High-precision installation control method for hyperboloid cold-bent glass curtain wall

Abstract

The present application relates to the technical field of building curtain wall construction, in particular to a high-precision installation control method based on a hyperbolic cold-bending glass curtain wall, comprising: obtaining a target hyperbolic curtain wall theoretical design model, the model containing a glass panel spatial curved surface equation, support structure coordinates and a theoretical joint width; then collecting actual surface point cloud data of the main structure by a three-dimensional laser scanning device, and spatially registering the data with the theoretical design model to obtain a construction installation error distribution map; using an improved point cloud matching algorithm with curvature feature constraints to perform high-precision matching of the actual point cloud and ideal point cloud of the support structure, calculate the three-dimensional spatial position correction amount of each support point, and generate a site leveling plan containing three-dimensional leveling instructions for each support point in combination with the error distribution map and the correction amount. The method realizes high-precision installation control of the hyperbolic cold-bending glass curtain wall, and optimizes the installation precision and panel adhesion effect of the curtain wall.

Description

Technical Field

[0001] This invention relates to the field of building curtain wall construction technology, and in particular to a high-precision installation control method based on hyperboloid cold-bent glass curtain walls. Background Technology

[0002] During the installation and construction of hyperboloid cold-formed glass curtain walls, 3D laser scanning devices are often used to collect point cloud data of the main structure surface. The point cloud is then matched with the theoretical design model using a conventional iterative nearest-point algorithm. A leveling scheme for the supporting structure is then developed based on the main structure's construction error data to complete the installation and positioning of the glass panels. Conventional techniques rely solely on the basic iterative nearest-point algorithm for point cloud matching, without considering the characteristics of the curved surface structure and setting constraints. The leveling schemes are often based on overall adjustments, without refining the adjustment parameters down to individual support points.

[0003] Conventional iterative nearest-point algorithms are prone to matching deviations during point cloud matching, failing to accurately adapt to the spatial morphology of hyperboloid support structures. The calculation results for the 3D spatial position correction of support points are inaccurate, and the construction and installation errors of the main structure cannot be linked to the adjustment parameters of the support points. On-site leveling operations lack corresponding 3D leveling commands for individual support points, making it difficult to meet the leveling accuracy requirements of hyperboloid cold-bent glass curtain walls. The width of glass panel joints deviates from the theoretical design value, and the overall installation fit of the curtain wall fails to meet the standards for spatial curved surface layout.

[0004] During the installation control of hyperboloid cold-bent glass curtain walls, there are problems such as insufficient matching accuracy due to the lack of curvature feature constraints in point cloud matching, and the inability to combine construction and installation errors with support point corrections, making it difficult to generate three-dimensional leveling commands for individual support points. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a high-precision installation control method based on hyperboloid cold-bent glass curtain walls.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a high-precision installation control method based on hyperboloid cold-formed glass curtain walls, comprising:

[0007] Obtain the theoretical design model of the target hyperboloid curtain wall. The theoretical design model includes the spatial surface equation of the glass panel, the spatial coordinates of the supporting structure, and the theoretical width of the joint between the panels.

[0008] The actual surface point cloud data of the existing main structure is collected using a three-dimensional laser scanning device, and the actual surface point cloud data is spatially registered with the theoretical design model to calculate the construction and installation error distribution map of the main structure.

[0009] An improved point cloud matching algorithm is invoked to perform high-precision matching between the actual surface point cloud data and the ideal point cloud of the support structure in the theoretical design model, and to calculate the three-dimensional spatial position correction of each support point. The improved point cloud matching algorithm is based on the principle of the iterative nearest point algorithm and introduces curvature feature constraints.

[0010] Based on the construction and installation error distribution map and the three-dimensional spatial position correction amount, an on-site leveling scheme for the support structure is generated, and the on-site leveling scheme includes three-dimensional leveling instructions for each support point.

[0011] As a further aspect of the present invention, the working principle of the improved point cloud matching algorithm includes:

[0012] Local curvature features of each point are extracted from the actual surface point cloud data, and ideal local curvature features of the corresponding points are extracted from the ideal point cloud of the supporting structure.

[0013] Based on the similarity between the local curvature features and the ideal local curvature features, one or more candidate corresponding points in the ideal point cloud of the supporting structure are initially matched for each point in the actual surface point cloud data.

[0014] In each iteration of the iterative nearest point algorithm, not only is the spatial distance error between corresponding point pairs calculated, but also the curvature feature difference between corresponding point pairs is calculated.

[0015] The spatial distance error and the curvature feature difference are weighted and fused to form a joint optimization objective function. The optimal spatial transformation parameters are solved by minimizing the joint optimization objective function.

[0016] After obtaining new spatial transformation parameters in each iteration, the matching weights of candidate corresponding points are dynamically adjusted based on the curvature feature differences to reduce the interference of outliers on the matching results.

[0017] When the value of the joint optimization objective function converges or the number of iterations reaches the upper limit, the iteration stops, and the optimal spatial transformation parameters obtained in the last iteration are applied to the actual surface point cloud data, thereby outputting the three-dimensional spatial position correction.

[0018] As a further aspect of the present invention, the step of using a three-dimensional laser scanning device to collect actual surface point cloud data of the existing main structure, and spatially registering the actual surface point cloud data with the theoretical design model to calculate the construction and installation error distribution map of the main structure includes:

[0019] The three-dimensional laser scanning device is controlled to scan the existing main structure from multiple stations to obtain multiple initial point cloud data.

[0020] The multiple initial point cloud data are denoised and stitched together to generate complete actual surface point cloud data.

[0021] In the theoretical design model, select theoretical feature points with obvious characteristics, including structural corner points and the center point of embedded parts;

[0022] In the actual surface point cloud data, geometric feature recognition technology is used to identify the actual feature points corresponding to the theoretical feature points;

[0023] Using the theoretical feature points and the actual feature points, calculate the optimal rigid body transformation matrix that aligns the actual surface point cloud data to the coordinate system of the theoretical design model;

[0024] The actual surface point cloud data after transformation by the optimal rigid body transformation matrix is ​​compared with the theoretical surface in the theoretical design model point by point to calculate the distance, thereby generating the construction and installation error distribution map that characterizes the magnitude of the deviation at each position.

[0025] As a further aspect of the present invention, based on the construction and installation error distribution map and the three-dimensional spatial position correction amount, a field leveling scheme for the support structure is generated, including:

[0026] Identify the critical areas in the construction and installation error distribution map where the deviation exceeds the allowable tolerance, and locate the supporting structures corresponding to the critical areas;

[0027] Extract the correction data of the support structure corresponding to the key area in the three-dimensional spatial position correction amount. The correction data includes two translation components in the horizontal plane and one translation component in the vertical direction.

[0028] Based on the correction data, calculate the thickness adjustment value of the bottom leveling shim of the support structure and the in-plane fine adjustment amount of the top connector;

[0029] The thickness adjustment value and the in-plane fine-tuning amount are combined with the construction procedures and adjustable component specifications to form a set of leveling instructions that includes the operation sequence and quantitative parameters, thus forming the on-site leveling scheme.

[0030] As a further aspect of the present invention, the method further includes:

[0031] Based on the adjusted support structure, the flat glass is gradually bent using a cold bending forming process, and the strain data of the glass panel is collected in real time during the bending process. The bending force and process are dynamically adjusted according to the strain data and the preset strain-curvature mapping relationship.

[0032] An optical positioning target is set up on a curved glass panel. The optical positioning target is tracked and measured in real time using a total station network. The actual coordinates of the measured target are compared with the corresponding theoretical coordinates in the theoretical design model to generate positioning deviation data of the glass panel.

[0033] Based on the positioning deviation data, the six-degree-of-freedom fine-tuning mechanism is driven to adjust the spatial pose of the glass panel until the positioning deviation data is less than a preset threshold, and then the final locking installation of the glass panel is performed.

[0034] As a further aspect of the present invention, based on the adjusted support structure, the flat glass is gradually bent using a cold bending forming process, and strain data of the glass panel is collected in real time during the bending process. The bending force and process are dynamically adjusted according to the strain data and a preset strain-curvature mapping relationship, including:

[0035] Position the flat glass at the predetermined installation position on the adjusted support structure, and temporarily fix its edges to the support points;

[0036] Multiple resistance strain gauges are attached to predetermined positions on a flat glass plate to form a strain monitoring network;

[0037] Start the actuator of the cold bending equipment so that the pressure head of the actuator contacts and pushes the glass panel at the initial pressure and speed, causing it to begin to produce elastic deformation;

[0038] During the pushing process, the strain values ​​of each monitoring point are collected in real time through the strain monitoring network. The real-time collected strain values ​​are input into the preset strain-curvature mapping relationship to deduce the current real-time curvature distribution of the glass panel.

[0039] The real-time curvature distribution at the same location is compared with the target curvature of the theoretical design model of the target hyperboloid curtain wall, and the curvature deviation is calculated.

[0040] Based on the magnitude and distribution of the curvature deviation, the thrust and speed of the actuator are adjusted in real time until the deviation between the real-time curvature distribution and the target curvature enters the allowable range.

[0041] As a further aspect of the present invention, the step of arranging optical positioning targets on a curved glass panel, using a total station network to perform real-time tracking and measurement of the optical positioning targets, and comparing the measured actual coordinates of the targets with the corresponding theoretical coordinates in the theoretical design model to generate positioning deviation data for the glass panel includes:

[0042] At least three non-collinear edge locations of the glass panel that has been cold-bent and whose stress is basically stable, optical positioning targets with unique codes are installed.

[0043] A total station measurement network consisting of at least three automatic tracking total stations is established at the construction site, and the total station measurement network is systematically oriented and calibrated.

[0044] The total station measurement network is controlled to simultaneously aim at and measure the three-dimensional coordinates of the center points of at least three optical positioning targets, and the actual spatial coordinates of each target are obtained by back intersection calculation.

[0045] From the theoretical design model, read the theoretical design coordinates corresponding to the edge positions where each optical positioning target is installed;

[0046] For each optical positioning target, the difference between its actual spatial coordinates and the theoretical design coordinates is calculated along each coordinate axis to form the position deviation vector of the optical positioning target.

[0047] By combining the positional deviation vectors of all optical positioning targets, the translation and rotation parameters describing the overall spatial pose deviation of the glass panel are calculated and fitted, thus forming the positioning deviation data.

[0048] As a further aspect of the present invention, based on the positioning deviation data, a six-degree-of-freedom fine-tuning mechanism is driven to adjust the spatial pose of the glass panel, including:

[0049] The translation and rotation parameters in the positioning deviation data are input into the control system of the six-degree-of-freedom fine-tuning mechanism;

[0050] The control system converts the translation and rotation parameters into extension and retraction commands for each electric cylinder in the six-degree-of-freedom fine-tuning mechanism.

[0051] Execute the telescopic command to cause the six-degree-of-freedom fine-tuning mechanism to drive the glass panel it adsorbs to produce translational and rotational motion in three-dimensional space;

[0052] During the movement, the total station measurement network continuously tracks and measures the actual spatial coordinates of the optical positioning target, and feeds back the updated actual spatial coordinates to the control system in real time;

[0053] The control system recalculates the real-time positioning deviation data based on the updated actual spatial coordinates, and generates new extension / retraction commands based on the recalculated real-time positioning deviation data, until the values ​​of translation and rotation parameters in the feedback positioning deviation data are both less than the preset allowable threshold.

[0054] As a further aspect of the present invention, the method further includes a step of verifying the accuracy of the seam between adjacent glass panels after the glass panels are finally locked and installed:

[0055] On both sides of the joint between the finally locked and installed glass panel and the adjacent installed glass panel, multiple joint measurement points are selected at equal intervals.

[0056] Use a digital caliper or laser rangefinder to measure the actual seam width at each seam measurement point;

[0057] Obtain the theoretical width of the seam at the corresponding position from the theoretical design model;

[0058] Calculate the difference between the actual joint width and the theoretical joint width at each joint measurement point to obtain a joint width deviation sequence. If the maximum value in the joint width deviation sequence exceeds the joint width tolerance, record the information of the out-of-tolerance point and assess whether further adjustments are needed.

[0059] As a further aspect of the present invention, the method further includes a step of predictively compensating for the subsequent installation of the glass panel based on the joint width deviation sequence:

[0060] Statistical analysis was performed on the seam width deviation sequence at all seam measurement points of the currently installed multiple glass panels to calculate the average deviation trend and dispersion of the seam width.

[0061] Based on the aforementioned average deviation trend and dispersion, the systematic bias and random fluctuation range of construction errors are determined;

[0062] Before installing the next glass panel, the theoretical installation position of the glass panel in the theoretical design model is pre-compensated and corrected based on its location and the systematic bias, thus generating a pre-compensated installation position.

[0063] After positioning and installation according to the pre-compensated installation position, the joint verification step is performed again, and the statistical information of the joint width deviation sequence is updated according to the new verification results for subsequent predictive compensation of the glass panel.

[0064] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0065] Based on the principle of the iterative nearest point algorithm, an improved point cloud matching algorithm is formed by introducing curvature feature constraints. The actual surface point cloud data of the main structure is matched with the ideal point cloud of the support structure in the theoretical design model. By relying on the curvature feature constraints to fit the spatial surface features of the hyperboloid support structure, abnormal matching points in the point cloud data are eliminated, and the spatial deviation in the point cloud matching process is reduced. The calculated three-dimensional spatial position correction of the support point can accurately correspond to the actual layout state of the support structure, and the correction value corresponds to the spatial position deviation of the support structure.

[0066] By integrating the error distribution map of the main structure construction and installation with the correction amount of the three-dimensional spatial position of the support points, an on-site leveling scheme containing three-dimensional leveling instructions for each support point is generated. The leveling instructions can directly correspond to the adjustment parameters of a single support point. The leveling operation can accurately adapt to the error distribution state of different areas of the main structure. The fit between the spatial position of the support structure after leveling and the theoretical design model is improved. The width of the glass panel joint can fit the theoretical design width parameters. The fit between the double-curved cold-bent glass panel and the support structure can match the spatial curved surface design shape. Attached Figure Description

[0067] Figure 1 This is a flowchart of the high-precision installation control method for hyperboloid cold-bent glass curtain walls described in this invention;

[0068] Figure 2 A flowchart illustrating the work done to improve the point cloud matching algorithm;

[0069] Figure 3 A flowchart generated for the construction and installation error distribution diagram. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0071] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0072] See Figure 1 This invention provides a high-precision installation control method for hyperboloid cold-formed glass curtain walls, the method comprising the following steps:

[0073] A theoretical design model of the target hyperboloid curtain wall is obtained, which includes the spatial surface equation of the glass panels, the spatial coordinates of the supporting structure, and the theoretical width of the joints between the panels. A 3D laser scanning device is used to collect actual surface point cloud data of the existing main structure, and this data is spatially registered with the theoretical design model to calculate the construction and installation error distribution map of the main structure. An improved point cloud matching algorithm is used to perform high-precision matching between the actual surface point cloud data and the ideal point cloud of the supporting structure in the theoretical design model, calculating the 3D spatial position correction amount for each support point. Based on the construction and installation error distribution map and the 3D spatial position correction amount, an on-site leveling scheme for the supporting structure is generated, which includes 3D leveling instructions for each support point.

[0074] In one embodiment of the present invention, the improved point cloud matching algorithm operates as follows: (See attached document) Figure 2 The algorithm extracts local curvature features from the actual surface point cloud data and ideal local curvature features from the ideal point cloud of the supporting structure. Based on the similarity between the local curvature features and the ideal local curvature features, it initially matches one or more candidate corresponding points in the ideal point cloud of the supporting structure for each point in the actual surface point cloud data. In each iteration of the iterative nearest point algorithm, it calculates the spatial distance error between corresponding point pairs and the curvature feature difference between corresponding point pairs. It then weights and fuses the spatial distance error and curvature feature difference to form a joint optimization objective function, and solves for the optimal spatial transformation parameters by minimizing this function. After obtaining new spatial transformation parameters in each iteration, it dynamically adjusts the matching weights of the candidate corresponding points based on the curvature feature difference. When the value of the joint optimization objective function converges or the number of iterations reaches the upper limit, the iteration stops, and the optimal spatial transformation parameters obtained in the last iteration are applied to the actual surface point cloud data to output the three-dimensional spatial position correction.

[0075] In practical implementation, the improved point cloud matching algorithm uses actual surface point cloud data containing several sampling points. The ideal point cloud of the supporting structure originates from discrete sampling of the supporting structure surface in the theoretical design model, with each point possessing three-dimensional coordinate attributes in the point cloud. Local curvature features are extracted from each sampling point in the actual surface point cloud data, obtained by calculating the rate of change of the normal vector within the point's neighborhood. Ideal local curvature features corresponding to the sampling point's position are extracted from the ideal point cloud of the supporting structure, pre-calculated and stored based on the geometric properties of the theoretical design model. Based on the numerical similarity between local and ideal local curvature features, for each sampling point in the actual surface point cloud data, candidate corresponding points with curvature differences less than a set threshold are searched in the ideal point cloud of the supporting structure. Each sampling point is matched with one or more candidate corresponding points, forming a preliminary candidate matching relationship.

[0076] In some embodiments, in a single iteration of the iterative nearest point algorithm, for each established candidate corresponding point pair, the Euclidean distance between the two points is calculated as the spatial distance error, and the magnitude of the difference in local curvature features between the two points is calculated as the curvature feature difference. The spatial distance error and the curvature feature difference are combined through a weighting factor to construct a joint optimization objective function to guide the matching process. The expression of the joint optimization objective function is as follows:

[0077]

[0078] in: This represents the total error of the joint optimization objective function. This represents the first point in the actual surface point cloud data. One sampling point, This represents the candidate corresponding point in the ideal point cloud of the supporting structure that matches it. This represents the Euclidean distance between two points. Indicates sampling point The local curvature eigenvalues, Indicate candidate points The ideal local curvature eigenvalue, and These are the weighting coefficients for the spatial distance error term and the curvature feature difference term, respectively. The values ​​of the weighting coefficients are determined based on the noise level and curvature sensitivity of the point cloud data. The optimal spatial transformation parameters, composed of the rotation matrix and translation vector, are solved by minimizing the joint optimization objective function, so that the transformed actual surface point cloud data and the ideal point cloud of the supporting structure simultaneously approximate each other in terms of spatial distance and curvature features.

[0079] In practice, after obtaining new spatial transformation parameters in each iteration, the curvature feature difference of each candidate corresponding point pair is recalculated. If the curvature feature difference of a point pair exceeds a set tolerance limit, the matching weight of that point pair is reduced to minimize its impact on subsequent iterations; conversely, the matching weight is increased. By dynamically adjusting the matching weight of candidate corresponding points, interference from abnormal points caused by surface attachments or local deformation of the main structure is suppressed. When the change in the value of the joint optimization objective function is less than the convergence threshold after multiple iterations, or when the total number of iterations reaches a preset upper limit, the iteration process is terminated. The spatial transformation parameters obtained in the last iteration are taken as the final result. These spatial transformation parameters describe the spatial offset of the actual surface point cloud data relative to the theoretical design model, and the three-dimensional spatial position correction of each support point is derived accordingly.

[0080] In one embodiment of the present invention, see [reference] Figure 3The system controls a 3D laser scanning device to scan the existing main structure from multiple stations, acquiring multiple initial point cloud data. These initial point cloud data are then denoised and stitched together to generate complete actual surface point cloud data. Distinctive theoretical feature points, including structural corner points and embedded component center points, are selected from the theoretical design model. Geometric feature recognition technology is used to identify the actual feature points corresponding to the theoretical feature points in the actual surface point cloud data. Using the theoretical and actual feature points, the optimal rigid body transformation matrix is ​​calculated to align the actual surface point cloud data to the coordinate system of the theoretical design model. Finally, the distance between the actual surface point cloud data transformed by the optimal rigid body transformation matrix and the theoretical surface in the theoretical design model is calculated point-by-point to generate a construction and installation error distribution map. Identify key areas in the construction and installation error distribution map where deviations exceed allowable tolerances, and locate the corresponding support structures. Extract correction data for the support structures corresponding to the key areas in three-dimensional spatial position correction, including two translation components in the horizontal plane and a translation component in the vertical direction. Based on the correction data, calculate the thickness adjustment value of the leveling shims at the bottom of the support structure, as well as the in-plane fine-tuning amount of the top connector. Combine the thickness adjustment value and the in-plane fine-tuning amount with the construction procedures and adjustable component specifications to compile a set of leveling instructions containing operation sequences and quantitative parameters, forming an on-site leveling scheme.

[0081] In practical implementation, the 3D laser scanning device employs a multi-station scanning mode to collect data from the existing main structure. Each station has an independent instrument position and scanning angle, acquiring multiple initial point cloud data pieces covering different areas of the main structure. After outlier removal and noise filtering, these initial point cloud data pieces are stitched together based on the same-name features of common overlapping areas to generate continuous and complete actual surface point cloud data. In the theoretical design model, easily identifiable structural corner points and embedded part center points are selected as theoretical feature points. These theoretical feature points have clear geometric definitions and stable spatial positions. In the actual surface point cloud data, a combination of curvature extremum detection and geometric shape matching is used to identify actual feature points corresponding to the theoretical feature points. The number of actual feature points is kept consistent with the number of theoretical feature points to ensure matching reliability.

[0082] In some embodiments, the optimal rigid body transformation matrix is ​​calculated using the correspondence between the theoretical feature point set and the actual feature point set, employing a least-squares fitting method. This optimal rigid body transformation matrix includes rotation and translation components, used to transform the actual surface point cloud data from the local scan coordinate system to the theoretical design model coordinate system. The transformed actual surface point cloud data is then compared point-by-point with the theoretical surface in the theoretical design model. The distance value at each point reflects the construction and installation error at that location. The distance values ​​of all points constitute an error value set, and the spatial visualization result of this error value set is the construction and installation error distribution map. This map visually displays the unevenness and deviation of the main structure surface.

[0083] It is understandable that areas in the construction and installation error distribution map where deviations exceed allowable tolerances are identified as critical areas. The criterion for determining a critical area is that the error value exceeds a set tolerance threshold. When locating the supporting structure corresponding to the critical area, the critical area is mapped to nearby support points or support components based on the relative positional relationship between the supporting structure and the main structure in the theoretical design model. Correction data for the supporting structure corresponding to the critical area in three-dimensional spatial position correction is extracted. This correction data includes the X-axis translation component, Y-axis translation component in the horizontal plane, and the Z-axis translation component in the vertical direction. The positive or negative sign of the translation component indicates whether the direction of movement is the same as or opposite to the positive direction of the coordinate axis.

[0084] Optionally, the in-plane fine-tuning amount of the top connector of the support structure is calculated based on the horizontal translation component. The in-plane fine-tuning amount is equal to the projection value of the horizontal translation component in the local coordinate system of the connector. The thickness adjustment value of the leveling shim at the bottom of the support structure is calculated based on the vertical translation component. The thickness adjustment value is equal to the vertical translation component divided by the elastic compression coefficient of the shim material. The thickness adjustment value and the in-plane fine-tuning amount are arranged according to the construction sequence, combined with the standard specifications of the shims and the adjustment range of the connectors, into a set of leveling instructions containing operation steps, adjustment directions, and quantitative values. The set of leveling instructions constitutes the core content of the on-site leveling scheme. On-site construction personnel perform precise adjustments to the support structure according to the set of leveling instructions.

[0085] In practice, the output of the leveling instruction set is in the form of a structured table or a digital list. Each instruction corresponds to one support point or a group of related support points. The instruction content includes the support point number, the increase or decrease in the thickness of the bottom shim, and the lateral and longitudinal displacement of the top connector. When multiple support points are involved in the same critical area, the leveling instructions for the support points located upstream of the structural load-bearing path are prioritized to avoid secondary interference during the adjustment process. After all support points have been adjusted, the actual surface point cloud data is re-collected for verification to confirm that the deviation values ​​in the construction and installation error distribution map are all within the allowable tolerance range. At this point, the on-site leveling plan is completed.

[0086] In one embodiment of the present invention, based on the adjusted support structure, a cold bending process is used to gradually bend the flat glass, and strain data of the glass panel is collected in real time during the bending process. The bending force and process are dynamically adjusted according to the strain data and a preset strain-curvature mapping relationship. The flat glass is positioned at a predetermined installation position on the adjusted support structure, and its edges are temporarily fixed to the support points. Multiple resistance strain gauges are attached to the preset positions of the flat glass to form a strain monitoring network. The actuator of the cold bending equipment is activated, so that the pressure head of the actuator contacts and pushes the glass panel at the initial pressure and speed, causing it to undergo elastic deformation. During the pushing process, the strain value of each monitoring point is collected in real time through the strain monitoring network. The real-time collected strain value is input into the preset strain-curvature mapping relationship to deduce the current real-time curvature distribution of the glass panel. The real-time curvature distribution at the same position is compared with the target curvature of the theoretical design model of the target hyperboloid curtain wall to calculate the curvature deviation. According to the magnitude and distribution of the curvature deviation, the magnitude and speed of the actuator's thrust are adjusted in real time until the deviation between the real-time curvature distribution and the target curvature is within the allowable range.

[0087] In practice, during cold bending operations based on the adjusted support structure, the flat glass is hoisted to its predetermined installation position on the support structure. Temporary clamps are used to fix the edges of the flat glass to the support points, ensuring initial positioning stability. Multiple resistance strain gauges are attached to preset positions on the surface of the flat glass, evenly distributed in the center and edge areas of the glass panel. All resistance strain gauges are connected to a data acquisition device via wires to form a strain monitoring network, used to capture the strain changes of the glass in real time during the bending process. After the actuator of the cold bending equipment is started, the pressure head of the actuator slowly contacts the surface of the glass panel with a set initial pressure and initial speed, applying a force perpendicular to the glass plane, causing the flat glass to begin elastic deformation. During the advancement of the pressure head, the strain monitoring network collects the real-time strain values ​​of each monitoring point at a fixed sampling frequency, and the real-time strain values ​​are transmitted to the control unit for processing.

[0088] In some embodiments, the control unit incorporates a pre-defined strain-curvature mapping relationship. This pre-calibrated relationship, established through materials mechanics experiments and finite element simulations, establishes a correspondence between strain values ​​and curvature values ​​for specific glass materials and thicknesses. The real-time acquired strain values ​​are substituted into the pre-defined strain-curvature mapping relationship to deduce the real-time curvature distribution of the glass panel in its current state. This real-time curvature distribution is stored in matrix form, with each element corresponding to the curvature value of a grid node on the glass surface. The real-time curvature distribution at the same location is compared with the target curvature of the theoretical design model of the target hyperboloid curtain wall. The target curvature is directly extracted from the curvature field derived from the spatial surface equation of the theoretical design model. The curvature deviation is calculated as the absolute difference between the real-time curvature value and the target curvature value, reflecting the degree of deviation between the current bending shape and the design target. Based on the magnitude and distribution characteristics of the curvature deviation, the control unit generates adjustment commands for the actuator: when the curvature deviation is large, the propulsion force is increased and the propulsion speed is reduced to enhance the precision of deformation control; when the curvature deviation is small, the propulsion parameters are maintained or fine-tuned to stabilize the forming process; this adjustment is repeated until the deviation between the real-time curvature distribution and the target curvature falls entirely within the allowable range, at which point the cold bending forming stage is complete. It can be understood that the mathematical expression of the preset strain-curvature mapping relationship can take different forms depending on the material properties and stress characteristics of the glass. In one specific implementation, for cold bending deformation within the linear elastic range, a linear mapping relationship is used as follows:

[0089]

[0090] in: The real-time curvature vector representing the glass panel. This represents the real-time strain vector acquired by the strain monitoring network. Represents the strain-curvature transformation matrix. The elements are determined by the elastic modulus, Poisson's ratio, and geometric dimensions of the glass, and specific values ​​are determined through preliminary calibration tests. Optionally, to visually demonstrate the strain-curvature relationship at key stages of the cold bending process, monitoring data at typical moments are recorded and compiled into the table below (see Table 1). The data in the table are only illustrative values, reflecting the correlation between strain and curvature:

[0091] Table 1: Schematic diagram of strain and calculated curvature at monitoring points during cold bending process

[0092]

[0093] In practical implementation, the actuator's thrust adjustment follows the gradient descent principle, and the amount of thrust adjustment... The adjustment amount of the propulsion speed is directly proportional to the average curvature deviation and is related to the rate of change of curvature deviation. The control unit updates the set values ​​of propulsion force and propulsion speed in real time, driving the hydraulic or electric servo system to execute parameter changes, ensuring that the bending process of the glass panel always progresses in a direction approaching the target curvature. When the curvature deviation at all monitoring points is less than the allowable threshold, the actuator stops propulsion and maintains the current pressure for a period of time. After the internal stress of the glass tends to balance, the temporary fixation is released, completing the cold bending forming operation.

[0094] In one embodiment of the present invention, optical positioning targets with unique codes are installed at at least three non-collinear edge positions of a glass panel that has completed cold bending and whose stress is basically stable; a total station measurement network consisting of at least three automatic tracking total stations is established at the construction site, and the network is systematically oriented and calibrated; the total station measurement network is controlled to simultaneously aim at and measure the three-dimensional coordinates of the center points of at least three optical positioning targets, and the actual spatial coordinates of each target are obtained through resection calculation; the design theoretical coordinates corresponding to the edge positions of each installed optical positioning target are read from the theoretical design model; for each optical positioning target, the difference between its actual spatial coordinates and the design theoretical coordinates in each coordinate axis direction is calculated to form a point position deviation vector; by combining the point position deviation vectors of all optical positioning targets, the translation parameters and rotation parameters describing the overall spatial pose deviation of the glass panel are fitted and calculated to form positioning deviation data. The translation and rotation parameters from the positioning deviation data are input into the control system of the six-degree-of-freedom fine-tuning mechanism. This system converts the translation and rotation parameters into extension and retraction commands for each electric cylinder in the six-degree-of-freedom fine-tuning mechanism. Executing the extension and retraction commands causes the six-degree-of-freedom fine-tuning mechanism to move the glass panel it is attached to in three-dimensional space, producing translational and rotational motions. During the motion, the total station measurement network continuously tracks and measures the actual spatial coordinates of the optical positioning target and feeds back the updated actual spatial coordinates to the control system in real time. The control system recalculates the real-time positioning deviation data based on the updated actual spatial coordinates and generates new extension and retraction commands based on the recalculated real-time positioning deviation data until the values ​​of the translation and rotation parameters in the feedback positioning deviation data are both less than the preset allowable threshold.

[0095] In practice, the cold-bent and stress-stabilized glass panel enters the installation and positioning stage. Uniquely coded optical positioning targets are installed at at least three non-collinear edge positions. The mounting bases of the optical positioning targets are fixed to the glass surface using vacuum suction cups or adhesive, ensuring that the relative position of the center of the optical positioning target to the reference point at the glass edge is known. A total station measurement network consisting of at least three automatic tracking total stations is deployed on the construction site. The total stations are distributed in an encircling pattern around the curtain wall installation area. The system orientation and calibration of the total station measurement network are completed by measuring known reference points, unifying the measurement coordinate system of all total stations. The total station measurement network is controlled to synchronously aim at at least three optical positioning targets. The automatic tracking function ensures that the measurement beam remains aligned with the center of the optical positioning targets as they move. Each total station measures the horizontal and vertical angles of the center of the optical positioning targets. The three-dimensional actual spatial coordinates of each optical positioning target in the global coordinate system are obtained through multi-station back intersection calculation.

[0096] In some embodiments, theoretical design coordinates corresponding to the glass edge positions where optical positioning targets are installed are extracted from the theoretical design model. These theoretical design coordinates originate from the boundary node coordinates of the glass panel in its design state. For each optical positioning target, the difference between its actual spatial coordinates and the theoretical design coordinates along the X, Y, and Z axes is calculated to form the position deviation vector of that optical positioning target. This vector records the translational offset of the optical positioning target in the three spatial dimensions. By combining the position deviation vectors of all optical positioning targets, a rigid body pose fitting algorithm is used to calculate the overall spatial pose deviation of the glass panel. The overall spatial pose deviation is described by three translation parameters and three rotation parameters, which together constitute the positioning deviation data of the glass panel.

[0097] It is understandable that the translation and rotation parameters from the positioning deviation data are input into the control system of the six-degree-of-freedom (DOF) fine-tuning mechanism. The six-DOF fine-tuning mechanism consists of six electric cylinders, upper and lower platforms, and a hinge mechanism. The glass panel is fixed to the upper platform of the six-DOF fine-tuning mechanism via a vacuum suction cup assembly. Based on the translation and rotation parameters, and combined with the kinematic model of the six-DOF fine-tuning mechanism, the control system calculates the required extension / retraction amount for each electric cylinder and generates extension / retraction commands for the electric cylinders. After executing the extension / retraction commands, the electric cylinders of the six-DOF fine-tuning mechanism extend or retract in tandem, causing the upper platform and the glass panel it adheres to to perform translational and rotational movements in three-dimensional space, thus changing the spatial orientation of the glass panel.

[0098] In practical implementation, the total station measurement network continuously tracks and measures the actual spatial coordinates of the optical positioning target during the movement of the six-degree-of-freedom fine-tuning mechanism, and the measurement data is transmitted to the control system in real time. The control system recalculates the positioning deviation data based on the updated actual spatial coordinates and compares it with a preset allowable threshold. If the absolute value of the translation or rotation parameter is greater than the threshold, a new extension / retraction command is generated for further adjustment. A comparison of typical measured values ​​of the optical positioning target coordinates with the theoretical design coordinates during the adjustment process is shown in Table 2. The data in the table is only used to illustrate the order of magnitude of the coordinate deviation.

[0099] Table 2: Comparison of measured and theoretical values ​​of optical positioning target coordinates

[0100]

[0101] Optionally, the update cycle of the positioning deviation data is matched with the sampling frequency of the total station measurement network. Deviation calculation and command generation are completed within each sampling interval to form continuous closed-loop control. When the values ​​of all translation and rotation parameters in the latest feedback positioning deviation data are less than the allowable threshold, the six-degree-of-freedom fine-tuning mechanism stops moving, the glass panel reaches the target pose, and the final locking installation can be performed at this time.

[0102] In one embodiment of the present invention, after the glass panel is finally locked and installed, multiple joint measurement points are selected at equal intervals on both sides of the joint between adjacent glass panels; the actual joint width at each joint measurement point is measured using a digital caliper or laser rangefinder; the theoretical joint width at the corresponding position is obtained from the theoretical design model; the difference between the actual joint width and the theoretical joint width at each joint measurement point is calculated to obtain a joint width deviation sequence; if the maximum value in the sequence exceeds the joint width tolerance, the information of the out-of-tolerance point is recorded, and it is evaluated whether subsequent adjustments are needed. Statistical analysis is performed on the joint width deviation sequence at all joint measurement points of the currently installed multiple glass panels to calculate the average deviation trend and dispersion of the joint width. Based on the average deviation trend and dispersion, the systematic bias and random fluctuation range of the construction error are determined. Before installing the next glass panel, the theoretical installation position of the glass panel in the theoretical design model is pre-compensated and corrected according to its location and systematic bias to generate a pre-compensated installation position. After positioning and installation according to the pre-compensated installation position, the joint verification step is performed again, and the statistical information of the joint width deviation sequence is updated according to the new verification results.

[0103] In practice, after the glass panels are finally locked and installed, accuracy verification is required at the joints between adjacent glass panels. Multiple joint measurement points are evenly spaced along the edge lines on both sides of the joint, with the spacing between adjacent measurement points set to a fixed value based on the joint length and accuracy requirements. A digital caliper or laser rangefinder is used to align the probe of each joint measurement point, measuring the actual joint width at that location. The actual joint width refers to the net distance between the edges of adjacent glass panels. The theoretical joint width at the corresponding location is extracted from the geometric information of the theoretical design model; this theoretical width is determined by the design drawings and curtain wall construction requirements. For each joint measurement point, the difference between the actual joint width and the theoretical joint width is calculated to obtain the joint width deviation at that point. The joint width deviations of all joint measurement points are arranged sequentially to form a joint width deviation sequence. If the maximum absolute value in the joint width deviation sequence exceeds the joint width tolerance, which is limited by the design specifications and the performance of the sealing material, the location number and deviation value of the location exceeding the tolerance are recorded to form the out-of-tolerance location information. Based on the out-of-tolerance location information, it is assessed whether the joint deviation needs to be corrected by loosening the connector and readjusting the glass position.

[0104] In some embodiments, statistical analysis is performed on the joint width deviation sequence at all joint measurement points of multiple currently installed glass panels to calculate the average deviation trend and dispersion of the joint width. The average deviation trend is obtained by calculating the arithmetic mean of all joint width deviations, reflecting the central tendency of the overall deviation. The dispersion is measured by the standard deviation or range, reflecting the fluctuation range of the deviation. Based on the numerical characteristics of the average deviation trend and dispersion, it is determined whether the construction error has a systematic bias or is dominated by random fluctuations. A systematic bias is manifested by a significant deviation of the average deviation from zero, while random fluctuations are manifested by a large degree of dispersion.

[0105] It is understandable that before installing the next glass panel, the theoretical installation position of the next glass panel in the theoretical design model is pre-compensated and corrected based on its location and the identified systematic bias. This pre-compensation correction is achieved by introducing an adjustment amount opposite to the systematic bias. The magnitude of this adjustment is related to the position of the next glass panel; the closer the position is to the area with the largest systematic bias, the larger the adjustment. After pre-compensation correction, a pre-compensated installation position is generated, containing the adjusted corner coordinates and rotation angle of the glass panel, serving as the basis for subsequent positioning and installation. Optionally, the mathematical expression of the pre-compensation correction can use a linear offset model, assuming that the average deviation of the systematic bias in a certain direction is... The pre-compensation amount for the next glass panel in that direction It can be calculated according to the proportional relationship:

[0106]

[0107] in: This represents the average deviation of the seam width deviation sequence in a specified direction. For compensation coefficient, The distance decay function is related to the position of the next glass panel. This represents the distance from the center of the next glass panel to the reference baseline. The pre-compensation amount calculated using this formula is superimposed on the theoretical installation position to obtain the pre-compensated installation position.

[0108] In practice, after positioning and installing the next glass panel according to the pre-compensated installation location, the joint verification step is performed again to measure the newly formed joint width and update the statistical information of the joint width deviation sequence. The newly statistically analyzed average deviation trend and dispersion are used to evaluate the pre-compensation effect. If the average deviation approaches zero and the dispersion decreases, the compensation is effective; otherwise, the compensation coefficient or distance attenuation function is adjusted. The updated joint width deviation sequence statistical information becomes the basis for a new round of predictive compensation, used for subsequent glass panel installation corrections, forming a continuously optimized closed-loop control process.

[0109] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A high-precision installation control method for hyperboloid cold-formed glass curtain walls, characterized in that, The method includes: Obtain the theoretical design model of the target hyperboloid curtain wall. The theoretical design model includes the spatial surface equation of the glass panel, the spatial coordinates of the supporting structure, and the theoretical width of the joint between the panels. The actual surface point cloud data of the existing main structure is collected using a three-dimensional laser scanning device, and the actual surface point cloud data is spatially registered with the theoretical design model to calculate the construction and installation error distribution map of the main structure. An improved point cloud matching algorithm is invoked to perform high-precision matching between the actual surface point cloud data and the ideal point cloud of the support structure in the theoretical design model, and to calculate the three-dimensional spatial position correction of each support point. The improved point cloud matching algorithm is based on the principle of the iterative nearest point algorithm and introduces curvature feature constraints. Based on the construction and installation error distribution map and the three-dimensional spatial position correction amount, an on-site leveling scheme for the support structure is generated, and the on-site leveling scheme includes three-dimensional leveling instructions for each support point.

2. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 1, characterized in that, The working principle of the improved point cloud matching algorithm includes: Local curvature features of each point are extracted from the actual surface point cloud data, and ideal local curvature features of the corresponding points are extracted from the ideal point cloud of the supporting structure. Based on the similarity between the local curvature features and the ideal local curvature features, one or more candidate corresponding points in the ideal point cloud of the supporting structure are initially matched for each point in the actual surface point cloud data. In each iteration of the iterative nearest point algorithm, not only is the spatial distance error between corresponding point pairs calculated, but also the curvature feature difference between corresponding point pairs is calculated. The spatial distance error and the curvature feature difference are weighted and fused to form a joint optimization objective function. The optimal spatial transformation parameters are solved by minimizing the joint optimization objective function. After obtaining new spatial transformation parameters in each iteration, the matching weights of candidate corresponding points are dynamically adjusted based on the curvature feature differences to reduce the interference of outliers on the matching results. When the value of the joint optimization objective function converges or the number of iterations reaches the upper limit, the iteration stops, and the optimal spatial transformation parameters obtained in the last iteration are applied to the actual surface point cloud data, thereby outputting the three-dimensional spatial position correction.

3. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 1, characterized in that, The process involves using a 3D laser scanning device to acquire actual surface point cloud data of the existing main structure, spatially registering the actual surface point cloud data with the theoretical design model, and calculating the construction and installation error distribution map of the main structure, including: The three-dimensional laser scanning device is controlled to scan the existing main structure from multiple stations to obtain multiple initial point cloud data. The multiple initial point cloud data are denoised and stitched together to generate complete actual surface point cloud data. In the theoretical design model, select theoretical feature points with obvious characteristics, including structural corner points and the center point of embedded parts; In the actual surface point cloud data, geometric feature recognition technology is used to identify the actual feature points corresponding to the theoretical feature points; Using the theoretical feature points and the actual feature points, calculate the optimal rigid body transformation matrix that aligns the actual surface point cloud data to the coordinate system of the theoretical design model; The actual surface point cloud data after transformation by the optimal rigid body transformation matrix is ​​compared with the theoretical surface in the theoretical design model point by point to calculate the distance, thereby generating the construction and installation error distribution map that characterizes the magnitude of the deviation at each position.

4. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 3, characterized in that, Based on the construction and installation error distribution map and the three-dimensional spatial position correction, a site leveling scheme for the support structure is generated, including: Identify the critical areas in the construction and installation error distribution map where the deviation exceeds the allowable tolerance, and locate the supporting structures corresponding to the critical areas; Extract the correction data of the support structure corresponding to the key area in the three-dimensional spatial position correction amount. The correction data includes two translation components in the horizontal plane and one translation component in the vertical direction. Based on the correction data, calculate the thickness adjustment value of the bottom leveling shim of the support structure and the in-plane fine adjustment amount of the top connector; The thickness adjustment value and the in-plane fine-tuning amount are combined with the construction procedures and adjustable component specifications to form a set of leveling instructions that includes the operation sequence and quantitative parameters, thus forming the on-site leveling scheme.

5. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 1, characterized in that, The method further includes: Based on the adjusted support structure, the flat glass is gradually bent using a cold bending forming process, and the strain data of the glass panel is collected in real time during the bending process. The bending force and process are dynamically adjusted according to the strain data and the preset strain-curvature mapping relationship. An optical positioning target is set up on a curved glass panel. The optical positioning target is tracked and measured in real time using a total station network. The actual coordinates of the measured target are compared with the corresponding theoretical coordinates in the theoretical design model to generate positioning deviation data of the glass panel. Based on the positioning deviation data, the six-degree-of-freedom fine-tuning mechanism is driven to adjust the spatial pose of the glass panel until the positioning deviation data is less than a preset threshold, and then the final locking installation of the glass panel is performed.

6. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 5, characterized in that, Based on the adjusted support structure, the flat glass is gradually bent using a cold bending forming process. During the bending process, strain data of the glass panel is collected in real time. The bending force and process are dynamically adjusted according to the strain data and a preset strain-curvature mapping relationship, including: Position the flat glass at the predetermined installation position on the adjusted support structure, and temporarily fix its edges to the support points; Multiple resistance strain gauges are attached to predetermined positions on a flat glass plate to form a strain monitoring network; Start the actuator of the cold bending equipment so that the pressure head of the actuator contacts and pushes the glass panel at the initial pressure and speed, causing it to begin to produce elastic deformation; During the pushing process, the strain values ​​of each monitoring point are collected in real time through the strain monitoring network. The real-time collected strain values ​​are input into the preset strain-curvature mapping relationship to deduce the current real-time curvature distribution of the glass panel. The real-time curvature distribution at the same location is compared with the target curvature of the theoretical design model of the target hyperboloid curtain wall, and the curvature deviation is calculated. Based on the magnitude and distribution of the curvature deviation, the thrust and speed of the actuator are adjusted in real time until the deviation between the real-time curvature distribution and the target curvature enters the allowable range.

7. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 6, characterized in that, The process involves deploying optical positioning targets on a curved glass panel, using a total station network to perform real-time tracking and measurement of the optical positioning targets, and comparing the measured actual coordinates of the targets with the corresponding theoretical coordinates in the theoretical design model to generate positioning deviation data for the glass panel, including: At least three non-collinear edge locations of the glass panel that has been cold-bent and whose stress is basically stable, optical positioning targets with unique codes are installed. A total station measurement network consisting of at least three automatic tracking total stations is established at the construction site, and the total station measurement network is systematically oriented and calibrated. The total station measurement network is controlled to simultaneously aim at and measure the three-dimensional coordinates of the center points of at least three optical positioning targets, and the actual spatial coordinates of each target are obtained by back intersection calculation. From the theoretical design model, read the theoretical design coordinates corresponding to the edge positions where each optical positioning target is installed; For each optical positioning target, the difference between its actual spatial coordinates and the theoretical design coordinates is calculated along each coordinate axis to form the position deviation vector of the optical positioning target. By combining the positional deviation vectors of all optical positioning targets, the translation and rotation parameters describing the overall spatial pose deviation of the glass panel are calculated and fitted, thus forming the positioning deviation data.

8. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 7, characterized in that, Based on the positioning deviation data, the six-degree-of-freedom fine-tuning mechanism is driven to adjust the spatial pose of the glass panel, including: The translation and rotation parameters in the positioning deviation data are input into the control system of the six-degree-of-freedom fine-tuning mechanism; The control system converts the translation and rotation parameters into extension and retraction commands for each electric cylinder in the six-degree-of-freedom fine-tuning mechanism. Execute the telescopic command to cause the six-degree-of-freedom fine-tuning mechanism to drive the glass panel it adsorbs to produce translational and rotational motion in three-dimensional space; During the movement, the total station measurement network continuously tracks and measures the actual spatial coordinates of the optical positioning target, and feeds back the updated actual spatial coordinates to the control system in real time; The control system recalculates the real-time positioning deviation data based on the updated actual spatial coordinates, and generates new extension / retraction commands based on the recalculated real-time positioning deviation data, until the values ​​of translation and rotation parameters in the feedback positioning deviation data are both less than the preset allowable threshold.

9. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 1, characterized in that, The method also includes a step of verifying the accuracy of the seams between adjacent glass panels after the glass panels are finally locked and installed: On both sides of the joint between the finally locked and installed glass panel and the adjacent installed glass panel, multiple joint measurement points are selected at equal intervals. Use a digital caliper or laser rangefinder to measure the actual seam width at each seam measurement point; Obtain the theoretical width of the seam at the corresponding position from the theoretical design model; Calculate the difference between the actual joint width and the theoretical joint width at each joint measurement point to obtain a joint width deviation sequence. If the maximum value in the joint width deviation sequence exceeds the joint width tolerance, record the information of the out-of-tolerance point and assess whether further adjustments are needed.

10. The high-precision installation control method for hyperboloid cold-formed glass curtain walls according to claim 9, characterized in that, The method further includes a step of predictively compensating for the installation of subsequent glass panels based on the joint width deviation sequence: Statistical analysis was performed on the seam width deviation sequence at all seam measurement points of the currently installed multiple glass panels to calculate the average deviation trend and dispersion of the seam width. Based on the aforementioned average deviation trend and dispersion, the systematic bias and random fluctuation range of construction errors are determined; Before installing the next glass panel, the theoretical installation position of the glass panel in the theoretical design model is pre-compensated and corrected based on its location and the systematic bias, thus generating a pre-compensated installation position. After positioning and installation according to the pre-compensated installation position, the joint verification step is performed again, and the statistical information of the joint width deviation sequence is updated according to the new verification results for subsequent predictive compensation of the glass panel.

Images