Intelligent building design generation system based on multi-source heterogeneous data and AI large model

Abstract

The invention discloses an intelligent building design generation system based on multi-source heterogeneous data and an AI large model, and the system comprises a multi-source heterogeneous data collection and processing module which is responsible for automatically collecting the multi-source heterogeneous data related to a project, and carrying out the data processing; the knowledge graph module is used for carrying out structured storage and association on multi-source heterogeneous data by constructing a knowledge network covering upper planning, historical culture, architectural design standards and law and regulation contents; the AI large model fine tuning module is used for performing fine tuning on the AI large model by using data in the special architectural design field; the scheme design module is responsible for generating various design schemes, providing choices of different design styles and function layouts and optimizing the schemes; and the visual interaction and feedback module is used for providing an interaction interface with a user. The method is remarkably superior to the prior art in the aspects of design substantial content processing, multi-source heterogeneous data integration, design generation and optimization, user interaction experience and the like, and a more comprehensive, intelligent and efficient solution is provided for building design.

Description

Technical Field

[0001] The present invention relates to the technical field of architectural design, and in particular to an intelligent architectural design generation system based on multi-source heterogeneous data and an AI large model. Background Art

[0002] With the increasing complexity of architectural design projects and the diversification of customer needs, traditional architectural design methods have gradually exposed their limitations. These methods often rely on the personal experience and subjective judgment of designers, resulting in long design cycles, low efficiency, and difficulty in quickly generating design solutions that meet the requirements under complex constraints. Especially when it comes to large-scale construction projects, design teams usually need to face a lot of data processing and coordination work, including but not limited to site condition analysis, understanding of regulatory requirements, and coordination of multiple needs. This not only increases the complexity of the design, but is also prone to human errors, which in turn affects the design quality and project progress.

[0003] In recent years, with the rapid development of artificial intelligence technology, especially the application of large models, new possibilities have been brought to the field of architectural design. Through artificial intelligence technology, AI large models can automatically generate multiple design solutions based on the language description and image input provided by the user, and optimize them according to user feedback. However, the current technology still has some shortcomings in practical applications. For example, there is a lack of in-depth intelligent analysis and processing understanding of design requirements such as architectural design task books. Large models are limited to architectural design knowledge and ignore other design influencing factors, making it difficult to form a comprehensive understanding and accurate analysis of design problems. It is difficult to build an effective multi-source heterogeneous data association and integration mechanism.

[0004] Existing architectural design technologies have the following major shortcomings in the implementation of complex projects, which limit their effectiveness and efficiency in practical applications: 1. Insufficient intelligent analysis of design briefs: Current design requirements are mostly derived from language descriptions and image inputs provided by users, and there is a lack of in-depth intelligent analysis and processing of architectural design briefs. However, design briefs usually contain the main source of project requirements, and insufficient analysis and interpretation of them will result in the design plan being unable to fully meet the key needs and goals of the project.

[0005] 2. Single data dimension: The existing intelligent design system only contains architectural design knowledge and lacks multi-source heterogeneous data involved in architectural design analysis (such as geography, climate, regulations, history and culture, etc.).

[0006] 3. Insufficient association of multi-source heterogeneous data: Existing systems have difficulty effectively integrating and associating multi-source heterogeneous data (such as geography, climate, regulations, history and culture, etc.), which makes it difficult for the system to form a comprehensive understanding of design problems, thus affecting the quality and accuracy of design solutions.

[0007] 4. Lack of in-depth processing of the substantive content of the design: Existing intelligent design systems mainly focus on the generation of building appearance, lacking a deep understanding and application of core content such as building structure, functional layout and regulatory compliance, resulting in the generated design solutions may lack practicality and feasibility in actual applications.

[0008] 5. Limited design generation and optimization capabilities: Existing technologies have problems with low automation and single optimization dimension when generating and optimizing design solutions. They are unable to provide diverse design options and are unable to meet the diverse needs of complex projects.

[0009] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0010] The purpose of the present invention is to provide an intelligent building design generation system based on multi-source heterogeneous data and AI big model, which can automatically extract key information from the design task book, and conduct site analysis by introducing knowledge graph knowledge base, combining multi-source heterogeneous data, and using fine-tuned AI big model to generate multiple building design solutions that meet the requirements. The system provides a more comprehensive, intelligent and efficient building design solution.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions: The present invention provides an intelligent building design generation system based on multi-source heterogeneous data and AI big model, comprising: Multi-source heterogeneous data acquisition and processing module: responsible for automatically acquiring multi-source heterogeneous data related to the project and performing data processing to support subsequent design analysis and generation; Knowledge graph module: By building a knowledge network covering higher-level planning, historical culture, architectural design standards, and legal and regulatory content, multi-source heterogeneous data is structured and associated; AI big model fine-tuning module: Use proprietary architectural design field data to fine-tune the AI ​​big model to ensure that the design solution output by the model is professional and practical; Scheme design module: responsible for generating multiple design schemes, providing options for different design styles and functional layouts, and optimizing schemes; Visual interaction and feedback module: provides an interactive interface with the user, displays the generated design scheme through visual tools, and allows users to provide real-time feedback and adjustments.

[0012] Furthermore, the multi-source heterogeneous data acquisition and processing module specifically includes: Demand data collection submodule: automatically extract project background, objectives, functional requirements, design standards, budget and duration information from the design task book; use natural language processing technology to deeply analyze the design task book to form structured data, providing a basis for design solution generation; Site analysis data collection submodule: responsible for collecting and processing the multi-dimensional data required for project site analysis to ensure the scientificity and feasibility of the design plan; the function of this submodule is to obtain key site information from different sources and integrate this information into basic data that can be used for design generation; specific data includes project background data, technical specifications and regulatory data, and site environment analysis data; through the collection and analysis of the above data, it is ensured that the design meets the planning objectives, project needs and standard requirements, providing a basis for the design; Architectural case data collection submodule: responsible for collecting completed architectural design cases and historical project data related to the project, providing reference for new architectural design plans, including the project background, design innovations, design drawings, and functional layout of the case. The data comes from well-known architectural design websites, architectural journals, and designers' public project displays.

[0013] Furthermore, the knowledge graph module supports dynamic updates and can automatically update the knowledge graph as new data is introduced, ensuring the real-time and accuracy of the data; in addition, the module also supports knowledge accumulation, and improves the intelligence of the system through continuous learning and expansion; combined with the semantic reasoning function, the module can support the deduction and solution generation of complex design logic, ensuring that the design solution comprehensively considers multiple factors such as technical specifications, environment, and culture, and provides more scientific, practical and innovative design support.

[0014] Furthermore, the AI ​​large model fine-tuning module specifically includes: Language model fine-tuning submodule: By inputting knowledge of the construction industry, the language AI model is fine-tuned so that it can accurately understand the requirements of the design brief and generate text descriptions and design plans that meet the requirements; Image generation large model fine-tuning submodule: By inputting the architectural design plan, including the general plan, the single functional space layout plan and the elevation rendering, the image is fine-tuned to generate the AI ​​large model to ensure that it can generate drawings that meet the architectural design standards and aesthetic requirements.

[0015] Furthermore, the scheme design module specifically includes: Scheme generation submodule: Generates multiple design schemes based on the design brief and site analysis data, including overall plan, detailed functional layout plan and elevation; Solution optimization submodule: Use optimization algorithms to perform multi-dimensional optimization on the generated design solutions and select the best solution.

[0016] Furthermore, the visual interaction and feedback module specifically includes: Visualization submodule: provides interactive visualization, where users can view the details of the design by dragging, dropping, and zooming; User feedback and adjustment submodule: allows users to adjust the design plan according to their own needs; secondly, supports users to modify design instructions, and the system calls the solution optimization submodule to perform real-time updates and optimizations based on user feedback.

[0017] By adopting the above technical solution, the present invention has the following beneficial effects: The present invention is significantly superior to existing technologies in terms of design content processing, multi-source heterogeneous data integration, design generation and optimization, and user interaction experience, providing a more comprehensive, intelligent and efficient solution for architectural design. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A structural diagram of an intelligent building design generation system based on multi-source heterogeneous data and AI big model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0022] Combination Figure 1As shown, this embodiment provides an intelligent building design generation system based on multi-source heterogeneous data and AI big models. The system is designed for the complexity and diversity of architectural design projects, and includes a multi-source heterogeneous data acquisition and processing module, a knowledge graph module, an AI big model fine-tuning module, a scheme design module, and a visual interaction and feedback module. The overall purpose is to provide a comprehensive and integrated architectural design solution. Through these interrelated modules, the system can efficiently generate architectural design solutions that meet project requirements and provide real-time optimization and adjustment functions during the design process. The specific functions and contents of each module are as follows: Multi-source heterogeneous data acquisition and processing module: responsible for automatically acquiring multi-source heterogeneous data related to the project and performing data processing to support subsequent design analysis and generation. This module is divided into three sub-modules: 1) Demand data collection submodule: Functional description: This submodule automatically extracts information such as project background, objectives, functional requirements, design standards, budget and duration from the design brief, and uses NLP technology for in-depth analysis to convert unstructured brief content into structured data, providing a basis for design solution generation.

[0023] Specific implementation steps: (1) Text processing of the task book: The system first inputs the text of the design brief into the NLP model. The content of the brief usually includes project goals, functional requirements, budget constraints, construction period requirements, etc.

[0024] The NLP model used is BERT (Bidirectional Encoder Representations from Transformers), which can accurately extract key information from the context. Through BERT's text classification function, the paragraphs in the task book are annotated by category (such as project goals, functional requirements, budget, etc.).

[0025] (2) Deep analysis and information extraction: After training, the BERT model can classify and extract different types of information. For example, for the following task book snippet: “Project goal: to build a modern office building that meets green building standards, including office areas, meeting rooms, parking lots and dining areas.

[0026] Functional requirements: The total building area is no less than 5,000 square meters, and must meet the office needs of 300 employees. The budget does not exceed 10 million yuan, and the construction period is 12 months. After parsing by the BERT model, the extracted structured information is: Project goals: Green building standards, modern office building; Functional requirements: building area ≥5000 square meters, accommodating ≥300 people; Budget: ≤ 10 million yuan; Construction period: 12 months.

[0027] (3) Structured data generation: The NLP model converts the extracted information into structured data in JSON format, for example: { "project_goal": "Modern office building with green building standards", "building_area": ​​"≥5000 m2", "capacity": "≥300 people", "budget": "≤10 million yuan", "duration": "12 months" } These structured data will serve as the basic data input for the subsequent design solution generation module.

[0028] 2) Site analysis data collection submodule: Functional description: This submodule is responsible for obtaining multi-dimensional site data from different sources, including project background data, technical specifications and regulatory data, and site environment analysis data, and integrating these data into basic data for design generation. Specific data include project background data (data source: higher-level planning documents issued by local governments), technical specifications and regulatory data (data source: design specifications and legal documents issued by local governments), site environment analysis data, such as geographical location, topography, climate, sunshine, wind direction, traffic, etc. (data source: Meteorological Bureau, Baidu, AutoNavi and other third-party data platforms, etc.). Through the collection and analysis of the above data, it can be ensured that the design meets the planning goals, project needs and standard requirements, providing a basis for the design.

[0029] Specific implementation steps: (1) Collection of multi-source site data The system uses GIS (Geographic Information System) to obtain the site's geographic data, including topography, landforms, climate, transportation, and natural environment information. Data sources include satellite remote sensing data, government open data, and site survey reports.

[0030] Taking terrain analysis as an example, the system extracts terrain features through remote sensing data, uses convolutional neural networks (CNN) to analyze images, and identifies the slope, aspect, and surface type of the terrain.

[0031] (2) Processing of site environmental data: The system uses the geographic data analysis module to convert the acquired terrain data into numerical data that can be used for design generation. For example, slope information is obtained through the calculation formula: ; in, represents the elevation difference; d represents the horizontal distance.

[0032] In areas where the terrain slope is greater than 15 degrees, the system will automatically mark it as a "steep slope area" and recommend adding slope protection facilities when the design is generated.

[0033] Integration of site regulatory data: The system uses the knowledge graph to analyze the correlation between technical specifications and regulatory data. For example, when the regulations stipulate that the building height shall not exceed 30 meters, the system will automatically apply this constraint when generating the design plan.

[0034] Taking the planning regulations of a certain project as an example, the regulations require that the building coverage ratio shall not exceed 50%. The system will calculate the maximum limit of the building area through the following formula: Max_Area=Total_Land_Area×0.5 Among them, Total_Land_Area is the total land area, and Max_Area is the maximum allowable building area.

[0035] 3) Building case data collection submodule Functional description: This submodule is responsible for collecting completed architectural design cases and historical project data related to the project, providing reference for new architectural design plans, including project background, design innovations, design drawings, functional layout, etc. The data comes from well-known architectural design websites, architectural journals, and designers' public project displays.

[0036] Specific implementation steps: (1) Collection of case data: The system obtains completed building case data from well-known architectural design websites (such as ArchDaily, Dezeen), architectural journals, and designers' public projects.

[0037] For example, the cases downloaded by the system from "ArchDaily" include different types of architectural design plans such as office buildings, residential communities, and commercial centers.

[0038] (2) Analysis and integration of case data: The system uses NLP technology to analyze the text description of the case and extract the project background and design innovation of the case. For example, for the description of "the project aims to promote communication between employees through an open space layout", the system marks it as "open space design".

[0039] The system groups similar cases through cluster analysis algorithms to form reference clusters for different types of projects. For example, office building projects with similar functional layouts and design styles are grouped into the same category to facilitate reference when designing new projects.

[0040] (3) Matching of reference cases: The system matches the most relevant cases from the case database based on the requirements of the current design task and site characteristics.

[0041] Through the above specific embodiments, the system can realize the automatic collection and processing of multi-source heterogeneous data in the three submodules of demand data collection, site analysis, and building case data analysis. The functions and algorithm applications of each submodule not only provide data support for the generation of design solutions, but also ensure the scientificity, feasibility and innovation of the design.

[0042] The knowledge graph module structures and associates multi-source heterogeneous data by building a knowledge network covering higher-level planning, historical culture, architectural design standards, laws and regulations, etc. The module supports dynamic updates and can automatically update the knowledge graph with the introduction of new data to ensure the real-time and accuracy of the data. In addition, the module also supports knowledge accumulation, and improves the intelligence of the system through continuous learning and expansion. Combined with the semantic reasoning function, the module can support the deduction of complex design logic and solution generation, ensuring that the design solution comprehensively considers multiple factors such as technical specifications, environment, and culture, and provides more scientific, practical and innovative design support.

[0043] In this embodiment, the implementation process of dynamic update in the knowledge graph module is as follows: The system acquires new data through a multi-source heterogeneous data collection module, including new regulations, construction cases, user feedback and other data, and automatically introduces them into the knowledge graph.

[0044] (1) Data update steps: Data import: Enter new regulatory texts or new design cases into the system.

[0045] Data parsing: The system uses NLP models to parse the content and structure of new data, for example, to parse the constraints of regulatory texts or the design features of new cases.

[0046] Graph node update: The system converts the parsed information into new nodes or edges in the knowledge graph and updates the attribute values ​​of existing nodes. For example, when new regulations stipulate that "the green coverage rate of residential buildings shall not be less than 30%", the system will add a new node in the knowledge graph to represent "green coverage rate limit" and establish a relationship with the "residential building design" node.

[0047] Consistency check: The system uses graph structure analysis algorithms (such as PageRank) to evaluate the relationship between new nodes and existing knowledge nodes to ensure the consistency and rationality of the data.

[0048] Data visualization: Through the graph visualization interface, users can view the latest regulatory nodes and their related information to ensure the intuitiveness of data updates.

[0049] The dynamic update algorithm is as follows: Node update algorithm: After introducing new data, the system calculates the correlation between the newly added nodes and the existing nodes, and uses cosine similarity to calculate the relationship: ; Among them, A and B are the feature vectors of two nodes; Sim (A, B) is the similarity between the two nodes.

[0050] If the similarity is higher than the set threshold (such as 0.8), the system will associate the new node with the existing node.

[0051] In this embodiment, the knowledge accumulation function in the knowledge graph module is implemented as follows: The realization process of knowledge accumulation: After each solution is generated, the system converts the generated design solutions and user feedback data into accumulative knowledge in the knowledge graph. For example, after generating "green building" design solutions many times, the system will extract common features (such as "optimization of solar panel layout") and add nodes to the knowledge graph to represent common strategies for green building design.

[0052] Knowledge expansion steps: Design solution analysis: Systematically analyze the structure, functional layout and innovative points of the completed design solution, and extract accumulative knowledge elements.

[0053] Feedback data analysis: The system analyzes user feedback data and adds user preferences and needs as new attributes to the knowledge graph.

[0054] Knowledge node expansion: The system establishes new knowledge nodes and relationship edges based on the commonalities and feedback of design solutions. For example, after repeatedly identifying that "office buildings need to have good lighting design", the system establishes a strategy node for "lighting design" in the knowledge graph.

[0055] Knowledge accumulation algorithm: New knowledge extraction and expansion algorithm: (1) The system groups the features in the design solutions through clustering analysis algorithms (such as K-Means), extracts common design strategies, and forms new knowledge nodes.

[0056] (2) Cluster calculation formula: ; Among them, J is the clustering objective function, which represents the closeness between nodes; represents the jth node in the i-th category; represents the center point of the i-th class.

[0057] Based on the clustering results, the system expands new knowledge nodes into the knowledge graph to improve the intelligence of the system.

[0058] In this embodiment, the semantic reasoning function implementation process in the knowledge graph module is as follows: The implementation process of semantic reasoning: The system uses the semantic reasoning ability of the knowledge graph to support the deduction of complex design logic in the process of generating design solutions. For example, when the system generates a design solution for the "Cultural Center Building", it needs to consider the cultural background, architectural style, functional layout and technical specifications at the same time.

[0059] Semantic reasoning steps: (1) Initial condition setting: The system sets initial conditions such as cultural background (such as Chinese style), functional requirements (such as exhibition space), and technical specifications (such as fire protection standards) based on the requirements of the design brief.

[0060] (2) Application of inference rules: The system deduces the design logic based on the inference rules in the knowledge graph. For example, when the system recognizes that a "cultural center building" must have "traditional elements" and is located in a "historical district", the inference rules will give priority to the use of traditional roof design and Chinese courtyard layout.

[0061] (3) Scheme generation and optimization: The system generates a preliminary design scheme based on the reasoning results and optimizes the scheme (such as functional layout and energy-saving efficiency) through a multi-objective optimization algorithm.

[0062] Semantic reasoning algorithm: (1) Rule-Based Reasoning (RBR): The system gradually deduces logical design strategies based on the inference rule set in the knowledge graph. For example, the inference rule for "Cultural Center Building" is: Rule 1: If the building type is "cultural center", open layout is preferred; Rule 2: If the cultural background is "Chinese", use Chinese architectural elements.

[0063] The system searches for the optimal solution in the knowledge graph based on the above rules and generates a design plan.

[0064] (2) Inference formula: IF (Condition1)∧(Condition2)∧… THEN (Action) Formula Description: The system reasons through a combination of logical conditions and takes corresponding design actions based on the degree to which the conditions are met.

[0065] Through the above embodiments, the knowledge graph module realizes intelligent design support in three aspects: dynamic update, knowledge accumulation and semantic reasoning. The system can not only update and expand the knowledge graph in real time, but also support complex design logic deduction through semantic reasoning, ensuring the scientificity and practicality of the generated design scheme under multiple factors such as technical specifications, environment and culture.

[0066] AI big model fine-tuning module: Use proprietary architectural design field data to fine-tune the AI ​​big model to ensure that the design solution output by the model is professional and practical. This module includes two sub-modules: Language model fine-tuning submodule: By inputting construction industry knowledge, the language AI model is fine-tuned so that it can accurately understand the requirements of the design brief and generate text descriptions and design plans that meet the requirements.

[0067] Image generation large model fine-tuning submodule: By inputting the architectural design plan, including the general plan, the single functional space layout plan and the elevation rendering, etc., the image generation AI large model is fine-tuned to ensure that it can generate drawings that meet the architectural design standards and aesthetic requirements, such as the general plan, detailed layout and elevation.

[0068] Preferably, the fine-tuning process of the language large model fine-tuning submodule is as follows: (1) Model selection: The system uses GPT (Generative Pre-trained Transformer) as the basic model of the language model, which has powerful text generation and comprehension capabilities.

[0069] (2) Fine-tuning steps: Enter the knowledge of the construction industry: The system inputs a large amount of construction industry knowledge text, including building specifications, design standards, technical requirements and regulatory provisions, as well as some specific design brief samples.

[0070] For example, the input sample may include text content such as: "The standard floor height of an office building shall not be less than 3.5 meters, the building area shall not be less than 5,000 square meters, and it shall comply with green building standards."

[0071] Fine-tune model parameters: The system uses the supervised fine-tuning (SFT) method to further train the pre-trained GPT model on construction industry knowledge.

[0072] Training process: By minimizing the cross entropy loss function, the model parameters are adjusted to improve the model's ability to understand and generate architectural professional language. The specific formula is: ; Among them, L represents the cross entropy loss; y i represents the actual label; Represents the model prediction probability.

[0073] Generate text description and design proposal: The fine-tuned GPT model can generate text descriptions that meet the requirements based on the input of the design brief. For example, when the input brief is "the office building must meet the green building standards", the text generated by the fine-tuned language model is: "The office building is designed in accordance with green building standards, including optimized building energy-saving systems, rainwater recycling devices and photovoltaic power generation facilities. In addition, the standard floor height is designed to be 4 meters to improve indoor lighting and ventilation performance." Practical application cases: Suppose the user enters a design brief: "Build an office building of no less than 10 floors in the central business district, which must meet green building certification standards and have a construction area of ​​no less than 8,000 square meters."

[0074] Summary of the design solution generated by the fine-tuned language large model: “The proposed office building is located in the central business district, with a total height of 12 floors and a total construction area of ​​approximately 8,500 square meters, meeting the requirements of green building certification. The design solution includes an optimized lighting system, energy-saving lighting, rainwater collection system and intelligent ventilation devices. The standard height of each floor is 3.8 meters to ensure a good sense of space and indoor air circulation.” Fine-tuning process of the image generation large model fine-tuning submodule: (1) Model selection: The system uses Stable Diffusion or DALL·E as the basic model for large-scale image generation, which has the ability to generate high-quality images.

[0075] (2) Fine-tuning steps: Enter the architectural design data: The system inputs the data of the architectural design plan into the image to generate a large model, including the general plan, the layout plan of the individual functional space and the elevation rendering.

[0076] Specific examples: The general plan shows the building's floor plan, roads and green space distribution; the single functional space layout plan shows the division of functional spaces such as office areas, conference rooms, and dining areas; the elevation rendering reflects the building's appearance and facade design.

[0077] Fine-tune model parameters: The system adopts a fine-tuning method based on image-text alignment to adjust the weight of the large model generated by the image, so that it can better meet the architectural design standards and aesthetic requirements when generating drawings.

[0078] Loss function: Use a loss function based on image feature matching, such as MSE (mean square error) or SSIM (structural similarity), to improve the quality and accuracy of the generated images.

[0079] ; Among them, L MSE represents the mean square error loss; I i Represents the real image pixel value; Represents the generated image pixel value.

[0080] Generate architectural drawings that meet standards: The fine-tuned image generation model generates high-quality architectural drawings based on the input design plan. For example, the input design plan includes "modern style office building with a total construction area of ​​8,000 square meters and a glass curtain wall design for the facade", and the renderings generated by the model are as follows: General plan: showing the relationship between the main building and the surrounding roads and green areas; Functional layout plan: showing the functional space division of different floors, such as office area, meeting room, toilet and dining area; Elevation rendering: shows the material and light and shadow effects of the glass curtain wall, which conforms to modern style and architectural aesthetic standards.

[0081] Practical application cases (1) Scenario description: The user wants to generate an office building facade that meets green building standards and has modern design elements.

[0082] (2) Description of the elevation image generated by the fine-tuned image generation model: "The office building adopts a glass curtain wall design, with sunshades added to the south facade and solar panels on the top. The overall appearance has simple and elegant lines, and the color tone is mainly cool, reflecting a modern style." Generated elevation effect: The elevation generated by the model shows the reflective effect of the glass curtain wall, the layout of the shading device, and the configuration of the solar panels on the top, which meets the green building design standards.

[0083] Through the above embodiments, the AI ​​large model fine-tuning module can achieve high-precision design support in terms of language understanding and image generation. The language large model fine-tuning submodule realizes accurate understanding and text generation of the design task book by inputting knowledge of the construction industry; the image generation large model fine-tuning submodule realizes high-quality generation of architectural drawings by inputting architectural design plans.

[0084] The fine-tuned language model and image generation model together constitute an intelligent building design model, which can generate text descriptions and image design plans based on the input of design task books and preliminary analysis data, and support multiple rounds of natural language interactions between users and the system to further optimize the design plan.

[0085] Solution design module: This module is the core functional module of the system, responsible for generating multiple design solutions, providing different design styles and functional layout options, and optimizing solutions. This module includes two sub-modules: 1) Solution generation submodule Functional description: Generation of multiple design schemes: This submodule uses the intelligent building design model to generate multiple building design schemes, including overall floor plans, detailed functional layout plans and elevations, based on the input design brief and site analysis data.

[0086] Implementation steps: (1) Input data preparation: The system receives the design brief and site analysis data. For example: The brief required: “Design a 10-story office building that meets green building standards with a total floor area of ​​8,000 square meters, including open office areas, meeting rooms and a roof garden.” Site analysis data showed: flat terrain, mild climate, and building height limit of 30 meters.

[0087] (2) Generate multiple design solutions: The system uses the intelligent building design model to generate multiple preliminary plans based on the input task book and site analysis data. Each plan includes an overall plan, a detailed functional layout plan and an elevation. For example: Option A: General plan: shows the layout of the building on the site, including parking and green areas.

[0088] Functional layout plan: The 1st to 3rd floors are open office areas, the 4th to 6th floors are conference rooms, the 7th to 10th floors are independent offices, and the roof has a garden and photovoltaic power generation equipment.

[0089] Elevation: Modern design, combining glass curtain walls and sunshades.

[0090] Option B: General Plan: The building is located close to the south side of the site to increase lighting.

[0091] Functional layout plan: Floors 1-4 are meeting rooms, floors 5-8 are open office areas, floors 9-10 are administrative office areas, and there is a garden on the roof.

[0092] Elevation: Horizontal shading devices and energy-efficient exterior wall materials are used.

[0093] (3) Solution generation strategy: The solution generation process is based on various strategies of intelligent building design models, such as: Space utilization strategy: Improve the building's utilization efficiency by maximizing the office area.

[0094] Green design strategies: Ensure the design meets green building standards and increase the area of ​​natural lighting and green roofs.

[0095] Practical application cases: (1) The user inputs the design brief and site analysis data: Design brief: "Design a 12-story commercial complex, including a shopping mall, office building and roof garden, with a total construction area of ​​no less than 10,000 square meters." Site analysis data: The terrain slope is large, and the allowed building height does not exceed 40 meters.

[0096] (2) Generate results: Plan A: The shopping mall is on the 1st to 4th floors, the office area is on the 5th to 12th floors, and there is a large garden and viewing platform on the roof.

[0097] Plan B: The shopping mall is on the 1st to 6th floors, the office area is on the 7th to 12th floors, and there is a roof garden and photovoltaic power generation equipment on the roof.

[0098] 2) Solution optimization submodule Function description: Multi-dimensional optimization design scheme: Use optimization algorithms to optimize the generated design scheme in multiple dimensions, including space utilization, energy efficiency, cost control and user experience, and select the best solution.

[0099] Implementation steps: (1) Optimize target setting: The system sets optimization goals based on design requirements and user feedback. For example: Goal 1: Maximize space utilization.

[0100] Goal 2: Improve the energy efficiency of buildings.

[0101] Objective 3: Control the overall cost of the building.

[0102] (2) Optimization algorithm selection and application: The system uses multi-objective optimization algorithms (such as genetic algorithms and particle swarm optimization) to optimize the design scheme in multiple dimensions.

[0103] Genetic algorithm optimization steps: Population initialization: The generated multiple design schemes are used as the initial population.

[0104] Fitness function calculation: Calculate the fitness of each solution based on the optimization goal. For example: ; Among them, F represents the fitness score; U represents the space utilization rate; E represents the energy saving efficiency; C represents the cost; α, β, and γ are the weights of different objectives.

[0105] Selection, crossover and mutation: Select the best solution according to fitness, perform crossover and mutation, and generate a new generation of solutions.

[0106] Iterative update: Repeat the above steps until the optimal solution is found or the iteration limit is reached.

[0107] Choose the best solution: After multiple rounds of optimization, the system selects the solution with the highest fitness as the final design solution. For example, the optimal solution after optimization is: (1) Overall plan: The green coverage rate was increased and the number of parking spaces was increased.

[0108] (2) Functional layout plan: The layout of the office area is optimized and natural lighting is enhanced.

[0109] (3) Elevation: The position of the sunshade has been adjusted to improve energy saving in summer.

[0110] Practical application cases: (1) User input optimization goal: The user hopes to optimize the solution to improve the green building score, increase the natural lighting area, and reduce the overall cost.

[0111] (2) Optimization process and results: The system uses genetic algorithms to perform multi-dimensional optimization on Plan A and Plan B based on the user's optimization goals. The optimal plan finally selected is: "The optimized design plan: the green roof area is increased, energy-efficient window materials are used, and the functional layout is adjusted to increase the daylighting area. The cost is expected to be reduced by 5% after optimization." Through the above embodiments, the scheme generation submodule and the scheme optimization submodule in the scheme design module can generate multiple architectural design schemes and select the optimal scheme through a multi-objective optimization algorithm. This module realizes the intelligence and efficiency of the system in the analysis of the design task book, the use of site analysis data, the generation and optimization of schemes.

[0112] Visual interaction and feedback module: provides an interactive interface with the user, displays the generated design scheme through visual tools, and allows users to provide real-time feedback and adjustments. This module includes two sub-modules: Visual display submodule: provides interactive visual display, and users can view the details of the design through operations such as dragging, zooming, etc.

[0113] User feedback and adjustment submodule: allows users to adjust the design plan according to their own needs, such as modifying the layout, adding functional areas, etc. Secondly, it supports users to modify design instructions, such as function type, space layout, entrances and exits, appearance design and rendering style, etc. The system will call the solution optimization submodule to update and optimize in real time based on user feedback.

[0114] Compared with the prior art, the invention has the following significant technical advantages and beneficial effects: 1. Strengthen the intelligent analysis and generation capabilities of the design task book: By enhancing the natural language processing capabilities in the demand data collection submodule, the present invention can deeply analyze the key information contained in the design task book, such as the project background, goals, functional requirements, design standards, etc. This not only improves the system's ability to understand the task book, but also ensures that the generated design solution can fully meet the key requirements and goals of the project, overcoming the problem of insufficient analysis of the task design book in the prior art.

[0115] 2. Improve the quality of the substantive content generation of architectural design: By introducing the knowledge graph module and the AI ​​large model fine-tuning module, the present invention can deeply understand the core elements of architectural design, such as building structure, functional layout, regulatory compliance, etc., making the generated design scheme more practical and feasible. Compared with the existing technology, the system is no longer limited to appearance design, but can comprehensively consider the overall function of the building and legal requirements, and provide a design scheme that is more in line with practical applications.

[0116] 3. Enhance the ability to associate and integrate multi-source heterogeneous data: The present invention realizes the effective association and integration of data from different sources and formats through the construction of multi-source heterogeneous data collection and processing modules and knowledge graph knowledge base. The system can fully understand and process multi-dimensional design data, including geography, climate, regulations, history and culture, etc., improve the accuracy of data processing and the scientific nature of design solutions, and solve the problem of insufficient data integration and association in the prior art.

[0117] 4. Improve the diversity and automation of design generation and optimization: The AI ​​large model fine-tuning module and solution design module of the present invention can automatically generate a variety of design solutions, and comprehensively optimize these solutions through multi-dimensional optimization algorithms, including space utilization, cost control, appearance design, etc. Compared with the existing technology, the system can provide rich design options to meet the diverse needs of users in complex projects, greatly improving the efficiency and quality of design generation.

[0118] 5. Improve user interaction experience and enhance design decision-making efficiency: The present invention provides intuitive 3D model display and real-time feedback functions through visual interaction and feedback modules, enabling users to participate in the design process flexibly and accurately. Users can make real-time adjustments and decisions at all stages of the design, which improves the scientific nature of the design process and user satisfaction, and solves the problem of poor interaction experience in the prior art.

[0119] In summary, the present invention is significantly superior to existing technologies in terms of design content processing, multi-source heterogeneous data integration, design generation and optimization, and user interaction experience, providing a more comprehensive, intelligent and efficient solution for architectural design.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent building design generation system based on multi-source heterogeneous data and AI big model, characterized by: include: Multi-source heterogeneous data collection and processing module, knowledge graph module, AI large model fine-tuning module, solution design module, and visual interaction and feedback module; among them, The multi-source heterogeneous data acquisition and processing module is responsible for automatically acquiring multi-source heterogeneous data related to the project and performing data processing to support subsequent design analysis and generation; and transmitting the acquired and processed data to the knowledge graph module and the AI ​​large model fine-tuning module; The knowledge graph module: by constructing a knowledge network covering higher-level planning, historical culture, architectural design standards, and legal and regulatory content, multi-source heterogeneous data is structuredly stored and associated, and then the data is transmitted to the scheme design module; The AI ​​big model fine-tuning module: uses proprietary architectural design field data to fine-tune the AI ​​big model to ensure that the design solution output by the model is professional and practical; then transmits the adjusted data to the solution design module; The scheme design module is responsible for generating multiple design schemes, providing options of different design styles and functional layouts, and optimizing the schemes; and then generating the optimized schemes and transmitting them to the visual interaction and feedback module; The visualization interaction and feedback module provides an interactive interface with the user, displays the generated design scheme through visualization tools, and allows the user to provide real-time feedback and adjustments; and feeds back the feedback and adjustment content to the scheme design module.

2. The intelligent building design generation system based on multi-source heterogeneous data and AI big model according to claim 1 is characterized in that: The multi-source heterogeneous data acquisition and processing module specifically includes: Demand data collection submodule: automatically extract project background, objectives, functional requirements, design standards, budget and duration information from the design task book; use natural language processing technology to deeply analyze the design task book to form structured data, providing a basis for design solution generation; Site analysis data collection submodule: responsible for collecting and processing the multi-dimensional data required for project site analysis to ensure the scientificity and feasibility of the design plan; the function of this submodule is to obtain key site information from different sources and integrate this information into basic data that can be used for design generation; specific data includes project background data, technical specifications and regulatory data, and site environment analysis data; through the collection and analysis of the above data, it is ensured that the design meets the planning objectives, project needs and standard requirements, providing a basis for the design; Architectural case data collection submodule: responsible for collecting completed architectural design cases and historical project data related to the project, providing reference for new architectural design plans, including the project background, design innovations, design drawings, and functional layout of the case. The data comes from well-known architectural design websites, architectural journals, and designers' public project displays.

3. The intelligent building design generation system based on multi-source heterogeneous data and AI big model according to claim 1 is characterized in that: The knowledge graph module supports dynamic updates and can automatically update the knowledge graph as new data is introduced to ensure the real-time and accuracy of the data. In addition, the module also supports knowledge accumulation and improves the intelligence of the system through continuous learning and expansion. Combined with semantic reasoning functions, this module can support the deduction of complex design logic and solution generation, ensuring that the design solution comprehensively considers multiple factors such as technical specifications, environment, and culture, and provides more scientific, practical and innovative design support.

4. The intelligent building design generation system based on multi-source heterogeneous data and AI big model according to claim 1 is characterized in that: The AI ​​large model fine-tuning module specifically includes: Language model fine-tuning submodule: By inputting knowledge of the construction industry, the language AI model is fine-tuned so that it can accurately understand the requirements of the design brief and generate text descriptions and design plans that meet the requirements; Image generation large model fine-tuning submodule: By inputting the architectural design plan, including the general plan, the single functional space layout plan and the elevation rendering, the image generation AI large model is fine-tuned to ensure that it can generate drawings that meet the architectural design standards and aesthetic requirements; The fine-tuned language model and image generation model together constitute an intelligent building design model, which can generate text descriptions and image design plans based on the input of design task books and preliminary analysis data, and support multiple rounds of natural language interactions between users and the system to further optimize the design plan.

5. The intelligent building design generation system based on multi-source heterogeneous data and AI big model according to claim 1 is characterized in that: The scheme design module specifically includes: Scheme generation submodule: Generates multiple design schemes based on the design brief and site analysis data, including overall plan, detailed functional layout plan and elevation; Solution optimization submodule: Use optimization algorithms to perform multi-dimensional optimization on the generated design solutions and select the best solution.

6. The intelligent building design generation system based on multi-source heterogeneous data and AI big model according to claim 1 is characterized in that: The visual interaction and feedback module specifically includes: Visualization submodule: provides interactive visualization, where users can view the details of the design by dragging, dropping, and zooming. User feedback and adjustment submodule: allows users to adjust the design plan according to their own needs; secondly, supports users to modify design instructions, and the system calls the solution optimization submodule to perform real-time updates and optimizations based on user feedback.

Images