Computer-aided design-based architectural design system and method

Abstract

The application discloses a building design system and method based on computer-aided design, relates to the technical field of computer-aided building design, and comprises the following modules: a dynamic demand perception module, which captures the design intention and function constraint of a user in real time through a natural language interaction interface; a multi-dimensional generation hub module, which generates a three-dimensional space model conforming to building specifications, a facade form scheme adapting to a site environment and a component connection graph meeting structural mechanics; a cross-dimension verification engine module, which performs real-time cross verification on the three-dimensional space model, the facade form scheme and the component connection graph to generate a multi-dimensional compatibility index; and an autonomous optimization feedback module, which automatically corrects based on the multi-dimensional compatibility index to form a closed-loop iterative optimization mechanism.The application realizes intelligent processing of the whole process from user demand input to final design scheme output, can effectively improve the efficiency and quality of building design, and reduces the labor cost and time cost in the design process.

Description

Technical Field

[0001] This invention relates to the field of computer-aided architectural design technology, and in particular to architectural design systems and methods based on computer-aided design. Background Technology

[0002] The architectural design field has long relied on a manually-driven design process. Designers had to manually draw up blueprints and coordinate structural, plumbing, and electrical design content through meetings or document transfers. Because the design tools of each discipline operated independently and their data formats were incompatible, design changes required repeated manual verification of data consistency, resulting in a significant extension of the design cycle.

[0003] While existing computer-aided design tools can achieve digital drafting, they are still limited to standalone operation and static drawing output. During the design process, key elements such as the structural feasibility of architectural schemes and the rationality of equipment and pipeline layouts rely on human experience for judgment, lacking a real-time collaborative verification mechanism. When multiple disciplines modify design schemes in parallel, design conflicts often arise due to the lack of version management and data synchronization delays, leading to rework during the construction phase.

[0004] A more prominent contradiction lies in the fact that quality control in the traditional design process relies heavily on individual experience. Differences in the understanding of standards and regulations among different design teams, coupled with information gaps caused by insufficient collaboration, make it difficult to reliably guarantee the safety and economy of architectural solutions. Summary of the Invention

[0005] In view of this, the present invention proposes a computer-aided design system and method for building design, in order to solve the problems of low efficiency and poor coordination in manual design in the prior art, which leads to unstable quality of building schemes.

[0006] The specific technical solution of this invention is as follows: A computer-aided architectural design system includes the following interconnected modules: The dynamic requirement perception module captures the user's design intent and functional constraints in real time through a natural language interaction interface, automatically identifies the spatial logic relationship and human environment elements of the architectural scene, and transforms unstructured requirements into structured design parameters. The multidimensional generation central module receives structured design parameters and synchronously drives the rule generation unit, image evolution unit, and topology map unit to generate a three-dimensional spatial model that conforms to building codes, a facade morphology scheme that adapts to the site environment, and a component connection map that meets structural mechanics requirements, respectively. The cross-dimensional verification engine module performs real-time cross-verification of the 3D spatial model, facade morphology scheme and component connection diagram. It detects lighting and ventilation efficiency through a physical environment simulator, evaluates structural stability through a stress field analyzer, and generates multi-dimensional compatibility indicators. The autonomous optimization feedback module, based on multi-dimensional compatibility indicators, automatically corrects the spatial segmentation logic of the rule generation unit, adjusts the morphological generation weight of the image evolution unit, and reconstructs the node connection strategy of the topological graph unit, forming a closed-loop iterative optimization mechanism.

[0007] Specifically, it also includes: an immersive collaborative interface module that synchronously maps optimized design data to the virtual construction environment, supports multiple users to adjust component properties in real time through spatial gestures and voice commands, and dynamically feeds all modifications back to the dynamic requirement perception module to trigger a re-optimization process.

[0008] Specifically, the dynamic demand perception module outputs a demand parameter vector, which includes a demand identifier field, a space type field, a functional attribute field, an environmental requirement field, a cultural characteristic field, and a constraint field. The demand identifier field uniquely identifies the demand record, the space type field records all space types and their area percentages, the functional attribute field records the specific functional requirements of each space, the environmental requirement field records quantitative indicators such as lighting or ventilation, the cultural characteristic field records cultural attribute labels, and the constraint field records budget restrictions or regulatory requirements. This vector is transmitted via a message queue and serialized in JSON format.

[0009] Specifically, the rule generation unit of the multidimensional generation central module generates a three-dimensional spatial model based on the rule library of building design specifications. The rule library is organized according to spatial type, including spatial area range, spatial shape constraints, spatial positional relationships, and spatial opening requirements. This unit determines the optimal spatial layout scheme under rule constraints through optimization algorithms and outputs a three-dimensional spatial model file in building information model data format, including three levels: overall building model, floor decomposition model, and spatial unit model.

[0010] Specifically, the image evolution unit of the multidimensional generation central module generates facade morphology schemes based on generative adversarial networks and style transfer technology. This unit extracts the volume features, color features, material features and detail features of the surrounding environment as generation constraints, and generates multiple facade morphology schemes through generative adversarial networks. The morphology generation weight is controlled to balance the user's aesthetic preferences and the harmony with the surrounding environment. The output facade morphology scheme file includes three levels: overall facade image, facade segmented image, and material annotation image.

[0011] Specifically, the topology graph unit of the multidimensional generation central module generates component connection graphs based on graph neural networks and topology optimization methods. This unit determines the structural system type based on building height, span, and function, generates structural component layout schemes based on spatial layout models, analyzes component connection relationships through graph neural networks, generates component connection graphs that meet mechanical equilibrium conditions, and automatically checks the stress state of components to ensure that the stress ratio is within a safe range. The output component connection graph file includes three core parts: a node list, an edge list, and an attribute mapping.

[0012] Specifically, the physical environment simulator of the cross-dimensional verification engine module performs lighting simulation, ventilation simulation, and thermal simulation. The lighting simulation uses a ray tracing algorithm to calculate the natural daylight coefficient, taking into account the window position, size, and shading effects. The ventilation simulation uses computational fluid dynamics to simulate airflow and calculate ventilation volume and air exchange rate. The thermal simulation calculates energy consumption indicators and thermal comfort based on a heat transfer model. The output physical environment verification report includes three parts: lighting verification results, ventilation verification results, and thermal verification results.

[0013] Specifically, the stress field analyzer of the cross-dimensional verification engine module performs static analysis, dynamic analysis, and stability analysis; the static analysis calculates the internal forces and deformations of the components based on the finite element method and evaluates the ultimate limit state of bearing capacity; the dynamic analysis performs modal analysis, seismic response analysis, and wind vibration response analysis; the stability analysis evaluates the buckling characteristics of the components and the second-order effects of the structure; the output structural safety verification report includes three parts: static analysis results, dynamic analysis results, and stability analysis results.

[0014] Specifically, the autonomous optimization feedback module performs problem diagnosis, parameter adjustment, and scheme regeneration based on multi-dimensional compatibility indicators. In the problem diagnosis stage, it identifies non-compliant indicators and associates them with the generation unit, outputting a diagnostic report. In the parameter adjustment stage, it automatically corrects the spatial segmentation logic of the rule generation unit, adjusts the morphological generation weight of the image evolution unit, and reconstructs the node connection strategy of the topology map unit. In the scheme regeneration stage, it drives the multi-dimensional generation hub module to regenerate the scheme, forming a closed-loop iterative optimization process. The iteration ends when the improvement of the compatibility indicators is less than the convergence threshold.

[0015] Computer-aided design-based architectural design methods include: Step S1: Receive the user's architectural design request, and use the dynamic requirement perception module to perform natural language parsing, core requirement identification, spatial logic reasoning, and cultural feature extraction on the request to generate a requirement parameter vector. Step S2: The multidimensional generation central module receives the demand parameter vector and executes in parallel the rule generation unit to generate a three-dimensional spatial model based on building design specifications, the image evolution unit to generate a facade morphology scheme based on generative adversarial network and site environment data, and the topology map unit to generate a component connection map based on graph neural network. Step S3: The cross-dimensional verification engine module performs lighting simulation, ventilation simulation and thermal analysis on the three-dimensional space model to generate a physical environment verification report, performs static analysis, dynamic analysis and stability analysis on the component connection diagram to generate a structural safety verification report, and summarizes and generates multi-dimensional compatibility indicators. Step S4: The autonomous optimization feedback module diagnoses non-compliant indicators based on multi-dimensional compatibility indicators, adjusts the spatial segmentation logic, morphological generation weight, or node connection strategy parameters, drives the multi-dimensional generation central module to regenerate the scheme, and performs closed-loop iteration until the compatibility indicators meet the standards. Step S5: The optimization scheme is mapped to the virtual reality environment through the immersive collaborative interface module, allowing users to interact through spatial gestures, voice commands or game controllers, and the collaborative operation log is fed back to the dynamic demand perception module. Step S6: The dynamic demand perception module updates the demand parameter vector according to the collaborative operation log and re-triggers the process that started from step S2. Step S7: Once the design scheme meets the verification requirements and is confirmed by the user, the architectural design results, including a 3D spatial model file, an elevation morphology scheme file, and a component connection diagram file, are output.

[0016] The beneficial effects of this invention are as follows: 1. The dynamic requirement perception module uses deep learning technology to build a semantic parsing model, identify the user's deep architectural design intentions, transform scattered requirements into standard design parameters, reduce the complexity of user expression, and improve the accuracy of requirement capture.

[0017] 2. The multi-dimensional generation central module generates standardized spatial models, environmentally adapted facade schemes, and safety component diagrams through multi-unit collaborative processing, enhancing the multi-dimensional generation capability of design schemes and optimizing the comprehensiveness of the schemes.

[0018] 3. The cross-dimensional verification engine module uses environmental simulation and structural analysis to evaluate the feasibility of the design scheme in terms of function, environment and structure. It integrates multi-disciplinary verification into the same engine and carries it out in parallel, avoiding verification bias caused by information gaps between different disciplines, improving verification coverage and ensuring the overall feasibility of the scheme.

[0019] 4. The autonomous optimization feedback module automatically diagnoses problems, adjusts parameters, and regenerates solutions based on the verification results, forming a closed-loop iterative process to achieve continuous optimization of the solution and avoid design defects.

[0020] 5. The immersive collaborative interface module presents design solutions through virtual reality technology, supports multi-party collaborative operations, synchronizes scenes and instructions in real time, achieves seamless integration of user interaction and system optimization, enhances the real-time nature of the design process, and reduces repeated modifications to the solution due to communication discrepancies. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the computer-aided design system for building design according to the present invention; Figure 2 This is a flowchart illustrating the computer-aided design method for architectural design according to the present invention. Detailed Implementation

[0023] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0024] This invention proposes a computer-aided design system and method for architectural design, such as... Figure 1 As shown, the system can accurately capture users' architectural design needs through the dynamic demand perception module, generate multi-dimensional design schemes in parallel through the multi-dimensional generation hub module, conduct comprehensive feasibility verification of the design schemes through the cross-dimensional verification engine module, realize closed-loop iterative optimization of the schemes through the autonomous optimization feedback module, and provide an immersive collaborative design environment for multiple users through the immersive collaborative interface module, thereby effectively improving the efficiency and quality of architectural design.

[0025] The dynamic demand perception module serves as the entry point and data source for the entire system, responsible for transforming users' vague and fragmented architectural design requests into standardized design parameters that the system can recognize. The core functions of this module include natural language parsing, core demand identification, spatial logic reasoning, and cultural feature extraction.

[0026] In terms of natural language parsing, the dynamic requirement perception module employs deep learning-based Natural Language Understanding (NLU) technology to construct a semantic parsing model specifically for the architectural design field. This model is pre-trained and fine-tuned on a large corpus of architectural design documents, design code clauses, and user requirement descriptions, enabling it to accurately understand the architectural design intentions expressed by users in everyday language. For example, when a user states, "I hope the living room has better natural light," the system can identify this as a specific requirement for natural lighting conditions; when a user states, "I feel the current apartment layout is a bit cramped," the system can infer that this is a potential need for spatial openness and smooth circulation. This ability to understand from surface semantics to deeper intentions is something that traditional form-based requirement collection methods lack.

[0027] In terms of identifying core needs, the dynamic needs perception module constructs an architectural needs ontology model, mapping various user expressions to corresponding architectural concept nodes. The needs ontology model includes four root categories: space type, functional attributes, environmental requirements, and cultural characteristics. Each root category is further subdivided into multiple subcategories. For example, space type includes specific spaces such as living room, bedroom, kitchen, and bathroom; functional attributes include area requirements, floor height requirements, storage requirements, and privacy requirements; environmental requirements include lighting requirements, ventilation requirements, sound insulation requirements, and landscape requirements; and cultural characteristics include architectural style, regional characteristics, and aesthetic preferences. When a user mentions multiple related concepts in their expression, the system can automatically identify the logical relationships between them, such as inclusion, parallel, and causal relationships, thereby more accurately grasping the user's complete needs.

[0028] In terms of spatial logic reasoning, the dynamic demand perception module can automatically infer the spatial logic relationships corresponding to the architectural scene based on the user's description. For example, in commercial building scenarios, the system will automatically associate spatial logic elements such as pedestrian flow, spatial nodes, and business distribution; in residential building scenarios, the system will automatically associate spatial logic elements such as active and quiet zoning, clean and dirty separation, and storage systems; in cultural building scenarios, the system will automatically associate spatial logic elements such as spatial sequence, cultural narrative, and regional expression. This automatic reasoning capability greatly reduces the burden of expression for users. Users do not need to list all design requirements one by one; the system can automatically supplement common spatial logic requirements based on the building type.

[0029] In terms of cultural feature extraction, the dynamic demand perception module can extract cultural attribute-related information from user descriptions by combining the cultural environment characteristics of the project's location. For example, when the project is located in a water town in southern China, the system will automatically focus on regional cultural elements such as water features and white-walled, black-tiled houses; when the project is located in an ancient northern city, the system will automatically focus on traditional architectural elements such as courtyard layouts and brick carvings. The system has a built-in cultural feature database covering major regions across the country, which can automatically match corresponding cultural tags based on the project location, providing cultural-dimensional design constraints for subsequent scheme generation.

[0030] The dynamic requirement perception module outputs a requirement parameter vector, which includes six primary fields: requirement identifier, space type, functional attribute, environmental requirements, cultural characteristics, and constraint conditions. The requirement identifier field uniquely identifies each requirement record and includes two subfields: requirement number and requirement source. The space type field is an array containing all space types involved by the user and their area percentages. The functional attribute field is a key-value pair field recording the user's specific functional requirements for each space. The environmental requirements field is an array containing the user's quantitative requirements for environmental indicators such as lighting, ventilation, temperature, humidity, and sound. The cultural characteristics field is a tag array containing cultural attribute tags extracted from user descriptions and project background. The constraint conditions field is a composite type, including rigid constraints such as budget limits, schedule requirements, and mandatory regulatory clauses. The requirement parameter vector is transmitted to the multi-dimensional generation hub module via a message queue and serialized using JSON format.

[0031] The multidimensional generation hub module is responsible for transforming structured requirement parameters into multidimensional architectural design schemes. This module consists of three generation units: a rule generation unit, an image evolution unit, and a topology graph unit. These generation units operate in parallel and interact with each other through a hierarchical bus and a real-time message queue.

[0032] The rule generation unit is responsible for generating 3D spatial models that conform to building design codes. Its core algorithm is based on Building Information Modeling (BIM) technology and parametric design methods. Internally, the unit maintains a complete rule library of building design codes, organized according to spatial type. Each type of space corresponds to a set of generation rules, including spatial area ranges, spatial shape constraints, spatial positional relationships, and spatial opening requirements. For example, rules such as bathrooms in residential buildings should avoid directly facing the living room, kitchens should be close to the dining room and far from bedrooms, and bedrooms should avoid being adjacent to elevator shafts are all encoded into the rule library. After receiving the requirement parameter vector, the rule generation unit first determines the required spatial combination based on the spatial type field, then determines the specific dimensions and positional relationships of each space based on the functional attribute field and environmental requirement field, and finally determines the optimal spatial layout scheme under the rule constraints through an optimization algorithm.

[0033] The data structure output by the rule generation unit is a 3D spatial model file. This file uses the industry-standard BIM data format and includes geometric and attribute information at three levels: the overall building model, the floor breakdown model, and the spatial unit model. The overall building model records macroscopic parameters such as the building's site location, area, and height. The floor breakdown model records information such as the floor plan outline, floor height, and structural system of each floor. The spatial unit model records detailed information such as the boundary geometry, door and window locations, and MEP (mechanical, electrical, and plumbing) points of each functional space. The models at different levels are linked through floor and spatial affiliations, supporting step-by-step queries from the overall model to the specific units. In the spatial unit model, each spatial unit includes attribute fields across four dimensions: spatial identifier, spatial type, functional attribute, and physical attribute. The spatial identifier field records the space's unique code and name; the spatial type field records the space's classification information; the functional attribute field records the space's usage function, pedestrian density, and storage requirements; and the physical attribute field records quantitative indicators such as the space's area, volume, daylight factor, and ventilation.

[0034] The image evolution unit is responsible for generating architectural facade design schemes adapted to the site environment. Its core algorithm is based on Generative Adversarial Network (GAN) and style transfer techniques. Internally, the unit maintains a sample library containing numerous excellent architectural facade examples, each labeled with site environment and style characteristics. Upon receiving the requirement parameter vector, the image evolution unit first acquires image data and 3D model data of the surrounding environment based on the project's site location. Then, it extracts information such as the volume, color, material, and detail features of surrounding buildings as generation constraints. Finally, under these constraints, the GAN generates multiple optional facade design schemes. During the generation process, the system controls the form generation weight parameters to ensure that the generated facade schemes both satisfy the user's aesthetic preferences and harmonize with the surrounding environment.

[0035] The data structure output by the image evolution unit is a facade morphology scheme file. This file includes three levels of image data and corresponding metadata: an overall facade image, segmented facade images, and material-annotated images. The overall facade image is a frontal view of each facade, with a resolution of at least 4K. The segmented facade images are grid-based diagrams of the facade, recording the position, size, and number of each grid cell. The material-annotated images are material-annotated diagrams of the facade, recording the material type and color parameters for each area. The metadata includes descriptive information such as the facade scheme number, design style tag, explanation of the source of inspiration, and a compatibility score with the surrounding environment. The facade morphology scheme file is stored in a layered image format, supporting lossless zoom and local modifications.

[0036] The topology graph unit is responsible for generating component connection graphs that meet structural safety requirements. Its core algorithm is based on graph neural network technology and topology optimization methods. Internally, the unit constructs a building structure knowledge graph covering various building structural systems, component types, connection methods, and stress characteristics. After receiving the requirement parameter vector, the topology graph unit first determines the recommended structural system type based on parameters such as the building's height, span, and function. Then, it generates an arrangement scheme for structural components based on a spatial layout model. Finally, it analyzes the connection relationships between components using a graph neural network to generate a component connection graph that satisfies mechanical equilibrium conditions. During the generation process, the system automatically checks the stress state of the components to ensure that the stress ratios of all components are within safe limits.

[0037] The topology graph unit outputs a component connection graph file, which consists of three core parts: a node list, an edge list, and an attribute mapping. The node list records the unique identifier, geometric parameters, section type, and material properties of all structural components. The edge list records the connection relationships between all components, with each edge including four fields: start node number, end node number, connection type, and constraint conditions. The attribute mapping records the stress analysis results for each node, including quantitative indicators such as axial force, shear force, bending moment, and stress ratio. The component connection graph is stored in a graph data format, supporting graph operations such as topology query, path analysis, and stress tracing.

[0038] After the requirement parameter vector is received from the input end of the multidimensional generation hub module, it is distributed to the input queues of the three generation units through the message routing component. After each generation unit completes its generation task, it writes the result file to the corresponding output storage area and sends a completion signal to the output end of the multidimensional generation hub module through the message notification component. After collecting the results from the three generation units, the output end of the multidimensional generation hub module performs format verification and version labeling, generates a complete design scheme package including three sub-schemes, and transmits it to the cross-dimensional verification engine module through the message queue.

[0039] The cross-dimensional verification engine module is responsible for comprehensively verifying and evaluating the multi-dimensional design schemes output by the multi-dimensional generation center. This module consists of two core components: a physical environment simulator and a stress field analyzer, which are responsible for physical environment performance verification and structural safety verification, respectively.

[0040] The physical environment simulator is responsible for simulating and analyzing the physical environmental performance of a building, including lighting, ventilation, and thermal performance. In lighting simulation, the simulator uses ray tracing algorithms to calculate the natural daylight coefficient of each area of ​​the building, considering window location, size, glass transmittance, and the shading effect of surrounding buildings, outputting daylight distribution maps and daylight hours statistics for each space. In ventilation simulation, the simulator uses computational fluid dynamics to simulate indoor and outdoor airflow, calculating the ventilation volume and efficiency of each vent, analyzing the rationality of airflow organization, and outputting wind speed distribution maps and air exchange rate statistics. In thermal simulation, the simulator calculates the building's energy consumption indicators based on steady-state or transient heat transfer models, evaluates the thermal performance of the building envelope, and outputs predicted energy consumption values ​​and thermal comfort indices.

[0041] The verification data structure output by the physical environment simulator is a physical environment verification report, which includes three main parts: lighting verification results, ventilation verification results, and thermal verification results. Lighting verification results include three subfields: a natural daylight coefficient distribution map, a daylight compliance rate index, and a list of major areas with insufficient daylight. Ventilation verification results include three subfields: a ventilation volume distribution map, a ventilation compliance rate index, and a list of major areas with poor ventilation. Thermal verification results include three subfields: a table of building envelope thermal parameters, predicted energy consumption values, and thermal comfort indexes. Each subfield includes a quantitative index value, a compliance determination result, and a problem description. The verification report is stored in a structured document format and supports conditional queries and problem tracing.

[0042] The stress field analyzer is responsible for assessing the overall structural stability of the building. In static analysis, the analyzer establishes a numerical model of the building structure based on the finite element method, applies design conditions such as self-weight, live load, wind load, and seismic load, calculates the internal forces and deformations of structural members, and evaluates the ultimate limit state and serviceability limit state of the structure. In dynamic analysis, the analyzer performs modal analysis to obtain the natural periods and mode shapes of the structure, performs seismic response analysis to assess the seismic performance of the structure, and performs wind-induced vibration response analysis to assess the wind-induced vibration characteristics of high-rise buildings. In stability analysis, the analyzer performs member stability analysis to assess the buckling characteristics of compression members and performs overall stability analysis to assess the second-order effects of the structure.

[0043] The verification data output by the stress field analyzer is structured as a structural safety verification report, which includes three main parts: static analysis results, dynamic analysis results, and stability analysis results. Static analysis results include displacement envelope diagrams, internal force envelope diagrams, stress distribution diagrams, and utilization rate indicators of key components under various working conditions. Dynamic analysis results include a natural vibration period table, mode shape diagrams, seismic response time history curves, and statistical values ​​of wind-induced vibration response. Stability analysis results include a buckling characteristic value table, instability mode diagrams, and stability verification results. Each part includes quantitative index values, safety judgment results, and a list of components requiring attention. The verification report is stored in a structured document format and linked bidirectionally with the component connection diagram file.

[0044] After receiving the complete design package output by the multidimensional generation center, the cross-dimensional verification engine module first performs data parsing and format conversion, converting the three-dimensional spatial model file into a geometric model format recognizable by the physical environment simulator, and converting the component connection diagram file into a finite element model format recognizable by the stress field analyzer; then, it starts two independent verification processes in parallel: physical environment simulation and structural analysis; finally, it summarizes the results of the two verification processes to generate a comprehensive multidimensional compatibility index.

[0045] The core data structure output by the module is a multi-dimensional compatibility index, which includes three primary indicators: functional compatibility, environmental compatibility, and structural compatibility, as well as several secondary indicators. Functional compatibility indicators reflect the degree to which the spatial layout meets functional requirements, including area compliance rate, spatial squareness, and circulation smoothness. Environmental compatibility indicators reflect the building's response quality to the physical environment, including daylight compliance rate, ventilation compliance rate, and thermal compliance rate. Structural compatibility indicators reflect the safety margin of the structural design scheme, including maximum stress ratio, displacement limit ratio, and stability safety factor. Each secondary indicator includes four fields: indicator name, indicator value, compliance threshold, and compliance determination. The compatibility indicators are stored in JSON format and serve as input for the autonomous optimization feedback module.

[0046] The autonomous optimization feedback module is responsible for automatically optimizing and adjusting the solution based on the compatibility metrics output by the cross-dimensional verification engine. The module's working mechanism includes three stages: problem diagnosis, parameter adjustment, and solution regeneration.

[0047] During the problem diagnosis phase, after receiving multi-dimensional compatibility indicators, the module first checks each secondary indicator item by item to identify those that fail to meet the standards. Then, it associates the indicator category with the corresponding generation unit to determine the root cause of the problem. For example, a failure to meet the lighting compliance rate is associated with the spatial segmentation logic of the rule generation unit; a failure to meet the ventilation compliance rate is also associated with the spatial segmentation logic of the rule generation unit; a failure to meet the facade and environmental adaptability standards is associated with the morphological generation weight of the image evolution unit; and an excessive component stress ratio is associated with the node connection strategy of the topology map unit. The problem diagnosis results are output in the form of a diagnosis report, which clearly defines the type, location, severity, and optimization direction of each problem.

[0048] During the parameter adjustment phase, the module makes targeted adjustments to the operating parameters of each generation unit based on the diagnostic report. For rule generation units, the system automatically adjusts the spatial segmentation logic, such as increasing the width of the lighting surface, adjusting the position of ventilation corridors, and increasing the area of ​​ventilation openings, while updating the relevant constraints in the rule base. For image evolution units, the system adjusts the morphological generation weights, such as increasing the reference weight for the style characteristics of surrounding buildings, reducing the weight of innovation, and tightening the volume control constraints, while updating the parameter configuration of the weight controller. For topology graph units, the system reconstructs the node connection strategy, such as increasing the cross-sectional dimensions of components, adjusting the placement of components, and adding supporting components, while updating the relevant rules in the structural knowledge graph. The parameter adjustment algorithm combines gradient-based optimization methods and rule-based heuristic methods to ensure the correctness of the adjustment direction and the rationality of the adjustment range.

[0049] During the scheme regeneration phase, the adjusted parameters are fed back to the rule base and weight controller of the multi-dimensional generation central module, driving the three generation units to regenerate the design scheme. The regenerated scheme will then re-enter the cross-dimensional verification engine module for verification, forming a complete closed-loop iterative process. During the iteration process, the system records the parameter adjustment amount and the change in verification indicators for each iteration. When the change in indicators is continuously less than the convergence threshold multiple times, the scheme is considered to have converged to an acceptable optimal solution, and the iteration process ends. When it is found that parameter adjustment cannot improve a certain indicator, the system will try a larger adjustment or switch to different optimization strategies to avoid getting trapped in local optima.

[0050] The core data structure output by the autonomous optimization feedback module is an optimization parameter package. This package includes three sub-packages: rule generation parameters, morphological evolution parameters, and structural topology parameters, plus an additional section: optimization history. Rule generation parameters include three fields: spatial segmentation logic adjustment amount, functional partition weight coefficient, and circulation organization constraints. Morphological evolution parameters include three fields: style matching weight, proportional control parameters, and detail generation parameters. Structural topology parameters include three fields: component cross-section adjustment table, connection method reconfiguration table, and node position offset. The optimization history includes a complete record of input metrics, parameter adjustments, output metrics, and improvement magnitude for each iteration, supporting scheme backtracking and parameter rollback. The optimization parameter package is stored in an incremental format, recording only the adjustments relative to the initial state, facilitating the tracking of specific actions in each optimization.

[0051] The autonomous optimization feedback module receives compatibility metrics from the cross-dimensional verification engine, diagnoses problems to generate a diagnostic report, and then generates parameter adjustment instructions based on the report. These instructions generate an optimized parameter package, which is fed back to the multi-dimensional generation central module to trigger scheme regeneration. The new scheme is output to the cross-dimensional verification engine module, generating new compatibility metrics. These new compatibility metrics then re-enter the autonomous optimization feedback module, forming a complete closed-loop iteration. The entire data cycle is controlled within the autonomous optimization feedback module. When the improvement in compatibility metrics for three consecutive iterations is less than the convergence threshold, the module automatically terminates the iteration process and outputs the final optimized scheme.

[0052] The immersive collaborative interface module is responsible for presenting the optimized design to the user in a virtual reality format and receiving user feedback. This module consists of three components: a virtual mapping engine, an interaction controller, and a collaborative synchronizer.

[0053] The virtual mapping engine is responsible for synchronously mapping the optimized design data to the virtual construction environment. The mapping process includes three levels: geometric mapping, material mapping, and attribute mapping. Geometric mapping converts the geometric data in the 3D spatial model into a mesh model that the virtual reality engine can recognize, while simplifying it at multiple levels of detail to ensure rendering performance. Material mapping maps the material annotation information in the facade form scheme to material definitions that the virtual reality engine can recognize, including parameters such as color, texture, gloss, and transparency. Attribute mapping maps the attribute data in the component connection diagram to queryable interactive attributes, including construction dimensions, material type, manufacturer, and price information. After the mapping is completed, users can put on a virtual reality headset to enter the virtual architectural space and view the actual effect of the design scheme in an immersive way.

[0054] The interaction controller is responsible for receiving and processing user input, supporting three interaction methods: spatial gesture recognition, voice command recognition, and handheld controller. Spatial gesture recognition, based on computer vision technology, analyzes the user's hand movement trajectory and posture to identify the user's operational intent, supporting basic operations such as click selection, drag movement, pinch zoom, and two-finger rotation. Voice command recognition, based on speech recognition and natural language understanding technology, converts user voice commands into executable system commands, supporting semantic actions such as opening, closing, moving, replacing, and adjusting. The handheld controller provides more precise positioning and selection capabilities, suitable for scenarios requiring precise adjustments. The interaction controller converts user input into standardized operation commands, transmitting them to the dynamic demand perception module to trigger a re-optimization process.

[0055] The collaborative synchronizer supports collaborative design scenarios involving multiple users. Based on a cloud simulation platform architecture, users in different geographical locations can simultaneously access the same virtual design environment for collaborative operations. The synchronizer maintains a real-time copy of the design scenario's state and synchronizes scenario changes, user locations, and operation commands among user terminals via low-latency network protocols. When a user modifies a design, the changes are broadcast to other users' terminals in real time, ensuring all participants see the latest design status. The synchronizer also provides user list management, access control, and conflict detection functions to guarantee the order and consistency of the collaborative design process.

[0056] A bidirectional data channel is established between the immersive collaborative interface module and the dynamic requirement awareness module, forming a complete requirement response loop. After user-generated modification commands via the virtual reality interface are transmitted to the dynamic requirement awareness module, the system parses them into adjustment amounts for the requirement parameter vector, re-triggering the entire process from requirement awareness to solution generation. The generated new solution is then mapped back onto the virtual environment and presented to the user. The entire response cycle is controlled within seconds, allowing users to see the adjusted effects in real time within the virtual environment, achieving a WYSIWYG design experience.

[0057] The module's data interaction structure is a collaborative operation log, which includes five core fields: operation user identifier, operation timestamp, operation type, operation object, and operation parameters, plus an additional field: operation result. The operation user identifier records the user ID and username who performed the operation; the operation timestamp is accurate to milliseconds for time-series reconstruction of multi-user operations; the operation type enumeration records the operation category, such as spatial movement, attribute modification, and viewpoint switching; the operation object reference records the unique identifier of the operated component; the operation parameters record the specific adjustment amounts, such as movement vector, scaling ratio, and color value; and the operation result records the system's response status to the operation, such as successful execution, parameter out-of-bounds adjustment required, or rejection due to specification violation. The collaborative operation log is stored using an append-only method, supporting complete playback and audit traceability of the design process.

[0058] like Figure 2 As shown, the workflow of the method of the present invention is as follows: Step one: The user inputs architectural design requirements into the dynamic requirements perception module through a natural language interaction interface, including unstructured descriptions such as functional layout requirements, site environmental constraints, and cultural attribute requirements. The requirements perception module performs natural language parsing and intent recognition on the input text, generating a requirement parameter vector that includes fields such as space type, functional attributes, environmental requirements, cultural characteristics, and constraints. The requirement parameter vector is then transmitted to the multi-dimensional generation central module via a message queue.

[0059] Step two: After receiving the requirement parameter vector, the multi-dimensional generation central module distributes it to three parallel processing queues—the rule generation unit, the image evolution unit, and the topology graph unit—through a message routing component. The rule generation unit generates a 3D spatial model based on a building code rule base, outputting spatial model files at three levels: the overall building model, the floor decomposition model, and the spatial unit model. The image evolution unit generates facade morphology schemes based on generative adversarial networks and site environment data, outputting facade scheme files including facade images, grid images, and material annotations. The topology graph unit generates a component connection graph based on graph neural networks and structural knowledge graphs, outputting a graph file including a node list, edge list, and attribute mappings. The three generation units maintain data synchronization through a hierarchical bus and a real-time message queue.

[0060] Step 3: The cross-dimensional verification engine module receives the output files from the three generation units in parallel, performs format conversion, and then sends them to the physical environment simulator and stress field analyzer respectively. The physical environment simulator performs lighting simulation, ventilation simulation, and thermal analysis on the 3D spatial model, and outputs a physical environment verification report. The stress field analyzer performs static analysis, dynamic analysis, and stability analysis on the component connection diagram, and outputs a structural safety verification report. The module summarizes the results of the two verification reports and generates a multi-dimensional compatibility index including functional compatibility, environmental compatibility, and structural compatibility.

[0061] Step four: After receiving multi-dimensional compatibility indicators, the autonomous optimization feedback module first diagnoses problems, identifies non-compliant indicators, and associates them with the corresponding generation units. Then, it generates targeted parameter adjustment instructions, including adjustments to spatial segmentation logic, morphological generation weights, and node connection strategies. Finally, it generates an optimization parameter package containing records of the adjustments. This optimization parameter package flows back to the rule base and weight controller of the multi-dimensional generation hub module, driving the three generation units to regenerate the scheme. The regenerated scheme re-enters the verification and evaluation phase, forming a closed-loop iteration. The iteration process continues until all compatibility indicators meet the standards or the maximum number of iterations is reached.

[0062] Step five: The optimized design scheme is synchronously mapped to the virtual construction environment of the immersive collaborative interface module. Users enter the virtual environment through virtual reality interactive devices to intuitively view and experience the spatial scale, lighting effects, ventilation, and other aspects of the design scheme. Users can adjust component attributes, such as moving wall positions, modifying door and window styles, and changing decoration materials, through spatial gestures, voice commands, or controllers. All user operations are converted into collaborative operation logs and synchronized in real time to the dynamic demand perception module.

[0063] Step Six: After receiving the collaborative operation logs from the collaborative interface module, the dynamic requirement awareness module parses the user's interaction modification intent into a new requirement parameter vector. If the user's modification involves functional layout adjustments, the space type and functional attribute fields are updated; if it involves environmental quality improvements, the environmental requirements field is updated; if it involves adding or deleting cultural elements, the cultural characteristics field is updated. The updated requirement parameter vector re-triggers the entire process starting from the parallel generation phase, achieving seamless integration between user interaction and system optimization.

[0064] Step seven: Once the design scheme passes all verifications and the user confirms their satisfaction, the system outputs a complete architectural design deliverable package, including a 3D spatial model file, facade form scheme file, component connection diagram file, and related design specification documents. The deliverable package uses an industry-standard data format for storage, is compatible with mainstream architectural design software and engineering management software, and can seamlessly integrate with subsequent business processes such as construction drawing design, project cost estimation, and construction management.

[0065] The beneficial effects of this invention are as follows: 1. The system significantly lowers the technical threshold for architectural design. Users do not need to have professional architectural design knowledge; they can simply express their design requirements through natural language, effectively expanding the scope of participants in architectural design.

[0066] 2. The system significantly improves the efficiency of architectural design. The dynamic demand perception module eliminates the manual conversion process in traditional demand docking. The parallel operation of the multi-dimensional generation hub module eliminates the time loss of sequential advancement of each stage in traditional design. The parallel verification of the cross-dimensional verification engine module eliminates the communication costs of traditional multi-disciplinary verification. The automatic iteration of the autonomous optimization feedback module eliminates the manual adjustment process in traditional optimization. The entire design cycle is greatly shortened.

[0067] 3. The system effectively improves the quality of the design. The rule generation unit ensures that the solution fully complies with all specifications and requirements, the image evolution unit ensures that the solution is coordinated and unified with the surrounding environment, the topology map unit ensures that the structure of the solution is safe and reliable, and the iterative optimization of the autonomous optimization feedback module ensures that the solution continuously approaches the optimal state.

[0068] 4. The system significantly improves the user experience. The immersive collaborative interface module allows users to feel and adjust the design scheme as if they were there. The real-time linkage mechanism ensures that every user request can be responded to quickly, effectively improving user satisfaction and the first-time pass rate of design schemes.

[0069] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A computer-aided architectural design system, characterized in that, Includes the following interconnected modules: The dynamic requirement perception module captures the user's design intent and functional constraints in real time through a natural language interaction interface, automatically identifies the spatial logic relationship and human environment elements of the architectural scene, and transforms unstructured requirements into structured design parameters. The multidimensional generation central module receives structured design parameters and synchronously drives the rule generation unit, image evolution unit, and topology map unit to generate a three-dimensional spatial model that conforms to building codes, a facade morphology scheme that adapts to the site environment, and a component connection map that meets structural mechanics requirements, respectively. The cross-dimensional verification engine module performs real-time cross-verification of the 3D spatial model, facade morphology scheme and component connection diagram. It detects lighting and ventilation efficiency through a physical environment simulator, evaluates structural stability through a stress field analyzer, and generates multi-dimensional compatibility indicators. The autonomous optimization feedback module, based on multi-dimensional compatibility indicators, automatically corrects the spatial segmentation logic of the rule generation unit, adjusts the morphological generation weight of the image evolution unit, and reconstructs the node connection strategy of the topological graph unit, forming a closed-loop iterative optimization mechanism.

2. The computer-aided design-based architectural design system as described in claim 1, characterized in that, Also includes: The immersive collaborative interface module synchronously maps the optimized design data to the virtual construction environment, allowing multiple users to adjust component properties in real time through spatial gestures and voice commands. All modifications are dynamically fed back to the dynamic requirement perception module to trigger the re-optimization process.

3. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The dynamic demand perception module outputs a demand parameter vector, which includes a demand identifier field, a space type field, a functional attribute field, an environmental requirement field, a cultural characteristic field, and a constraint field. The demand identifier field uniquely identifies the demand record, the space type field records all space types and their area percentages, the functional attribute field records the specific functional requirements of each space, the environmental requirement field records quantitative indicators such as lighting or ventilation, the cultural characteristic field records cultural attribute tags, and the constraint field records budget restrictions or regulatory requirements. This vector is transmitted via a message queue and serialized in JSON format.

4. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The rule generation unit of the multidimensional generation central module generates a three-dimensional spatial model based on the rule library of building design specifications. The rule library is organized according to spatial type, including spatial area range, spatial shape constraints, spatial positional relationships and spatial opening requirements. This unit determines the optimal spatial layout scheme under rule constraints through optimization algorithms and outputs a three-dimensional spatial model file in Building Information Modeling (BIM) data format, including three levels: overall building model, floor decomposition model, and spatial unit model.

5. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The image evolution unit of the multidimensional generation central module generates facade morphology schemes based on generative adversarial networks and style transfer technology. This unit extracts the volume features, color features, material features and detail features of the surrounding environment as generation constraints, and generates multiple facade morphology schemes through generative adversarial networks. The morphology generation weight is controlled to balance the user's aesthetic preferences and the harmony with the surrounding environment. The output facade morphology scheme file includes three levels: overall facade image, facade segmented image, and material annotation image.

6. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The topology graph unit of the multidimensional generation central module generates component connection graphs based on graph neural networks and topology optimization methods. This unit determines the structural system type according to the building height, span, and function, generates a structural component layout scheme based on the spatial layout model, analyzes the component connection relationships through graph neural networks, generates a component connection graph that meets the mechanical equilibrium conditions, and automatically checks the stress state of the components to ensure that the stress ratio is within a safe range. The output component connection graph file includes three core parts: a node list, an edge list, and an attribute mapping.

7. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The physical environment simulator of the cross-dimensional verification engine module performs lighting simulation, ventilation simulation, and thermal simulation. The lighting simulation uses a ray tracing algorithm to calculate the natural daylight coefficient, taking into account the window position, size, and shading effects. The ventilation simulation uses computational fluid dynamics to simulate airflow and calculate ventilation volume and air exchange rate. The thermal simulation calculates energy consumption indicators and thermal comfort based on a heat transfer model. The output physical environment verification report includes three parts: lighting verification results, ventilation verification results, and thermal verification results.

8. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The stress field analyzer of the cross-dimensional verification engine module performs static analysis, dynamic analysis and stability analysis; the static analysis calculates the internal forces and deformations of the component based on the finite element method and evaluates the ultimate limit state of bearing capacity. Dynamic analysis includes modal analysis, seismic response analysis, and wind-induced vibration response analysis; Stability analysis assesses the buckling characteristics of components and the second-order effects of the structure; The output structural safety verification report includes three parts: static analysis results, dynamic analysis results, and stability analysis results.

9. The computer-aided design-based architectural design system as described in claim 1, characterized in that, The autonomous optimization feedback module performs problem diagnosis, parameter adjustment, and scheme regeneration based on multi-dimensional compatibility indicators; in the problem diagnosis stage, it identifies non-compliant indicators and associates them with the generation unit, and outputs a diagnostic report. The parameter adjustment stage automatically corrects the spatial segmentation logic of the rule generation unit, adjusts the morphological generation weight of the image evolution unit, and reconstructs the node connection strategy of the topological map unit. The scheme regeneration phase drives the multi-dimensional generation central module to regenerate the scheme, forming a closed-loop iterative optimization process. The iteration ends when the improvement of the compatibility index is less than the convergence threshold.

10. A computer-aided design-based architectural design method, characterized in that, include: Step S1: Receive the user's architectural design requirements, and use the dynamic requirement perception module to perform natural language parsing, core requirement identification, spatial logic reasoning, and cultural feature extraction on the requirements to generate a requirement parameter vector; Step S2: The multidimensional generation central module receives the demand parameter vector and executes in parallel the following steps: the rule generation unit generates a three-dimensional spatial model based on building design specifications, the image evolution unit generates a facade morphology scheme based on generative adversarial network and site environment data, and the topology map unit generates a component connection map based on graph neural network. Step S3: The three-dimensional spatial model is subjected to lighting simulation, ventilation simulation and thermal analysis through the cross-dimensional verification engine module to generate a physical environment verification report. The component connection diagram is subjected to static analysis, dynamic analysis and stability analysis to generate a structural safety verification report. Multi-dimensional compatibility indicators are then generated. Step S4: The autonomous optimization feedback module diagnoses the non-compliant indicators based on the multi-dimensional compatibility indicators, adjusts the spatial segmentation logic, morphological generation weight, or node connection strategy parameters, drives the multi-dimensional generation hub module to regenerate the scheme, and performs closed-loop iteration until the compatibility indicators meet the standards. Step S5: The optimization scheme is mapped to the virtual reality environment through the immersive collaborative interface module, allowing users to interact through spatial gestures, voice commands or game controllers, and the collaborative operation log is fed back to the dynamic demand perception module. Step S6: The dynamic demand perception module updates the demand parameter vector according to the collaborative operation log and re-triggers the process that started from step S2. Step S7: Once the design scheme meets the verification requirements and is confirmed by the user, the architectural design results, including a 3D spatial model file, an elevation morphology scheme file, and a component connection diagram file, are output.

Images