Digital twinning-adaptive assembly correction method and system for prefabricated segments of composite structure

Abstract

The invention relates to the technical field of digital twinning control, and discloses a composite structure prefabricated segment digital twinning-adaptive assembly correction method and system, and the method comprises the steps: reading BIM geometric model data, and constructing an assembly reference coordinate system and a digital twinning geometry of a prefabricated segment; composite data are collected in real time, and end tooth groove boundary feature point cloud is extracted; matching the actually measured point cloud with the digital twinborn geometry through a dynamic point cloud registration technology, calculating a six-degree-of-freedom pose error vector, and generating a predicted total pose error vector in combination with a pre-constructed pose drift prediction model; based on the error vector, a mechanical fine adjustment jack is driven to execute position and posture adjustment until the position and posture are converged, and then tooth groove precise meshing and mechanical locking are completed; according to the method, through the synergistic effect of the dynamic mapping of the digital twinborn model and the self-adaptive correction algorithm, high-precision dynamic correction of the multi-combination structure segment assembly process is achieved on the premise that an original mechanical connection structure is not changed.

Description

Technical Field

[0001] The present invention relates to the field of digital twin control technology, and more specifically, to a digital twin-adaptive assembly correction method and system for prefabricated segments of a composite structure. Background Art

[0002] In the field of precise assembly of prefabricated segments of composite structures, the integration of digital twin technology and real-time posture correction algorithms provides a new path to solve the problems of insufficient precision and low efficiency in traditional assembly. Through dynamic mapping of virtual models and entities, it can realize digital control of the assembly process.

[0003] In the prior art, the Chinese patent application with publication number CN119397844A discloses a method for modeling and calculating the geometric digital twin of an assembly based on deep learning. Through point cloud measurement and neural network training, a mapping model between geometric distribution error and assembly posture is established, focusing on the error modeling of precision mechanical products in static scenes, and providing a data processing basis for high-precision assembly; the Chinese patent application with publication number CN118673619A proposes an online prediction method and system for the assembly accuracy of complex products based on digital twins. By integrating the point cloud geometry model and the deformation mechanism model, a multi-dimensional error digital twin assembly model is constructed, and assembly accuracy prediction is realized based on the Jacobi spinor error transfer model, focusing on the analysis of error transfer laws in the assembly process of complex products.

[0004] However, the above-mentioned existing technologies still have targeted limitations: the deep learning modeling in the Chinese patent application with publication number CN119397844A relies on static point cloud data and batch training samples, and does not involve the prediction and closed-loop correction of real-time pose drift during dynamic assembly, and is difficult to cope with the dynamic changes of six degrees of freedom during the lifting of prefabricated segments of combined structures; although the accuracy prediction model in the Chinese patent application with publication number CN118673619A integrates multi-dimensional errors, it does not construct a real-time point cloud registration closed loop of digital twin geometry and physical segments, and cannot directly convert the prediction error into an instruction to drive the mechanical fine-tuning device, resulting in the cumulative error during the serial assembly of multiple segments cannot be proactively compensated through real-time data processing, especially there is a fault in the link between the tooth groove boundary feature extraction and the jack drive instruction generation, which makes it difficult to meet the dynamic control requirements of millimeter-level assembly accuracy of prefabricated segments of combined structures. Summary of the Invention

[0005] The present invention is applicable to the assembly and construction of prefabricated segments of steel-concrete composite beams, steel-wood composite components, precast concrete composite structures, etc. in the fields of bridges, buildings, municipal administration, etc., especially for the high-precision docking requirements when assembling multiple segments in series, and can achieve real-time correction and locking of the segment posture during the lifting process. In order to overcome the above-mentioned defects of the prior art, the present invention provides a digital twin-adaptive assembly correction method and system for prefabricated segments of composite structures. By constructing a closed-loop control system of digital twin models and real-time perception of composite sensors, high-precision dynamic correction of prefabricated segments of composite structures is achieved, significantly improving assembly accuracy and efficiency. Its integration of dynamic point cloud registration and posture drift prediction technology can effectively eliminate the cumulative errors of multiple segments, ensure accurate docking of connection parts, and provide a reliable solution for the assembled intelligent construction of large-scale composite structures.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Digital twin of prefabricated segments of composite structures - adaptive assembly correction method, including:

[0008] Read BIM geometric model data and establish the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment;

[0009] A composite sensor assembly is deployed at the four corners of the prefabricated segment to collect composite data of the prefabricated segment in real time and extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment;

[0010] Dynamically align the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculate the six-degree-of-freedom pose error vector, and perform pose drift prediction based on the pre-built pose drift prediction model to obtain the predicted total pose error vector;

[0011] Generate driving instructions for the jack based on the predicted total posture error vector and perform posture adjustment;

[0012] Determine whether the posture adjustment reaches the posture convergence condition. If the posture convergence condition is reached, perform precise tooth-slot meshing and mechanical locking.

[0013] Furthermore, the composite sensor combination includes a laser radar sensor and an IMU inertial measurement unit; the composite data of the prefabricated segment includes three-dimensional point cloud data of the prefabricated segment surface and six-degree-of-freedom posture data of the prefabricated segment; the six-degree-of-freedom posture data includes three-axis acceleration data and three-axis angular velocity data;

[0014] The method for extracting the feature point cloud of the tooth groove boundary of the end of the prefabricated segment includes: performing statistical filtering to remove noise from the three-dimensional point cloud data and converting it into the prefabricated segment assembly reference coordinate system to form integrated point cloud data; and automatically extracting the feature point cloud of the tooth groove boundary of the end of the prefabricated segment from the integrated point cloud data using a supervoxel-principal curvature segmentation algorithm.

[0015] Furthermore, the method of automatically extracting the tooth groove boundary feature point cloud of the end of the prefabricated segment from the integrated point cloud data using the supervoxel-principal curvature segmentation algorithm includes:

[0016] The octree decomposition algorithm is used to divide the space of the integrated point cloud data into a multi-scale voxel grid to generate an initial voxel set; the point cloud within the initial voxel set is clustered by local features to form multiple supervoxel units;

[0017] Calculate the covariance matrix of all points inside each supervoxel unit, obtain the curvature values ​​of the three main directions through eigenvalue decomposition, and define the principal curvature κ according to the curvature values ​​of the three main directions;

[0018] Set a curvature threshold and select supervoxel units whose principal curvature κ is greater than the curvature threshold as high-curvature supervoxel units;

[0019] Adjacent high-curvature supervoxel units are aggregated to form a continuous tooth groove boundary feature region, and all point cloud data within the tooth groove boundary feature region are extracted as the tooth groove boundary feature point cloud of the prefabricated segment end.

[0020] Furthermore, the curvature values ​​of the three main directions are the curvature value λ1 of the first main direction, the curvature value λ2 of the second main direction and the curvature value λ3 of the third main direction; the main curvature κ is the ratio of the curvature value of the first main direction to the sum of the curvature values ​​of the three main directions, where λ1≥λ2≥λ3.

[0021] Furthermore, the method for dynamically registering the feature point cloud of the tooth groove boundary of the end portion of the prefabricated segment with the digital twin geometry of the prefabricated segment includes:

[0022] The feature point cloud of the tooth groove boundary at the end of the prefabricated segment is layered by spatial octree to construct a multi-scale point cloud pyramid structure; the multi-scale point cloud pyramid structure includes coarse-scale layer, medium-scale layer and fine-scale layer;

[0023] At the coarse-scale level, the FPFH feature descriptor is used to perform initial registration between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment to obtain the coarse registration transformation matrix T coarse ;

[0024] Based on the coarse registration transformation matrix T coarse, the ICP iterative closest point algorithm is executed at the mesoscale level to optimize the correspondence between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment;

[0025] During the ICP iteration process, the six-degree-of-freedom pose data provided by the IMU inertial measurement unit is integrated and the state is predicted and updated through the EKF extended Kalman filter;

[0026] At the fine-scale level, the point-to-plane distance metric is used instead of the point-to-point distance to output the final precise registration transformation matrix T fine .

[0027] Furthermore, the calculation method of the six-degree-of-freedom posture error vector includes:

[0028] The fine registration transformation matrix T fine and the design target pose matrix T of the prefabricated segment target Perform matrix operations to calculate the posture deviation matrix ΔT;

[0029] Perform singular value decomposition on the pose deviation matrix ΔT to extract the 3×3 rotation matrix R and the 3×1 translation vector t;

[0030] Based on the rotation matrix R, the Rodriguez formula is used to calculate the rotation axis vector n and the rotation angle θ;

[0031] Decompose the rotation axis vector n and the rotation angle θ into the X, Y, and Z axes of the prefabricated segment assembly reference coordinate system to obtain the three rotation components of the posture error;

[0032] Directly extract the three translation components of the pose error from the translation vector t;

[0033] The three rotational components of the pose error and the three translational components of the pose error are combined to form the complete six-DOF pose error vector ε.

[0034] Furthermore, the method for performing posture drift prediction includes:

[0035] Synchronously collect environmental factor data and fuse them with the six-degree-of-freedom posture error vector to form a comprehensive state vector;

[0036] The comprehensive state vector is input into the pre-built pose drift prediction model to obtain the predicted pose error vector.

[0037] Furthermore, the method for obtaining the predicted total pose error vector includes: adding the predicted pose error vector to the six-degree-of-freedom pose error vector of the current position to obtain the predicted total pose error vector.

[0038] Furthermore, the method for determining whether the posture adjustment reaches the posture convergence condition includes:

[0039] Obtain the real-time pose error vector, determine whether the three translation components of the real-time pose error vector have converged, and determine whether the three rotation components of the real-time pose error vector have converged. If both the translation convergence condition and the rotation convergence condition are met, the pose is determined to be converged.

[0040] A digital twin-adaptive assembly correction system for prefabricated segments of a combined structure is used to implement the above-mentioned digital twin-adaptive assembly correction method for prefabricated segments of a combined structure. The system includes:

[0041] Twin model construction module; used to read BIM geometric model data and establish the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment;

[0042] Feature point cloud extraction module: used to deploy a composite sensor combination at the four corner points of the prefabricated segment, collect composite data of the prefabricated segment in real time and extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment;

[0043] Pose prediction module: used to dynamically register the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculate the six-degree-of-freedom pose error vector, and perform pose drift prediction based on the pre-built pose drift prediction model to obtain the predicted total pose error vector;

[0044] Posture adjustment drive module: generates the driving instructions of the jack according to the predicted total posture error vector and performs posture adjustment;

[0045] Posture convergence judgment module: used to judge whether the posture adjustment has reached the posture convergence condition. If the posture convergence condition is reached, the tooth groove precision meshing and mechanical locking are executed.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention, through the combination of composite sensor combination and dynamic point cloud registration technology, can capture the three-dimensional spatial posture deviation of the tooth groove boundary at the end of the prefabricated segment in real time, and realize the forward-looking compensation of the cumulative error in combination with the posture drift prediction model, which significantly improves the accuracy and stability of the serial assembly of multiple segments. The driving instructions of the mechanical fine-tuning jack are generated based on the six-degree-of-freedom posture error vector, which realizes the multi-dimensional coordinated adjustment of the segment's spatial posture, avoiding the local stress concentration and channel eccentricity problems caused by traditional manual knocking or single-point pushing. The software-hardware coupled closed-loop control system integrates the single-segment "scanning-algorithm-fine-tuning-locking" process into an automated operation unit, greatly shortening the assembly time and reducing the risk of manual intervention. At the same time, the posture convergence judgment mechanism ensures the effectiveness of each step of adjustment, and ultimately achieves the precise docking of the tooth groove meshing surface and the reliable anchoring of the mechanical locking device, providing an intelligent solution with both efficiency and precision for the assembly of combined structures in the fields of bridges, buildings, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] 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.

[0049] Figure 1 This is a flow chart of the method for the digital twin-adaptive assembly correction method of the composite structure prefabricated segment in the present invention;

[0050] Figure 2 This is a schematic diagram of the principle of deploying composite sensors at the four corner points of a prefabricated segment according to the present invention;

[0051] Figure 3 A flow chart of the method for dynamically registering the feature point cloud of the tooth groove boundary at the end of a prefabricated segment with the digital twin geometry of the prefabricated segment according to the present invention;

[0052] Figure 4 This is a flow chart of the principle of determining whether posture adjustment has reached posture convergence in the present invention;

[0053] Figure 5 This is a flow chart of the principle of judging whether the precision meshing accuracy has been verified in the present invention;

[0054] Figure 6 This is a flow chart of the principle of determining whether the hole coaxiality accuracy has been verified in the present invention;

[0055] Figure 7 This is a functional module diagram of the digital twin-adaptive assembly correction system for prefabricated segments of the combined structure in the present invention. DETAILED DESCRIPTION

[0056] 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.

[0057] Example 1:

[0058] See also Figure 1 As shown, this embodiment provides a digital twin-adaptive assembly correction method for prefabricated segments of a composite structure, including:

[0059] Step S10, reading the BIM geometric model data and establishing the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment;

[0060] Furthermore, step S10 includes:

[0061] Step S11, reading BIM geometric model data including prefabricated segment 3D geometric information, material properties and assembly constraint relationships and establishing a prefabricated segment assembly reference coordinate system;

[0062] Step S12: construct a three-dimensional virtual space environment based on the prefabricated segment assembly reference coordinate system, load the BIM geometric model data into the virtual space environment, and form a digital twin geometry of the prefabricated segment.

[0063] Specifically, the three-dimensional geometric information of the precast segments includes outline dimensions, tooth shape, and channel location; material properties include density and elastic modulus; and assembly constraints include the connection sequence of each segment and relative position restrictions. Composite structures encompass precast component assemblies in bridges, buildings, and other fields using steel-concrete, steel-wood, and other composite forms. A right-handed coordinate system is established with the composite structure design origin as the virtual coordinate origin, where the X-axis points to the longitudinal direction of the composite structure, such as the extension direction of the bridge; the Y-axis points to the transverse direction of the composite structure, such as the width direction of the bridge; and the Z-axis points vertically upward, that is, perpendicular to the ground. This coordinate system is set because in engineering practice, the assembly of composite structures is mostly carried out in sequence along the longitudinal direction, with the transverse and vertical directions as the main adjustment directions. Using this as a benchmark can provide a unified spatial reference for the assembly process of all precast segments, ensuring that the relative position relationship between different segments can be accurately expressed and avoiding the accumulation of assembly errors caused by inconsistent coordinate systems. After establishing the prefabricated segment assembly reference coordinate system, a three-dimensional virtual space environment is constructed based on this coordinate system. The aforementioned BIM geometric model data is loaded into the constructed three-dimensional virtual space environment to form a digital twin geometry of the prefabricated segment. The construction of the three-dimensional virtual space environment requires simulating the parameters of the real construction environment, including basic physical parameters such as the gravity field and temperature field, so that the virtual environment and the physical construction environment are physically consistent. Loading BIM geometric model data is not a simple import, but requires format conversion, such as converting the commonly used .ifc format to an internal format suitable for the virtual space environment. It also requires accuracy verification to ensure that the deviation between the model's geometric dimensions and the design values ​​is within the allowable range, and topological relationships are reconstructed to clarify the connection relationship between the various components in the model, such as the spatial position association between the tooth groove and the channel, to ensure that the digital twin geometry can accurately map the physical properties and geometric characteristics of the prefabricated segment.

[0064] In the existing technology, traditional static BIM can only perform pre-verification and cannot reflect the assembly process in real time. Different segments may use their own independent coordinate systems, resulting in multiple coordinate conversions during assembly. Each conversion will introduce errors. As the number of segments increases, the errors continue to accumulate, affecting the precise docking of the connection parts. Step S10 establishes a unified prefabricated segment assembly reference coordinate system so that all prefabricated segments use the same coordinate system as a reference during the assembly process, eliminating the source of error caused by coordinate conversion, making the position information of different segments directly comparable, and providing a consistent benchmark for subsequent posture error calculations, solving the problem of easy error superposition in multi-segment serial assembly. At the same time, the construction of digital twin geometry converts the static BIM model into a dynamic model that can be synchronously mapped with the entity in the virtual space, providing an accurate virtual reference body for the subsequent real-time collected point cloud data, so that subsequent steps such as point cloud registration have a clear comparison standard. The unified reference coordinate system is not only applicable to a single type of composite structure, but can also be used in the assembly of composite structures in multiple fields, such as bridges, buildings, and municipal engineering. This is because the definition of its coordinate axes aligns with the main extension direction, horizontal and vertical directions of the structure in each field, eliminating the need to redefine the coordinate system for different fields, thereby improving the versatility of the method. The simulation of physical parameters during the construction of the digital twin geometry ensures that the virtual model is not only geometrically consistent with the entity, but also approximately reflects the entity's state in different environments in terms of physical properties. For example, slight deformations caused by temperature changes can be reflected in the virtual model through corresponding physical parameter adjustments, which provides a basis for subsequent consideration of the impact of environmental factors on the posture. Without step S10, the subsequent point cloud acquisition and registration will lack a unified coordinate reference. The point cloud data of different segments will be in their own coordinate systems and cannot be directly compared to calculate the pose error. The point cloud registration process will become extremely complicated and have low accuracy. At the same time, without the digital twin geometry as a reference, the point cloud data collected in real time lacks a comparison object, and the deviation between the physical segment and the design state cannot be determined. The entire adaptive assembly correction process will lose its foundation, resulting in the inability to achieve the goal of dynamically capturing pose errors and driving the fine-tuning device for correction.

[0065] Step S20: deploying a composite sensor assembly at the four corner points of the prefabricated segment to collect composite data of the prefabricated segment in real time and extracting a feature point cloud of the tooth groove boundary at the end of the prefabricated segment;

[0066] Furthermore, step S20 includes:

[0067] Step S21: deploying a composite sensor combination including a laser radar sensor and an IMU inertial measurement unit at the four corner points of the prefabricated segment to collect composite data of the prefabricated segment in real time. The composite data of the prefabricated segment includes three-dimensional point cloud data of the prefabricated segment surface and six-degree-of-freedom pose data of the prefabricated segment; the six-degree-of-freedom pose data includes three-axis acceleration data and three-axis angular velocity data;

[0068] Step S22 , performing statistical filtering on the three-dimensional point cloud data to remove noise and converting it into a prefabricated segment assembly reference coordinate system to form integrated point cloud data;

[0069] Step S23 : using a supervoxel-principal curvature segmentation algorithm to automatically extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment from the integrated point cloud data.

[0070] Furthermore, step S23 includes:

[0071] Step S231 , using an octree decomposition algorithm to divide the space of the integrated point cloud data into a multi-scale voxel grid to generate an initial voxel set;

[0072] Step S232 , performing local feature clustering on the point cloud within the initial voxel set to form a plurality of supervoxel units;

[0073] Step S233 , calculating the covariance matrix of all points inside each supervoxel unit, and obtaining the curvature values ​​λ1, λ2, and λ3 of the three principal directions through eigenvalue decomposition, where λ1 ≥ λ2 ≥ λ3, and defining the principal curvature κ = λ1 / (λ1+λ2+λ3); where λ1 is the curvature value of the first principal direction, λ2 is the curvature value of the second principal direction, and λ3 is the curvature value of the third principal direction;

[0074] Step S234, setting the curvature threshold κ thre , the supervoxel units with a principal curvature κ greater than the curvature threshold are selected as high-curvature supervoxel units, and the high-curvature supervoxel units correspond to the geometric mutation area of ​​the alveolar boundary;

[0075] In step S235, spatial connectivity analysis is performed on the screened high-curvature supervoxel units, and the adjacent high-curvature supervoxel units are aggregated using a region growing algorithm to form a continuous alveolar boundary feature region. All point cloud data within the alveolar boundary feature region are extracted as the alveolar boundary feature point cloud of the prefabricated segment end.

[0076] Specifically, if Figure 2As shown in the figure, composite sensor assemblies consisting of lidar sensors and inertial measurement units (IMUs) are deployed at the four corners of the prefabricated segment. There are four composite sensor assemblies: composite sensor assembly 1, composite sensor assembly 2, composite sensor assembly 3, and composite sensor assembly 4. The lidar sensors are used to obtain the three-dimensional spatial coordinate information of the prefabricated segment surface, and the IMUs are used to sense the prefabricated segment's motion state. The two work together to achieve omnidirectional monitoring of the prefabricated segment's six-degree-of-freedom posture. The configuration of the four corner points can fully capture the segment's translational and rotational states because the four vertices of the quadrilateral structure uniquely determine the position and orientation of a plane through spatial geometric relationships. Multi-point measurement can reduce the impact of random errors in single-point measurement through data redundancy, thereby improving overall measurement accuracy. When the tower crane begins to lift the precast segment, the four deployed composite sensing units are activated to synchronously acquire 3D point cloud data of the segment's surface. Each lidar sensor generates a point cloud dataset containing tens of thousands of 3D coordinate points in a single scan. The simultaneous operation of the four sensors yields dense point cloud coverage of the segment's surface, with a point cloud density sufficient to capture millimeter-level geometric details. Simultaneously, the inertial measurement unit (IMU) outputs the segment's three-axis acceleration and angular velocity data in real time. The three-axis acceleration data includes linear acceleration along the X-axis, the Y-axis, and the Z-axis, reflecting the segment's translational acceleration along the three coordinate axes, respectively. The three-axis angular velocity data includes angular velocity around the X-axis, the Y-axis, and the Z-axis, reflecting the segment's rotational velocity around the three coordinate axes, respectively. Together, these data constitute the segment's six-degree-of-freedom pose data, fully describing the segment's instantaneous motion.

[0077] After acquiring the 3D point cloud data, it is statistically filtered to remove noise and converted to the prefabricated segment assembly reference coordinate system to form integrated point cloud data. The specific process of statistical filtering noise removal is as follows: for each point in the point cloud data, the average distance to the surrounding points within a set radius is calculated. Points that deviate from the average distance by more than a set distance threshold are identified as noise points and removed. The distance threshold is determined based on the inherent noise level of the lidar sensor. For example, the noise distribution characteristics are obtained through multiple static scanning experiments, and the threshold is set at three times the noise standard deviation to ensure that more than 99% of valid points are retained. Coordinate transformation is achieved by converting the point cloud data in the lidar sensor's own coordinate system to the prefabricated segment assembly reference coordinate system through the extrinsic parameter matrix obtained during the sensor calibration process. The extrinsic parameter matrix is ​​solved by collecting data in a calibration field with known precise coordinates. The calibration field coordinates must be strictly aligned with the prefabricated segment assembly reference coordinate system. The least squares method is used to solve the transformation relationship between the sensor coordinate system and the reference coordinate system to ensure that the converted point cloud data is in the same coordinate frame as the digital twin geometry.

[0078] A supervoxel-principal curvature segmentation algorithm is used to automatically extract feature point clouds of the alveolar boundary at the end of the prefabricated segment from the integrated point cloud data. The supervoxel-principal curvature segmentation algorithm first uses an octree decomposition algorithm to partition the integrated point cloud data space into a multi-scale voxel grid, generating an initial voxel set. The voxel size is selected based on the geometric abruptness of the alveolar boundary. Too small voxels increase the computational complexity, while too large voxels may lose detailed features. Geometric abruptness of the alveolar boundary is typically evident at the millimeter scale; for example, a voxel size of 5 mm can be used. The point clouds within the initial voxel set are then clustered based on local features to form multiple supervoxel units. Each supervoxel unit contains a subset of the point cloud in a local region and its centroid coordinates. The clustering process is based on the spatial distance between voxels and the angle between the point cloud normal vectors. Voxels with close distances and consistent normal vector directions are grouped into a supervoxel unit to preserve local geometric continuity. The covariance matrix of all points within each supervoxel is calculated. This covariance matrix is ​​obtained by statistically calculating the coordinate values ​​of all points within the unit and reflects the distribution characteristics of the point cloud in space. The curvature values ​​of the three principal directions are obtained through eigenvalue decomposition, where the curvature value of the first principal direction is greater than or equal to the second, and the second is greater than or equal to the third. The principal curvature is defined as the ratio of the curvature value of the first principal direction to the sum of the curvature values ​​of the three principal directions. A curvature threshold is set. The curvature threshold is determined by statistical analysis of the principal curvature of a large number of alveolar boundary samples and is set to 0.7, for example. The alveolar boundary region has a larger principal curvature value due to the drastic change in geometric shape, while the principal curvature value of the non-boundary region is smaller. Supervoxel units with a principal curvature greater than the curvature threshold are screened as high-curvature supervoxel units. The spatial connectivity analysis of the screened high-curvature supervoxel units was performed, and the adjacent high-curvature supervoxel units were aggregated using the region growing algorithm to form a continuous alveolar boundary feature region. The adjacency was determined based on the Euclidean distance between the supervoxel centroids being less than the set adjacent space threshold, which was determined based on the average size of the supervoxel units. After aggregation, all point cloud data within the alveolar boundary feature region were extracted as the alveolar boundary feature point cloud of the prefabricated segment end.

[0079] In the existing technology, traditional in-situ correction relies on manual tapping or inefficient jack fine-tuning, and lacks real-time posture monitoring means, resulting in low correction accuracy and time-consuming; simple laser ranging can only obtain position information of a limited number of points, and cannot fully reflect the overall posture and surface features of the segment. Step S20 deploys a combination of composite sensors. The dense point cloud data provided by the lidar can fully reflect the three-dimensional geometric information of the segment surface, and the six-degree-of-freedom posture data provided by the IMU can capture the motion state of the segment in real time. The combination of the two realizes dynamic and comprehensive monitoring of the segment posture, solving the problem of incomplete and delayed information acquisition in traditional methods. Statistical filtering and coordinate transformation ensure the accuracy and uniformity of point cloud data, laying the foundation for subsequent alignment with digital twin geometry. The supervoxel-principal curvature segmentation algorithm can accurately extract tooth-slot boundary features from complex surface point clouds through multi-scale voxel division and principal curvature analysis, avoiding interference from point clouds in non-critical areas and improving the pertinence and efficiency of feature extraction. The four-corner point deployment method of the composite sensor combination is not only applicable to prefabricated segments of regular shapes, but is also effective for irregular composite structures. This is because the four vertices of the quadrilateral can adapt to segments of different cross-sectional shapes through spatial geometric relationships. It is only necessary to adjust the installation angle of the sensor according to the actual contour of the segment, without the need to redesign the sensor layout, which enhances the adaptability of the method.

[0080] Step S20 works in conjunction with the prefabricated segment assembly reference coordinate system established in step S10. The point cloud data after coordinate conversion can be directly compared with the digital twin geometry in the same spatial framework, ensuring the accuracy of the subsequent alignment process. The tooth groove boundary feature point cloud obtained by feature extraction forms an accurate correspondence with the tooth groove model in the digital twin geometry, so that the pose error calculation can focus on the key connection parts, improving the relevance of the error assessment. Without step S20, the subsequent pose error calculation will lack reliable original data support, and it will be impossible to accurately obtain the deviation between the actual pose of the prefabricated segment and the design target pose, resulting in a loss of basis for adaptive correction. At the same time, the lack of tooth groove boundary features will make the alignment process based only on the overall contour, unable to reflect the subtle deviations of the key connection parts, thereby affecting the accuracy and reliability of mechanical locking. Step S20 provides real-time and accurate feature data, enabling the digital twin model to dynamically map the state of the physical segment, providing a key feedback link for closed-loop control, and promoting the transformation of the entire assembly process from static planning to dynamic adjustment.

[0081] Step S30, dynamically aligning the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculating the six-degree-of-freedom pose error vector, and performing pose drift prediction based on the pre-built pose drift prediction model to obtain a predicted total pose error vector;

[0082] Furthermore, step S30 includes:

[0083] Step S31, using a multi-scale ICP-EKF fusion algorithm to dynamically align the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment;

[0084] See also Figure 3 As shown, further, step S31 includes:

[0085] Step S311, performing spatial octree layering on the feature point cloud of the tooth groove boundary at the end of the prefabricated segment to construct a multi-scale point cloud pyramid structure; the multi-scale point cloud pyramid structure includes a coarse-scale layer, a mesoscale layer, and a fine-scale layer;

[0086] Step S312: At the coarse-scale level, the FPFH feature descriptor is used to perform an initial registration between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment to obtain the coarse registration transformation matrix T coarse ;

[0087] Step S313: Based on the coarse registration transformation matrix T coarse , the ICP iterative closest point algorithm is executed at the mesoscale level to optimize the correspondence between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment;

[0088] Step S314: During the ICP iteration process, the six-degree-of-freedom pose data provided by the IMU inertial measurement unit is integrated and state prediction and update are performed through the EKF extended Kalman filter.

[0089] Step S315: At the fine-scale level, point-to-plane distance metric is used instead of point-to-point distance to refine the registration result and output the final refined registration transformation matrix T fine .

[0090] Specifically, the octree layering is achieved by recursively dividing the three-dimensional space of the feature point cloud of the tooth groove boundary at the end of the prefabricated segment into 8 sub-cubes. The side length of each layer of cubes is half of the previous layer, forming a coarse-scale layer, a medium-scale layer and a fine-scale layer. The coarse-scale layer retains the overall outline of the point cloud, the medium-scale layer contains medium details, and the fine-scale layer retains the subtle features of the tooth groove boundary. This structure can reduce the amount of registration calculation while ensuring accuracy. At the coarse-scale level, the FPFH feature descriptor is used to perform initial registration of the feature point cloud and the digital twin geometry. The FPFH feature descriptor is generated by calculating the angle and distance distribution of the normal vector of the point and its neighboring points. It can remain invariant under rotation and translation transformations. The corresponding point pairs are found through feature matching, and the coarse registration transformation matrix is ​​calculated. Based on the coarse registration transformation matrix, an ICP iterative closest point algorithm is executed at the mesoscale level. This algorithm iteratively finds the closest point on the digital twin geometry for each point in the point cloud, calculates the transformation matrix that minimizes the sum of squared distances between corresponding points, and optimizes the correspondence between the point cloud and the model. The number of iterations is determined by the coarse registration error and is typically 10 to 20 to balance efficiency and accuracy. During the ICP iteration, the six-degree-of-freedom pose data provided by the IMU inertial measurement unit is integrated, and state prediction and update are performed using an extended Kalman filter (EKF). The EKF state vector contains the position and pose of the segment. The position of the segment is the coordinate on the X, Y, and Z axes, and the pose is the rotation angle around the X, Y, and Z axes. The prediction step uses the IMU motion model, namely, the integration of acceleration and angular velocity, to estimate the state and covariance matrix at the next moment. The update step uses the pose correction calculated by the ICP as the observation value, and the predicted value and observation value are fused through the Kalman gain to correct the state estimate. This fusion mechanism can compensate for the lag of the ICP algorithm in dynamic scenarios. For example, when a segment swings rapidly due to wind loads, the real-time IMU data can predict the pose change in advance, enabling faster convergence of ICP iterations and avoiding matching errors caused by point cloud motion blur. At the fine-scale level, the point-to-plane distance metric is used instead of the point-to-point distance. The point-to-plane distance is achieved by calculating the perpendicular distance from the point to the corresponding triangular facet. Compared with the point-to-point distance, it is more resistant to noise and local shape deviations. It refines the registration results and outputs the final fine registration transformation matrix, which describes the optimal spatial transformation relationship from the feature point cloud to the digital twin geometry.

[0091] Multi-scale registration can accurately focus on key connection parts, avoid interference from point clouds in non-feature areas, and make the registration error mainly come from the tooth-groove joint surface. When processing dynamic point clouds, the traditional ICP algorithm is prone to convergence to the local optimum due to excessive initial pose deviation. Simply relying on geometric registration cannot cope with real-time movement during the lifting process, resulting in low registration accuracy and time-consuming. The multi-scale strategy uses a coarse-to-fine hierarchical registration to quickly narrow the range of pose deviations and then gradually refine the details, solving the problem of low efficiency and easy divergence of traditional algorithms when searching over a large range. The fusion of EKF and IMU data organically combines geometric information with kinematic information, allowing the registration process to dynamically adapt to the motion state of the segment and remain stable under interference such as wind loads and cable vibrations, improving accuracy compared to pure geometric registration. The multi-scale structure of the octree layer is robust to local defects in the tooth socket boundary. For example, when there are small gaps in the tooth socket caused by construction, the coarse-scale layer's registration is not affected by the local defect, and the fine-scale layer can still maintain overall accuracy by interpolating the surrounding complete features. Traditional single-scale registration can produce significant errors due to local defects. The state prediction mechanism of the EKF can compensate for the inertial motion of the segment in advance. For example, at the moment of starting and stopping the tower crane, the slight sway caused by the segment's inertia can be captured by the IMU and corrected in advance by the EKF, so that the registration result is ahead of the actual posture change, providing time redundancy for subsequent error prediction and correction.

[0092] Step S32, based on the registration result, calculating the six-degree-of-freedom posture error vector ε between the current posture of the prefabricated segment and the design target posture by transformation matrix decomposition;

[0093] Furthermore, step S32 includes:

[0094] Step S321: transform the fine registration transformation matrix T fine and the design target pose matrix T of the prefabricated segment target Perform matrix operations to calculate the pose deviation matrix ;

[0095] Step S322: perform singular value decomposition on the posture deviation matrix ΔT to extract the 3×3 rotation matrix R and the 3×1 translation vector t;

[0096] Step S323, based on the rotation matrix R, use the Rodrigues formula to calculate the rotation axis vector n and the rotation angle θ;

[0097] Step S324 , decomposing the rotation axis vector n and the rotation angle θ into the X, Y, and Z axes of the prefabricated segment assembly reference coordinate system to obtain three rotation components of the posture error;

[0098] Step S325, directly extracting the three translation components of the pose error from the translation vector t;

[0099] Step S326 , combining the three rotational components of the pose error and the three translational components of the pose error to form a complete six-degree-of-freedom pose error vector ε.

[0100] Step S33, synchronously collecting environmental factor data and fusing the data with the six-degree-of-freedom posture error vector to form a comprehensive state vector;

[0101] In step S34, the comprehensive state vector is input into a pre-built pose drift prediction model to obtain a predicted pose error vector, and the predicted pose error vector is added to the six-degree-of-freedom pose error vector of the current position to obtain a predicted total pose error vector.

[0102] Specifically, the design target pose matrix is ​​the theoretical position matrix of the digital twin geometry in the reference coordinate system. The pose deviation matrix is ​​the product of the fine-registration transformation matrix and the inverse of the design target pose matrix, reflecting the spatial transformation relationship between the actual pose and the theoretical pose. The pose deviation matrix is ​​subjected to singular value decomposition to extract a 3×3 rotation matrix and a 3×1 translation vector. Singular value decomposition is achieved by decomposing the matrix into the product of an orthogonal matrix, a diagonal matrix, and an orthogonal matrix. The rotation matrix describes the segment's rotational deviation, while the translation vector describes the segment's translational deviation. Based on the rotation matrix, the Rodriguez formula is used to calculate the rotation axis vector and rotation angle. The Rodriguez formula is implemented by converting the rotation matrix into a rotation vector, with the direction of the rotation axis and the modulus of the rotation angle. The rotation axis vector and rotation angle are decomposed onto the X, Y, and Z axes of the prefabricated segment assembly reference coordinate system to obtain the three rotational components of the pose error. The decomposition process calculates the angle between the rotation axis and each coordinate axis and proportionally distributes the rotation angle to the corresponding axis. The three translation components of the pose error are directly extracted from the translation vector, that is, the projection values ​​of the translation vector on the X, Y, and Z axes. The three rotational components and the three translational components of the pose error are combined to form a complete six-degree-of-freedom pose error vector.

[0103] Environmental factor data is collected synchronously and fused with the six-degree-of-freedom pose error vector to form a comprehensive state vector. This environmental factor data includes parameters such as real-time wind speed data, cable tension data, and crane rotation angle data. Real-time wind speed data is obtained using an anemometer mounted on the crane's top, reflecting the impact of wind loads on segment stability. Cable tension data is measured using tension sensors on the cables, reflecting the load state of the lifting system. Crane rotation angle data is obtained using the crane control system's angle encoder, reflecting changes in the lifting position. After normalization, these data are concatenated with the six-degree-of-freedom pose error vector by dimension, forming a comprehensive state vector that includes pose error and environmental information. Normalization is achieved by converting the data to a normal distribution with a mean of 0 and a standard deviation of 1, ensuring effective fusion of data of varying magnitudes.

[0104] The pose drift prediction model uses a lightweight GNN-Transformer neural network. The GNN graph neural network captures the spatial relationships between segments through node embedding. Node features include segment geometry and historical pose data, while edge features represent the relative positional relationships between segments. The Transformer network uses a self-attention mechanism to process temporal variations and learn the temporal evolution of pose error. The GNN-Transformer neural network is trained using training data from historical assembly processes. This training data includes comprehensive state vectors and corresponding actual pose drift values ​​under different environmental conditions. The network parameters are optimized using a gradient descent algorithm, with the loss function being the mean squared error between the predicted drift and the actual drift value. Given the current comprehensive state vector as input, the neural network outputs a predicted pose error vector for a preset time in the future, such as three seconds in the future. This vector consists of three translational and three rotational components. Each component of the predicted total pose error vector is the sum of the current pose error component and the corresponding predicted drift component, reflecting the total pose deviation after accounting for future drift.

[0105] In existing technologies, traditional static BIM cannot reflect the assembly process in real time. Simple laser ranging or manual correction methods are difficult to handle pose changes in dynamic environments, and errors easily accumulate during multi-segment assembly. Step S30 implements dynamic point cloud registration using a multi-scale ICP-EKF fusion algorithm. This coarse-to-fine registration strategy improves computational efficiency while maintaining accuracy, resolving the conflict between real-time performance and accuracy in dynamic scenarios. The EKF integrates IMU data, reducing the lag in point cloud registration during rapid motion and ensuring that pose calculations more accurately reflect actual motion. The decomposition of pose error vectors is achieved through rigorous matrix operations and geometric transformations, ensuring the physical meaning of the error components is clear and providing a precise quantitative basis for subsequent corrections. The integration of environmental data considers the impact of external factors on segment pose, avoiding prediction bias caused by relying solely on geometric information, and making pose drift prediction more consistent with actual construction environments. The lightweight GNN-Transformer neural network combines spatial correlation with temporal learning to better capture error evolution in complex environments compared to traditional temporal models, improving prediction accuracy. The multi-scale registration strategy is not only applicable to regular features such as tooth grooves, but can also maintain high registration accuracy for segments with surface construction defects. This is because the coarse-scale layer can filter out the impact of local defects, while the fine-scale layer focuses on key feature areas, making the algorithm robust. After environmental data fusion, the pose drift prediction model can adapt to different weather conditions. By learning error patterns under different wind speeds from historical data, it can still provide reliable predictions during sudden wind changes, reducing the correction lag caused by sudden environmental changes. The lightweight design of GNN-Transformer enables the model to run on edge computing devices without relying on cloud computing power, reducing the impact of network instability at the construction site and improving the independence and response speed of the system.

[0106] Dynamic registration is implemented based on a unified coordinate system, ensuring the baseline consistency of pose error calculations. The precise extraction of feature point clouds provides high-quality input for registration, reduces the interference of redundant data, and makes the registration results more reliable. Without step S30, the subsequent jack drive instructions will lack accurate error basis, and the correction process can only be carried out blindly, leading to problems such as insufficient engagement of the connection parts and eccentricity of the channel. At the same time, the inability to predict pose drift will cause the correction action to lag behind the actual error change, forming an "adjustment-drift-readjustment" cycle, reducing construction efficiency. Step S30 establishes a quantification and prediction mechanism for pose errors, dynamically linking the digital twin model with the physical segment, providing a decision-making core for adaptive correction.

[0107] Step S40, generating a driving instruction for the jack according to the predicted total posture error vector, and performing posture adjustment;

[0108] Furthermore, step S40 includes:

[0109] Step S41: Initialize micro hydraulic jack devices with three-degree-of-freedom adjustment capabilities at the four corner points at the bottom of the prefabricated segment; each micro hydraulic jack device has a built-in magnetostrictive displacement sensor for real-time monitoring of the telescopic displacement of the jack piston rod and a pressure sensor for monitoring changes in the oil pressure in the hydraulic cylinder;

[0110] Step S42, converting the predicted total posture error vector into initial driving instructions for four micro hydraulic jack devices through a kinematic inverse solution algorithm;

[0111] Step S43: Sending a drive instruction to the hydraulic control valve array to perform posture adjustment;

[0112] Step S44: during the posture adjustment process, obtain the real-time posture error vector ε real , the real-time pose error vector ε real The adjustment effect deviation is calculated by comparing it with the target posture adjustment amount; if the adjustment effect deviation exceeds the preset adjustment threshold, the conversion of the drive instruction is re-executed to generate the corrected drive instructions for the four micro hydraulic jack devices and send them for execution.

[0113] Specifically, the micro-hydraulic jack device is equipped with a built-in magnetostrictive displacement sensor and a pressure sensor. The magnetostrictive displacement sensor monitors the piston rod's telescopic displacement in real time by measuring the propagation time difference of the magnetostrictive wave. The pressure sensor senses the oil pressure change in the hydraulic cylinder and converts it into an electrical signal, reflecting the supporting force output by the jack. The four jack devices and the main sling system together form a virtual Stewart platform mechanism. The Stewart platform is a six-degree-of-freedom parallel mechanism that achieves precise control of any spatial position by adjusting the length of each branch chain. The four-corner point layout enables each jack to independently undertake horizontal push and pull along the X and Y axes and vertical lifting and lowering along the Z axis. The installation position coordinates are measured by a total station before the prefabricated segment is hoisted and recorded in the digital twin system as the basic parameters of the kinematic model.

[0114] Based on the predicted total posture error vector, it is converted into the driving instructions of four micro hydraulic jack devices through the kinematic inverse solution algorithm. The specific process is: construct the kinematic model of the Stewart platform, establish the conversion relationship between the platform coordinate system and the prefabricated segment assembly reference coordinate system, the platform coordinate system takes the segment center of gravity as the origin, and each axis is parallel to the reference coordinate system; decompose the predicted total posture error vector into the expected displacement and expected rotation of the platform center of mass, the expected displacement is the three translation components of the predicted total posture error vector, and the expected rotation is the three rotation components; based on the inverse kinematics equation of the Stewart platform, according to the jack installation position coordinates and the expected posture parameters, calculate the length change of each jack branch chain, the inverse kinematics equation is established through geometric derivation, and the spatial posture change is mapped to the linear displacement of each branch chain; combined with the mass analysis of the prefabricated segment The distribution and center of gravity position are determined by static analysis, and the supporting force that each jack needs to bear is calculated. The mass distribution is obtained by calculating the material properties and geometric dimensions in the BIM model. The center of gravity position is directly extracted by the 3D modeling software. The supporting force distribution follows the principle of moment balance to ensure that the segment does not overturn during the adjustment process; considering the dynamic response characteristics of the hydraulic system, the flow-displacement response curves under different oil temperatures and pressures are measured experimentally, and the change in branch chain length and the supporting force are converted into target displacement instructions and target pressure instructions for the hydraulic cylinder. The target displacement instruction corresponds to the piston rod extension and contraction, and the target pressure instruction corresponds to the hydraulic pump output pressure; a complete drive instruction set including the timing control strategy is generated. The timing control avoids the additional torque caused by synchronous adjustment by setting the sequence and rate of the actions of each jack, ensuring that the segment is balanced.

[0115] Drive commands are sent to the hydraulic control valve array to perform position adjustment. The hydraulic control valve array includes proportional directional valves, flow control valves, and pressure control valves. The proportional directional valves adjust the flow of hydraulic oil according to the target displacement command, controlling the extension and retraction of the jack. The flow control valves control the piston rod's speed by adjusting the oil flow rate per unit time, ensuring a smooth adjustment process. The pressure control valves limit the maximum oil pressure according to the target pressure command to prevent overload damage to the structure. During the execution process, a magnetostrictive displacement sensor collects the actual displacement of the piston rod in real time, while a pressure sensor collects the actual pressure. This data is transmitted to the control unit via an industrial bus, forming a closed-loop feedback loop. During the posture adjustment process, the real-time posture error vector is obtained, the real-time posture error vector is compared with the target posture adjustment amount, and the adjustment effect deviation is calculated. The real-time posture error vector is obtained by repeatedly executing the point cloud acquisition of step S20 and the alignment calculation of step S30. The target posture adjustment amount is the theoretical adjustment value corresponding to the predicted total posture error vector, and the adjustment effect deviation is the difference between the two. If the adjustment effect deviation exceeds the preset adjustment threshold, the conversion of the drive instruction is re-executed, and the corrected drive instruction is generated and issued. The adjustment threshold is determined through multiple debugging, for example, the adjustment accuracy is tested under different error scenarios, and it is finally set to a translation deviation of no more than 0.1 mm and a rotation deviation of no more than 0.005 degrees. The adjustment threshold ensures the adjustment accuracy and avoids system oscillation caused by frequent corrections.

[0116] In the existing technology, traditional jack fine-tuning relies on manual operation, with low adjustment accuracy and inability to cope with dynamic errors. During multi-segment assembly, additional deformation is easily generated due to uneven force. Step S40 achieves precise six-degree-of-freedom control of posture through the coordination of the three-degree-of-freedom adjustment of the micro-hydraulic jack and the Stewart platform mechanism, solving the problem that traditional single-degree-of-freedom adjustment cannot compensate for complex spatial posture errors. The kinematic inverse algorithm converts abstract posture errors into specific jack action instructions, enabling the virtual adjustments of the digital twin model to be accurately mapped to the physical segments. The real-time feedback and dynamic correction mechanism compensates for the influence of nonlinear factors such as hydraulic system hysteresis and component elastic deformation through a closed-loop sensor data loop, ensuring that the adjustment effect is consistent with the theoretical target. The retractable micro-hydraulic jacks are fully retractable after adjustment. The depth of the recess reserved for their installation location is greater than the jack's maximum extension length. After retraction, they are flush with the bottom surface of the segment, which not only avoids interference with the permanent stress on the structure but also adapts to the continuous assembly of subsequent segments without the need for additional disassembly. The coordinated monitoring of magnetostrictive displacement sensors and pressure sensors can identify the contact status between the segment and the installed structure in real time during the adjustment process. When the pressure sensor detects an abnormal pressure change, it automatically pauses the adjustment and issues an alarm to prevent damage from tooth-slot collision. This passive protection function is not available in traditional adjustment methods. The parallel mechanism characteristics of the Stewart platform make the adjustment force of the four jacks redundant. When a single jack experiences a minor fault, the system can compensate by adjusting the output of the other three jacks to maintain adjustment accuracy and improve the fault tolerance of the construction. Without step S40, the digital twin model's pose error analysis cannot be translated into actual adjustment actions, breaking the entire closed-loop control chain. Segment assembly still relies on manual operation, failing to achieve millimeter-level precision. The lack of a dynamic correction mechanism causes the deviation between the initial adjustment instructions and the actual error to increase over time, ultimately causing problems such as insufficient joint engagement and eccentric holes to recur. By converting error information in the virtual space into adjustment actions in the physical space, step S40 establishes a real-time mapping and data exchange mechanism between the digital twin model and the physical construction process, achieving precise correlation and coordinated control between the virtual and physical spaces.

[0117] Step S50: determine whether the posture adjustment reaches posture convergence. If posture convergence is reached, perform tooth-slot precise meshing and mechanical locking.

[0118] Furthermore, step S50 includes:

[0119] Step S51, see Figure 4As shown, the method for judging whether the posture adjustment reaches posture convergence is as follows: judging whether the three translation components of the real-time posture error vector have translational convergence, and whether the three rotation components of the real-time posture error vector have rotational convergence. If both the translation convergence condition and the rotation convergence condition are met, the posture is judged to be converged; otherwise, the posture is judged to be not converged.

[0120] Step S52: After reaching the convergence of the posture, the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment are used to verify whether the precision meshing accuracy and the coaxiality of the channel meet the set accuracy standards;

[0121] The method for verifying whether the precision meshing accuracy and the hole coaxiality meet the set accuracy standards includes: judging whether the precision meshing accuracy is verified and whether the hole coaxiality accuracy is verified; if the precision meshing accuracy is verified and the hole coaxiality accuracy is verified, it is determined that the set accuracy standards are met.

[0122] See also Figure 5 As shown in FIG, the method for determining whether the precision meshing accuracy has been verified includes: extracting the three-dimensional contour line of the tooth groove contact surface based on the feature point cloud of the tooth groove boundary of the prefabricated segment end after the posture convergence is achieved; calculating the spatial distance between the extracted three-dimensional contour line and the corresponding tooth groove contour of the adjacent installed segment to generate a contact surface gap distribution map; performing statistical analysis on the contact surface gap distribution map and calculating the gap standard deviation as the gap uniformity deviation σ gap , identify the maximum value point in the gap distribution diagram as the maximum local gap, and perform precision meshing accuracy verification; if the gap uniformity deviation is less than the uniformity limit and the maximum local gap is less than the maximum gap limit, the precision meshing accuracy verification is passed.

[0123] See also Figure 6 As shown, the method for determining whether the hole coaxiality accuracy verification has passed includes: extracting the prestressed hole centerline data of the current segment and the adjacent segments from the digital twin geometry of the prefabricated segment, and calculating the hole coaxiality error; verifying the hole coaxiality accuracy based on the hole coaxiality error; if the hole coaxiality error is less than the set coaxiality standard, the hole coaxiality accuracy verification has passed.

[0124] Step S53: If the set accuracy standard is reached, a locking preparation instruction is issued to retract the micro hydraulic jack device. After the retraction is completed, the prefabricated segment is permanently mechanically locked and a locking completion signal is fed back.

[0125] Specifically, the translational convergence condition requires that the absolute values ​​of the three translational components are less than or equal to the set translational error, while the rotational convergence condition requires that the absolute values ​​of the three rotational components are less than or equal to the set rotational error. The translational and rotational errors are determined by engineering practice requirements for assembly accuracy of composite structures. For example, the translational error can be referenced by the minimum clearance requirement for tooth-slot fit in mechanical joints to ensure no interference during mating. The rotational error can be derived based on the allowable deviation of the bore coaxiality to avoid bore misalignment due to rotational deviation. The judgment process also requires temporal stability, requiring that the convergence condition be met for five consecutive data acquisition cycles to prevent misjudgments caused by transient errors that meet the standard but fail to stabilize. The data acquisition cycle is determined based on the sensor sampling frequency to ensure that dynamic fluctuations of the segment are captured. If the pose error does not meet the convergence condition, meaning the pose has not converged, the dynamic correction process continues. If it does, the mechanical locking preparation phase begins.

[0126] After reaching the pose convergence, the tooth groove boundary feature point cloud of the prefabricated segment end and the digital twin geometry are used to verify whether the precision meshing accuracy and the channel coaxiality meet the set accuracy standards. The specific process is: based on the tooth groove boundary feature point cloud after reaching the pose convergence, the three-dimensional contour line of the tooth groove contact surface is extracted. The extraction method is realized by performing three-dimensional curve fitting on the feature point cloud, and the mathematical equation of the contour line is fitted using the least squares method to ensure that the contour line can accurately reflect the actual shape of the tooth groove; the spatial distance between the extracted three-dimensional contour line and the corresponding tooth groove contour of the adjacent installed segment is calculated to generate a contact surface gap distribution map. The spatial distance is calculated as the Euclidean distance of the corresponding point pairs on the contour line, and the distribution map represents the gap with a color gradient. The size of the gap directly reflects the contact uniformity; the contact surface gap distribution diagram is statistically analyzed, and the gap standard deviation is calculated as the gap uniformity deviation. The standard deviation is obtained by calculating the variance of all gap values ​​and then taking the square root, reflecting the degree of dispersion of the gap. The maximum value point in the gap distribution diagram is identified as the maximum local gap, and the precision meshing accuracy is verified; if the gap uniformity deviation is less than the uniformity limit and the maximum local gap is less than the maximum gap limit, the precision meshing accuracy verification is passed, and the uniformity limit is set to 0.2 mm. The basis is that excessive gap unevenness will cause stress concentration when force is applied. The maximum gap limit is set to 0.3 mm, and the effective range of the mechanical locking device is referred to ensure that the gap can be eliminated after locking. The prestressed channel centerline data for the current segment and adjacent segments is extracted from the digital twin geometry. The channel centerline is generated using the channel parameters in the BIM model as a spatial straight line equation. The channel coaxiality error is calculated by evenly selecting several points along the length of the two channels, calculating the distance between the two axis lines at corresponding points, and taking the maximum value as the coaxiality error. If the channel coaxiality error is less than the set coaxiality standard, such as 1 mm, the channel coaxiality accuracy is verified to pass. This standard is based on the diameter of the prestressed steel strand and the allowable deviation, ensuring that the strand does not rub against the channel wall during insertion. If both the precision meshing accuracy and channel coaxiality accuracy are verified to pass, the set accuracy standard is determined.

[0127] If the set accuracy standard is reached, a locking preparation instruction is issued and the micro hydraulic jack device is recovered. During the recovery process, the hydraulic system controls the contraction of the piston rod until the jack is completely retracted into the reserved groove at the bottom of the prefabricated segment. The groove depth is greater than the maximum extension length of the jack to ensure that it does not protrude from the segment surface after recovery. After the recovery is completed, the prefabricated segment is permanently mechanically locked and a locking completion signal is fed back. The mechanical locking uses a locking device that matches the original mechanical connection structure, such as a prestressed tooth groove mechanical locking device. The locking bolt is tightened with a hydraulic wrench until the designed preload force is reached. The preload force value is determined based on the structural force calculation to ensure the connection strength. After permanent mechanical locking, the final assembly pose matrix containing the segment position information and posture information is recorded. The matrix is ​​generated through the coordinates and posture parameters of the current segment in the reference coordinate system; the prefabricated segment assembly reference coordinate system and the digital twin geometry are updated, and the actual pose of the locked segment is included in the reference as a reference for subsequent segment assembly; the cumulative geometric error of the assembled segment sequence is calculated by superimposing the pose error vectors of each segment, monitoring the cumulative coaxial error of the prestressed duct, and ensuring that the overall error is within the allowable range; the accuracy of the deep learning network pose drift prediction is monitored, and by comparing the deviation between the predicted value and the actual value, the incremental learning algorithm is used to optimize the GNN-Transformer neural network model parameters. Incremental learning fine-tunes the network weights by introducing new assembly data to avoid forgetting historical data; after completing data recording and model updating, the updated digital twin model is called as the reference to start the next prefabricated segment assembly task.

[0128] In the existing technology, traditional assembly lacks clear convergence judgment standards, often resulting in insufficient accuracy due to premature locking, or wasting time due to excessive adjustment, and there is a lack of monitoring of error accumulation after mechanical locking, which affects the subsequent segment assembly. Step S50 uses strict convergence judgment, combined with the set translation error and rotation error and time stability requirements, to ensure that the segment is locked after it is stable and meets the standards, avoiding the accuracy rebound caused by instantaneous errors and solving the problem of inaccurate locking timing in traditional methods. The verification of precise meshing and hole coaxiality specifically checks the key accuracy indicators of the connection parts to ensure that the tooth grooves are fully engaged and the hole is connected, providing a reliable foundation for subsequent construction. The model update and error monitoring after mechanical locking enable the digital twin system to reflect the actual status of the assembled structure in real time, provide a dynamic benchmark for the serial assembly of multiple segments, and reduce error superposition. The temporal stability requirement not only ensures the assembly accuracy of the current segment, but also indirectly reflects the degree of influence of environmental interference. If it remains stable for multiple consecutive cycles, it means that the influence of environmental factors such as wind load and cable tension has weakened. At this time, locking can reduce the impact of external disturbances on the connection parts. This implicit judgment of environmental adaptability is not available in traditional static judgment methods. The design of the retractable jack ensures that the permanent stress of the structure is not affected. At the same time, its reserved groove can be used as an observation point for subsequent inspection after locking, which facilitates non-destructive inspection of the connection parts in the later stage and improves the maintainability of the structure. The application of incremental learning algorithms enables the posture drift prediction model to be continuously optimized as the construction progresses, and maintains high-precision predictions under different working conditions. Especially in the later stages of multi-segment assembly, the model's ability to predict cumulative errors is significantly improved, further reducing the number of adjustments.

[0129] Convergence judgment provides a clear termination condition for dynamic adjustment to avoid over- or under-adjustment, while dynamic correction ensures that the error can reach the convergence condition through continuous approximation. The two form a closed loop of "adjustment-judgment-locking", making precision control more reliable. Precision meshing verification is based on the final state of the error vector, combined with the detailed features of the point cloud data. It can better reflect the actual connection quality than relying solely on the error vector, and improves the comprehensiveness of precision verification. Without step S50, the timing of segment locking cannot be scientifically determined, which may lead to substandard connection accuracy or low construction efficiency, and the lack of a model update mechanism will cause subsequent segment assembly to use the initial benchmark, and the cumulative error will continue to expand, ultimately affecting the overall structural performance. Step S50 achieves real-time synchronization between the digital twin model and the physical structure by completing the transition from dynamic adjustment to static locking, providing a key link for closing and continuing the closed-loop control of the entire assembly process, ensuring the consistency and reliability of the multi-segment assembly accuracy.

[0130] Example 2:

[0131] This embodiment provides a digital twin-adaptive assembly correction system for prefabricated segments of a combined structure based on embodiment 1, such as Figure 7 Shown, including:

[0132] Twin model construction module; used to read BIM geometric model data and establish the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment;

[0133] Feature point cloud extraction module: used to deploy a composite sensor combination at the four corner points of the prefabricated segment, collect composite data of the prefabricated segment in real time and extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment;

[0134] Pose prediction module: used to dynamically register the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculate the six-degree-of-freedom pose error vector, and perform pose drift prediction based on the pre-built pose drift prediction model to obtain the predicted total pose error vector;

[0135] Posture adjustment drive module: generates the driving instructions of the jack according to the predicted total posture error vector and performs posture adjustment;

[0136] Posture convergence judgment module: used to judge whether the posture adjustment has reached posture convergence. If posture convergence is reached, precise tooth-slot meshing and mechanical locking are performed.

[0137] In the feature point cloud extraction module, the method of deploying a composite sensor assembly at the four corner points of the prefabricated segment, collecting composite data of the prefabricated segment in real time, and extracting the feature point cloud of the tooth groove boundary at the end of the prefabricated segment includes:

[0138] Step S21: deploying a composite sensor combination including a laser radar sensor and an IMU inertial measurement unit at the four corner points of the prefabricated segment to collect composite data of the prefabricated segment in real time. The composite data of the prefabricated segment includes three-dimensional point cloud data of the prefabricated segment surface and six-degree-of-freedom pose data of the prefabricated segment; the six-degree-of-freedom pose data includes three-axis acceleration data and three-axis angular velocity data;

[0139] Step S22 , performing statistical filtering on the three-dimensional point cloud data to remove noise and converting it into a prefabricated segment assembly reference coordinate system to form integrated point cloud data;

[0140] Step S23 : using a supervoxel-principal curvature segmentation algorithm to automatically extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment from the integrated point cloud data.

[0141] Furthermore, step S23 includes:

[0142] Step S231 , using an octree decomposition algorithm to divide the space of the integrated point cloud data into a multi-scale voxel grid to generate an initial voxel set;

[0143] Step S232 , performing local feature clustering on the point cloud within the initial voxel set to form a plurality of supervoxel units;

[0144] Step S233 , calculating the covariance matrix of all points inside each supervoxel unit, and obtaining the curvature values ​​λ1, λ2, and λ3 of the three principal directions through eigenvalue decomposition, where λ1 ≥ λ2 ≥ λ3, and defining the principal curvature κ = λ1 / (λ1+λ2+λ3); where λ1 is the curvature value of the first principal direction, λ2 is the curvature value of the second principal direction, and λ3 is the curvature value of the third principal direction;

[0145] Step S234, setting the curvature threshold κ thre , the supervoxel units with a principal curvature κ greater than the curvature threshold are selected as high-curvature supervoxel units, and the high-curvature supervoxel units correspond to the geometric mutation area of ​​the alveolar boundary;

[0146] In step S235, spatial connectivity analysis is performed on the screened high-curvature supervoxel units, and the adjacent high-curvature supervoxel units are aggregated using a region growing algorithm to form a continuous alveolar boundary feature region. All point cloud data within the alveolar boundary feature region are extracted as the alveolar boundary feature point cloud of the prefabricated segment end.

[0147] In the pose prediction module, the method for dynamically registering the feature point cloud of the tooth groove boundary of the end portion of the prefabricated segment with the digital twin geometry of the prefabricated segment includes:

[0148] Step S311, performing spatial octree layering on the feature point cloud of the tooth groove boundary at the end of the prefabricated segment to construct a multi-scale point cloud pyramid structure; the multi-scale point cloud pyramid structure includes a coarse-scale layer, a mesoscale layer, and a fine-scale layer;

[0149] Step S312: At the coarse-scale level, the FPFH feature descriptor is used to perform an initial registration between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment to obtain the coarse registration transformation matrix T coarse ;

[0150] Step S313: Based on the coarse registration transformation matrix T coarse , the ICP iterative closest point algorithm is executed at the mesoscale level to optimize the correspondence between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment;

[0151] Step S314: During the ICP iteration process, the six-degree-of-freedom pose data provided by the IMU inertial measurement unit is integrated and state prediction and update are performed through the EKF extended Kalman filter.

[0152] Step S315: At the fine-scale level, point-to-plane distance metric is used instead of point-to-point distance to refine the registration result and output the final refined registration transformation matrix Tfine .

[0153] The methods and systems of the present application may be implemented in many ways. For example, the methods and systems of the present application may be implemented using software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps used in the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified.

[0154] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0155] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Digital twin of prefabricated segments of composite structures - adaptive assembly correction method, characterized by: The method comprises: Read BIM geometric model data and establish the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment; A composite sensor assembly is deployed at the four corners of the prefabricated segment to collect composite data of the prefabricated segment in real time and extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment; Dynamically align the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculate the six-degree-of-freedom pose error vector, and perform pose drift prediction based on the pre-built pose drift prediction model to obtain the predicted total pose error vector; Generate driving instructions for the jack based on the predicted total posture error vector and perform posture adjustment; Determine whether the posture adjustment reaches the posture convergence condition. If the posture convergence condition is reached, perform precise tooth-slot meshing and mechanical locking.

2. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 1 is characterized in that: The composite sensor combination includes a laser radar sensor and an IMU inertial measurement unit; the composite data of the prefabricated segment includes three-dimensional point cloud data of the prefabricated segment surface and six-degree-of-freedom posture data of the prefabricated segment; the six-degree-of-freedom posture data includes three-axis acceleration data and three-axis angular velocity data; The method for extracting feature point clouds of tooth groove boundaries at the end of a prefabricated segment comprises: performing statistical filtering on three-dimensional point cloud data to remove noise and converting the data into a prefabricated segment assembly reference coordinate system to form integrated point cloud data; The supervoxel-principal curvature segmentation algorithm is used to automatically extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment from the integrated point cloud data.

3. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 2 is characterized in that: The method for automatically extracting the tooth groove boundary feature point cloud of the end of the prefabricated segment from the integrated point cloud data using the supervoxel-principal curvature segmentation algorithm includes: The octree decomposition algorithm is used to divide the space of the integrated point cloud data into a multi-scale voxel grid to generate an initial voxel set; the point cloud within the initial voxel set is clustered by local features to form multiple supervoxel units; Calculate the covariance matrix of all points inside each supervoxel unit, obtain the curvature values ​​of the three main directions through eigenvalue decomposition, and define the principal curvature κ according to the curvature values ​​of the three main directions; Set a curvature threshold and select supervoxel units whose principal curvature κ is greater than the curvature threshold as high-curvature supervoxel units; Adjacent high-curvature supervoxel units are aggregated to form a continuous tooth groove boundary feature region, and all point cloud data within the tooth groove boundary feature region are extracted as the tooth groove boundary feature point cloud of the prefabricated segment end.

4. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 3 is characterized in that: The curvature values ​​of the three main directions are respectively the curvature value λ1 of the first main direction, the curvature value λ2 of the second main direction and the curvature value λ3 of the third main direction; the main curvature κ is the ratio of the curvature value of the first main direction to the sum of the curvature values ​​of the three main directions, where λ1≥λ2≥λ3.

5. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 4 is characterized in that: The method for dynamically registering the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment includes: The feature point cloud of the tooth groove boundary at the end of the prefabricated segment is layered by spatial octree to construct a multi-scale point cloud pyramid structure; the multi-scale point cloud pyramid structure includes coarse-scale layer, medium-scale layer and fine-scale layer; At the coarse-scale level, the FPFH feature descriptor is used to perform initial registration between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment to obtain the coarse registration transformation matrix T coarse ; Based on the coarse registration transformation matrix T coarse , the ICP iterative closest point algorithm is executed at the mesoscale level to optimize the correspondence between the feature point cloud of the tooth groove boundary at the end of the prefabricated segment and the digital twin geometry of the prefabricated segment; During the ICP iteration process, the six-degree-of-freedom pose data provided by the IMU inertial measurement unit is integrated and the state is predicted and updated through the EKF extended Kalman filter; At the fine-scale level, the point-to-plane distance metric is used instead of the point-to-point distance to output the final precise registration transformation matrix T fine .

6. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 5 is characterized in that: The calculation method of the six-degree-of-freedom posture error vector includes: The fine registration transformation matrix T fine and the design target pose matrix T of the prefabricated segment target Perform matrix operations to calculate the posture deviation matrix ΔT; Perform singular value decomposition on the pose deviation matrix ΔT to extract the 3×3 rotation matrix R and the 3×1 translation vector t; Based on the rotation matrix R, the Rodriguez formula is used to calculate the rotation axis vector n and the rotation angle θ; Decompose the rotation axis vector n and the rotation angle θ into the X, Y, and Z axes of the prefabricated segment assembly reference coordinate system to obtain the three rotation components of the posture error; Directly extract the three translation components of the pose error from the translation vector t; The three rotational components of the pose error and the three translational components of the pose error are combined to form the complete six-DOF pose error vector ε.

7. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 6 is characterized in that: The method for performing posture drift prediction includes: Synchronously collect environmental factor data and fuse them with the six-degree-of-freedom posture error vector to form a comprehensive state vector; The comprehensive state vector is input into the pre-built pose drift prediction model to obtain the predicted pose error vector.

8. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 7 is characterized in that: The method for obtaining the predicted total pose error vector includes: adding the predicted pose error vector to the six-degree-of-freedom pose error vector of the current position to obtain the predicted total pose error vector.

9. The digital twin-adaptive assembly correction method for prefabricated segments of a composite structure according to claim 8, characterized in that: The method for determining whether the posture adjustment reaches the posture convergence condition includes: Obtain the real-time pose error vector, determine whether the three translation components of the real-time pose error vector have converged, and determine whether the three rotation components of the real-time pose error vector have converged. If both the translation convergence condition and the rotation convergence condition are met, the pose is determined to be converged.

10. A digital twin-adaptive assembly correction system for prefabricated segments of a combined structure, which is used to implement the digital twin-adaptive assembly correction method for prefabricated segments of a combined structure according to any one of claims 1 to 9, characterized in that: The system comprises: Twin model construction module; used to read BIM geometric model data and establish the prefabricated segment assembly reference coordinate system and the digital twin geometry of the prefabricated segment; Feature point cloud extraction module: used to deploy a composite sensor combination at the four corner points of the prefabricated segment, collect composite data of the prefabricated segment in real time and extract the feature point cloud of the tooth groove boundary at the end of the prefabricated segment; Pose prediction module: used to dynamically register the feature point cloud of the tooth groove boundary at the end of the prefabricated segment with the digital twin geometry of the prefabricated segment, calculate the six-degree-of-freedom pose error vector, and perform pose drift prediction based on the pre-built pose drift prediction model to obtain the predicted total pose error vector; Posture adjustment drive module: generates the driving instructions of the jack according to the predicted total posture error vector and performs posture adjustment; Posture convergence judgment module: used to judge whether the posture adjustment has reached the posture convergence condition. If the posture convergence condition is reached, the tooth groove precision meshing and mechanical locking are executed.

Images