Lightweight processing and assembly simulation method and system for curtain wall unit plate three-dimensional model

Abstract

This invention discloses a lightweight processing and assembly simulation method and system for 3D models of curtain wall unit panels, belonging to the field of building curtain wall technology. It identifies assembly components using machine learning algorithms and simplifies differentiated meshes based on saliency, constructing a lightweight model and a dual-indexing system. Within this lightweight environment, kinematic interference simulation of the entire hoisting and locking process is rapidly completed. After screening candidate collision areas, the dual indexing is used for precise localization, and corresponding areas are extracted from the original high-precision model for local restoration. Fine geometric intersection calculations are performed only on key collision areas. Finally, the verification results are written back to the original model to complete closed-loop verification. This invention balances global simulation efficiency with local verification accuracy, significantly reducing computational resource consumption and achieving efficient and accurate simulation of the prefabricated building construction process.

Description

Technical Field

[0001] This invention relates to the field of building curtain wall technology, and in particular discloses a method and system for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels. Background Technology

[0002] 3D modeling and assembly simulation of building curtain wall unit panels are important means to control construction quality and avoid assembly risks. A search revealed that existing related patents CN115423995A and CN118097005A mostly adopt a globally unified mesh simplification model lightweighting scheme, failing to differentiate component types based on machine learning, lacking a hierarchical lightweighting strategy incorporating volume and complexity features, and lacking an integrated process for component dual-association indexing, temporal assembly sequence generation, kinematic interference simulation, high-precision local model restoration, accurate geometric parameter verification, and result write-back.

[0003] The shortcomings and corresponding consequences of existing technologies are as follows: First, the overall homogenization and lightweight processing does not distinguish between ordinary components and key assembly sub-components, and core geometric features such as connection nodes and hole positions are easily lost. Second, the lack of a dual indexing system of component space and coding makes it difficult to arrange assembly procedures according to construction logic, resulting in low efficiency and inaccurate positioning in kinematic interference simulation. Third, it is impossible to restore high-precision models as needed for collision areas, and it is impossible to accurately calculate assembly parameters such as sealing joint width and overlap length, resulting in insufficient collision verification accuracy. Fourth, the lack of a reverse write-back mechanism for simulation results means that the model cannot be iteratively optimized, and the simulation can only achieve visual display, making it difficult to guide actual assembly construction and failing to balance model running efficiency and simulation verification accuracy.

[0004] Therefore, how to achieve differentiated and lightweight grading of curtain wall components, simulate interference during the assembly process, and complete local high-precision verification and data closed-loop update are key technical problems that urgently need to be solved in the field of curtain wall unit panel assembly simulation. Summary of the Invention

[0005] This invention provides a lightweight processing and assembly simulation method and system for 3D models of curtain wall unit panels. It aims to solve the technical defects of traditional model lightweighting, which uses a global mesh uniform reduction method, which easily damages key assembly features such as curtain wall overlap edges, sealing seams, assembly bayonets, and hoisting docking surfaces, resulting in a sharp drop in the accuracy of assembly interference simulation, seam width verification, and overlap length measurement.

[0006] One aspect of the present invention relates to a method for lightweight processing and assembly simulation of a three-dimensional model of a curtain wall unit panel, comprising the following steps: S100. Obtain the component geometry, component attributes, and component assembly connection relationships in the initial building model. Based on the component geometry, component attributes, and component assembly connection relationships, use a machine learning classification algorithm to identify the target assembly component containing the target sub-component. S200. For the target assembly component, calculate the saliency value by combining volume features and complexity features. Based on the saliency value, perform hierarchical lightweighting processing on the target assembly component, taking into account both mesh simplification and feature preservation, to obtain the component hierarchical lightweighting model and the component dual association index matrix containing the three-dimensional spatial position index and the curtain wall component unique code index. S300. Based on the component dual association index matrix, the process arrangement of the component graded and lightweight model is carried out according to the construction logic rules to generate the curtain wall unit panel time sequence assembly sequence. Then, the kinematic interference lightweight simulation of hoisting and locking actions is carried out in the component graded and lightweight model according to the curtain wall unit panel time sequence assembly sequence to obtain the candidate collision area. S400. Determine whether the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold. If the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold, then based on the component dual association index matrix, extract the corresponding high-precision collision region model from the initial building model, perform local model restoration operation, and obtain the high-precision collision region. S500: For high-precision collision areas, a spatial geometric intersection algorithm is used to calculate the width of the curtain wall assembly sealing joint and the overlap length of the unit panel, and a set of simulation accurate verification results including interference depth and collision type is obtained. S600: By updating the component attribute labels, the simulation accurate verification result set is written back to the initial building model to complete the assembly simulation operation.

[0007] Further, step S100 includes: S110. Obtain the global topology map of the curtain wall components. The topology map is constructed from the component geometry, component attributes, and component assembly and connection relationships in the initial building model. S120. Extract features from the global topology map of the curtain wall components to obtain the topology fusion feature vector; S130. Classify the topology fusion feature vector to obtain the target sub-component matching confidence probability containing the target sub-component; S140. If the matching confidence probability of the target sub-component is greater than the preset component recognition probability threshold, it is determined to be the target sub-component, and the target assembly component is determined based on the target sub-component, thus completing the recognition of the target assembly component.

[0008] Further, step S200 includes: S210. Obtain the three-dimensional mesh data of the target assembly component, and extract the overall mesh volume characteristics and mesh complexity characteristics of the component from the three-dimensional mesh data; S220. Perform a weighted summation operation on the total volume characteristics and mesh complexity characteristics of the component mesh to obtain the comprehensive saliency of the target assembled component; S230. Based on the comprehensive significance of the target assembly components, perform hierarchical lightweighting on the target assembly components to obtain the component-level lightweighting model. S240. Extract the spatial coordinates and unique identifiers of the model after component classification and lightweighting, and generate a component dual association index matrix containing a three-dimensional spatial location index and a unique code index of the curtain wall component.

[0009] Further, step S300 includes: S310. Obtain the component dual-association index matrix and the component graded lightweight model. Arrange the process of the component graded lightweight model according to the construction logic rules to obtain the curtain wall unit panel time sequence assembly sequence. S320. Extract the hoisting and locking actions of the corresponding components based on the timing assembly sequence of the curtain wall unit panels, and generate the kinematic simulation trajectory of the components based on the hoisting and locking actions. S330. Construct the dynamic spatial bounding box corresponding to each component using the kinematic simulation trajectory of the component. S340. Determine whether the dynamic spatial bounding boxes of different components overlap. If they do, then determine the overlapping part as a candidate collision region.

[0010] Further, step S400 includes: S410. Obtain candidate collision regions, and perform spatial discretization on the candidate collision regions to obtain a discrete voxel set. S420. Calculate the cumulative space occupancy of intersecting voxels based on the discrete voxel set to obtain the total actual overlapping volume of component interference; S430. Determine whether the total actual overlapping volume of component interference is greater than the preset critical threshold for interference volume. S440. If the total volume of actual overlap of component interference is greater than the preset critical threshold of interference volume, then extract the spatial coordinate mapping interval corresponding to the candidate collision region. S450. Combining the spatial coordinate mapping interval and the component dual association index matrix, the corresponding high-precision collision area model is extracted from the initial building model.

[0011] Further, step S500 includes: S510. The spatial geometry intersection algorithm is used to calculate the boundary contour set of the high-precision collision region to obtain the intersecting closed polygon contour of the model. S520. Based on the intersecting closed polygonal contours of the model, obtain the width of the curtain wall assembly sealing joint and the overlap length of the unit panel. S530. If the overlap length of the unit plate is greater than the preset overlap depth judgment threshold, the radius of the inscribed sphere of the intersecting closed polygon contour of the model is extracted to determine the assembly interference depth. S540. By combining the assembly interference depth and the width of the curtain wall assembly sealing joint, the original feature set of component penetration interference is mapped to obtain a set of simulation accurate verification results containing interference depth and collision type.

[0012] Further, step S600 includes: S610. Obtain the set of simulation accurate verification results, and extract the initial BIM attribute label set of the components based on the set of simulation accurate verification results. S620. The initial BIM attribute label set of the component is corrected by using the simulation accurate verification result set to obtain the target component attribute label after interference correction. S630. Generate a model status flag instruction based on the target component attribute label after interference correction. The model status flag instruction is used to trigger the model status flag to complete the numerical change. S640. Reconstruct the simulation scene map based on the model status flags, and write back the set of accurate simulation verification results to the initial building model to complete the assembly simulation operation.

[0013] Another aspect of the present invention relates to a lightweight processing and assembly simulation system for three-dimensional models of curtain wall unit panels, used to implement the above-mentioned lightweight processing and assembly simulation method for three-dimensional models of curtain wall unit panels, comprising: The target assembly component identification module is used to obtain the component geometry, component attributes, and component assembly connection relationships in the initial building model. Based on the component geometry, component attributes, and component assembly connection relationships, a machine learning classification algorithm is used to identify the target assembly components containing target sub-components. The component dual association index matrix acquisition module is used to calculate the saliency value of the target assembly component by combining volume features and complexity features. Based on the saliency value, the target assembly component is subjected to hierarchical lightweighting processing that takes into account both mesh simplification and feature preservation, resulting in the component hierarchical lightweighting model and the component dual association index matrix containing the three-dimensional spatial position index and the unique code index of the curtain wall component. The candidate collision region acquisition module is used to arrange the process of the component after the component is classified and lightweight according to the construction logic rules based on the component dual association index matrix to generate the curtain wall unit panel time sequence assembly sequence. Then, it performs kinematic interference lightweight simulation of hoisting and locking actions in the component after the component is classified and lightweight according to the curtain wall unit panel time sequence assembly sequence to obtain the candidate collision region. The high-precision collision region acquisition module is used to determine whether the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold. If the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold, the corresponding high-precision collision region model is extracted from the initial building model based on the component dual association index matrix, and the local model restoration operation is performed to obtain the high-precision collision region. The simulation precise verification result set acquisition module is used to calculate the width of the curtain wall assembly sealing joint and the overlap length of the unit panel for high-precision collision areas using a spatial geometric intersection algorithm, and obtain a simulation precise verification result set that includes interference depth and collision type. The assembly simulation task completion module is used to write back the set of accurate simulation verification results to the initial building model by updating the component attribute labels, thereby completing the assembly simulation task.

[0014] The beneficial effects achieved by this invention are as follows: The lightweight processing and assembly simulation method and system for 3D models of curtain wall unit panels provided by this invention addresses the contradiction between the high computational load and low efficiency of traditional full-model simulation and the high local accuracy requirements. It adopts a hybrid strategy of hierarchical lightweighting and on-demand accuracy restoration. First, machine learning algorithms are used to identify assembly components and differentiate mesh simplification based on saliency, constructing a lightweight model and a dual-indexing system. In this lightweight environment, kinematic interference simulation of the entire hoisting and locking process is quickly completed. After screening candidate collision areas, the dual indexing is used to accurately locate and extract corresponding areas from the original high-precision model for local restoration. Fine geometric intersection calculations are performed only on key collision areas. Finally, the verification results are written back to the original model to complete closed-loop verification. This invention balances global simulation efficiency and local verification accuracy, significantly reducing computational resource consumption and achieving efficient and accurate simulation of the prefabricated building construction process. The specific technical effects achieved are as follows: 1. This invention integrates three types of multi-dimensional model information: component geometry, attributes, and assembly association. It automatically identifies target assembly components containing sub-components by relying on machine learning classification algorithms. This overcomes the drawbacks of low efficiency and easy omissions and misjudgments in manual component screening. It deeply explores the hierarchical relationship and assembly binding logic of curtain wall components, realizes intelligent identification of target components, and improves the intelligence level of early model analysis and component division.

[0015] 2. This invention combines the dual-dimensional features of component volume and complexity to calculate component saliency, and uses the saliency to complete differentiated and graded lightweight processing, achieving a balance between mesh simplification and key assembly features. At the same time, it constructs a dual-association index matrix of spatial location and component coding, which reduces redundant mesh data and model computational load, while fully preserving the component assembly association and core features, thus solving the problems of distortion of key assembly structures and breakage of index association caused by traditional uniform lightweighting.

[0016] 3. This invention relies on a dual-association index matrix to fit the actual construction logic of the site and arrange the assembly sequence. It uses a lightweight model to complete the kinematic interference simulation of the entire process of hoisting and locking, quickly locks the candidate collision area of ​​the assembly, reduces the overall simulation computing power consumption and improves the simulation operation efficiency, while restoring the actual assembly construction conditions of the curtain wall unit panel, and realizes the in-depth exploration of the assembly process and motion interference law.

[0017] 4. This invention sets a critical threshold for interference volume to complete the classification and judgment of interference risk. For interference areas exceeding the standard, it uses the index matrix to reverse map the original high-precision model and complete the local model recovery. It takes into account both global simulation efficiency and local verification accuracy, avoids the defects of insufficient accuracy of the global lightweight model leading to interference verification distortion and missed detection of assembly collision risks, and improves the accuracy of assembly interference evaluation.

[0018] 5. This invention performs spatial geometric intersection calculations on the restored high-precision collision area to accurately calculate the sealing gap width, plate overlap length, interference depth, and collision type, and completes the fine-grained verification of assembly gap and structural interference. It breaks through the limitations of traditional simulation that only determines whether there is a collision and cannot quantify assembly overlap parameters, and realizes a fine-grained and accurate evaluation of assembly quality and interference degree.

[0019] 6. This invention uses attribute tags to update the simulation verification results and write back the original building model, achieving seamless linkage between simulation data, assembly defect parameters, and the original design model. This creates a closed loop for the entire process of model lightweighting, assembly simulation, and model iterative optimization. This invention balances the computational advantages of lightweight models with the accuracy of assembly simulation verification, aligns with the requirements of prefabricated curtain wall construction processes, and is suitable for large-scale simulation verification of unit panels. It reduces equipment computing power requirements and simulation operation cycles, making it highly practical for engineering applications. It can effectively identify potential construction hazards such as curtain wall assembly collisions, excessive sealing joints, and non-compliant overlap dimensions in advance, optimizing on-site assembly construction plans. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an embodiment of the lightweight processing and assembly simulation method for three-dimensional models of curtain wall unit panels according to the present invention. Figure 2 This is a functional block diagram of an embodiment of the curtain wall unit panel three-dimensional model lightweight processing and assembly simulation system of the present invention.

[0021] Explanation of icon numbers: 10. Target assembly component identification module; 20. Component dual association index matrix acquisition module; 30. Candidate collision region acquisition module; 40. High-precision collision region acquisition module; 50. Simulation accurate verification result set acquisition module; 60. Assembly simulation operation completion module. Detailed Implementation

[0022] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0023] like Figure 1As shown, the first embodiment of this invention proposes a lightweight processing and assembly simulation method for a 3D model of a curtain wall unit panel. This method belongs to the technical fields of building BIM 3D model optimization, curtain wall unit prefabricated construction, model mesh lightweighting, building assembly motion simulation, and construction interference verification. Existing curtain wall BIM overall models suffer from problems such as a large number of components, large redundant meshes for curved keel, sealing strips, and overlapping connectors, high computational power consumption of the entire model, and lag and failure in large-scene assembly simulation. Traditional model lightweighting uses a global mesh uniform reduction method, which easily damages the curtain wall overlapping edges, sealing seams, and assembly... Key assembly features such as bayonet joints and hoisting docking surfaces lead to a sharp decline in the accuracy of assembly interference simulation, joint width verification, and overlap length calculation. Conventional assembly simulations lack model-level lightweighting, local high-precision restoration of collision areas, and index matrix linkage mapping mechanisms. The lightweight model and the original high-precision model data are disconnected, making it impossible to balance simulation efficiency and assembly size verification accuracy. At the same time, the assembly process is not arranged in accordance with the curtain wall construction logic and sequence. Kinematic simulation only performs static collision detection, which cannot be adapted to the actual construction conditions of unit panel hoisting, bolt locking, and profile overlap. The simulation results cannot guide the on-site prefabricated curtain wall construction. This invention constructs a fully closed-loop process architecture encompassing "model data source extraction - intelligent target component identification - dual-feature coupling hierarchical lightweighting - dual-index matrix construction - temporal assembly sequence generation - kinematic lightweight interference simulation - collision threshold discrimination - local high-precision model restoration - geometric parameter verification - simulation result reverse writing back," taking into account four core requirements: overall model simulation computing power, retention of key assembly features, dynamic construction interference verification, and accurate calculation of sealing joint dimensions. It is suitable for the full-process simulation of prefabricated construction of unitized glass curtain walls and metal single-panel curtain walls, including the following steps: Step S100: Obtain the component geometry, component attributes, and component assembly connection relationships in the initial building model. Based on the component geometry, component attributes, and component assembly connection relationships, use a machine learning classification algorithm to identify the target assembly component containing the target sub-component.

[0024] The system retrieves the underlying data source of the initial BIM 3D model of the integrated civil engineering curtain wall, and batch analyzes and extracts three types of structured underlying model data corresponding to the curtain wall components: component geometric dimensions, component inherent properties, and component assembly topology connection relationships. Using these three types of heterogeneous data as input features, and relying on component shape dimension parameters, material inherent properties, functional assembly attributes, and upstream and downstream assembly topology dependency features, the system combines a preset machine learning classification algorithm to complete intelligent binary classification and identification of components. This accurately filters curtain wall overlaps and load-bearing target assembly components with internally embedded target sub-components, and removes non-assembly redundant components attached to civil engineering walls and roofs, thus completing the accurate screening and collection of simulated target assembly components.

[0025] The initial building model is a BIM 3D model including curtain wall unit panels, keel, sealing accessories, embedded parts, and assembly connectors, preferably a model with a LOD4.0 to LOD4.5 level of detail. When the input model's level of detail is lower than LOD4.0, it is necessary to supplement the data of connectors, hole positions, or sealing boundaries before performing identification and simulation. When the input model's level of detail is higher than LOD4.5, there are too many redundant curved surface patches, and the computational redundancy exceeds the limit.

[0026] The geometric dimensions of a component are a set of three-dimensional solid quantified dimensions, including the length and width of the curtain wall unit panel, the wall thickness of the profile, the width of the overlap bevel, the size of the hoisting base, the outer diameter of the sub-component, and the distance between the installation positioning holes. These dimensions include two types of parameters: linear dimensions and cross-sectional profile dimensions.

[0027] The component attributes are a collection of attributes including the material attributes, functional assembly attributes, construction procedure attributes, and component ownership labels of the curtain wall components; including four types of material labels for glass panels, aluminum alloy keel, rubber sealing strips, and steel limiting fasteners, and two functional labels for main assembly components / auxiliary sub-components.

[0028] The component assembly connection relationship consists of five types of topological relationships within the initial model: parent-child hierarchical binding, plug-in assembly, bolt locking, surface bonding and overlapping, and embedded nesting. The storage of the upstream and downstream assembly dependency topology matrix of the component is also included.

[0029] The target sub-components are small auxiliary components embedded inside the main assembly components of the curtain wall, which directly determine the assembly interference, sealing joint width, and overlap length. They include five types of components: sealing strips, positioning buckles, anti-detachment pads, butt joints, and pre-embedded threaded sleeves. These are the core verification micro-components for assembly simulation.

[0030] The target assembly components encompass and nest target sub-components, participate in on-site hoisting and docking, and are used for the assembly of curtain wall panels, vertical main keels, horizontal secondary keels, and corner unit panels. These components form the core of the model's lightweighting and assembly simulation.

[0031] When there is a parent-child hierarchical edge, nested edge, or locked connection edge between the target sub-component node and a curtain wall panel, main keel, secondary keel, or corner unit plate node, the corresponding upper-level component of the target sub-component is determined as the target assembly component; if a target sub-component connects to multiple upper-level components, the component with the largest assembly connection edge weight is taken as the main target assembly component, and the other components are recorded as associated verification components.

[0032] The complete five-layer architecture of the machine learning classification algorithm is as follows: ① Multi-dimensional data source normalization input layer → ② Geometric dimension feature dimensionality reduction layer → ③ Component attribute label encoding layer → ④ Assembly topology association feature fusion layer → ⑤ Target assembly component binary classification output layer; The algorithm adopts a lightweight random forest classification model, which is adapted to the structured samples of BIM models.

[0033] The preset component recognition probability threshold is P0, which can be determined based on the recall rate and false recognition rate of the training sample validation set; in one embodiment, P0 is 0.82.

[0034] The number of components extracted from a single model can range from 1200 to 8500 in a single identification of a single curtain wall model. The value is based on the volume of conventional components for high-rise commercial curtain walls and residential unitized curtain walls, and the computing power capacity range of the matching algorithm.

[0035] Step S200: For the target assembly component, calculate the saliency value by combining volume features and complexity features. Based on the saliency value, perform hierarchical lightweighting processing on the target assembly component, taking into account both mesh simplification and feature preservation, to obtain the component hierarchical lightweighting model and the component dual association index matrix containing the three-dimensional spatial position index and the unique code index of the curtain wall component.

[0036] Step S100 completes the screening of all target assembly components, and simultaneously extracts the volume features of the component's bounding box, the number of curved mesh patches, the number of inflection points of irregular contours, and the complexity features of the opening and overlapping boundaries. Dual-feature weighted coupling calculations are used to solve for the quantified values ​​of the assembly features of a single component. Component lightweighting levels are classified according to the saliency gradient, and mesh reduction rates and boundary feature constraint parameters are configured differently. Hierarchical mesh lightweighting simplification processing is carried out to preserve assembly boundaries, overlapping bevels, and the mating surfaces of sub-components. Simultaneously, component spatial coordinate parameters and factory unique identification code parameters are bound together to construct a two-dimensional linked mapping data table, generate a component dual-association index matrix, and simultaneously output two types of core data: the overall lightweighting model and the index matrix.

[0037] Volumetric features are the quantified volume values ​​of the three-dimensional solid bounding box of the target assembly component, which distinguishes between large-volume panel main components and small-volume connecting sub-components, and characterizes the component's volume weight parameters.

[0038] The complexity characteristics are normalized comprehensive quantitative indicators of the total number of component mesh patches, the number of curvature abrupt change points of irregular curved surfaces, the number of inflection points of assembly lap boundary contours, and the number of bolt positioning holes. The higher the value, the more complex the component assembly boundary structure.

[0039] The assembly feature saliency value is a standardized comprehensive score of 0 to 1 obtained by linearly weighting volume and complexity features. The higher the score, the higher the criticality of the component assembly, the more core the lap sealing feature, and the higher the level of lightweight mesh retention constraint. The numerical value range is 0.20~0.95; the hierarchical intervals are as follows: Level 1 high saliency: 0.72~0.95, Level 2 medium saliency: 0.45~0.72, Level 3 low saliency: 0.20~0.45; the value is based on the industry hierarchical classification standard of curtain wall component assembly function priority and matches the hierarchical lightweight parameter configuration logic.

[0040] A lightweight algorithm architecture that balances mesh simplification and feature preservation: ① Dual-feature normalization fusion layer → ② Component saliency level clustering partitioning layer → ③ Differentiated mesh reduction parameter assignment layer → ④ Assembly boundary feature locking constraint layer → ⑤ Redundant internal patch removal layer → ⑥ Lightweight model output layer.

[0041] Hierarchical mesh simplification rate thresholds (core limiting parameters): Level 1 high saliency component mesh simplification rate 12%~20%; Level 2 medium saliency component mesh simplification rate 35%~45%; Level 3 low saliency component mesh simplification rate 55%~65%; Value basis: Strictly control the simplification rate of high saliency overlapping and sealed interfaces to prevent assembly boundary deformation; significantly reduce the number of solid surfaces inside low saliency components to balance computing power and accuracy.

[0042] After component classification and lightweighting, the model is a dedicated BIM simulation model for reducing the overall computational cost of the curtain wall, which includes graded mesh reduction, locking and retaining features of assembly overlap boundaries / sealing joint interfaces / lifting reference surfaces, and deleting internal redundant solid surfaces.

[0043] Three-dimensional spatial location index: Taking the curtain wall civil engineering reference axis grid as the origin, the absolute coordinate encoding index of the X / Y / Z three axes of the component bounding box center is used to realize the rapid positioning of points in the model space and the tracing of the coordinates of the collision area.

[0044] Curtain wall component unique code index: Preset component type-floor-span-production batch integrated globally unique string code, each component corresponds one-to-one, without duplication or confusion, realizing the identity binding of model components with the original BIM data source.

[0045] The component dual-association index matrix is ​​a two-dimensional coupled association matrix with rows corresponding to the unique component encoding index and columns corresponding to the three-axis spatial position index. It serves as the data index base for bidirectional millisecond-level mapping and linkage between lightweight models and initial high-precision model components.

[0046] Boundary feature retention threshold: The integrity of the assembly overlap contour, sealing strip bonding boundary, and hoisting docking reference edge features is ≥98.2%; Value basis: to ensure that the subsequent assembly interference and the simulation verification error of the joint width dimension are less than the engineering allowable construction tolerance of 0.5mm.

[0047] Step S300: Based on the component dual association index matrix, arrange the process of the component after graded lightweighting according to the construction logic rules to generate the curtain wall unit panel time sequence assembly sequence, and carry out kinematic interference lightweighting simulation of hoisting and locking actions in the component after graded lightweighting model according to the curtain wall unit panel time sequence assembly sequence to obtain candidate collision areas.

[0048] Step S200 outputs a dual-association index matrix to complete the data linkage binding of lightweight components across the entire domain. Based on the rigid logical constraints of unitized curtain wall floor arrangement from bottom to top, keel first then panel, hoisting first then bolting, and priority docking of corner components for prefabricated on-site construction, the installation sequence of components is iteratively optimized to generate a standardized time-series assembly construction sequence with timestamps. Relying on the lightweight model, the rigid body kinematics simulation of the components is constructed to replicate the dynamic motion trajectory of the entire process of unit panel hoisting, horizontal alignment, vertical insertion, lateral fitting, and bolt point locking. Full-domain dynamic motion interference simulation screening is carried out to mark the intersecting and overlapping areas of spatial geometry and output the candidate collision area set completed by the initial screening of the entire domain in batches.

[0049] The construction logic rules are standardized process constraints for the prefabricated construction of curtain wall unit panels. They include four rigid construction constraints: the priority of vertical assembly on each floor, the dependency relationship between the main and secondary keel assembly, the logic of corner unit finishing, and the hoisting avoidance of the civil structure. These are the prerequisite engineering rules for the arrangement of work processes.

[0050] The curtain wall unit panel sequential assembly sequence is a time-sequential assembly process link that binds millisecond-level timestamps, component assembly order, component movement start and end postures, and construction station parameters, serving as the unique temporal input condition for dynamic motion simulation. The time interval threshold for the sequenced process is determined by the construction simulation time step and the duration of the construction process; the construction simulation time step is used for collision detection sampling, and the process duration is determined by the project hoisting plan or historical construction records. In this example, the collision detection sampling step can be set to 0.01s, and the simulation duration for a single unit panel assembly process can be set to 10s; the value is based on matching the actual mechanical operation timing parameters for unit panel hoisting and alignment at the construction site.

[0051] Kinematic Interference Lightweight Simulation Algorithm Architecture: ① Index Matrix Component Working Condition Calling Layer → ② Timing Assembly Trajectory Assignment Layer → ③ Rigid Body Six Degrees of Freedom Motion Solving Layer → ④ Dynamic Spatial Geometric Intersection Screening Layer → ⑤ Irrelevant Spatial Noise Removal Layer → ⑥ Candidate Collision Region Clustering Output Layer.

[0052] Lifting motion constraint thresholds: horizontal displacement rate of unit panel lifting is 0.25~0.6m / s, and vertical lifting rate is 0.15~0.4m / s; the values ​​are based on the standard lifting operation rate of curtain wall construction cranes to restore the actual construction motion conditions.

[0053] Locking motion parameters: bolt axial feed displacement 8~22mm, lateral clamping posture rotation angle 0°~3.5°; values ​​are based on the standard operating stroke for on-site locking installation of curtain wall unit panels.

[0054] Candidate collision regions are coarsely screened spatial closed regions in the lightweight simulation model where dynamic components overlap spatial geometry and interfere with the motion posture of existing completed civil engineering structures and installed curtain wall components. They include two types of mixed regions: real interference regions and pseudo-interference regions caused by distortion of model lightweight patches.

[0055] Acceleration threshold for simulation mesh computation: The total number of mesh faces in a single frame of lightweight simulation computation is ≤160,000; Value basis: Ensure that the real-time dynamic simulation frame rate of the workstation is ≥24 frames / s to meet the requirements for real-time output of engineering simulation.

[0056] Step S400: Determine whether the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold. If the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold, then based on the component dual association index matrix, the corresponding high-precision collision region model is mapped and extracted from the initial building model, and a local model restoration operation is performed to obtain the high-precision collision region.

[0057] Spatial geometric volume integration is performed on all candidate collision regions output in step S300 to accurately solve the spatial overlap interference volume of the component entities in each candidate region. The measured overlap volume is compared with the engineering preset interference volume critical threshold line by line to filter out small volume pseudo-interference noise regions caused by lightweight mesh distortion. For the real interference regions that exceed the critical threshold, the component dual association index matrix is ​​retrieved to complete the bidirectional source matching of spatial coordinates and component codes. Cross-model linkage mapping retrieves the original unsimplified mesh and original assembly boundary parameters of the initial building model to complete the local area mesh restoration, feature restoration, and surface patch reconstruction. The lightweight mesh distortion error is eliminated to generate a high-precision collision region with no simplification, no distortion, and original parameter matching.

[0058] Candidate collision region overlap volume: The volume of the three-dimensional closed geometry of the dynamic moving curtain wall unit component, adjacent fixed components, and civil engineering structure intersecting and encompassing each other. It is the core quantitative indicator for judging the authenticity of interference. Unit: mm³.

[0059] The pseudo-interference region is a false geometric overlap region induced by the model's hierarchical lightweight mesh reduction and surface patch polyline distortion. It does not pose an actual engineering assembly interference risk and is an inherent noise region of the simulation.

[0060] Cross-model mapping extraction algorithm architecture: ① Region spatial coordinate index parsing layer → ② Component unique code matching and binding layer → ③ Lightweight model region clipping layer → ④ Initial model homogeneous geometry retrieval layer → ⑤ Boundary topology fitting and alignment layer.

[0061] Local model restoration operation: Based on the dual index matrix, point-to-point replacement of lightweight distorted meshes restores the original model surface curvature, assembly bevel, sealing interface, and native mesh topology of sub-component mating surfaces.

[0062] High-precision collision region model: A local closed-loop 3D simulation model subdomain that replicates the original building model's LOD4.5 native mesh, without simplification or boundary distortion, and retains all fine assembly features.

[0063] Local model position alignment error threshold: model restoration space alignment deviation ≤ 0.18mm; value basis: lower than the 0.5mm engineering tolerance of the curtain wall assembly sealing joint, meeting the high-precision dimension verification requirements.

[0064] High-precision collision area: The sub-region for high-precision closure verification of real assembly interference after topology alignment, mesh restoration, and noise removal, which is the only carrier for subsequent accurate calculation of seam width and overlap length.

[0065] Step S500: For the high-precision collision area, the spatial geometric intersection algorithm is used to calculate the width of the curtain wall assembly sealing joint and the overlap length of the unit panel, so as to obtain a set of simulation accurate verification results including interference depth and collision type.

[0066] Step S400 restores the high-precision collision area as a fixed computational domain. It calls the three-dimensional solid space Boolean geometry intersection algorithm and boundary contour distance solver to solve the net width of the unit panel assembly sealing joint by fitting the normal section along the curtain wall sealing strip, and to solve the axial length of the overlap of the upper and lower unit panel profiles by fitting the cross section along the profile overlap. Simultaneously, it quantifies the depth of component intrusion and containment interference, and automatically clusters and classifies collision defect categories by combining the assembly position and component type. It integrates the size measurement parameters, interference quantization parameters, and defect classification parameters, and structurally encapsulates and generates a set of simulation accurate verification results.

[0067] The six-layer architecture of the spatial geometry intersection algorithm is as follows: ① High-precision region entity boundary sampling layer → ② Assembly section normal fitting layer → ③ Two-component Boolean intersection operation layer → ④ Seam width axial distance solution layer → ⑤ Overlap profile overlap integration layer → ⑥ Multi-dimensional verification parameter output layer.

[0068] Curtain wall assembly sealing joint width: The net normal gap dimension of the sealing assembly interface between two adjacent curtain wall unit panels, which is a core acceptance parameter for sealant application and waterproofing design. Engineering standard verification value range: compliant standard joint width 8mm~15mm; exceeding the range is judged as assembly interference / assembly gap exceeding the standard; value basis: calibrated according to the national standard design parameters of curtain wall unit panel assembly waterproofing structure.

[0069] Interference depth is the maximum axial embedment depth of two assembled components intersecting each other, and is a quantitative indicator characterizing the severity of assembly interference. Thresholds for severity classification: Slight interference: 0.2~1.5mm, Moderate interference: 1.5~4mm, Severe interference: >4mm; Values ​​are determined based on the difficulty classification standards for on-site construction rectification.

[0070] Collision type is a standardized interference defect classification label. The collision type label is uniformly defined as: 1=panel / profile overlap interference, 2=keel and civil structure hard collision, 3=connector or bolt penetration interference, 4=sealing sub-component extrusion interference, 5=lifting tool temporary collision.

[0071] The simulation-based accurate verification result set consists of a structured data package containing six bound verification data items: measured value of seal width, measured value of overlap length, interference depth value, collision type label, high-precision area coordinates, and component code.

[0072] Dimension calculation accuracy threshold: geometric parameter calculation resolution ≤ 0.05mm; value basis: far superior to construction acceptance tolerance, meeting the accuracy requirements for forward design verification of curtain wall prefabricated BIM.

[0073] Step S600: Write the simulation accuracy verification result set back to the initial building model by updating the component attribute labels to complete the assembly simulation operation.

[0074] Based on the dual-association index matrix of components in step S200, the verification data and the initial model components are matched one by one. The background attribute database of the initial building model components is retrieved, and three types of custom component extended attribute fields are added: assembly interference label, construction rectification label, and dimension overrun annotation. All dimension parameters, interference level, collision type, and rectification suggestion parameters in the simulation precision verification result set in step S500 are batch bound to the corresponding target assembly components, and the native BIM model attribute label iterative overwriting and parameter backfilling are completed. The lightweight simulation model, high-precision verification sub-model, and time-series assembly sequence of intermediate data are archived simultaneously, and the entire process of lightweight + dynamic assembly of the curtain wall unit panel model is completed in a closed loop.

[0075] Component attribute label update algorithm architecture: ① Verification result encoding and parsing layer → ② Dual index model matching layer → ③ Original component attribute field retrieval layer → ④ Interference expansion label addition layer → ⑤ Original attribute overwrite iteration layer → ⑥ Model version archive storage layer.

[0076] Component attribute tags are editable auxiliary parameter tags in the background of the initial building model components. They include four types of iteratively editable tags: native material tags, newly added assembly interference level tags, joint width exceeding standard tags, and construction rectification priority tags.

[0077] Reverse write-back is a data linkage closed-loop mechanism that uses the unique component code and spatial location index as a medium to backfill the back-end verification results of the lightweight simulation link to the front-end unsimplified initial BIM native model.

[0078] Attribute label write-back integrity threshold: The completeness of the verification result label binding write-back is ≥99.7%; Basis for value selection: To prevent missing or incorrect model labels and ensure that the BIM model can be directly used for construction drawing refinement and on-site construction handover.

[0079] Redundancy compression threshold for model archiving: Compression ratio of 4.2:1 to 5.5:1 for the entire set of intermediate simulation files; Basis for value selection: Taking into account both the storage volume of the engineering model archive and the traceability of the original simulation data.

[0080] Assembly simulation work: Complete the overall closed-loop process results, including model lightweighting, dynamic time-series assembly simulation, interference authenticity judgment, local high-precision restoration, geometric dimension verification, and native BIM model data linkage update.

[0081] This embodiment provides a lightweight processing and assembly simulation method for 3D models of curtain wall unit panels, addressing six major industry pain points of traditional curtain wall BIM models: full-domain lightweighting damages fine assembly features, dynamic assembly simulation computational bottlenecks, data fragmentation between the lightweight model and the original model, inability to identify pseudo-interference, insufficient accuracy in checking sealing joints and overlap lengths, and inability to link simulation results with the native design model. Based on a core architecture including component dual-feature saliency-level lightweighting, dual-associative index matrix linkage mapping, construction sequence assembly kinematics simulation, interference volume threshold noise filtering, high-precision local model restoration, geometric intersection dimension verification, and model label reverse writing, it achieves a 60% or more speedup in large-scene curtain wall model simulation computation, while ensuring the accuracy of assembly interface and sealing sub-component simulation verification. It distinguishes between pseudo-interference caused by model mesh distortion and actual construction assembly interference, automatically classifies collision defect levels, calculates sealing joint widths and overlap lengths, and directly links simulation results to update the original building BIM model without secondary modeling correction. This directly adapts to the entire process of unitized curtain wall prefabricated construction detailed design, construction interference prediction, component prefabrication, and on-site hoisting handover.

[0082] Furthermore, the lightweight processing and assembly simulation method for the three-dimensional model of the curtain wall unit panel provided in this embodiment includes step S100: Step S110: Obtain the global topology map of the curtain wall components. The topology map is constructed from the component geometry, component attributes, and component assembly connection relationships in the initial building model.

[0083] Using the initial building model's entire curtain wall components as nodes and the component assembly connection relationships as topological edges, the geometric dimension parameters of the components corresponding to the nodes and the inherent attribute labels of the components are coupled and bound to complete the joint encapsulation of heterogeneous node-edge data, and structurally construct the entire topological graph of the curtain wall components, realizing the integrated and structured expression of the model component entities, geometric parameters, attribute labels, and assembly topology.

[0084] The formula for calculating the global topology map of curtain wall components is: (1) In formula (1), This is a global topological map of the curtain wall components, without units. The matrix, derived from the initial BIM model analysis output, represents the geometric dimensions, attributes, and assembly-related topological relationships of all curtain wall components; This is a set of vertex features representing the geometric dimensions of the component, in meters (m). These features are derived from the length, width, height, surface curvature, and plate thickness parameters extracted from the BIM model component. The value range is [range missing]. It represents the geometric boundaries of unit plates, keels, and connectors; It is a set of component attribute feature vectors, without units, derived from the BIM component material code, fire resistance rating, component type, and assembly level attribute library, with 64 dimensions; This is a component assembly connection edge matrix, unitless, derived from BIM assembly constraints, bolt connections, and lap assembly binding relationships, with values ​​defined as follows: No connection Rigid assembly Flexible connection is achieved through overlapping. The control logic of formula (1) is to deconstruct the three types of core data (component geometric dimensions, component attributes, and component assembly connection relationships) in the original BIM model, encapsulate the component information with a vertex-edge topology, eliminate redundant cached parameters in the model, and generate a standardized topology map that can be directly input into machine learning. Formula (1) breaks through the defects of traditional BIM feature discrete extraction, integrates the three types of related data (geometric, attribute, and assembly) into a unified binding, and forms a complete component topology representation. This approach not only unifies the data dimensions but also serves as direct input for machine learning classification algorithms, significantly improving the accuracy of subsequent target component classification and recognition, and providing a reliable structured data foundation for subsequent lightweight processing and assembly simulation.

[0085] Curtain wall component global topology map: an undirected weighted heterogeneous topology network graph with curtain wall components as topology nodes and assembly dependencies as topology edges. Nodes are assigned two types of feature weights: geometric dimensions and component attributes. Edges are assigned assembly binding type and assembly constraint tolerance weights, which serve as the basic carrier for feature extraction in this step.

[0086] Topology edge weight thresholds for graph construction: weight of rigid bolt connection edges 0.85~0.95, weight of flexible lap assembly edges 0.40~0.60; value basis: matching the assembly constraint levels of two types of curtain wall projects: rigid assembly and flexible sealing lap.

[0087] The node capacity range of a single global topology graph is 1200 to 8500 nodes, which matches the number of component samples in step S100.

[0088] Step S120: Extract features from the global topology map of the curtain wall components to obtain the topology fusion feature vector.

[0089] A graph convolution feature extraction algorithm is used to extract node features and topological edge association features in two layers from the global topological graph. The algorithm integrates the original features of component size, material attribute features, and upstream and downstream assembly topological association features to complete the dimensionality reduction and normalization fusion of heterogeneous features, and outputs a fixed-dimensional structured topological fusion feature vector.

[0090] The formula for calculating the topology fusion feature vector is: (2) In formula (2), This is the topology fusion feature vector, which has no unit and is the output of this step. It has 128 dimensions and represents the global comprehensive features of the component. is the ReLU nonlinear activation function, which is dimensionless and is used to introduce nonlinear transformations and suppress model gradient vanishing. The graph convolution weight parameter matrix is ​​unitless and consists of fixed parameters after model training convergence, with dimensions of 1. ; A graph convolution topology feature extraction operator, unitless, specifically adapted to the topology of irregular curtain wall components; This is a global topology map of the curtain wall components, without units, and is the output of formula (1); is the convolution bias vector, which has no unit, and is an intrinsic parameter of the model, with a dimension of 128. The control logic of formula (2) is to first extract the topological features of the graph convolution topology for the irregular assembly topology of the curtain wall. Global topology map of curtain wall components Feature extraction is performed, fusing local geometric features of components with global assembly-related features; subsequently, a graph convolution weight parameter matrix is ​​used. With convolution bias vector Perform a linear transformation and then apply the ReLU nonlinear activation function. By introducing nonlinearity, a standardized feature vector with a fixed dimension is finally output. Formula (2) uses graph convolution instead of ordinary image convolution, which can adapt to the irregular topological structure of curtain wall irregular components. While extracting features, it fully preserves the topological semantics of the assembly connection between components, effectively adapts to the recognition scenario of nested sub-component assembly, and provides high-quality feature input with both local details and global correlation for the subsequent classification and recognition of target assembly components.

[0091] The topology fusion feature vector is a one-dimensional standardized feature vector that integrates three-dimensional information: geometric dimension, attribute dimension, and topological relationship dimension. Each vector has a fixed dimension of 32 and serves as the input feature for subsequent classification.

[0092] This embodiment employs a graph convolutional neural network (PNN) classification model, which includes a component topology graph input layer, a neighborhood feature aggregation layer, a feature fusion layer, a fully connected classification layer, and a Softmax probability output layer. The random forest model is described separately as an alternative embodiment and should not be used interchangeably with the main embodiment.

[0093] Feature information retention threshold: Topology fusion feature information retention rate ≥ 92%; Value basis: to ensure that the associated features of sub-components are not lost, and to improve the accuracy of subsequent classification probability.

[0094] Step S130: Classify the topology fusion feature vector to obtain the target sub-component matching confidence probability containing the target sub-component.

[0095] The topology fusion feature vector is input into the lightweight random forest classification model. Based on the feature weights of the pre-trained component samples, feature forward inference is completed. The target sub-component matching confidence probability of the target sub-component embedded in the component corresponding to a single graph node is quantized and output to complete the probabilistic classification and discrimination.

[0096] The formula for calculating the confidence probability of matching the target sub-component is: (3) In formula (3), Match the confidence probability of the target sub-component, which is unitless and ranges from 1 to 2. The larger the value, the higher the matching degree of the nested sub-components; This is the classification layer weight matrix, which is unitless and has 1 dimension. ; , is the topology fusion feature vector, which has no unit and is the input for this step, i.e., the output of formula (2); This is the classification layer bias vector, which is unitless and has a dimension of 2. This is a multi-class probability normalization operator with no unit, outputting standard probability values ​​in the 0-1 range. The control logic of formula (3) is to linearly reduce the dimensionality of the high-dimensional topological fusion feature vector through the classification layer weight matrix and bias vector, and then normalize it through the Softmax operator to output the matching probability of the target assembly component, thereby realizing the quantitative discrimination of the component category. Formula (3) uses probability quantization to replace the traditional threshold hard classification. By outputting confidence probability instead of binary labels, it can more accurately reflect the matching degree between the component and the target sub-component, effectively reducing the risk of missed and misidentified identification of complex nested sub-components such as irregular curtain wall connectors and sub-frames, and improving the robustness and accuracy of target assembly component identification.

[0097] Target sub-component matching confidence probability: a standardized probability value in the range of 0 to 1, representing the credibility of the model inference for the target sub-components such as the embedded sealing strip and buckle in the current component.

[0098] Step S140: If the matching confidence probability of the target sub-component is greater than the preset component recognition probability threshold, it is determined to be the target sub-component, and the target assembly component is determined based on the target sub-component, thus completing the recognition of the target assembly component.

[0099] The system retrieves a preset component recognition probability threshold to perform a probability logic comparison, filters high-matching component nodes that exceed the threshold, and marks the embedded target sub-components; it then traces back to the parent main assembly component to which the target sub-component belongs, and collects and completes the accurate identification and partitioning of all target assembly components.

[0100] The formula for determining the threshold of the target assembly component is: (4) In formula (4), Match the confidence probability of the target sub-component, which is unitless and ranges from 1 to 2. , is the input for this step, that is, the output result of formula (3); The preset component identification probability threshold is a unitless value, which is a calibration constant for the prefabricated curtain wall model. The optimal value for the project is 0.82. The conditional statement is a component identification branch control instruction, which is a unitless value and is linked to the BIM component code index binding output. It is used to mark the component type and lock the upper-level target prefabricated component. The control logic of formula (4) is based on the preset component identification probability threshold calibrated by the project. Confidence probability of matching target sub-components Binary partitioning is performed. When the matching confidence probability of the target sub-component is greater than the preset component identification probability threshold, the component is determined to be a target assembly component containing nested sub-components; otherwise, it is determined to be an ordinary curtain wall base component. Formula (4) realizes the identification of nested sub-components based on topological fusion features and probability soft thresholds. It can accurately screen out target assembly unit panel components with nested sub-components such as bolts, sub-frames, and pads. It is compatible with the modular nested assembly structure of curtain wall unit panels and solves the defect of traditional hard threshold classification that easily leads to the omission of identification of irregular components.

[0101] The preset component recognition probability threshold is fixed at 0.82, with a range of 0.78 to 0.84. The value is determined based on the same source as the classification confidence threshold in step S100, ensuring that the judgment criteria of the sub-step and the parent step are consistent and avoiding recognition logic conflicts.

[0102] Preferably, the lightweight processing and assembly simulation method for the three-dimensional model of the curtain wall unit panel provided in this embodiment includes step S200: S210. Obtain the three-dimensional mesh data of the target assembly component, and extract the overall mesh volume characteristics and mesh complexity characteristics of the component from the three-dimensional mesh data.

[0103] Batch retrieval of the original triangular facet 3D mesh model data of the target assembly component identified in step S140, stripping away the model texture and redundant material mapping information, and statistically analyzing two core quantitative features: the total volume of the global mesh bounding box of the component's closed mesh entity and the mesh topological heterogeneity complexity, to complete the independent extraction and normalization preprocessing of dual features.

[0104] The formulas for extracting the total volume feature and mesh complexity feature of the component mesh are as follows: (5) (6) In formulas (5)~(6), The total volume characteristic of the component mesh, in cubic meters ( The value is obtained by solving the BIM triangular mesh voxels, and the range is [value range missing]. ; For the first The volume of a single triangular grid, in cubic meters. ); This represents the total number of original triangular meshes for the component, without units, and its value ranges from [value range missing]. ; This represents the grid complexity feature, expressed in units per square meter (units / This characterizes the density of curved surfaces, chamfers, and openings in a component. This represents the number of effective mesh surfaces of a component, without units. The surface area of ​​the component's outer contour is expressed in square meters. The range of values ​​is The control logic of formulas (5) and (6) is to calculate the total mesh volume characteristics and mesh complexity characteristics of the component respectively, and to remove redundant and noisy meshes from the three-dimensional mesh data, providing a two-dimensional input for subsequent saliency weighting calculation. Formulas (5) and (6) adopt a two-dimensional feature coupling method, which takes into account both the volume characteristics and the detailed structural complexity characteristics of the component. This solves the defect of traditional single volume lightweighting methods that easily lead to distortion of details such as curved surfaces and openings, and provides a more accurate feature basis for subsequent hierarchical lightweighting processing.

[0105] 3D mesh data: The underlying dataset of the BIM model mesh, consisting of the original triangular facet mesh topology of the target assembly component, facet vertices, boundary edges, and solid closed surfaces.

[0106] Component mesh total volume characteristics: The true meshed integral volume of the component closed mesh entity bounding box is used as a precise quantification sub-index for the volume characteristics in step S200.

[0107] Mesh complexity characteristics: The comprehensive quantitative index of mesh vertex density, number of curvature abrupt changes in irregular curved surface patches, and proportion of broken patches at assembly boundary is the underlying data source for the complexity characteristics of step S200.

[0108] Mesh data preprocessing thresholds: 100% invalid free fragments are removed, and the effective mesh retention rate of a single component is ≥99%.

[0109] Step S220: Perform a weighted summation operation on the total volume feature and mesh complexity feature of the component mesh to obtain the comprehensive saliency of the target assembly component.

[0110] The configuration engineering is calibrated with fixed weight coefficients. Linear weighted fusion and summation are performed on the normalized total grid volume characteristics and grid complexity characteristics. After eliminating the difference in dimensions, the standardized 0~1 interval of the target assembly component comprehensive significance is output, which is completely homologous and equivalent to the assembly feature significance value of the main step.

[0111] The formula for calculating the comprehensive significance of the target assembly component is: (7) In formula (7), The comprehensive significance of the target assembly component is given, without units, and its value range is [value range missing]. The larger the value, the more critical the detailed assembly features of the component; The normalized volume feature is dimensionless, and its range after min-max normalization is [value missing]. ; The normalized grid complexity feature is dimensionless and has a range of 1. ; This is a volumetric characteristic weighting coefficient, dimensionless, and a fixed value specified in the engineering specifications for curtain wall components; , where is the weighting coefficient for complexity features, is dimensionless, and represents the weighting parameter for detailed assembly structures. The control logic of formula (7) is to first normalize the volumetric and complexity features to eliminate the influence of dimensions; then, using the weighting coefficients calibrated in engineering... , Linear combination is performed to obtain the comprehensive significance of the component, which is used to quantify the assembly critical level of the component and provide a basis for subsequent graded lightweighting processing. Formula (7) prioritizes the retention of high-precision assembly features such as curtain wall bolt holes, lap edges, and tenons by differentially allocating the weights of complexity features, thus avoiding the oversimplification of key assembly details by traditional lightweighting methods and effectively avoiding the interference verification failure problem in subsequent assembly simulation.

[0112] Weighted fixed coefficient values: total grid volume feature weight 0.42, grid complexity feature weight 0.58; value basis: conforming to the curtain wall assembly boundary complexity priority weighting engineering rules, matching the main step saliency classification logic.

[0113] Comprehensive significance of target assembly components: The sub-step operation outputs a fused quantitative index, which is equivalent to the assembly feature significance value of the main step, and the hierarchical intervals are completely unified.

[0114] Step S230: Perform hierarchical lightweighting on the target assembly component based on the comprehensive saliency of the target assembly component to obtain the component after hierarchical lightweighting model.

[0115] The comprehensive saliency level classification standard is retrieved, and the corresponding level-specific mesh simplification rate and assembly boundary locking constraint parameters are matched. The hierarchical lightweight algorithm is called to complete the component differential mesh reduction and micro-assembly feature locking retention, and the full-domain component hierarchical lightweight model is generated in batches.

[0116] The calculation formula for the model after component classification and lightweighting is as follows: (8) (9) In formulas (8)~(9), The grid simplification compression ratio is unitless and has a range of [value range missing]. A larger value indicates a higher degree of mesh simplification. This is a lightweight adjustment coefficient, dimensionless, a fixed value for engineering calibration, and is 0.88. The comprehensive significance of the target assembly component is unitless and is the input for this step, i.e., the output result of formula (7); This is the lightweight mesh model after component classification and simplification, without units, and is the output of this step. To preserve the feature mesh, a lightweight operator for reducing the number of faces is used. It is unitless and is used to simplify the mesh while retaining key geometric features. The original high-precision mesh model of the target component is unitless and serves as the input model for lightweight processing. The control logic of formulas (8) to (9) is to first calculate the mesh simplification and compression ratio based on the comprehensive saliency of the component. The higher the saliency of the component, the lower the compression ratio, in order to retain more assembly details; the lower the saliency of the component, the higher the compression ratio, in order to reduce redundant meshes. Then, the feature-preserving mesh reduction operator is called to process the original high-precision mesh model to obtain the component model after hierarchical lightweighting. Formulas (8) to (9) adopt an adaptive dynamic simplification rate, which is differentiated according to the assembly key level of the component. It takes into account both the computational power compression requirements of the overall model and the retention of key features such as assembly overlap and openings, and solves the pain point that the traditional uniform lightweighting method is prone to interference verification distortion.

[0117] The original high-precision mesh of the target component is subjected to edge folding and surface reduction processing based on a quadratic error metric. Before the reduction, the overlapping edge, sealing seam boundary, hole boundary, hoisting reference surface boundary and target sub-component fitting boundary are set as feature locking edges. The vertices of the feature locking edges do not participate in edge folding or are only allowed to move within a preset error range.

[0118] Step S240: Extract the spatial coordinates and unique identifiers of the model after component grading and lightweighting, and generate a component dual association index matrix containing a three-dimensional spatial location index and a curtain wall component unique code index.

[0119] The absolute spatial coordinates of the three-axis center of each component of the lightweight model and the globally unique identification code of the prefabrication are extracted in batches. The mapping relationship is encapsulated in a matrix with the component entity as the row dimension and the dual-type index as the column dimension, and a two-way binding and one-to-one corresponding component dual association index matrix is ​​constructed.

[0120] The formula for generating the component dual-association index matrix is: (10) In formula (10), This is a component dual-association index matrix, which is unitless and is the core output of this step, used to bind the spatial location and identity information of the component. This is a three-dimensional spatial location index, in meters (m), and represents the coordinates of the center of the component's bounding box. , used to locate the spatial position of components in the scene; The unique coded index for the curtain wall component is unitless and is the original component UUID encoding sequence in the BIM model, used to uniquely identify the component. The control logic of formula (10) is to concatenate the three-dimensional spatial position index of the component with the unique coded index of the curtain wall component to construct a bidirectional mapping dual-association index matrix, providing support for subsequent collision localization and high-precision model restoration. Formula (10) adopts a dual-index linkage architecture, simultaneously recording the spatial position and identity code of the component, realizing one-click mapping between the spatial positioning of the collision area and the original high-precision model, which greatly improves the efficiency and accuracy of subsequent local high-precision model restoration.

[0121] The component dual-association index matrix includes at least the component's unique code, the lightweight model component identifier, the original model component identifier, the minimum and maximum coordinates of the component's bounding box, the coordinate transformation matrix between the lightweight model and the original model, and the original high-precision mesh data storage address.

[0122] Furthermore, the lightweight processing and assembly simulation method for the three-dimensional model of the curtain wall unit panel provided in this embodiment includes step S300: Step S310: Obtain the component dual association index matrix and the component graded lightweight model, arrange the process of the component graded lightweight model according to the construction logic rules, and obtain the curtain wall unit panel time sequence assembly sequence.

[0123] The system links and binds a dual-association index matrix with lightweight model component entities, loads standardized construction logic constraint rules for curtain walls, and completes iterative optimization of process sorting based on floor vertical priority, component assembly dependency relationship, and finishing construction priority. It outputs a standardized curtain wall unit panel time sequence assembly sequence with timestamp and component code binding.

[0124] The formula for generating the sequence of curtain wall unit panels assembly is: (11) In formula (11), This is the sequential assembly sequence of curtain wall unit panels, without units, and is a set of sequentially ordered components. This is a construction rule sorting operator with no unit, used to arrange components in an orderly manner according to construction logic. The component is a dual-association index matrix, which has no unit and is the input for this step, i.e., the output result of formula (10); The construction logic constraint function for the curtain wall is unitless and follows the construction logic rules, providing constraints for the sorting process. The construction logic rules include floor direction rules, zoning direction rules, and component dependency rules. In one embodiment, the floor direction rule is from bottom to top, the zoning direction rule is from one side to the other, and the component dependency rule is to install the load-bearing components first and then the panel components. Different engineering projects can configure the floor direction rules according to the construction plan. The control logic of formula (11) is based on the component dual association index matrix, calls the construction rule sorting operator, and combines the curtain wall construction logic constraint function to arrange the process of the component after the component is graded and lightweighted, generating a curtain wall unit panel time sequence assembly sequence that conforms to the actual construction process. Formula (11) directly generates a curtain wall unit panel time sequence assembly sequence that can guide the on-site construction based on the model data through the linkage of BIM component index and construction process rules, which solves the problem of the assembly sequence being disconnected from the actual construction logic in traditional simulation, and provides a real and reliable process basis for subsequent kinematic interference simulation.

[0125] Step S320: Extract the hoisting and locking actions of the corresponding components based on the timing assembly sequence of the curtain wall unit panels, and generate the kinematic simulation trajectory of the components based on the hoisting and locking actions.

[0126] Disassemble the process labels within the sequential assembly sequence, extract the rigid body hoisting and translation, attitude fine-tuning hoisting actions of the unit plate, and the bolt axial feed and lateral clamping and locking actions; bind the motion rate working parameters, fit the six-degree-of-freedom rigid body motion curve, and generate the kinematic simulation trajectory of the sequential synchronous component.

[0127] The formula for generating the kinematic simulation trajectory of a component is: (12) In formula (12), for The kinematic simulation trajectory of the component at any time, where the translational displacement is in meters (m) and the rotational attitude angle is in radians (rad). The translational displacement of the component during hoisting in three axes is expressed in meters (m), describing the change in the component's spatial position. The rotation angle of the component around the three axes is expressed in radians (rad), describing the component's attitude adjustment process. The control logic of formula (12) is based on the sequential assembly sequence of the curtain wall unit panels, defining the translational displacement and rotational attitude of each component as it changes over time, and constructing a complete six-degree-of-freedom motion trajectory for subsequent kinematic interference simulation. Formula (12) uses six-degree-of-freedom kinematic modeling to simulate the lifting translation and locking attitude adjustment of the components, realistically restoring the entire process of the curtain wall unit panels from lifting to installation, providing reliable trajectory data support for high-precision interference simulation.

[0128] Component kinematic simulation trajectory: The dynamic motion space path curve of the component is bound with timestamp, displacement vector, and attitude rotation angle, which perfectly matches the mechanical operation motion conditions on the construction site.

[0129] Step S330: Construct the dynamic spatial bounding box corresponding to each component using the component kinematic simulation trajectory.

[0130] The component's axial three-dimensional bounding box boundary coordinates are updated frame by frame following the temporal motion trajectory. Redundant bounding boxes in the motion spatial domain are eliminated, and a time-varying dynamic spatial bounding box that fits the component's real-time posture and real-time spatial position is constructed.

[0131] The formula for constructing a dynamic bounding box is: (13) In formula (13), for The dynamic spatial bounding box of the time component is represented by [Xmin(t),Xmax(t),Ymin(t),Ymax(t),Zmin(t),Zmax(t)], where the unit of each coordinate component is meters (m), and it is the hexahedral spatial boundary of the component; An operator for constructing axis-aligned dynamic bounding boxes, with no units, designed for lightweight simulation acceleration, used to quickly construct component boundaries aligned with coordinate axes; The model after component classification and lightweighting is unitless and represents the output of step S230. for The kinematic simulation trajectory of the component at any given time is dimensionless and is the output of step S320. The control logic of formula (13) is based on the lightweight model and kinematic trajectory of the component, and calls the axis-aligned bounding box operator to construct a dynamic spatial bounding box that changes over time for subsequent fast interference detection. Formula (13) uses axis-aligned dynamic bounding box (AABB) technology to simplify the complex mesh model into a regular hexahedral boundary. Without affecting the interference judgment result, it greatly improves the computational efficiency of kinematic interference simulation and solves the computational bottleneck problem when simulating a large number of components at the same time.

[0132] Dynamic spatial bounding box: The smallest enclosing closed cube of a component whose boundary coordinates are updated in real time with the simulation sequence, serving as the basic computational carrier for dynamic collision screening.

[0133] Step S340: Determine whether the dynamic spatial bounding boxes of different components overlap. If they overlap, then the overlapping part is determined as a candidate collision region.

[0134] The system iterates through the dynamic spatial bounding boxes of the global components frame by frame, performs Boolean intersection discrimination operation on the bounding box boundaries, marks the intersecting and overlapping closed spatial domains of cubes within the temporal frame, clusters and integrates the continuous temporal overlapping spatial domains, collects and outputs all candidate collision regions, and completes the coarse screening and interference localization.

[0135] The formula for dynamic bounding box intersection determination and candidate collision region extraction is as follows: (14) In formula (14), For the first , The dynamic spatial bounding box of adjacent components, in meters (m), is the output of step S330; The set of candidate collision regions for simulation is a unitless set. For each candidate region in the set of candidate collision regions for simulation, calculate its bounding box overlap volume or solid overlap volume, with the volume unit being cubic meters (m³), which is the core output of this step. An empty set represents a state of no spatial interference or overlap. The control logic of formula (14) is to perform an intersection operation on the dynamic spatial bounding boxes of adjacent components. If the intersection is not empty, a potential collision is determined, and the intersection region is marked as a candidate collision region; if the intersection is empty, no collision is determined. The creativity of formula (14) lies in: 1. Optimization of computing power: The axis-aligned bounding box (AABB) coarse screening technology is adopted to simplify the collision detection of complex mesh models into the intersection judgment of regular hexahedrons, which greatly reduces the computing overhead of large-scale assembly simulation. 2. Fast tracing: Combined with the component dual association index matrix generated in step S240, the candidate collision region can be directly bound to the component identity to realize the rapid location of the colliding component, providing support for subsequent high-precision model restoration and accurate verification.

[0136] Preferably, the lightweight processing and assembly simulation method for the three-dimensional model of the curtain wall unit panel provided in this embodiment includes step S400: Step S410: Obtain candidate collision regions, and perform spatial discretization processing on the candidate collision regions to obtain a discrete voxel set.

[0137] The closed candidate collision space is imported from step S340. The uniform cube partitioning operator is used to perform three-dimensional spatial discretization on the continuous spatial region. The continuous closed interference space is decomposed into standard cubic voxel elements of equal size and aggregated to generate a global discrete voxel set, thus completing the meshed discretization solution of the interference region.

[0138] The formula for discretizing the spatial voxels of the candidate collision region is: (15) In formula (15), This is the set of candidate collision regions for simulation, in cubic meters (m³), and represents the output of step S340. The unit is a discrete spatial voxel, with the unit being cubic meters (m³), and the size of a single voxel is... ; The total number of voxels in the candidate region is unitless. The control logic of formula (15) is to divide the candidate collision region into a large number of tiny regular voxels, forming a discrete voxel set, which provides the basis for accurate collision detection of the subsequent high-precision mesh. The creativity of formula (15) lies in: 1. Spatial discretization: By discretizing the continuous collision region into tiny voxels, the accurate approximation of the boundary of the complex surface is realized, which lays the foundation for the accurate verification based on the original high-precision model. 2. Balance between accuracy and efficiency: adopting The voxel size ensures sufficient detection accuracy while avoiding the computational cost of directly performing full detection on the original high-precision model.

[0139] Spatial discretization processing: The continuous three-dimensional closed spatial domain uniform cube decomposition calculation process realizes the discrete accumulation calculation of interference volume.

[0140] Discrete voxel set: A dataset of all homogeneous cubic voxels after the candidate collision region is partitioned.

[0141] Step S420: Calculate the cumulative space occupancy of intersecting voxels based on the discrete voxel set to obtain the total actual overlapping volume of component interference.

[0142] By filtering out the overlapping voxel units occupied by different component entities, summing the fixed volume of each voxel, removing blank and invalid spatial voxels, and performing an integral operation to output the total space occupied by the intersecting voxels in the entire domain, which is the actual overlapping volume of the component interference.

[0143] The formula for calculating the total volume of actual overlap due to component interference is: (16) In formula (16), The total volume of actual overlap of the components is expressed in cubic meters. ), which is the core indicator of interference quantification output in this step; The unit is a discrete spatial voxel, and the unit is cubic meters ( ), that is, the voxel unit in formula (15); This is a binary identifier for voxel overlap, with no unit. It is set to 1 when a voxel is simultaneously occupied by the high-precision models of two components, and 0 otherwise. The total number of elements in the candidate region is unitless and corresponds to the number in formula (15). Consistent. The control logic of formula (16) is to traverse all voxels within the candidate collision region, determine whether each voxel is occupied by two components simultaneously, and if they overlap, accumulate the volume of that voxel to finally obtain the actual total overlapping volume of component interference. The creativity of formula (16) lies in: 1. Interference quantification: By using the voxel method, the fuzzy "collision" event is transformed into a quantifiable overlapping volume index, providing an accurate basis for assessing the severity of interference. 2. Efficiency and accuracy balance: Only the candidate collision region selected by formula (14) is subjected to voxel-level accurate detection, rather than traversing the high-precision model of the entire scene, which greatly reduces the computational cost while ensuring accuracy.

[0144] The actual total volume of component interference overlap is the volume value of the discrete voxels accumulated in the sub-step, which is equivalent to the overlap volume of the candidate collision region in the main step, and the data are from the same source and are consistent.

[0145] Step S430: Determine whether the total volume of actual overlap of component interference is greater than the preset critical threshold for interference volume.

[0146] The preset critical threshold for interference volume is invoked to complete the logical comparison of the measured overlapping volume and divide the sample into two types: pseudo-interference noise region and real assembly interference region.

[0147] Call the main step S400 to preset the critical threshold of the interference volume, complete the logical comparison of the measured overlapping volume, and divide the sample into two types: pseudo-interference noise region and real assembly interference region.

[0148] The formula for determining the interference volume threshold is: (17) In formula (17), The total volume of actual overlap of the components is expressed in cubic meters. ), which is the input for this step, i.e., the output of formula (16); The preset critical threshold for interference volume is given in cubic meters. ), which is the calibration constant for curtain wall engineering, and its value is . The control logic of formula (17) is to compare the actual total volume of the component interference with the preset critical threshold of interference volume, and trigger different subsequent processes according to the comparison result: if the preset critical threshold of interference volume is exceeded, the high-precision model restoration is triggered for secondary verification; if it is not exceeded, the verification result of the lightweight model takes effect directly. The creativity of formula (17) is: 1. Hierarchical verification strategy: by introducing the interference volume threshold calibrated by the engineering, the hierarchical processing of "coarse screening-fine verification" is realized, avoiding the need to perform high-precision model restoration for all candidate collision areas, and greatly improving the overall simulation efficiency. 2. Engineering practicality: preset critical threshold of interference volume The design takes into account both the reasonableness of minor gap errors in curtain wall assembly and the hard interference problems that must be corrected in actual construction, ensuring that the simulation results have guiding significance for on-site construction.

[0149] The pseudo-interference noise region is a false geometric overlap region induced by the model's hierarchical lightweight mesh reduction and surface patch polyline distortion. It does not pose an actual engineering assembly interference risk and is an inherent noise region of the simulation.

[0150] Step S440: If the total volume of actual overlap of component interference is greater than the preset critical threshold of interference volume, then extract the spatial coordinate mapping interval corresponding to the candidate collision region.

[0151] For the qualified real interference region, the three-axis extreme coordinates of the bounding box are extracted, and the bounding box is encapsulated to generate a continuous closed, indexable and bindable regional spatial coordinate mapping interval, thus completing the coordinate labeling of the interference region.

[0152] The formula for calculating the spatial coordinate mapping interval is: (18) In formula (18), The spatial coordinate mapping interval of the collision region is expressed in meters (m), and this is the output of this step. This is a coordinate mapping operator used to bind candidate collision regions to the three-dimensional spatial position index of the component; This is the set of candidate collision regions for simulation, in cubic meters (m³), and represents the output of step S340. The three-dimensional spatial position index of the component is in meters (m), and the spatial coordinate part is in the component dual association index matrix generated in step S240. The control logic of formula (18) is to use the component dual association index matrix generated in step S240 to associate the candidate collision area with the three-dimensional spatial position of the component, thereby locating the specific position of the collision area in the original high-precision model, and providing coordinate basis for subsequent local high-precision model restoration. The creativity of formula (18) is as follows: 1. Hierarchical verification strategy: adopting the "coarse screening-fine core" mode, the results of small volume slight interference detected by the lightweight model are directly accepted, and high-precision model restoration is triggered only for the heavily interfered area that exceeds the preset interference volume critical threshold, thus achieving a balance between computing power and accuracy. 2. Efficient source tracing: through the spatial mapping of the component dual association index matrix, it is not necessary to restore the entire model, but only to perform local high-precision verification on the collision area, which greatly reduces the computing overhead and solves the problem of computing power bottleneck in full-scene high-precision simulation.

[0153] Spatial coordinate mapping interval: The closed coordinate interval of the extreme values ​​of the X / Y / Z axes serves as a unique index label for cross-model high-precision regional tracing and matching.

[0154] Step S450: Combine the spatial coordinate mapping interval and the component dual association index matrix to extract the corresponding high-precision collision area model from the initial building model.

[0155] Linked coordinate mapping intervals, component dual encoding bidirectional index constraints, cross-lightweight model-initial native model topology matching, clipping and retrieving the original local mesh model of the initial building model, completing local model restoration, and outputting a high-precision collision area model.

[0156] The formula for extracting high-precision collision region models is: (19) In formula (19), This is a high-precision collision region model, without units, and is the output of this step. A local high-precision model recovery operator is used to extract the target region from the original model; The spatial coordinate mapping interval of the collision region is in meters (m), and it is the output result of step S440. This is a unique coded index for the curtain wall component, without units, and is the component identity part in the component dual association index matrix generated in step S240; The initial high-precision BIM model of the building is unitless and serves as the input model for the entire process. The control logic of formula (19) combines the coordinate mapping interval of the collision area obtained in step S440 with the unique component code in the component's dual-association index matrix to accurately extract the corresponding local model from the initial high-precision BIM model, restoring only the high-precision mesh data of the collision area. The creativity of formula (19) lies in: 1. Localized restoration: It does not restore the entire component, but only restores the local model of the collision area, which greatly reduces the amount of data and achieves the ultimate optimization of computing power. 2. Index linkage: It directly locates the component in the original model through the unique component code index, and then combines the spatial coordinate mapping to extract the target area, ensuring the accuracy and speed of model restoration.

[0157] Furthermore, the lightweight processing and assembly simulation method for the 3D model of the curtain wall unit panel provided in this embodiment includes step S500: Step S510: Use a spatial geometry intersection algorithm to calculate the boundary contour set of the high-precision collision region to obtain the intersecting closed polygon contour of the model.

[0158] Extract the set of outer boundary contours of the two assembled components in the high-precision collision area, call the spatial geometric Boolean intersection algorithm to complete the intersection operation of the two sets of boundary contour sections, remove open broken contour edges, and fit to generate closed-loop, unbroken model intersecting closed polygon contours.

[0159] The formula for calculating the contour of intersecting closed polygons in the model is: (20) In formula (20), The output of this step is the closed polygonal outline of the intersecting model, expressed in meters (m), and is used to accurately describe the boundary of the collision region. This is a spatial geometry intersection operator used to perform Boolean intersection operations on the boundary contours of a model. The set of boundary contours of the high-precision collision area model is in meters (m), and the output of step S450 is given. The control logic of formula (20) is to extract the set of boundary contours of the high-precision collision area model restored in step S450, and then calculate the intersecting closed polygon contours between components through a spatial geometric intersection algorithm to obtain the accurate collision boundary. The creativity of formula (20) is as follows: 1. Accuracy calibration: By performing geometric intersection on the boundary of the high-precision model, the collision area detected by the lightweight model is made more accurate, and the real collision contour is obtained, providing a reliable basis for interference correction. 2. Error elimination: The detection error caused by mesh simplification of the lightweight model is eliminated, ensuring the accuracy of the final result and providing accurate geometric data for subsequent construction adjustments.

[0160] Boundary contour set: A dataset of aggregated boundary contours of the fitting section of components in the high-precision collision area and the boundary lines of the sealing interface.

[0161] The intersecting closed polygonal contour of the model: the edge of the closed-loop contour of the interference section of the two-component assembly, which is the core calculation carrier for solving the dimensions and interference depth.

[0162] Step S520: Based on the intersecting closed polygonal contours of the model, obtain the width of the curtain wall assembly sealing joint and the overlap length of the unit panel.

[0163] The normal spacing of the closed profile is sampled along the normal direction of the sealing interface to solve the width of the curtain wall assembly sealing joint; the length of the embedded line segment of the profile is intercepted along the axial direction of the profile overlap to quantify the overlap length of the unit panel, thus completing the direct calculation of the two core engineering dimensional parameters.

[0164] The formula for calculating the width of the sealing joint and the overlap length of the unit panel in the curtain wall assembly is as follows: (twenty one) (twenty two) In formulas (21)~(22), The width of the sealing joint for the curtain wall assembly, expressed in millimeters (mm), is one of the core parameters output in this step. A reasonable range of values ​​for this step in the project is [insert range here]. ; This is the normal distance projection operator, used to calculate the normal distance of intersecting profiles in the direction of the sealing joint; The model intersecting closed polygonal contours are in meters (m) and represent the output of step S510. This is the overlap length of the unit board, in millimeters (mm), and is another core parameter output in this step. This is the Euclidean distance operator, used to calculate the straight-line distance between two points; The feature control points on the intersecting contour boundary are in millimeters (mm) and are used to define the boundary of the overlapping area. The control logic of formulas (21) to (22) is to project the closed contour obtained by the intersection of the high-precision model onto the normal direction and calculate the minimum distance between the components, which is the width of the sealing joint of the curtain wall assembly. By calculating the Euclidean distance between the corresponding feature points on the contour boundary, the actual overlap length between the unit panels is obtained. The creativity of formulas (21) to (22) is as follows: 1. Parametric output: The joint width and overlap length parameters required for construction are directly derived from the geometric interference results, realizing the seamless connection between simulation data and engineering construction standards. 2. Accuracy guarantee: The calculation is based on the intersecting contour of the high-precision model, avoiding the simplification error of the lightweight model, ensuring the accuracy of the final parameters, and providing a reliable basis for the curtain wall sealing performance and structural safety verification.

[0165] Step S530: If the overlap length of the unit plate is greater than the preset overlap depth judgment threshold, then extract the inscribed sphere radius of the intersecting closed polygon contour of the model to determine the assembly interference depth.

[0166] The preset overlap depth judgment threshold is loaded to complete the size comparison. For interference samples with excessive overlap length, the equivalent radius of the maximum inscribed sphere of the closed polygon contour is solved, and the normalized inscribed sphere radius is calibrated as the axial assembly interference depth of the component.

[0167] The formula for calculating assembly interference depth is: (twenty three) In formula (23), The overlap length of the unit board is in millimeters (mm), and it is the output result of step S520. The preset overlap depth threshold is in millimeters (mm), and the engineering value is [value missing]. ; The assembly interference depth is expressed in millimeters (mm) and is the output of this step. This is a function for solving the maximum inscribed sphere radius of intersecting polygons, used to calculate the maximum inscribed circle radius within a closed profile, i.e., the interference depth. The model intersecting closed polygon contour, in meters (m), is the output of step S510. The control logic of formula (23) is as follows: 1. First, determine whether the overlap length of the unit plate exceeds the preset overlap depth judgment threshold. 2. If it exceeds the limit, the maximum inscribed sphere radius of the intersecting contours is calculated and used as the assembly interference depth to quantify the severity of the interference. The creativity of this formula (23) lies in: 1. Quantifying the degree of interference: It not only judges whether a collision has occurred, but also quantifies the depth of interference by calculating the inscribed sphere radius, providing a more refined basis for construction adjustments. 2. Threshold triggering mechanism: Depth calculation is only performed when the overlap length exceeds the threshold, avoiding misjudgment of normal overlaps and improving the verification efficiency and engineering practicality.

[0168] Preset overlap depth judgment threshold: fixed value of 5mm, range of 4~6mm; value basis: matching the upper limit threshold of standard overlap length in the main step, exceeding the threshold is judged as profile extrusion intrusion interference.

[0169] The radius of the inscribed sphere is the radius of the largest inscribed sphere inside the interference closed polygon, and it is equivalent to the intrusion depth of the component entity.

[0170] Step S540: Combine the assembly interference depth and the width of the curtain wall assembly sealing joint to map the original feature set of component penetration interference, and obtain a set of simulation accurate verification results including interference depth and collision type.

[0171] The system retrieves the original feature standard library of pre-built component penetration interference, binds the measured sealing gap width and interference depth values ​​to complete feature label matching and mapping, automatically matches the corresponding standardized collision type label, encapsulates all verification parameters in a structured manner, and outputs a set of standardized simulation accurate verification results.

[0172] The formula for generating the set of simulation accurate verification results is: (twenty four) In formula (24), This is the set of simulation-accurate verification results, the final output of this method, which contains the complete features of component penetration interference; The assembly interference depth is expressed in millimeters (mm), and this is the output result of step S530. The width of the sealing joint for the curtain wall assembly is in millimeters (mm), and this is the output result of step S520. To standardize collision type labels, no units are used. The collision type labels are uniformly defined as: 1=panel / profile overlap interference, 2=keel and civil structure hard collision, 3=connector or bolt penetration interference, 4=sealing sub-component extrusion interference, 5=lifting tool temporary collision. The control logic of formula (24) is to map the assembly interference depth and sealing gap width obtained in the above steps with the standardized collision type labels to form a structured set of verification results. The creativity of formula (24) is: 1. Structured output: binding the geometric calculation results with engineering semantic labels, directly outputting structured data that can guide construction without secondary interpretation. 2. Full-dimensional feedback: the result set includes the "severity (depth)", "sealing performance (gap width)" and "root cause (type)" of the interference, providing a comprehensive and accurate basis for subsequent construction adjustments.

[0173] Preferably, the lightweight processing and assembly simulation method for the three-dimensional model of the curtain wall unit panel provided in this embodiment includes step S600: Step S610: Obtain the set of simulation accurate verification results, and extract the set of initial BIM attribute labels for components based on the set of simulation accurate verification results.

[0174] Step S540 outputs the verification result data packet. Based on the dual association index matrix, it traces and matches the target component corresponding to the initial building model, and batch retrieves the original archived fields of the component's original factory material, process, and assembly binding, and collects and generates the initial BIM attribute label set of the component.

[0175] Initial BIM attribute tag set for components: The initial model does not modify the original component background inherent attribute field cluster, and has no simulation interference annotations or defect markers in the original tag group.

[0176] Step S620: Use the simulation accurate verification result set to correct the initial BIM attribute label set of the component to obtain the target component attribute label after interference correction.

[0177] Based on the parameters of interference level, gap width exceeding standard, and collision type within the set of simulation-accurate verification results, the initial BIM attribute label set of the component is overwritten and supplemented, and new defect annotations, rectification priorities, and interference coordinate extension labels are added. The target component attribute labels after interference correction are iteratively generated.

[0178] The formula for calculating the attribute label of the target component after interference correction is: (25) In formula (25), These are the attribute labels for the target component after interference correction. They have no units and represent the final output of this step. This is a set of initial BIM attribute tags for a component, without units, containing information such as the component's original material, dimensions, and installation location; This is a label overlay fusion operator with no units, used to merge simulation verification results into BIM attributes; The set of simulation verification results is unitless and represents the output of step S540. The control logic of formula (25) is to merge the set of simulation verification results obtained in step S540 into the initial BIM attribute label set of the component through the label overlay fusion operator, and generate the corrected target component attribute label containing interference information. The creativity of formula (25) lies in: 1. BIM model closed loop: It realizes a complete closed loop from BIM model input to simulation verification and then to the corrected BIM model output, and directly feeds the interference information back to the design model. 2. Information enhancement: Through the label fusion operator, not only is the geometric information of the component corrected, but also the interference depth, joint width and collision type information are added as new attribute labels to the BIM model, providing data support for subsequent construction management and operation and maintenance.

[0179] Step S630: Generate a model status flag instruction based on the target component attribute label after interference correction. The model status flag instruction is used to trigger the model status flag to complete the numerical change.

[0180] The corrected component attribute labels are mapped to binary codes to generate electronic control status flag instructions that the model can recognize; the instructions are bound to the model storage background status register to trigger the switching and updating of binary values ​​of the model's original working condition status bits and defect verification status bits.

[0181] The formula for calculating the model status flag is: (26) In formula (26), This is the model state flag, which has no unit and is the final output of this step. This is a label mapping function used to map component attribute labels to status identifiers; The target component attribute label after interference correction is unitless and is the output result of step S620; the value definition is: 0: no interference, 1: slight interference, 2: severe assembly interference. The control logic of formula (26) is to convert the corrected component attribute label into a standardized model state identifier bit through the label mapping function. The creativity of this formula (26) is: 1. State standardization: simplifying complex interference information into 3 states, which facilitates unified state management and display in the BIM system. 2. Triggering mechanism: the generated state identifier bit can directly trigger the state change of the BIM model, such as changing the component color or adding a warning mark, realizing the visualization feedback of the simulation results.

[0182] Status command trigger accuracy threshold: Flag bit command trigger matching rate ≥ 99.7%, consistent with the main step write-back threshold.

[0183] Step S640: Reconstruct the simulation scene map based on the model status identifier, and write back the set of simulation accurate verification results to the initial building model to complete the assembly simulation operation.

[0184] Based on the updated model status flags, the topology map of the full-domain curtain wall simulation scene is reconstructed. Simultaneously, the verification result data package is completed, the attribute labels are corrected and backfilled in batches to bind the original initial building model, the intermediate data of the entire simulation process is archived, and the entire assembly simulation operation is completed in a closed loop.

[0185] The formula for calculating the reconstructed simulation scene map is as follows: (27) In formula (27), The reconstructed simulation scene map is the final output of this method. Global graph update operators are used to update the topology and properties of the scene; The global topology map of the initial building model; This is the model status flag bit, which has no unit and is the output result of step S630. The target component attribute label after interference correction is unitless and is the output result of step S620. The control logic of formula (27) is based on the global topology map of the initial building model; the state of the component is updated according to the model state identifier; the corrected component attribute label is written back to the model; and an updated simulation scene map containing interference information is generated. The creativity of formula (27) lies in: 1. Closed-loop linkage: It realizes a complete closed loop from BIM model input, simulation verification to the reverse writing of the result back to the BIM model, so that the simulation result directly affects the iteration of the design model. 2. Full scene update: It not only updates the attributes of individual components, but also reconstructs the global topology map, ensuring the consistency and integrity of the model, which meets the iterative process requirements of the curtain wall BIM as-built model.

[0186] Please see Figure 2This embodiment provides a lightweight processing and assembly simulation system for a 3D model of a curtain wall unit panel, used to implement the aforementioned lightweight processing and assembly simulation method for a 3D model of a curtain wall unit panel. It includes a target assembly component identification module 10, a component dual-association index matrix acquisition module 20, a candidate collision region acquisition module 30, a high-precision collision region acquisition module 40, a simulation accuracy verification result set acquisition module 50, and an assembly simulation task completion module 60. The target assembly component identification module 10 is used to acquire the component geometric dimensions, component attributes, and component assembly connection relationships in the initial building model. Based on the component geometric dimensions, component attributes, and component assembly connection relationships, a machine learning classification algorithm is used to identify target assembly components containing target sub-components. The component dual-association index matrix acquisition module 20 is used to calculate the saliency value of the target assembly component by combining volume features and complexity features. Based on the saliency value, the target assembly component undergoes hierarchical lightweight processing that balances mesh simplification and feature preservation, resulting in a hierarchically lightweight model of the component and a component dual-association index matrix containing a 3D spatial position index and a unique encoding index of the curtain wall component. The candidate collision region... The module 30 is used to arrange the process of the component-level lightweight model according to the construction logic rules based on the component dual association index matrix to generate the curtain wall unit panel time sequence assembly sequence. It also performs kinematic interference lightweight simulation of hoisting and locking actions in the component-level lightweight model according to the curtain wall unit panel time sequence assembly sequence to obtain candidate collision areas. The high-precision collision area acquisition module 40 is used to determine whether the overlap volume of the candidate collision area is greater than the preset interference volume critical threshold. If the overlap volume of the candidate collision area is greater than the preset interference volume critical threshold, the corresponding high-precision collision area model is extracted from the initial building model based on the component dual association index matrix, and a local model restoration operation is performed to obtain the high-precision collision area. The simulation accurate verification result set acquisition module 50 is used to calculate the curtain wall assembly sealing joint width and unit panel overlap length using a spatial geometric intersection algorithm for the high-precision collision area to obtain a simulation accurate verification result set including interference depth and collision type. The assembly simulation operation completion module 60 is used to write back the simulation accurate verification result set to the initial building model by updating the component attribute tags to complete the assembly simulation operation.

[0187] The following detailed description uses specific embodiments to illustrate the lightweight processing and assembly simulation method and system for three-dimensional models of curtain wall unit panels provided by the present invention: I. Implementation Scenarios Taking a super high-rise unitized glass curtain wall project as an example, this method is used to simulate the assembly process of a standard unit panel (1200mm×3000mm).

[0188] The initial high-precision BIM model has a total of 1,200,000 triangular faces.

[0189] The total number of triangles in the simplified model is 150,000 (simplification rate 87.5%).

[0190] Assembly path: vertical hoisting + horizontal fine adjustment, step size 0.01s, simulation duration 10s.

[0191] II. Explanation of Key Formula Parameters and Implementation Data (1) Formula (1-13): Model lightweighting and dynamic spatial bounding box construction Table 1

[0192] Control logic verification: The key feature retention rate of the simplified model is ≥98.2%, and the calculation time of the dynamic bounding box is only 1 / 12 of that of the high-precision model, which effectively reduces the computing power consumption in the coarse screening stage.

[0193] (2) Formula (14-18): Collision screening and candidate region extraction Table 2

[0194] Implementation data: Table 3

[0195] (3) Formula (19-24): Local restoration and interference verification of high-precision model Table 4

[0196] Implementation data: Table 5

[0197] (4) Formula (25-27): BIM model update and status feedback Table 6

[0198] III. Comparison Table of Beneficial Effects of the Invention and Traditional Solutions Table 7

[0199] IV. Verification Conclusions of Examples Feasibility: All parameter values ​​conform to the actual scenario of curtain wall engineering, and the formula logic can be directly implemented in mainstream BIM / CAE software without additional customized development.

[0200] Authenticity: All implementation data comes from actual project simulation tests. The computing power consumption and detection accuracy are significantly better than traditional solutions, and the data is authentic and reliable.

[0201] Creative verification: By adopting a hierarchical strategy of "lightweight coarse screening - local high-precision verification - model closed-loop update", the simulation efficiency is improved by about 100 times without losing detection accuracy, thus solving the computing power bottleneck problem of large-scale curtain wall assembly simulation.

[0202] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for lightweight processing and assembly simulation of a three-dimensional model of a curtain wall unit panel, characterized in that, Includes the following steps: S100: Obtain the component geometric dimensions, component attributes, and component assembly connection relationships in the initial building model; based on the component geometric dimensions, component attributes, and component assembly connection relationships, use a machine learning classification algorithm to identify the target assembly component containing the target sub-component. S200. For the target assembly component, calculate the saliency value by combining volume features and complexity features. Based on the saliency value, perform hierarchical lightweighting processing on the target assembly component, taking into account both mesh simplification and feature preservation, to obtain the component hierarchical lightweighting model and the component dual association index matrix containing the three-dimensional spatial position index and the curtain wall component unique code index. S300. Based on the dual association index matrix of the component, the process arrangement is carried out on the component after graded lightweighting model according to the construction logic rules to generate the curtain wall unit panel time sequence assembly sequence. Kinematic interference lightweighting simulation of hoisting and locking actions is carried out in the component after graded lightweighting model according to the curtain wall unit panel time sequence assembly sequence to obtain candidate collision areas. S400. Determine whether the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold. If the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold, then based on the component dual association index matrix, map and extract the corresponding high-precision collision region model from the initial building model, perform a local model restoration operation, and obtain the high-precision collision region. S500. For the high-precision collision area, the spatial geometric intersection algorithm is used to calculate the width of the curtain wall assembly sealing joint and the overlap length of the unit panel to obtain a set of simulation accurate verification results including interference depth and collision type. S600. The simulation accuracy verification result set is written back to the initial building model by updating the component attribute labels to complete the assembly simulation operation.

2. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 1, characterized in that, Step S100 includes: S110. Obtain the global topology map of the curtain wall components. The topology map is constructed from the component geometric dimensions, component attributes, and component assembly and connection relationships in the initial building model. S120. Extract features from the global topology map of the curtain wall components to obtain a topology fusion feature vector; S130. Classify the topology fusion feature vector to obtain the target sub-component matching confidence probability containing the target sub-component; S140. If the matching confidence probability of the target sub-component is greater than the preset component recognition probability threshold, it is determined to be a target sub-component, and the target assembly component is determined based on the target sub-component to complete the recognition of the target assembly component.

3. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 1, characterized in that, Step S200 includes: S210. Obtain the three-dimensional mesh data of the target assembly component, and extract the total mesh volume feature and mesh complexity feature of the component from the three-dimensional mesh data; S220. Perform a weighted summation operation on the total volume feature of the component mesh and the mesh complexity feature to obtain the comprehensive saliency of the target assembly component; S230. Based on the comprehensive saliency of the target assembly component, perform hierarchical lightweighting processing on the target assembly component to obtain the component hierarchical lightweighting model. S240. Extract the spatial coordinates and unique identifier of the component after graded lightweighting model, and generate a component dual association index matrix containing a three-dimensional spatial location index and a curtain wall component unique code index.

4. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 3, characterized in that, Step S300 includes: S310. Obtain the component dual association index matrix and the component graded lightweight model, and arrange the process of the component graded lightweight model according to the construction logic rules to obtain the curtain wall unit panel time sequence assembly sequence. S320. Extract the hoisting and locking actions of the corresponding components based on the timing assembly sequence of the curtain wall unit panel, and generate the kinematic simulation trajectory of the components based on the hoisting and locking actions. S330. Construct the dynamic spatial bounding box corresponding to each component using the kinematic simulation trajectory of the component; S340. Determine whether the dynamic spatial bounding boxes of different components overlap. If they overlap, then determine the overlapping portion as a candidate collision region.

5. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 1, characterized in that, Step S400 includes: S410. Obtain candidate collision regions, and perform spatial discretization processing on the candidate collision regions to obtain a discrete voxel set. S420. Calculate the cumulative space occupancy of intersecting voxels based on the discrete voxel set to obtain the total actual overlapping volume of component interference; S430. Determine whether the total actual overlapping volume of the component interference is greater than the preset critical threshold for interference volume. S440. If the total actual overlapping volume of the component interference is greater than the preset interference volume critical threshold, then extract the spatial coordinate mapping interval corresponding to the candidate collision region. S450. Combining the spatial coordinate mapping interval and the component dual association index matrix, extract the corresponding high-precision collision area model from the initial building model.

6. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 5, characterized in that, Step S500 includes: S510. The spatial geometry intersection algorithm is used to calculate the boundary contour set of the high-precision collision region to obtain the intersecting closed polygon contour of the model. S520. Based on the intersecting closed polygonal contour of the model, obtain the width of the curtain wall assembly sealing joint and the overlap length of the unit panel. S530. If the overlap length of the unit plate is greater than the preset overlap depth judgment threshold, then extract the inscribed sphere radius of the intersecting closed polygon contour of the model to determine the assembly interference depth. S540. By combining the assembly interference depth and the width of the curtain wall assembly sealing joint, the original feature set of the component penetration interference is mapped to obtain a set of simulation accurate verification results containing the interference depth and collision type.

7. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 1, characterized in that, Step S600 includes: S610. Obtain the set of simulation accurate verification results, and extract the initial BIM attribute label set of the component based on the set of simulation accurate verification results; S620. The initial BIM attribute label set of the component is corrected using the simulation accurate verification result set to obtain the target component attribute label after interference correction. The formula for calculating the attribute label of the target component after interference correction is: ； in, The attribute labels for the target component after interference correction. For the initial BIM attribute label set of the component, For the label coverage fusion operator, This is a set of simulation-based accurate verification results; S630. Generate a model status identifier instruction based on the target component attribute label after interference correction. The model status identifier instruction is used to trigger the model status identifier to complete the numerical change. S640. Reconstruct the simulation scene map based on the model status identifier, and write the simulation accurate verification result set back to the initial building model to complete the assembly simulation operation.

8. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 7, characterized in that, In step S630, the formula for calculating the model status flag is: ； in, For model state flag bits, This is a label mapping function.

9. The method for lightweight processing and assembly simulation of three-dimensional models of curtain wall unit panels according to claim 8, characterized in that, In step S640, the formula for calculating the reconstructed simulation scene map is: ； in, To reconstruct the simulation scene map, For global graph update operators, This is the global topology map of the initial building model.

10. A lightweight processing and assembly simulation system for three-dimensional models of curtain wall unit panels, used to implement the lightweight processing and assembly simulation method for three-dimensional models of curtain wall unit panels as described in any one of claims 1 to 9, characterized in that, include: The target assembly component identification module is used to obtain the component geometric dimensions, component attributes, and component assembly connection relationships in the initial building model, and to identify the target assembly components containing target sub-components based on the component geometric dimensions, component attributes, and component assembly connection relationships using a machine learning classification algorithm. The component dual association index matrix acquisition module is used to calculate the saliency value of the target assembly component by combining volume features and complexity features, and to perform hierarchical lightweighting processing on the target assembly component based on the saliency value, taking into account both mesh simplification and feature preservation, to obtain the component hierarchical lightweighting model and the component dual association index matrix containing the three-dimensional spatial position index and the curtain wall component unique code index. The candidate collision region acquisition module is used to rely on the component's dual association index matrix, arrange the process of the component's graded and lightweight model according to the construction logic rules to generate the curtain wall unit panel's time sequence assembly sequence, and carry out kinematic interference lightweight simulation of hoisting and locking actions in the component's graded and lightweight model according to the curtain wall unit panel's time sequence assembly sequence to obtain candidate collision regions. The high-precision collision region acquisition module is used to determine whether the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold. If the overlapping volume of the candidate collision region is greater than the preset interference volume critical threshold, the corresponding high-precision collision region model is extracted from the initial building model based on the component dual association index matrix, and a local model restoration operation is performed to obtain the high-precision collision region. The simulation accurate verification result set acquisition module is used to calculate the width of the curtain wall assembly sealing joint and the overlap length of the unit panel for the high-precision collision area using a spatial geometric intersection algorithm, and obtain a simulation accurate verification result set including interference depth and collision type. The assembly simulation task completion module is used to write back the set of simulation precision verification results to the initial building model by updating the component attribute labels, thereby completing the assembly simulation task.

Images