Fusion knowledge retrieval enhanced building curtain wall structure automated construction drawing system and method

Abstract

The application belongs to the field of computer aided design, and provides an automatic construction drawing drawing system and method of building curtain wall structure fused with knowledge retrieval enhancement, which converts curtain wall three-dimensional components into attributed component objects, and constructs a construction drawing scene description based on component categories, materials, spatial geometry, local connection topology and drawing tasks; performs version validity screening and semantic-topology-geometry mixed retrieval from an engineering knowledge base containing specification provisions, standard drawing sets, enterprise rules and historical curtain wall node cases, to generate machine executable construction drawing constraints bound with components. Actual drawing expression characteristics and constraints are checked for consistency, and inconsistent areas are updated with expression parameters or regenerated locally after re-retrieval. The application can maintain consistency between three-dimensional model data and two-dimensional drawing expression, and improve the unity, standardization and engineering applicability between different drawing types.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided design and artificial intelligence-aided engineering design, and in particular relates to an automated construction drawing system and method for building curtain wall structures that integrates knowledge retrieval-enhanced generation and construction drawing rule constraints. Background Technology

[0002] Current curtain wall designs typically begin by creating a 3D model of the curtain wall using 3D modeling software such as Rhino and Crown. This model represents the spatial location, dimensions, layout, and connection relationships of components such as columns, beams, panels, doors and windows, framing, connectors, and details. Based on this model, 2D construction drawings, including floor plans, elevations, sections, detail drawings, and large-scale drawings, are then created. Currently, these construction drawings still largely rely on designers to perform secondary drawing and editing in software such as AutoCAD and Tianzheng Architecture. This involves a large amount of manual work, a long drawing cycle, and is prone to problems such as omissions, errors, and inconsistencies between different drawings.

[0003] While existing CAD and BIM software possess certain capabilities for converting 3D models into 2D models and assisting in the generation of construction drawings, the methods for automatically generating general architectural construction drawings are typically based on relatively standardized component types, geometric relationships, and drafting rules. Compared to main structural components such as beams, columns, walls, and slabs, building curtain walls are composed of a large number of components, including panels, columns, beams, joists, transition pieces, hangers, opening components, and other connectors. These components are attached to the main structure and are characterized by a large number of components, significant dimensional differences, complex spatial hierarchy, a high proportion of non-standard components, non-standardized local connection structures, and strong project customization. Therefore, it is difficult to directly apply uniform building component templates and fixed drawing rules.

[0004] Especially in areas such as corners, edges, openings, inter-layer spaces, and curved, folded, and inclined surfaces, the local coordinate directions, spatial orientations, and connection methods of components in different curtain wall areas may change depending on their location. Furthermore, there are dense and multi-layered local connections between different components. For the aforementioned irregular curtain wall structures, using a fixed viewing direction based on the overall coordinate system, a uniform sectioning method, or a single projection template can easily lead to unreasonable projection directions of local components, distorted occlusion relationships, omissions of key connecting components, and incomplete representation of local structures.

[0005] Furthermore, the engineering representation in curtain wall construction drawings does not have a simple one-to-one correspondence with the geometry of 3D components. Some smaller components that play a crucial connecting role may not have prominent geometric features in the 3D model, but they need to be highlighted in detail drawings or large-scale drawings, or replaced with standard drawings. The same type of component may also use different 2D representations depending on its location, adjacent components, installation direction, and local structure. Therefore, the automatic generation of curtain wall construction drawings requires not only geometric projection but also determining the view direction, sectioning range, drawing selection, and corresponding drawing representation method based on component attributes, local geometry, and connection relationships.

[0006] To accommodate these differences, rules for component classification, drawing specifications, and construction drawing representation can be pre-established, and the corresponding rules can be automatically generated based on component type or preset conditions. However, curtain wall engineering is highly customized, with significant differences in construction methods and drawing requirements for different curtain wall systems, material systems, connection types, and engineering parts. A fixed rule library cannot fully cover engineering scenarios such as irregular components, special connection relationships, and non-standard structures. When new irregular components, special connection types, or customized drawing requirements arise, the original rules often cannot be directly applied, requiring designers to re-compile, supplement, or modify the corresponding rules and drawings. As project types and construction forms continue to increase, the rule library requires continuous manual maintenance and updates, which not only increases system configuration and maintenance costs but also makes it difficult for existing rules to quickly adapt to new engineering scenarios. At the same time, a large number of node diagrams, construction methods, and drawing experience accumulated in historical projects are usually stored in the form of drawings, documents, or project files, making them difficult to reuse directly in the current automatic drawing generation process.

[0007] Existing knowledge retrieval and AI-assisted technologies can obtain information related to design tasks from standard documents, standard drawing sets, enterprise rules, or historical engineering data. However, the retrieval results are usually returned in the form of text fragments, rule descriptions, or historical cases. It is difficult to directly establish a correspondence with specific components, local geometric features, and connection relationships in the curtain wall 3D model. It is also difficult to further convert them into drawing constraints such as view parameters, section parameters, drawing selection, line type, fill, annotation, and index that can be directly executed by the construction drawing generation program. Summary of the Invention

[0008] To address at least one of the technical problems mentioned above, this invention provides an automated construction drawing system and method for building curtain wall structures that integrates knowledge retrieval enhancement. While retaining the geometric calculation and attribute-based drawing mechanisms of the 3D model, it incorporates project scenarios, component topology, and current engineering knowledge into the drawing decision-making process. This allows specifications, atlases, company rules, and historical curtain wall node cases to be retrieved, filtered, and converted into machine-executable structured construction drawing constraint parameters, thereby controlling the generation of 2D drawings and verifying the consistency of the results. This not only improves the specification adaptability, traceability, data consistency, and engineering reliability of automatically generated curtain wall construction drawings but also helps reduce the risk of errors and omissions caused by manual specification checks, repeated drawing, and modifications.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A first aspect of the present invention provides an automated construction drawing system for building curtain wall structures that integrates knowledge retrieval enhancement, comprising: The target component dataset construction layer is used to read the target layer in the 3D model of the building curtain wall and obtain the component instance geometry in the corresponding layer; establish a classification representation model of the curtain wall components; establish the association between the 3D geometric objects of the curtain wall and the component representation parameters; and convert the original geometric objects into attributed component objects to obtain the target component dataset. The knowledge retrieval enhancement decision layer is used to construct construction drawing scene description data based on the target construction dataset and the task parameters of the drawings to be generated. Combined with candidate knowledge selected from the engineering knowledge base, it performs hybrid retrieval and re-ranking that combines semantic similarity, local connectivity topological similarity and geometric parameter matching degree to obtain a set of evidence knowledge that matches the current construction drawing scene and generate a set of machine-executable construction drawing constraints bound to the target component. The 2D construction drawing generation layer uses a unified 3D-to-2D conversion framework to call the target component dataset and uses the executable construction drawing constraint set as control input to convert attributed component objects into 2D basic line drawings of different types. It includes a plan view generation module, an elevation view generation module, a node view generation module, and a detail view generation module. Each module generates corresponding 2D construction drawing basic line drawings for attributed component objects by combining preset view directions, cutting ranges, and projection rules. The unified 3D-to-2D conversion framework includes five steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing expression mapping. The drawing post-conformity verification layer is used to process the section filling, annotation and drawing frame of the basic line drawings of the generated two-dimensional construction drawings of different types. It extracts the actual expression features from the processed drawings and compares them with the set of constraints of the executable construction drawings. When there are inconsistencies, it updates the expression parameters of the corresponding components, triggers a second search, or locally regenerates the affected drawing area. When the constraints are met, a standardized two-dimensional construction drawing result is formed, and the source identifier corresponding to the evidence knowledge used is retained.

[0010] Furthermore, in the target component dataset construction module, the classification representation model includes a component category table, a material category table, and a drawing representation rule table; the drawing representation rule table includes at least the following rules: Door and window component representation rules: For components identified as doors and windows, a local representation coordinate system for door and window symbols is established based on the opening direction information in their attribute parameters and the local enclosing range of the component, and the outer contour and opening direction symbols of doors and windows are generated in the corresponding two-dimensional drawings; Elevation identification rules: For components with elevation identification attributes, the spatial position parameters of the component are used as the elevation reference position, and the corresponding elevation positioning point or elevation reference line is generated in the elevation projection result. Node pattern replacement rules: For components with pattern expression attributes, the corresponding pattern is called from the preset pattern library according to the component's projection positioning point, main direction and pattern name, and the pattern is then placed into the node diagram after attitude matching according to the component's direction. Material labeling and positioning rules: For components with material names, extract candidate labeling points based on the spatial mapping results of the component's projected line segments, and group the candidate labeling points according to the material or material name; Line type representation rules: For projected line segments that are determined to be occluded after the visibility solution, they are classified into the hidden line representation set and assigned a hidden line type; for projected line segments that are determined to be visible or clipped to visible, they are classified into the visible line representation set.

[0011] Furthermore, in the target component dataset construction module, the association between the three-dimensional geometric objects of the curtain wall and the component expression parameters is established, including: performing semantic parsing on the component attribute expressions, converting continuous attribute description information into an attribute parameter set composed of attribute names and attribute values, using geometric objects as the basic binding unit, establishing a correspondence between each geometric object and the attribute parameter set, and forming an attributed component object including geometric data units and attribute parameter units.

[0012] Furthermore, in the two-dimensional construction drawing generation layer module, in the view parameter construction step, the viewing direction and the direction on the drawing are determined according to the drawing type, and combined with the overall spatial enclosure range of the target component dataset, a parallel view calculation model for two-dimensional projection is established. In the local clipping domain construction step, spatial clipping constraints are established based on the clipping plane, cutting range, or local selection range set by the user, so that only components or parts of components located within the target drawing range participate in the two-dimensional transformation. During component projection preprocessing, it is determined whether symbolic representation, drawing replacement, or elevation positioning processing is required based on the component attribute parameters; for ordinary components that do not require special representation, their geometric contours are directly included in the projection solution object. In the occlusion relationship solving step, the visibility of the two-dimensional line segments formed after projection is classified based on the viewing direction, clipping constraints and the spatial relationship between components; In the drawing representation mapping step, based on the visibility classification results and component attribute parameters, the two-dimensional line segments are mapped to visible lines, hidden lines, door and window symbol lines, elevation points, node patterns or material annotation positioning points, thereby forming the basic line drawing of the corresponding two-dimensional construction drawing.

[0013] Furthermore, the plan view generation module is used to generate a two-dimensional planar basic line drawing of the curtain wall structure based on the target component dataset, including: Establish a planar projection coordinate system according to the drawing requirements of the plan view. The planar projection coordinate system includes the planar viewing direction, the direction on the drawing, and the projection reference point. The projection calculation domain is determined based on the spatial enclosing range of each component in the target component dataset, and a parallel projection model is established within this projection calculation domain. The parallel projection model is used to map the spatial contour of the three-dimensional component to the two-dimensional drawing plane, while preserving the front-to-back occlusion relationship between different components along the viewing direction. Before projection, the target component is evaluated for its attributes. For ordinary curtain wall components, their geometric contours are used as the projection solution objects. For components with door and window attributes, symbolic representation units for doors and windows are constructed based on the local geometric range and opening direction parameters of the component. The symbolic representation units for doors and windows include door and window boundary representation and opening trajectory representation. Line type mapping is performed based on visibility status. Two-dimensional line segments that are determined to be visible or clipped to be visible are written into the visible line set, and two-dimensional line segments that are determined to be occluded are written into the hidden line set and assigned a hidden line type expression. Based on the input reference point, the visible line set and the hidden line set are located as a whole, and the basic line drawing of the plan view is output.

[0014] Furthermore, the elevation drawing generation module is used to convert the target component dataset into a two-dimensional basic line drawing in the direction of the curtain wall elevation, including: A facade projection coordinate system is established based on the facade orientation. The facade projection coordinate system includes the facade viewing direction and the vertical direction. The facade projection field of view is determined based on the overall spatial range of the target component dataset. A parallel projection calculation model is constructed so that the three-dimensional component is projected onto the two-dimensional facade expression plane along the facade viewing direction. Before component projection, corresponding expression processing is performed on ordinary components, components with door and window attributes, and components with elevation recognition attributes according to the attribute parameters. In the process of solving the elevation projection, the clipping plane is used as the constraint of the elevation drawing range, the elevation viewing direction is used as the depth judgment direction, the front and back occlusion relationship between the three-dimensional components is solved, and the two-dimensional line segments obtained by projection are divided into visible line segments, clipping visible line segments and occlusion line segments. The visible line segments and clipping visible line segments are entered into the elevation visible line set, and the occlusion line segments are entered into the elevation hidden line set and assigned a hidden line type. The set of visible lines on the facade, the set of hidden lines on the facade, and the elevation positioning points together constitute the basic line drawing data of the facade drawing.

[0015] Furthermore, the node diagram generation module is used to represent the local connection areas of the curtain wall using two-dimensional nodes, specifically including: A local node coordinate system is established based on the node image range specified by the user. The local node coordinate system is determined by the image range plane, the node viewing direction, and the node image direction. A local spatial clipping domain is formed based on the image range, and the components in the target component dataset that have spatial intersection with the local spatial clipping domain are extracted as a set of node candidate components. For the set of candidate node components, two types of processing are performed according to the attribute parameters: the first type is geometric projection processing, and the second type is graphical representation processing. In the local projection solution stage, the node viewing direction is taken as the projection direction and the local spatial clipping domain is taken as the projection boundary. The set of candidate node components is processed into two dimensions, and the occlusion relationship between component outlines within the local area is solved. For ordinary component line segments in the projection results, the attributes of their source components are synchronously written into the two-dimensional line segments, and material annotation candidate points are generated according to their material or material properties; for pattern positioning objects in the projection results, the corresponding node patterns are called from the pattern library according to their pattern expression attributes, and the pattern is matched with the orientation according to the main direction of the component to make the pattern consistent with the actual arrangement direction of the component in the node area. Finally, the node pattern set, the node positioning point set, and the annotation positioning point set grouped by material or material properties are output. Alternatively, the detailed drawing generation module is used to enlarge and represent a local area of ​​the curtain wall in two dimensions, including: A large-scale drawing projection calculation model is established based on the local magnified area, the viewing direction, the direction on the drawing, and the cutting plane. The set of components participating in the local magnification expression is extracted based on the target component dataset. The projection field of view of the magnified area is determined based on the spatial enclosed range of the local component set, and the parallel projection relationship is established. The clipping plane constrains the scope of the components involved in the projection, so that the detailed drawing only expresses the structural information within the specified local area; for components with door and window representation attributes, symbolic representations of doors and windows in the detailed drawing are generated based on their local geometric scope and opening direction parameters; for ordinary components, their three-dimensional contours are preserved for projection and occlusion solutions. After completing the projection and occlusion solutions, the visible or clipped visible line segments are added to the visible line set of the large-scale drawing, and the occluded line segments are added to the hidden line set of the large-scale drawing. The component attributes are then synchronously written to the corresponding two-dimensional line segments to obtain the basic line drawing of the large-scale drawing.

[0016] A second aspect of the present invention provides a method for automated construction drawing of building curtain wall structures that integrates knowledge retrieval enhancement, comprising the following steps: Read the target layer from the 3D model of the building curtain wall and obtain the component instance geometry within the corresponding layer; establish a classification representation model for the curtain wall components; establish the association between the 3D geometric objects of the curtain wall and the component representation parameters, and convert the original geometric objects into attributed component objects to obtain the target component dataset; Based on the target dataset and the task parameters of the drawings to be generated, construction drawing scene description data is constructed. Combined with candidate knowledge selected from the engineering knowledge base, a hybrid retrieval and re-ranking process combining semantic similarity, local connectivity topological similarity, and geometric parameter matching degree is performed to obtain a set of evidence knowledge that matches the current construction drawing scene and generate a set of machine-executable construction drawing constraints bound to the target component. A unified 3D-to-2D conversion framework is adopted, which calls the target component dataset and uses the executable construction drawing constraint set as the control input to convert attributed component objects into 2D basic line drawings of different types. It includes a plan view generation module, an elevation view generation module, a node view generation module, and a detail view generation module. Each module generates 2D construction drawing basic line drawings of the corresponding type for the attributed component objects by combining preset view directions, cutting ranges, and projection rules. The unified 3D-to-2D conversion framework includes five steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing expression mapping. The basic line drawings of two-dimensional construction drawings of different types that have been generated are processed by section filling, annotation and drawing frame processing. The actual expression features are extracted from the processed drawings and compared with the set of constraints of executable construction drawings. When there are inconsistencies, the expression parameters of the corresponding components are updated according to the inconsistencies, triggering a second search or locally regenerating the affected drawing area. When the constraints are met, a standardized two-dimensional construction drawing result is formed and the source identifier corresponding to the evidence knowledge used is retained.

[0017] A third aspect of the present invention provides a computer-readable storage medium.

[0018] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for automated construction drawing of building curtain wall structures with enhanced knowledge retrieval.

[0019] A fourth aspect of the present invention provides a computer device.

[0020] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the above-described method for automated construction drawing of building curtain wall structures with enhanced knowledge retrieval.

[0021] Compared with the prior art, the beneficial effects of the present invention are: This invention addresses the challenges of numerous and spatially dispersed curtain wall components, significant differences in local coordinate directions and spatial orientations of components in different areas, non-standardized local connection structures, and the non-one-to-one correspondence between 3D geometry and 2D engineering representation. It combines the component attributes, spatial geometric features, local connection relationships, and specific drawing tasks of the current curtain wall model. It retrieves relevant and effective engineering knowledge from standards, atlases, company rules, and historical curtain wall node cases, transforming the retrieval results into executable construction drawing constraints. This synergizes with the processes of 3D curtain wall model analysis, 2D construction drawing generation, post-processing, and result verification, improving the standard compatibility, traceability, data consistency, and engineering reliability of automatically generated curtain wall construction drawings.

[0022] This invention, based on a 3D model of a building curtain wall, automatically generates various types of construction drawings, including floor plans, elevations, detail drawings, and large-scale drawings, through steps such as layer acquisition, component identification, attribute binding, automatic generation of 2D drawings, and post-processing. It further enables section filling, automatic annotation, and frame layout, effectively reducing the workload of manual secondary drafting and improving the efficiency and accuracy of converting 3D models to 2D construction drawings. Simultaneously, this invention maintains consistency between 3D model data and 2D drawing representations, improving the uniformity, standardization, and engineering applicability of different drawing types. This has positive significance for promoting the digitalization, automation, and standardization of curtain wall construction drawing production processes.

[0023] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0024] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0025] Figure 1 This is an architecture diagram of an automated construction drawing system for building curtain wall structures that integrates knowledge retrieval and is enhanced, provided in an embodiment of the present invention. Figure 2 This is a flowchart of the automated construction drawing method for building curtain wall structures with integrated knowledge retrieval enhancement provided in this embodiment of the invention; Figure 3 This is the effect of the plan view output provided in the embodiment of the present invention; Figure 4 This is the side elevation drawing effect provided by the embodiment of the present invention; Figure 5 This is the elevation view output effect provided in the embodiment of the present invention; Figure 6 This is the node diagram output effect provided in the embodiment of the present invention; Figure 7 This is the output effect of the large-scale image provided in the embodiment of the present invention. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0029] like Figure 1 As shown, this embodiment provides an automated construction drawing system for building curtain wall structures that integrates knowledge retrieval enhancement, including: The target component dataset construction module is used to read the target layer in the 3D model of the building curtain wall and obtain the geometry of the component instances in the corresponding layer; establish a classification representation model of the curtain wall components, establish the association between the 3D geometric objects of the curtain wall and the component representation parameters, and convert the original geometric objects into attributed component objects to obtain the target component dataset; Specifically, it includes the following modules: The layer acquisition module is used to read the target layer in the 3D model of the building curtain wall and obtain the geometry of the component instances within the corresponding layer; In this embodiment, considering the situation where different designers typically manage layers according to component category, material category, or construction habits during the actual curtain wall modeling process, this embodiment supports selective reading of one or more target layers, thereby enabling the centralized extraction of similar or attributed components to obtain a set of geometric objects; The component library module is used to build a classification and representation model for curtain wall components; In the component library module, the classification and representation model includes a component category table, a material category table, and a drawing representation rule table. The component category table records the type codes of different curtain wall components; the material category table records the material or material information corresponding to different components; and the drawing representation rule table records the mapping relationship between component attributes and two-dimensional drawing representation methods.

[0030] As a further implementation, the drawing representation rule table shall include at least the following rules: First, the rules for representing door and window components. For components identified as doors and windows, the system establishes a local coordinate system for representing door and window symbols based on the opening direction information in their attribute parameters and the local enclosing range of the component, and generates the outer contour of the door and window and the opening direction symbol in the corresponding two-dimensional drawing.

[0031] Second, elevation identification rules. For components with elevation identification attributes, the system uses the spatial position parameters of the component as the elevation reference position and generates the corresponding elevation positioning point or elevation reference line in the elevation projection result.

[0032] Third, node pattern replacement rules. For components with pattern representation attributes, the entire projected outline of the component is not directly used as the node pattern representation. Instead, the corresponding pattern is called from the preset pattern library according to the component's projection positioning point, main direction, and pattern name. The pattern is then placed into the node pattern after being orientation-matched according to the component's direction.

[0033] Fourth, material labeling and positioning rules. For components with material names, candidate labeling points are extracted based on the spatial mapping results of the component's projected line segments, and these candidate points are grouped according to the material or material name for subsequent material labeling or numbering.

[0034] Fifth, line type representation rules. For projected line segments that are determined to be occluded after visibility resolution, the system classifies them into the hidden line representation set and assigns them a hidden line type; for projected line segments that are determined to be visible or clipped to visible status, they are classified into the visible line representation set.

[0035] The attribute binding module is used to combine the geometric object collection and component attribute expressions to establish the association between the three-dimensional geometric objects of the curtain wall and the component expression parameters, convert the original geometric objects into attributed component objects, and construct the target component dataset; The attribute binding module receives a set of geometric objects and component attribute expressions as input. The component attribute expressions record the component category, material category, drawing representation method, and subsequent drawing control parameters according to the preset attribute description syntax.

[0036] First, semantic parsing is performed on the component attribute expressions to convert continuous attribute description information into an attribute parameter set consisting of attribute names and attribute values. Then, using geometric objects as the basic binding unit, a correspondence is established between each geometric object and the attribute parameter set to form an attributed component object that includes geometric data units and attribute parameter units.

[0037] Let the target component dataset be: , Among them, the Each target component is represented as: , Indicates the first Three-dimensional geometric data of each component, This represents the set of attribute parameters bound to this component. The attribute parameter set can be represented as: , in, For attribute names, The attribute name includes at least one or more of the following: component type, material type, material name, door and window expression parameters, elevation identification parameters, drawing expression parameters, line type parameters, or fill parameters.

[0038] This embodiment obtains the aforementioned set of attribute parameters by parsing the component attribute expressions and then compares them with the corresponding three-dimensional geometric data. Binding to form target components .

[0039] As a further implementation, the attribute parameters include at least component identification parameters, material identification parameters, and drawing representation parameters. Component identification parameters are used to distinguish columns, beams, panels, doors and windows, connectors, keels, or auxiliary components; material identification parameters are used to indicate the material type or material name of the component; drawing representation parameters are used to indicate how the component is represented in the two-dimensional construction drawing, including whether it participates in door and window symbol generation, whether it participates in elevation identification, whether it participates in node drawing replacement, whether it participates in material annotation point generation, and one or more of the corresponding line type and corresponding fill style.

[0040] During the binding process, the spatial geometric relationships of the original 3D model are not altered. Instead, geometric objects are treated as independent entities, and the parsed attribute parameter set is attached to each independent geometric object. The resulting target component dataset can simultaneously provide geometric calculation and drawing representation basis in the subsequent generation of plan views, elevation views, node views, and detail views. This allows the same 3D component to be expressed in different drawing types with different projections, clippings, line types, symbols, and annotations based on its attribute parameters.

[0041] This invention can complete the unified identification, classification, attribute assignment and data organization of curtain wall components before the three-dimensional model enters the two-dimensional construction drawing generation process, thereby solving the problems of scattered three-dimensional model data, difficulty in directly using component attributes for two-dimensional expression and inconsistent data calling between different drawing types in the prior art.

[0042] As a further implementation, this embodiment sets up a knowledge retrieval enhancement decision layer between the target component dataset and the two-dimensional construction drawing generation layer. The knowledge retrieval enhancement decision layer includes an engineering knowledge base construction module, a construction drawing scene description module, a hybrid retrieval and re-ranking module, a constraint generation and binding module, and a consistency verification feedback module. Each module is connected to the attributed component objects and the two-dimensional drawing calculation module through a structured data interface, enabling engineering knowledge matching the current construction drawing scene to be converted into definite drawing parameters or constraints, and participating in subsequent three-dimensional to two-dimensional conversion and drawing post-processing.

[0043] The knowledge retrieval enhancement decision layer adopts a retrieval enhancement generation mechanism, which uses external engineering knowledge that matches the current construction drawing scenario as the evidence context for semantic parsing and constraint generation. The generated results are represented by a preset structured constraint pattern, and a correspondence is established between the source identifier and the retrieved knowledge unit, so that the construction drawing constraints entering the drawing process have a clear knowledge source.

[0044] The engineering knowledge base construction module is used to structure and organize the specifications, standard drawing instructions, enterprise internal standardized drawing rules, and historical curtain wall node cases that have been manually verified, which are related to the construction drawings of building curtain walls.

[0045] For text-based knowledge, clauses, rules, or content units that can independently express constraint semantics are used as knowledge units. One or more metadata elements are stored, including source identifier, document name, clause number, version number, effective date, expiration date or validity status, applicable region, curtain wall system type, component category, material category, drawing type, and keywords. Semantic vector index and structured metadata index are also established.

[0046] For historical curtain wall node cases, in addition to saving node drawings, source projects, and applicable conditions, a connection topology description is also established based on the set of three-dimensional local components corresponding to the node. This connection topology uses columns, beams, panels, connectors, embedded parts, opening components, and other local components as nodes, and uses connections, contacts, adjacencies, supports, or dependencies as the relationships between components. It also records the type, material, local dimensions, main direction, and relative position parameters of the corresponding components, thereby forming topological and geometric features for historical node retrieval.

[0047] The construction drawing scene description module is used to convert the current drawing task into a structured scene description. For node drawing tasks, the target component set is determined based on the local drawing range selected by the user. The component category, material, local enclosure range, main direction, relative position, and local connection topology between target components are extracted. This is combined with the drawing type, viewing direction, drawing orientation, drawing scale, project location, and design reference date to form a structured query object. For floor plans, elevations, and detail drawings, the module extracts component grids, door and window opening parameters, elevations, materials, and index task parameters related to the corresponding drawing type.

[0048] The hybrid retrieval and reordering module first filters candidate knowledge in the engineering knowledge base based on applicable region, design reference date, knowledge validity status, curtain wall system type, and drawing type to exclude knowledge units that are invalid or do not match the current project conditions. Then, it calculates semantic relevance for the filtered text knowledge, calculates local connection topological similarity and geometric parameter matching degree for historical curtain wall node cases, and reorders candidate knowledge based on metadata matching degree and version validity.

[0049] The overall matching degree of candidate knowledge d with respect to the current construction drawing scenario query q can be expressed as: , in, Indicates semantic relevance. Indicates the topological similarity of local connections. Indicates the degree of geometric parameter matching. Indicates the degree of metadata matching. This indicates the version validity score. This indicates a penalty for conflict of source or conflict of applicable conditions. to Non-negative weights, either preset or configured according to different map types.

[0050] When multiple candidate rules correspond to the same target object and the same constraint attribute, conflict resolution is carried out according to the knowledge validity status, the degree of matching of the applicable scope, the rule level, the source priority, and the comprehensive matching degree. For conflicts that cannot be resolved automatically, the corresponding candidate knowledge is marked as items to be manually confirmed and will not be used as mandatory drawing constraints in the subsequent automatic generation process.

[0051] The constraint generation and binding module is used to perform structured semantic parsing on the evidence knowledge set obtained after retrieval and reordering, forming machine-executable construction drawing constraint records. The j-th constraint record can be represented as: , in, Identify the target component or set of target components. For constraint attributes, For comparison, selection, or matching relationships, For constraint values ​​or parameter ranges, For the corresponding drawing action, For the identification of the source of evidence, To analyze credibility.

[0052] A defined parameter mapping relationship is established between the drawing action and the 2D construction drawing generation layer and post-processing process. Specifically, section or visible range constraints are used to update local clipping domain parameters; door / window or node representation constraints are used to determine symbolic representation methods or node patterns; line type and material constraints are used to update drawing representation mapping and section fill parameters; dimension and elevation constraints are used to determine annotation objects and annotation rules; and index constraints are used to establish drawing numbers and pointing relationships between main drawings, node drawings, and detail drawings. This allows knowledge retrieval results to enter the original drawing calculation process through a defined parameter interface, rather than merely being output as natural language prompts.

[0053] For the reuse of historical curtain wall node cases, the system prioritizes searching for candidate nodes with similar local connection topology and matching geometric parameters to the current target. After determining the candidate historical node drawings, a correspondence is established between the reference components in the candidate nodes and the current target components. The system then calculates the rotation, translation, and allowable parametric size adaptation parameters based on the main direction, local enclosure range, positioning reference point, and relative position of the reference and target components, so that the historical node drawings can be adapted to the current node area. When there is an unmapped critical connection relationship between the candidate node and the target node, automatic drawing replacement is not performed, and a second search or manual confirmation is triggered.

[0054] The consistency verification and feedback module is used to extract actual expressive features from the 2D construction drawings after post-processing, such as line type, fill, symbol type, drawing source, annotation value, elevation, index relationship, and drawing scale. It then compares these features item by item with the machine-executable construction drawing constraint records according to the target object identifier. When the actual expressive features do not meet the corresponding construction drawing constraints, a conflict record is generated, including the object identifier, constraint attribute, expected value, actual value, and evidence source. For inconsistencies that can be resolved through parameter correction, the expression parameters of the corresponding component are updated, and only the affected local graphics are regenerated. For inconsistencies caused by knowledge mismatch, the knowledge retrieval is performed again after supplementing the search conditions based on the conflict record.

[0055] The 2D construction drawing generation layer module is used to generate drawings using a unified 3D-to-2D conversion framework. The system retrieves geometric and attribute information from the target component dataset and, in conjunction with preset view orientation, sectioning range, and projection rules, generates basic line drawings for 2D construction drawings of attributed component objects in different drawing types, including plan views, elevation views, node views, and detail views. In this embodiment, the 3D-to-2D conversion framework includes five sub-steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing representation mapping. When constructing view parameters, the viewing direction and the direction on the drawing are determined according to the drawing type, and a parallel view calculation model for two-dimensional projection is established by combining the overall spatial enclosure range of the target component dataset. Specifically, this includes: In this embodiment, firstly, based on the target component dataset... Calculate the overall spatial enclosed area.

[0056] For the Each component has a 3D bounding box denoted as: , of which: The minimum and maximum points of each component are as follows: , , and These are the corresponding three-dimensional coordinates; The overall enclosure box for all components is: Its minimum and maximum points are respectively: , , and For the corresponding three-dimensional coordinates; The center point of the overall enclosure is: , The diagonal length of the entire enclosure is: , in, Represents the vector magnitude and center point. The diagonal length is used as the view reference point for the projection calculation model. Scale parameters used to determine the size and depth of the projected field of view.

[0057] For any drawing to be generated, let the observation direction vector be... The direction vector on the drawing is First, normalize both to obtain the corresponding unit vectors. and : , , To establish a coordinate system for the two-dimensional drawing, the system further constructs the horizontal basis vectors of the drawing. : , in, This represents the cross product of vectors. Therefore, , , These respectively constitute the horizontal direction, vertical direction, and projection depth direction of the two-dimensional drawing.

[0058] Using parallel projection to represent three-dimensional points Convert to 2D drawing points. Let the projection reference point be... Then three-dimensional points The coordinates on the two-dimensional drawing plane are: , , Therefore, three-dimensional points The two-dimensional projection point can be represented as: , in, This represents the parallel projection operator. For components... Space curves on Its two-dimensional projection curve Represented as: , in, These are the curve parameters. The system uses the aforementioned projection calculation model to map the 3D component outline into a set of candidate line segments in the 2D construction drawing.

[0059] For different map types and The values ​​are different: in a planar diagram, The preferred viewing direction is vertical. Take the top direction in the plan view; in the elevation view, Viewing direction from the curtain wall facade. Take the vertical direction; in the node diagram and the detailed drawing, It is determined by the local map area or the sectioning area.

[0060] In the local clipping domain construction step, the system establishes spatial clipping constraints based on the clipping plane, cutting range, or local selection range set by the user, ensuring that only components or parts of components located within the target output range participate in the 2D transformation. Specifically, this includes: Camera position and projection field of view calculation: To ensure that the target component enters the two-dimensional projection field of view completely, the calculation is based on the center of the overall bounding box. Observation direction and scale parameters Determine the location of the projection camera: , in, For the position of the projection camera, This is the projection distance coefficient, preferably set to 1 or determined adaptively according to the map type.

[0061] For elevation views, the projection distance can also be compensated based on the distance from the center of the component's bounding box to the global coordinate origin, i.e.: , in, The origin of the global coordinate system. This refers to the elevation projection distance parameter. The corresponding camera position is: , The left and right boundaries, top and bottom boundaries, near clipping distance, and far clipping distance of the projected field of view can be represented as: , in, Indicates the projection frustum parameters. It is a positive number, preferably a small distance close to 0, to avoid the near-cutting surface from coinciding with the model.

[0062] For partial projection requirements in floor plans, elevations, detail drawings, and node drawings, the spatial range participating in the projection is limited by cropping the plane.

[0063] Let the first The cutting planes are: ,in, To cut a point on the plane, This is the normal vector of the clipping plane.

[0064] For any three-dimensional point Its directed distance relative to the cutting plane is: , When the half-space preservation rule is applied, the point satisfy: If so, then the point is considered to be located at the th position. The remaining side of each cutting plane. If there are a total of If there is a cutting plane, then the point Located in the local clipping region The internal conditions are: , For components If its surrounding box With clipping area If there is an intersection, then the component will participate in the projection calculation of the current drawing: , For the rectangular area to be captured in the node graph, let the four corner points of the rectangle be: The normal direction of the plane containing the rectangle is . No. The direction vector of the edge is: ,in, This represents the next corner point adjacent to the h-th corner point according to the arrangement of the corner points of the rectangle, which is the endpoint of the h-th side. This represents the h-th corner point, which is the starting point of the h-th edge.

[0065] The normal to the clipping plane corresponding to this edge It can be represented as: Alternatively, the direction of the image range can be taken in the opposite direction.

[0066] The four clipping planes generated by the four edges together form the local clipping region of the node graph: , Therefore, the node diagram does not project the entire curtain wall model, but rather first uses local clipping domains. Extract candidate components, and then perform local two-dimensional calculations on the candidate components.

[0067] During component projection preprocessing, it is determined whether symbolic representation, drawing replacement, or elevation positioning processing is required based on the component attribute parameters; for ordinary components that do not require special representation, their geometric contours are directly included in the projection solution object.

[0068] In the occlusion relationship solving step, based on the viewing direction, clipping constraints, and the spatial relationship between components, the visibility of the two-dimensional line segments formed after projection is classified, specifically including: Occlusion Calculation: The visibility of the projected 2D line segments is determined. Let points on the 3D component contour be... Projection as two-dimensional points The depth value of this point along the observation direction is defined as: , in, Point Depth parameters along the viewing direction. For the same two-dimensional projection position. If there are two component points and satisfy: ,and:

[0069] Then in In the projection model for the viewing direction, Compared to Closer to the observation side, The corresponding projection line segment is obscured at that position.

[0070] For projection line segments Its visibility state function is defined as: , in, Corresponding to the visible line, Corresponding to the visible lines of the clipping. Corresponding to the hidden line, This indicates that no output is needed.

[0071] Based on this, line segment classification rules are established: , , in, For the set of visible lines, This is a set of hidden lines. Line segments within this set are assigned a hidden linetype representation. .

[0072] In the drawing representation mapping step, based on the visibility classification results and component attribute parameters, two-dimensional line segments are mapped to visible lines, hidden lines, door and window symbol lines, elevation points, node patterns, or material annotation positioning points, thereby forming the basic line drawing of the corresponding two-dimensional construction drawing, which specifically includes the following: Based on attribute parameter set Mapping the two-dimensional representation of components. Define the representation mapping function: ,in, To represent the mapping function in the drawing, For components The corresponding two-dimensional representation result.

[0073] For ordinary components: That is, to directly project and represent its geometric contour.

[0074] For door and window components, if: Then the system executes the symbolic representation of doors and windows: , in, For the function representing doors and windows, For enclosing boxes of door and window components. These are the reference parameters for doors and windows. Enables the direction parameter.

[0075] In the plan view, let the width and thickness of the enclosure box for the door and window components in the local coordinate system of the drawing be: , , Then the outer frame of the door and window can be determined by the reference point. unit vector in the width direction unit vector in the thickness direction The construction is as follows: , , , , The opening curve of doors and windows can be represented as: ,in: Alternatively, the reverse interval can be obtained based on the opening direction parameter.

[0076] In the elevation drawing, the center point of the door and window components is set as: , The local lateral vector of the facade is: , The local vertical vector of the facade is: , The half-width and half-height of doors and windows are as follows: , The four points of the outer frame of the door and window are: , , , , The opening direction line connects the vertices of the outer frame and the midpoints of the corresponding sides according to the opening direction parameters, thereby forming door and window opening symbols for left-opening, right-opening, top-opening, or bottom-opening.

[0077] Elevation point calculation: when the attribute parameter set When elevation identification parameters are included, the system generates an elevation reference object based on the component's spatial location. Let the component center be: , The elevation reference line can be represented as: , in, The reference line direction vector, , This is half the length of the reference line.

[0078] The elevation reference line is obtained after projection: , The elevation positioning point is taken as the midpoint of the projection reference line: , Or equivalently: , The elevation positioning point Used for subsequent elevation label generation.

[0079] For components with node-based drawing attributes, the system determines the drawing insertion position and orientation based on the component's local bounding area. Let the bounding box size of the component in the node's local coordinate system be: , , The component principal direction determination function is: , The rotation angle of the pattern is: , Let the original geometric set of the node pattern be... Its original center is The target location point in the node graph is Then the pattern transformation relationship is: , in, Indicates translation transformation. This represents a rotational transformation about the normal of the node diagram. Therefore, the node diagram can automatically perform attitude matching based on the component orientation.

[0080] For components with material or component properties, the system projects them as line segments. The midpoint is used as a candidate point for material labeling: , And according to the material or material name Grouping: , To ensure the generated results fall within the specified location on the drawing, the system performs overall positioning of the 2D output results. Let the current set of 2D results be: , in, This represents a set of door and window symbols, node symbols, or other symbolized results. The system calculates the bounding box center of the two-dimensional result set: , in, and These are the minimum and maximum points of the bounding box of the two-dimensional output result, respectively.

[0081] Let the user-specified drawing reference point be... Then the overall translation vector is: , For any output two-dimensional point The result after location is: , As a result, the set of visible lines, the set of hidden lines, the set of symbols, and the set of annotation positioning points are all moved to the reference position of the drawing.

[0082] As a further embodiment, by combining preset view direction, section range, and projection rules, basic line drawings for two-dimensional construction drawings of attributed component objects are generated in different drawing types, including plan views, elevation views, detail views, and large-scale views. Specifically, this includes: The process of generating a two-dimensional plan view of a building curtain wall structure includes: receiving the geometric objects of the curtain wall components after attribute binding as input, and setting the corresponding plan view direction and clipping plane parameters to limit the viewing direction and projection depth of the plan view; on this basis, performing coordinate extraction and projection calculation on the three-dimensional geometry of the input components, performing visibility analysis on the curve contours located in the set view direction to distinguish between visible and invisible contour lines, and outputting them to the corresponding result sets respectively.

[0083] Furthermore, a two-dimensional planar basic line drawing of the curtain wall structure is generated based on the target component dataset, including: firstly, establishing a planar projection coordinate system according to the drawing requirements of the plan view, which includes the planar viewing direction, the direction on the drawing, and the projection reference point; secondly, determining the projection calculation domain based on the spatial enclosure range of each component in the target component dataset, and establishing a parallel projection model within this projection calculation domain; the parallel projection model is used to map the spatial contour of the three-dimensional components to the two-dimensional drawing plane, while preserving the front-to-back occlusion relationship between different components along the viewing direction.

[0084] Before projection, the target component's attributes are determined. For ordinary curtain wall components, their geometric contours are used as the projection solution object; for components with door and window attributes, symbolic representation units for doors and windows are constructed based on the component's local geometric range and opening direction parameters. These symbolic representation units include door and window boundary representations and opening trajectory representations, and are used as the projection solution object in subsequent calculations.

[0085] Line type mapping is performed based on visibility status: 2D line segments determined to be visible or clipped are written into the visible line set; 2D line segments determined to be occluded are written into the hidden line set and assigned a hidden line type representation. For each output 2D line segment, the system simultaneously writes the attribute parameters of its source component, enabling subsequent annotation, filling, layer management, and drawing organization modules to continue identifying the component and material categories corresponding to that line segment. Finally, the system performs overall positioning of the visible and hidden line sets based on the input reference points, outputting the basic line drawing of the plan view.

[0086] In this process, not only was the two-dimensional planar contour of the three-dimensional curtain wall components extracted, but the component attribute information corresponding to each curve was also retained when outputting the two-dimensional line drawing, so that the subsequent annotation, line type control, color control and drawing frame organization modules could continue to call it.

[0087] Furthermore, the plan view generation module can adapt to different plan view output requirements in curtain wall engineering. For curtain wall models of different floors, different axis segments, or different areas, the target area can be oriented and generated by adjusting the view direction parameters and clipping plane parameters, thereby obtaining a two-dimensional plan view that meets the construction expression requirements. Since this module retains component attributes and component category information simultaneously when generating two-dimensional contours, it can further support the differentiated expression of components such as doors, windows, grids, keels, and panels in the plan view, improving the consistency between the plan view output results and the actual curtain wall structure.

[0088] The elevation drawing generation module converts the target component dataset into a two-dimensional basic line drawing in the direction of the curtain wall facade. This module establishes a facade projection coordinate system based on the curtain wall facade orientation, including the facade viewing direction and the vertical direction. The system determines the facade projection field of view based on the overall spatial extent of the target component dataset and constructs a parallel projection calculation model, projecting the three-dimensional components onto the two-dimensional facade representation plane along the facade viewing direction.

[0089] Before component projection, the elevation drawing generation module performs elevation representation preprocessing on the components based on attribute parameters. For ordinary components, the system retains their three-dimensional outline for elevation projection; for components with door and window attributes, the system establishes a local representation coordinate system for doors and windows based on the local enclosure range of the door and window components, the elevation viewing direction, and the vertical direction, and generates the outer frame lines and opening direction lines of doors and windows according to the opening direction parameters. This ensures that door and window components are no longer simply represented as ordinary projected outlines in the elevation drawing, but are transformed into symbolic graphics that conform to the expression habits of curtain wall construction drawings.

[0090] For components with elevation identification attributes, the elevation drawing generation module generates an elevation reference object based on the component's spatial positioning information and incorporates this elevation reference object into the elevation projection solution process. After projection is completed, the system extracts elevation positioning points from the two-dimensional results corresponding to the elevation reference object, which serve as input for the subsequent elevation annotation module.

[0091] During the elevation projection solution process, the system uses the clipping plane as the constraint for the elevation drawing range and the elevation viewing direction as the depth determination direction. It solves for the front-to-back occlusion relationships between 3D components and divides the projected 2D line segments into visible line segments, clipped visible line segments, and occluded line segments. Visible and clipped visible line segments are added to the elevation visible line set, while occluded line segments are added to the elevation hidden line set and assigned a hidden line type. The elevation visible line set, the elevation hidden line set, and the elevation positioning points together constitute the basic line drawing data for the elevation drawing.

[0092] In some implementations, after outputting the basic elevation line drawing, the elevation drawing generation module can also work collaboratively with the section filling module, legend generation module, and elevation grid display module in the subsequent drawing enhancement layer. This allows for further functions in the elevation drawing, such as material filling, generation of door and window opening symbols, and automatic addition of elevation lines and axes. Therefore, the elevation drawing generation module not only automatically generates the basic elevation outline but also serves as the foundation for standardized expression in subsequent construction drawings, providing support for the formation of complete, clear, and engineering-applicable curtain wall elevation construction drawings.

[0093] The node diagram generation module is used to represent two-dimensional nodes in the local connection areas of the curtain wall. This module first establishes a local node coordinate system based on the user-specified node drawing range. This local node coordinate system is determined by the drawing range plane, the node viewing direction, and the direction on the node diagram. The system forms a local spatial clipping domain based on the drawing range and extracts components from the target component dataset that have spatial intersection with this local spatial clipping domain as a set of candidate node components.

[0094] For the set of candidate node components, the node diagram generation module performs two types of processing based on attribute parameters. The first type is geometric projection processing, which preserves the geometric contour of ordinary components and incorporates them into the local projection solution model. The second type is pattern representation processing, which, for components with pattern representation attributes, determines their principal direction and positioning reference based on the component's local bounding range and converts the component into a pattern positioning object. The pattern positioning object is used to determine the pattern insertion position and pattern orientation in the node diagram projection result, rather than directly using the original 3D contour as the final node diagram representation.

[0095] During the local projection solution phase, the node diagram generation module uses the node viewing direction as the projection direction and the local spatial clipping domain as the projection boundary to perform two-dimensional processing on the set of candidate node components and solve the occlusion relationship between component outlines within the local area. For ordinary component line segments in the projection result, the system synchronously writes the attributes of their source components into the two-dimensional line segments and generates material annotation candidate points based on their material or material properties. For pattern positioning objects in the projection result, the system calls the corresponding node pattern from the pattern library based on its pattern expression attributes and performs posture matching on the pattern according to the main direction of the component to ensure that the pattern is consistent with the actual arrangement direction of the component within the node area.

[0096] The node diagram generation module ultimately outputs a set of node patterns, a set of node positioning points, and a set of labeled positioning points grouped by material or material properties. Therefore, node diagram generation is no longer just a simple projection of a local 3D model, but a node representation algorithm that integrates local clipping, occlusion solving, pattern replacement, and material labeling and positioning.

[0097] In this process, the node diagram generation module not only extracts the two-dimensional node outlines of the three-dimensional components of the curtain wall node, but also retains the component attribute information and node association information corresponding to each curve when outputting the two-dimensional line drawing, so that the subsequent annotation, section filling, line type control and drawing frame organization modules can continue to call it.

[0098] Furthermore, the node diagram generation module can adapt to the drawing requirements of different node details in curtain wall engineering. For different types of local structures such as column-beam connection nodes, panel-stud connection nodes, transition node nodes, embedded part nodes, edge finishing nodes, and operable sash nodes, users can adjust the section position parameters, local selection range parameters, and viewing direction parameters to orient the target node area and obtain a two-dimensional node diagram that meets the construction expression requirements. Since this module retains the component attributes, material attributes, and node position association information simultaneously when generating the two-dimensional contour, it can further support the differentiated expression of connectors, hardware, profiles, and auxiliary components in the node diagram, improve the ability of the node diagram output to reflect the local structural details and connection relationships of the curtain wall, and make it more suitable for construction installation and node detailing design.

[0099] The enlarged-scale drawing generation module is used to enlarge and represent a local area of ​​the curtain wall in two dimensions. This module establishes a projection calculation model of the enlarged-scale drawing based on the local enlarged area, the viewing direction, the direction on the drawing, and the clipping plane, and extracts the set of components involved in this local enlarged representation based on the target component dataset.

[0100] In the large-scale drawing projection calculation model, the system first determines the projection field of view of the enlarged area based on the spatial enclosure range of the local component set and establishes parallel projection relationships. Then, it constrains the range of the components participating in the projection based on the clipping plane, so that the large-scale drawing only expresses the structural information within the specified local area. For components with door and window representation attributes, the system generates symbolic representations of doors and windows in the large-scale drawing based on their local geometric range and opening direction parameters; for ordinary components, their three-dimensional contours are preserved for projection and occlusion solutions.

[0101] The enlarged-scale drawing generation module can also generate index-related representations based on node sectioning objects. Specifically, the system incorporates node sectioning objects as index positioning objects into the projection solution process, and determines the start and end ranges and label positions of index symbols based on their projection results; subsequently, it combines node index information to generate index expression units including leader lines, index identifiers, and local pointing relationships. Thus, the enlarged-scale drawing not only outputs enlarged local line drawings but also maintains the index association with the node diagram or main diagram.

[0102] After completing the projection and occlusion solutions, the system categorizes visible or clipped visible line segments into the visible line set of the detail drawing, and occluded line segments into the hidden line set of the detail drawing, while simultaneously writing the component attributes into the corresponding 2D line segments. Finally, the detail drawing generation module performs overall positioning of the output results based on the input reference points, forming a localized enlarged basic line drawing that can be used for subsequent annotation, filling, and frame arrangement.

[0103] Furthermore, after outputting the basic line drawing of the detailed drawing, the detailed drawing generation module can work collaboratively with the section filling module, annotation module, legend generation module, and index association module in the subsequent drawing enhancement layer. This allows for further functions such as filling in local components, dimensioning, material identification, and generating index correspondences with the main drawing or node drawing within the detailed drawing. Thus, the detailed drawing generation module can automatically generate enlarged outlines of local curtain wall structures, providing support for creating complete, clear, and highly readable detailed construction drawings of curtain walls. Compared to plan views and elevation views, detailed drawings focus more on the enlarged expression and detailed display of local areas, providing more refined construction drawing representations for complex curtain wall structures, irregular connection parts, and key construction methods.

[0104] Post-processing layer: This layer processes the generated floor plans, elevations, detail drawings, and large-scale drawings by adding sections, annotations, and drawing frames to create standardized 2D construction drawings.

[0105] The post-processing layer of drawings includes at least a section filling module, an annotation module, and a drawing frame module. Each module shares the two-dimensional line drawing data and its corresponding component attribute information output by the aforementioned two-dimensional construction drawing generation layer. Based on preset filling rules, annotation rules, and layout rules, each module performs section material expression, dimension and index information supplementation, and drawing frame organization and layout output processing on the basic line drawing, thereby forming a standardized two-dimensional construction drawing result that can be directly used for curtain wall engineering design, detailed design, and construction drawing.

[0106] The section filling module is used to automatically fill the section areas in the two-dimensional construction drawings of the building curtain wall structure. This module receives the section boundary curves output from the two-dimensional construction drawing generation layer as input, and reads the component attribute information, component type information, and preset section filling style information corresponding to the boundary curves to determine the filling expression method for different components in the construction drawings. Based on this, the section filling module classifies the target boundary curves according to the section filling style, and performs area identification and filling generation based on the classified boundary curves, thereby outputting section filling results adapted to different curtain wall components, different material categories, and different construction expression requirements. In this process, the section filling module not only realizes automatic material expression of the section area, but also retains the correspondence between the filling results and the original component attributes, so that subsequent layer management, color control, and drawing output modules can continue to use them.

[0107] Furthermore, the section filling module can adapt to the differentiated requirements of section expression for different components in curtain wall engineering.

[0108] For steel, profiles, keels, connectors, embedded parts, and other components with local holes, the section filling module can determine whether to retain the hole boundary based on the component type attributes, and generate the corresponding filling area according to the combination relationship between the outer contour boundary and the inner contour boundary. For components that do not need to retain holes, the filling result can be directly generated based on the closed boundary. To improve the filling stability under complex section boundaries, the section filling module can also perform intersection recognition, curve breaking, duplicate edge elimination, and closed area construction processing on the target curve, thereby improving the accuracy and completeness of the section filling result in complex curtain wall structures and irregular component scenarios. Thus, the section filling module can not only complete the section style expression of general components, but also serve as an important support module for the partial section expression of curtain wall node drawings, detail drawings, and elevation drawings, improving the clarity of material differentiation and structural expression in construction drawings.

[0109] The annotation module is used to automatically generate annotation information for target components and locations in 2D construction drawings of building curtain wall structures. This module receives the 2D line drawing output from the 2D construction drawing generation layer and its corresponding component attribute information as input. Based on the annotation rules of different drawing types, it identifies the dimensions, component locations, node locations, elevation locations, and grid locations that need to be annotated, automatically generating dimension annotations, component annotations, node index annotations, elevation annotations, and grid annotations corresponding to the target drawing. Furthermore, the annotation module can also apply different annotation methods and display rules to different annotation objects based on component category, drawing scale, and drawing type, thereby outputting standardized annotation content that meets the requirements of curtain wall construction drawings. In this process, the annotation module not only provides supplementary explanations for the line drawing information but also preserves the association between the annotation object and component attributes, so that the corresponding annotation results are updated synchronously when the drawing is modified or partially updated.

[0110] Furthermore, the annotation module can adapt to the differentiated annotation requirements of plan views, elevation views, detail views, and large-scale drawings. For plan views, the annotation module can automatically extract plan dimension boundaries, opening parts, and grid positions to generate corresponding plan dimension annotations and grid annotations; for elevation views, it can extract elevation positions, grid boundaries, and key component positions to generate elevation annotations and component identifiers; for detail views and large-scale drawings, it can automatically generate connector annotations, node index annotations, and local dimension annotations based on local component relationships. Therefore, the annotation module not only improves the efficiency of drawing annotation generation and reduces repetitive manual annotation work, but also enhances the uniformity and accuracy of annotation expression across different drawing types, providing support for the standardized output of curtain wall construction drawings.

[0111] The drawing frame module is used for layout organization and drawing output processing of 2D construction drawings after section filling and annotation. This module receives the post-processed target drawing data as input and automatically calls the corresponding drawing frame template according to preset drawing type, drawing size, scale parameters, and title block rules. It performs drawing frame matching, title block information arrangement, and drawing layout processing for plan views, elevation views, detail views, and large-scale drawings. Furthermore, the drawing frame module can integrate drawing information such as drawing name, drawing number, scale, project name, and index description with the target graphic result, thereby outputting a standardized drawing page that meets the requirements of curtain wall construction drawing management. In this process, the drawing frame module not only achieves unified organization of the drawing graphic content and layout content but also retains the drawing number association and detail drawing index relationship between different drawing types for subsequent drawing management and project archiving.

[0112] In some implementations, the drawing frame module can automatically adjust the layout, scaling, and detail numbering of drawing content within the drawing frame according to different drawing types. It also automatically generates detailed drawing reference information corresponding to the main drawing, based on the index relationship between node drawings, large-scale drawings, and the main drawing. Therefore, the drawing frame module can not only complete the layout organization of a single construction drawing but also maintain consistency in format, numbering, and referencing across all drawing pages during the output of the entire set of curtain wall construction drawings, improving the standardization of the entire set of drawings.

[0113] like Figure 2 As shown, as an embodiment of the present invention, this embodiment provides a method for automated construction drawing of building curtain wall structures that integrates knowledge retrieval enhancement, including the following steps: Read the target layer from the 3D model of the building curtain wall and obtain the component instance geometry within the corresponding layer; establish a classification representation model for the curtain wall components; establish the association between the 3D geometric objects of the curtain wall and the component representation parameters, and convert the original geometric objects into attributed component objects to obtain the target component dataset; Based on the target dataset and the task parameters of the drawings to be generated, construction drawing scene description data is constructed. Combined with candidate knowledge selected from the engineering knowledge base, a hybrid retrieval and re-ranking process combining semantic similarity, local connectivity topological similarity, and geometric parameter matching degree is performed to obtain a set of evidence knowledge that matches the current construction drawing scene and generate a set of machine-executable construction drawing constraints bound to the target component. A unified 3D-to-2D conversion framework is adopted, which calls the target component dataset and uses the executable construction drawing constraint set as the control input to convert attributed component objects into 2D basic line drawings of different types. It includes a plan view generation module, an elevation view generation module, a node view generation module, and a detail view generation module. Each module generates 2D construction drawing basic line drawings of the corresponding type for the attributed component objects by combining preset view directions, cutting ranges, and projection rules. The unified 3D-to-2D conversion framework includes five steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing expression mapping. The basic line drawings of two-dimensional construction drawings of different types that have been generated are processed by section filling, annotation and drawing frame processing. The actual expression features are extracted from the processed drawings and compared with the set of constraints of executable construction drawings. When there are inconsistencies, the expression parameters of the corresponding components are updated according to the inconsistencies, triggering a second search or locally regenerating the affected drawing area. When the constraints are met, a standardized two-dimensional construction drawing result is formed and the source identifier corresponding to the evidence knowledge used is retained.

[0114] It should be noted that the specific implementation of the knowledge retrieval-enhanced automated construction drawing method for building curtain wall structures in this embodiment of the invention is similar to the specific implementation of the knowledge retrieval-enhanced automated construction drawing system for building curtain wall structures in this embodiment of the invention. For details, please refer to the description in the method section. To reduce redundancy, it will not be repeated here.

[0115] The invention will be further illustrated below with reference to a specific embodiment, in which a partial 3D model of a curtain wall is selected as the original input model. This partial 3D model of the curtain wall is composed of multiple types of curtain wall components, specifically including 450 stone hangers, 245 stone slabs, 20 aluminum plates, 25 glass panels, and auxiliary components such as steel handles. These components are modeled in the Rhino 3D modeling environment and categorized according to preset layer rules for subsequent automatic identification and processing.

[0116] In this embodiment, the data acquisition and preprocessing process is performed first. Specifically, the system reads the target layer from the partial 3D model of the curtain wall and extracts the geometric objects of the component instances within each layer. For the extracted stone hangers, stone slabs, aluminum plates, glass, steel handles, and other auxiliary components, the system further calls a preset component library to assign corresponding component type attributes, material type attributes, and drawing representation attributes to each component. The drawing representation attributes include at least line type attributes, color attributes, whether to participate in annotation attributes, whether to participate in section filling attributes, and corresponding filling style attributes. After attribute assignment, the system uses an attribute binding module to establish a connection between the 3D geometric data of each component and its corresponding component attributes, material attributes, and drawing representation attributes, thereby forming a target component dataset that can be automatically generated and used in subsequent 2D construction drawings.

[0117] Before the aforementioned local model is processed into 2D drawings, the system further establishes a construction drawing scene description based on the current drawings to be generated. Taking the generation of a node diagram for a local connection area of ​​a stone slab, hanger, and keel as an example, the system extracts the type, material, main direction, local dimensions, relative position, and interconnection relationship of the stone slab, hanger, and adjacent supporting components based on the node drawing range, and forms a structured search query by combining the node diagram scale, viewing direction, project location, and design reference date.

[0118] The knowledge retrieval enhancement decision-making layer first eliminates knowledge units that do not meet the version status or applicable conditions based on the project location, design reference date, knowledge validity status, and drawing type. Then, it retrieves evidence knowledge matching the current local connection topology and geometric parameters from specifications, standard drawing sets, company rules, and historical curtain wall node cases. For historical curtain wall node cases, the system simultaneously compares the local connection topology, component type, main direction, local dimensions, and relative position of candidate nodes with the current node to determine candidate cases that can establish a component correspondence with the current node.

[0119] The system performs structured parsing of the evidence knowledge, binding the obtained node pattern selection, material filling, line type, annotation, and indexing constraints to the corresponding target component or local component set. For historical node cases that meet the topology mapping conditions, the rotation, translation, and size adaptation parameters of the node pattern are determined based on the main direction, local dimensions, positioning reference point, and relative position between the reference component and the current target component. For candidate nodes that cannot establish a key connection mapping, automatic pattern replacement is not performed, and a second search or manual confirmation is triggered.

[0120] After completing the construction of the target component dataset, the system proceeds to the two-dimensional construction drawing generation process. For plan view generation, the system sets the corresponding planar viewing direction, clipping plane position, and projection depth parameters. It then performs planar projection calculations on the target components in the curtain wall's local model and analyzes the visibility of the projected two-dimensional contours to distinguish between visible and invisible contour lines in the plan view, generating the basic line drawing for the plan view, such as... Figure 3 As shown.

[0121] For elevation drawing generation, the system adjusts the viewing direction to correspond to the curtain wall facade, and sets corresponding projection parameters and occlusion judgment rules according to the elevation drawing requirements of the target area. It then performs elevation projection and contour extraction on the target components to obtain the basic line drawing of the elevation, such as... Figure 4 As shown and Figure 5 As shown.

[0122] For node diagram generation, the system first determines the sectioning and drawing range of the node region based on preset node positions or user-specified local structural positions, and then extracts the set of components within the node region from the target component dataset. Subsequently, the system performs local sectioning and 2D projection processing on the components within the node region, performs visibility analysis on the contour lines within the sectioning range, and outputs the basic line drawing of the node diagram, such as... Figure 6 As shown.

[0123] For generating large-scale drawings, the system extracts the corresponding set of local components based on the local area to be enlarged, the section position, or the index position. It then projects and extracts the outlines of the components within that local area to generate the basic line drawing for the large-scale drawing. Figure 7 As shown.

[0124] If necessary, the index relationship information between the detailed drawing and the node drawing or the main drawing can also be retained so as to facilitate the related expression of subsequent drawings.

[0125] After generating the basic line drawings for plan views, elevation views, detail views, and large-scale drawings, the system further performs post-processing. First, in the section filling process, the system reads the boundary curves from the 2D line drawings that require section representation and, combined with the corresponding component type attributes and fill style attributes, automatically fills the section areas of different components. For components requiring hole representation, such as some metal connectors, hangers, or perforated profiles, the system retains their inner boundaries when constructing the filling area to form a perforated filling result that meets construction expression requirements. For components that do not require retaining holes, the system directly generates the corresponding filling pattern based on the closed boundary. To ensure filling accuracy in complex boundary situations, the system can also perform intersection recognition, curve breaking, duplicate edge elimination, and closed area reconstruction processing on boundary curves.

[0126] Secondly, in the annotation process, the system identifies the dimension boundaries, node positions, component positions, elevation positions, and grid positions in the target drawing according to the annotation rules of different drawing types, and automatically generates corresponding dimension annotations, component annotations, node index annotations, elevation annotations, and grid annotations. Specifically, for node drawings and detail drawings, the system focuses on outputting the annotation results of local connection relationships and detailed dimensions; for plan drawings and elevation drawings, it focuses on outputting overall dimensions, grid dimensions, elevations, and grid information.

[0127] Finally, in the drawing frame processing, the system calls the corresponding drawing frame template according to the target drawing type, and combines information such as drawing scale, drawing name, drawing number, and project name to organize and output the layout of the completed and annotated drawing content. For node drawings and large-scale drawings, the system can also simultaneously arrange the detail drawing numbers and their index reference relationships with the main drawing, thereby forming a complete and unified two-dimensional construction drawing page for curtain walls. The model and calculation results of this example are shown in the attached drawings of the instruction manual.

[0128] After completing the above-mentioned post-processing of drawings, the consistency verification feedback module further extracts the actual expressive features of the generated drawings, such as line types, material fills, source of node drawings, annotations, elevations, index relationships, and drawing scale, and compares them item by item with the machine-executable construction drawing constraints bound to the target component. Taking the stone slab-hanger-keel node area as an example, the system checks whether the node drawings, material expression methods, annotation content, and node indexes used meet the constraints formed by the corresponding evidence knowledge; when there is inconsistency in a certain component or local drawing area, a corresponding conflict record is generated.

[0129] For inconsistencies that can be resolved by adjusting drawing parameters, the system updates the representation parameters of the corresponding target component and only re-performs projection, representation mapping, or post-processing on the affected local graphic areas, without needing to regenerate the entire set of construction drawings. For inconsistencies caused by a mismatch between the current retrieved knowledge and the target structure, the system supplements the scene retrieval conditions using conflict records and performs knowledge retrieval again. Thus, while maintaining the original 3D-to-2D geometric conversion process, the retrieved engineering knowledge can participate in the generation of construction drawings and form a closed-loop verification of the final output results.

[0130] Through the above implementation process, the system can automatically identify, bind attributes, convert 2D graphics, and post-process various components such as stone hangers, stone panels, aluminum panels, glass, and steel handles based on the partial 3D model of the curtain wall. This results in the generation of plan views, elevation views, detail views, and large-scale drawings that meet the application requirements of curtain wall engineering. Compared to manually redrawing 2D construction drawings based on a 3D model, this embodiment significantly improves drawing efficiency, reduces the risk of human error, and enhances data consistency and standardized expression between the 3D model and the 2D construction drawings.

[0131] As an embodiment of the present invention, this embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the above-described method for automated construction drawing of building curtain wall structures with enhanced knowledge retrieval.

[0132] As an embodiment of the present invention, this embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the above-described method for automated construction drawing of building curtain wall structures with enhanced knowledge retrieval.

[0133] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An automated construction drawing system for building curtain wall structures that integrates knowledge retrieval and enhancement, characterized in that: include: The target component dataset construction layer is used to read the target layer in the 3D model of the building curtain wall and obtain the geometry of the component instances in the corresponding layer; Establish a classification and representation model for curtain wall components; establish the association between the three-dimensional geometric objects of the curtain wall and the component representation parameters, and convert the original geometric objects into attributed component objects to obtain the target component dataset; The knowledge retrieval enhancement decision layer is used to construct construction drawing scene description data based on the target construction dataset and the task parameters of the drawings to be generated. Combined with candidate knowledge selected from the engineering knowledge base, it performs hybrid retrieval and re-ranking that combines semantic similarity, local connectivity topological similarity and geometric parameter matching degree to obtain a set of evidence knowledge that matches the current construction drawing scene and generate a set of machine-executable construction drawing constraints bound to the target component. The 2D construction drawing generation layer uses a unified 3D-to-2D conversion framework to call the target component dataset and uses the executable construction drawing constraint set as control input to convert attributed component objects into 2D basic line drawings of different types. It includes a plan view generation module, an elevation view generation module, a node view generation module, and a detail view generation module. Each module generates corresponding 2D construction drawing basic line drawings for attributed component objects by combining preset view directions, cutting ranges, and projection rules. The unified 3D-to-2D conversion framework includes five steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing expression mapping. The drawing post-conformity verification layer is used to process the section filling, annotation and drawing frame of the basic line drawings of the generated two-dimensional construction drawings of different types. It extracts the actual expression features from the processed drawings and compares them with the set of constraints of the executable construction drawings. When there are inconsistencies, it updates the expression parameters of the corresponding components, triggers a second search, or locally regenerates the affected drawing area. When the constraints are met, a standardized two-dimensional construction drawing result is formed, and the source identifier corresponding to the evidence knowledge used is retained.

2. The automated construction drawing system for building curtain wall structures with enhanced knowledge retrieval as described in claim 1, characterized in that, In the target component dataset construction module, the classification representation model includes a component category table, a material category table, and a drawing representation rule table; the drawing representation rule table includes at least the following rules: Door and window component representation rules: For components identified as doors and windows, a local representation coordinate system for door and window symbols is established based on the opening direction information in their attribute parameters and the local enclosing range of the component, and the outer contour and opening direction symbols of doors and windows are generated in the corresponding two-dimensional drawings; Elevation identification rules: For components with elevation identification attributes, the spatial position parameters of the component are used as the elevation reference position, and the corresponding elevation positioning point or elevation reference line is generated in the elevation projection result. Node pattern replacement rules: For components with pattern expression attributes, the corresponding pattern is called from the preset pattern library according to the component's projection positioning point, main direction and pattern name, and the pattern is then placed into the node diagram after attitude matching according to the component's direction. Material labeling and positioning rules: For components with material names, extract candidate labeling points based on the spatial mapping results of the component's projected line segments, and group the candidate labeling points according to the material or material name; Line type representation rules: For projected line segments that are determined to be occluded after the visibility solution, they are classified into the hidden line representation set and assigned a hidden line type; for projected line segments that are determined to be visible or clipped to visible, they are classified into the visible line representation set.

3. The automated construction drawing system for building curtain wall structures with enhanced knowledge retrieval as described in claim 1, characterized in that, In the target component dataset construction module, the association between the three-dimensional geometric objects of the curtain wall and the component expression parameters is established, including: semantic parsing of the component attribute expressions, converting continuous attribute description information into an attribute parameter set composed of attribute names and attribute values, using geometric objects as the basic binding unit, establishing a correspondence between each geometric object and the attribute parameter set, and forming an attributed component object including geometric data units and attribute parameter units.

4. The automated construction drawing system for building curtain wall structures with enhanced knowledge retrieval as described in claim 1, characterized in that, In the 2D construction drawing generation layer module, in the view parameter construction step, the viewing direction and the direction on the drawing are determined according to the drawing type, and a parallel view calculation model for 2D projection is established in combination with the overall spatial enclosure range of the target component dataset. In the local clipping domain construction step, spatial clipping constraints are established based on the clipping plane, cutting range, or local selection range set by the user, so that only components or parts of components located within the target drawing range participate in the two-dimensional transformation. During component projection preprocessing, it is determined whether symbolic representation, drawing replacement, or elevation positioning processing is required based on the component attribute parameters; for ordinary components that do not require special representation, their geometric contours are directly included in the projection solution object. In the occlusion relationship solving step, the visibility of the two-dimensional line segments formed after projection is classified based on the viewing direction, clipping constraints and the spatial relationship between components; In the drawing representation mapping step, based on the visibility classification results and component attribute parameters, the two-dimensional line segments are mapped to visible lines, hidden lines, door and window symbol lines, elevation points, node patterns or material annotation positioning points, thereby forming the basic line drawing of the corresponding two-dimensional construction drawing.

5. The automated construction drawing system for building curtain wall structures with integrated knowledge retrieval enhancement as described in claim 1, characterized in that, The plan view generation module is used to generate a two-dimensional planar basic line drawing of the curtain wall structure based on the target component dataset, including: Establish a planar projection coordinate system according to the drawing requirements of the plan view. The planar projection coordinate system includes the planar viewing direction, the direction on the drawing, and the projection reference point. The projection calculation domain is determined based on the spatial enclosing range of each component in the target component dataset, and a parallel projection model is established within this projection calculation domain. The parallel projection model is used to map the spatial contour of the three-dimensional component to the two-dimensional drawing plane, while preserving the front-to-back occlusion relationship between different components along the viewing direction. Before projection, the target component is evaluated for its attributes. For ordinary curtain wall components, their geometric contours are used as the projection solution objects. For components with door and window attributes, symbolic representation units for doors and windows are constructed based on the local geometric range and opening direction parameters of the component. The symbolic representation units for doors and windows include door and window boundary representation and opening trajectory representation. Line type mapping is performed based on visibility status. Two-dimensional line segments that are determined to be visible or clipped to be visible are written into the visible line set, and two-dimensional line segments that are determined to be occluded are written into the hidden line set and assigned a hidden line type expression. Based on the input reference point, the visible line set and the hidden line set are located as a whole, and the basic line drawing of the plan view is output.

6. The automated construction drawing system for building curtain wall structures with enhanced knowledge retrieval as described in claim 1, characterized in that, The elevation drawing generation module is used to convert the target component dataset into a two-dimensional basic line drawing in the direction of the curtain wall facade, including: A facade projection coordinate system is established based on the facade orientation. The facade projection coordinate system includes the facade viewing direction and the vertical direction. The facade projection field of view is determined based on the overall spatial range of the target component dataset. A parallel projection calculation model is constructed so that the three-dimensional component is projected onto the two-dimensional facade expression plane along the facade viewing direction. Before component projection, corresponding expression processing is performed on ordinary components, components with door and window attributes, and components with elevation recognition attributes according to the attribute parameters. In the process of solving the elevation projection, the clipping plane is used as the constraint of the elevation drawing range, the elevation viewing direction is used as the depth judgment direction, the front and back occlusion relationship between the three-dimensional components is solved, and the two-dimensional line segments obtained by projection are divided into visible line segments, clipping visible line segments and occlusion line segments. The visible line segments and clipping visible line segments are entered into the elevation visible line set, and the occlusion line segments are entered into the elevation hidden line set and assigned a hidden line type. The set of visible lines on the facade, the set of hidden lines on the facade, and the elevation positioning points together constitute the basic line drawing data of the facade drawing.

7. The automated construction drawing system for building curtain wall structures with enhanced knowledge retrieval as described in claim 1, characterized in that, The node diagram generation module is used to represent the local connection areas of the curtain wall using two-dimensional nodes, specifically including: A local node coordinate system is established based on the node image range specified by the user. The local node coordinate system is determined by the image range plane, the node viewing direction, and the node image direction. A local spatial clipping domain is formed based on the image range, and the components in the target component dataset that have spatial intersection with the local spatial clipping domain are extracted as a set of node candidate components. For the set of candidate node components, two types of processing are performed according to the attribute parameters: the first type is geometric projection processing, and the second type is graphical representation processing. In the local projection solution stage, the node viewing direction is taken as the projection direction and the local spatial clipping domain is taken as the projection boundary. The set of candidate node components is processed into two dimensions, and the occlusion relationship between component outlines within the local area is solved. For ordinary component line segments in the projection results, the attributes of their source components are synchronously written into the two-dimensional line segments, and material annotation candidate points are generated according to their material or material properties; for pattern positioning objects in the projection results, the corresponding node patterns are called from the pattern library according to their pattern expression attributes, and the pattern is matched with the orientation according to the main direction of the component to make the pattern consistent with the actual arrangement direction of the component in the node area. Finally, the node pattern set, the node positioning point set, and the annotation positioning point set grouped by material or material properties are output. Alternatively, the detailed drawing generation module is used to enlarge and represent a local area of ​​the curtain wall in two dimensions, including: A large-scale drawing projection calculation model is established based on the local magnified area, the viewing direction, the direction on the drawing, and the cutting plane. The set of components participating in the local magnification expression is extracted based on the target component dataset. The projection field of view of the magnified area is determined based on the spatial enclosed range of the local component set, and the parallel projection relationship is established. The clipping plane constrains the scope of the components involved in the projection, so that the detailed drawing only expresses the structural information within the specified local area; for components with door and window representation attributes, symbolic representations of doors and windows in the detailed drawing are generated based on their local geometric scope and opening direction parameters; for ordinary components, their three-dimensional contours are preserved for projection and occlusion solutions. After completing the projection and occlusion solutions, the visible or clipped visible line segments are added to the visible line set of the large-scale drawing, and the occluded line segments are added to the hidden line set of the large-scale drawing. The component attributes are then synchronously written to the corresponding two-dimensional line segments to obtain the basic line drawing of the large-scale drawing.

8. A method for an automated construction drawing system for building curtain wall structures that integrates knowledge retrieval and is characterized by: Includes the following steps: Read the target layer from the 3D model of the building curtain wall and obtain the component instance geometry within the corresponding layer; establish a classification representation model for the curtain wall components; establish the association between the 3D geometric objects of the curtain wall and the component representation parameters, and convert the original geometric objects into attributed component objects to obtain the target component dataset; Based on the target dataset and the task parameters of the drawings to be generated, construction drawing scene description data is constructed. Combined with candidate knowledge selected from the engineering knowledge base, a hybrid retrieval and re-ranking process combining semantic similarity, local connectivity topological similarity, and geometric parameter matching degree is performed to obtain a set of evidence knowledge that matches the current construction drawing scene and generate a set of machine-executable construction drawing constraints bound to the target component. A unified 3D-to-2D conversion framework is adopted, which calls the target component dataset and uses the executable construction drawing constraint set as the control input to convert attributed component objects into 2D basic line drawings of different types. It includes a plan view generation module, an elevation view generation module, a node view generation module, and a detail view generation module. Each module generates 2D construction drawing basic line drawings of the corresponding type for the attributed component objects by combining preset view directions, cutting ranges, and projection rules. The unified 3D-to-2D conversion framework includes five steps: view parameter construction, local clipping domain construction, component projection preprocessing, occlusion relationship solving, and drawing expression mapping. The basic line drawings of two-dimensional construction drawings of different types that have been generated are processed by section filling, annotation and drawing frame processing. The actual expression features are extracted from the processed drawings and compared with the set of constraints of executable construction drawings. When there are inconsistencies, the expression parameters of the corresponding components are updated according to the inconsistencies, triggering a second search or locally regenerating the affected drawing area. When the constraints are met, a standardized two-dimensional construction drawing result is formed and the source identifier corresponding to the evidence knowledge used is retained.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the automated construction drawing method for building curtain wall structures enhanced by fusion knowledge retrieval as described in claim 8.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the automated construction drawing method for building curtain wall structures with enhanced knowledge retrieval as described in claim 8.

Images