating sensor array, which is distributed in a spiral pattern along the axis of the main reinforcement. The shear force-bending moment coupling characteristic spectrum of the key section is extracted by the strain modal decomposition algorithm.

[0071] Environmental coupling parameters include real-time environmental parameters obtained by temperature and humidity composite sensors and anemometers deployed on the hoisting interface, and a wind-structure dynamic interference model based on computational fluid dynamics is constructed;

[0072] The dynamic response parameter set adopts a multi-physics field joint analysis method, and through the fusion of the time-frequency domain characteristics of the vibration acceleration sensor and the acoustic emission sensor, the transfer function matrix between the structural resonance mode and the hoisting mechanical excitation is established;

[0073] The time-space correlation parameters are synchronized through the distributed clock mechanism of the edge nodes to establish the timestamp mapping relationship of the multi-sensor data, and the time-space evolution prediction model of the lifting trajectory is constructed based on the improved LSTM network;

[0074] Compared with traditional control methods that rely only on a single parameter or a simple combination of parameters, this embodiment achieves all-round perception of the spatial position, structural status, and environmental impact of componentized steel bars through multi-dimensional parameter fusion. Through multi-parameter collaborative optimization, the control error is reduced to meet the requirements of high-precision bridge construction.

[0075] In one embodiment, step S3 further includes the following:

[0076] S301: The dual model consists of a digital twin benchmark model and a reinforcement learning dynamic model running collaboratively in the cloud, where:

[0077] The digital twin benchmark model is based on the BIM model and constructed with historical lifting data. It uses the finite element-multibody dynamics coupling algorithm to simulate the rigid-flexible coupled motion characteristics of componentized steel bars in real time.

[0078] The reinforcement learning dynamic model is trained through a deep deterministic policy gradient algorithm to establish an end-to-end mapping relationship between the fusion parameter set and the control quantity of the lifting machinery;

[0079] The digital twin benchmark model is used as a virtual simulation, and the reinforcement learning dynamic model is used for real lifting parameters. The lifting is first run through virtual simulation, and the virtual lifting data is output to the reinforcement learning model for learning. The simulated lifting process of componentized steel bars can be applied to actual lifting. By inputting fusion parameters, the six-degree-of-freedom control quantities of the lifting machinery, such as translation in the X, Y, and Z directions, rotation in the R, P, and Y directions, and auxiliary parameters such as hydraulic system pressure and motor speed, are output to monitor the lifting parameters of componentized lifting.

[0080] S302: Model difference value generation adopts a double closed-loop feedback mechanism:

[0081] The first closed loop injects the real-time fusion parameter set into the digital twin benchmark model through the online simulation engine, and outputs the theoretical motion trajectory and stress distribution cloud map;

[0082] The second closed loop sends the control instructions output by the reinforcement learning dynamic model to the lifting mechanical actuator to obtain feedback on the actual motion state.

[0083] In one embodiment, step S4 includes the following:

[0084] S401: Use the dynamic time warping algorithm to nonlinearly align the theoretical motion trajectory output by the digital twin model with the actual execution feedback trajectory, and calculate the difference between the two. This difference is mainly reflected in two aspects;

[0085] Spatial posture deviation: By comparing the position and posture of the theoretical trajectory and the actual trajectory in three-dimensional space, the specific deviation values ​​are calculated, such as lateral deviation, longitudinal deviation, vertical deviation, and rotation deviation around each axis;

[0086] Stress Exceedance Rate: Count the percentage of time during the actual lifting process when the stress of the componentized steel bars exceeds the material safety threshold to assess the safety of the lifting process. Based on this, a difference vector containing information such as spatial posture deviation and stress exceedance rate is generated.

[0087] S402: When the deviation of any dimension in the difference vector exceeds a preset threshold, the system will initiate an incremental learning mechanism based on a generative adversarial network (GAN) to correct and update the model parameters;

[0088] S403: To improve the efficiency and accuracy of model training for the new project, this embodiment adopts a cross-project transfer learning architecture and migrates the trained model parameters to the new project model through knowledge distillation technology. The specific process is as follows:

[0089] Knowledge distillation: The trained digital twin benchmark model and reinforcement learning model are used as teacher models, and the new project initialization model is used as the student model. The knowledge of the teacher model is transferred to the student model through the distillation loss function, enabling the student model to quickly achieve a high performance level.

[0090] Dynamic parameter library update: During the hoisting process, parameter sets, such as posture offsets, are continuously stored and regularly uploaded to the cloud knowledge base to form a cross-project parameter matrix. When a new project is initialized, appropriate parameters can be called from the cloud parameter library for initialization based on project requirements, thereby accelerating the model training process for the new project.

[0091] In one embodiment, step S5 includes the following:

[0092] S501: The traditional A* algorithm only considers path length and obstacle avoidance. However, this application introduces the structural stress risk coefficient output by the digital twin benchmark model to evaluate the stress impact of the path on the steel structure and the wind load impact factor, as well as the interference of real-time wind pressure on lifting stability, and dynamically adjust the planned path.

[0093] S502: Collect environmental perception data in real time through edge computing nodes, such as changes in obstacle position, wind speed fluctuations, and equipment posture offsets, and dynamically update the cost map. For example, when a gust of wind causes the hoisted object to swing, the edge node immediately adjusts the path curvature to avoid collision with obstacles.

[0094] In one embodiment, a laser scanner collects point cloud data of the steel bar end faces in real time, the data format including three-dimensional coordinates and reflection intensity; the theoretical geometric information of the steel bars to be connected is extracted from the bridge tower BIM model and converted into a point cloud format to generate a theoretical point cloud model;

[0095] ICP point cloud matching: Perform a rough match between the measured point cloud and the theoretical point cloud to obtain the initial rotation matrix and translation vector. It iterate the initial rotation matrix and translation vector and calculate the final matching error mean and standard deviation to evaluate the matching quality.

[0096] Based on the ICP matching results, the position deviation between the measured and theoretical componentized steel bars is calculated, and a correction trajectory is generated to ensure that the componentized steel bars are hoisted along the optimal path.

[0097] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for controlling the intelligent lifting of componentized steel bars of a bridge tower, characterized in that: Includes the following: S1: Obtaining the lifting data of the componentized steel bars through the multimodal sensing unit deployed on the componentized steel bars; S2: Based on the edge computing node, the lifting data is temporally and spatially aligned and feature-level fused to obtain the fusion parameter set of componentized steel bars; S3: Drives the dual models built on the cloud based on the fusion parameter set and outputs the model difference value: S4: Based on the real-time comparison of model difference values ​​by the self-correction module, a model parameter correction value is generated; S5: The adaptive path planner generates a dynamic obstacle avoidance path based on the model parameter corrections; It also includes the butt joint of componentized steel bars, as follows: S601: Embed a conical permanent magnet array module circumferentially on the end face of the componentized steel bar, and a corresponding annular electromagnet array module on the end face of the embedded steel bar. The electromagnet is energized after obtaining the spacing value between the embedded steel bar and the componentized steel bar based on the fusion parameter set in step S2. S602: Closed-loop feedback adjustment: Using a multimodal sensing unit, the position deviation and contact pressure between the prefabricated and embedded steel bars are monitored in real time, and the data is transmitted back to the edge computing node. S603: Dynamic optimization of electromagnetic parameters. The current intensity of the annular electromagnet array is dynamically adjusted according to the posture deviation and spacing value to enhance the magnetic attraction. The real-time dynamic magnetic attraction is calculated as follows: F(t)=the αd(t) I(t) 2 d(t) is the real-time distance, I(t) is the dynamic current, k is the magnetic coupling coefficient, and α is the distance sensitivity coefficient.

2. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 1 is characterized in that: The outer side of the annular electromagnet array module on the end face of the embedded steel bar is covered with a magnetic shielding layer, the inner diameter of which matches the outer diameter of the electromagnet array; the magnetic shielding layer is used to confine the magnetic field generated by the electromagnet to the docking area, reducing the magnetic induction docking interference of the surrounding steel bars.

3. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 1 is characterized in that: The electromagnet array generates a gradient magnetic field with a polarity complementary to that of the permanent magnet array according to a preset magnetic pole arrangement pattern, so that the componentized steel bars generate a rotational self-alignment torque.

4. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 1 is characterized in that: The fusion parameter set includes multi-dimensional space posture parameters, stress and strain characteristic parameters, environmental coupling parameters, dynamic response parameter set and time-space correlation parameters.

5. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 1 is characterized in that: Step S3 also includes the following: S301: The dual model consists of a digital twin benchmark model and a reinforcement learning dynamic model running collaboratively on the cloud, wherein: The digital twin benchmark model is constructed based on the BIM model and historical hoisting data, and uses a finite element-multibody dynamics coupling algorithm to simulate the rigid-flexible coupled motion characteristics of componentized steel bars in real time. The reinforcement learning dynamic model is trained through a deep deterministic policy gradient algorithm to establish an end-to-end mapping relationship between the fusion parameter set and the control quantity of the lifting machinery; S302: Model difference value generation adopts a double closed-loop feedback mechanism: The first closed loop injects the real-time fusion parameter set into the digital twin benchmark model through the online simulation engine, and outputs the theoretical motion trajectory and stress distribution cloud map; The second closed loop sends the control instructions output by the reinforcement learning dynamic model to the lifting mechanical actuator to obtain feedback on the actual motion state.

6. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 5 is characterized in that: The step S4 includes the following contents: S401: Difference value calculation is based on the dynamic time warping algorithm, which performs multi-dimensional alignment matching on the theoretical motion trajectory and the actual motion trajectory to generate a difference vector including spatial posture deviation, stress excess rate, and energy loss coefficient; S402: Establish an online correction channel for model parameters. When any dimension in the difference vector exceeds a preset threshold, an incremental learning mechanism based on a generative adversarial network is triggered to convert the difference vector into a correction value for the boundary conditions of the twin model and an adjustment value for the reinforcement learning reward function. S403: The dual model adopts a cross-project transfer learning architecture, uses the knowledge distillation technology to use the trained model parameters as the initialization weights of the new project model, and dynamically updates the model parameter library during the lifting process.

7. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 6 is characterized in that: The step S5 includes the following contents: S501: Global path planning, using the A* algorithm, introduces the structural stress risk coefficient and wind load impact factor output by the digital twin benchmark model to generate the optimal path; S502: Establish a dynamic cost map update mechanism, integrate environmental perception data in real time through edge computing nodes, and make progressive corrections to the planned path to ensure that the global path always matches the current working conditions.

8. The intelligent lifting control method for componentized steel bars of a bridge tower according to claim 7 is characterized in that: A laser scanner is used to construct a point cloud model of the steel bar end face, and ICP point cloud matching is performed with the BIM data of the digital twin benchmark model.

Images