An artificial intelligence-based building structure automatic design method and system

Abstract

The application relates to the technical field of data processing, and discloses a building structure automatic design method and system based on artificial intelligence. The method comprises the following steps: constructing a collaborative building information model by collecting building design task data, performing load analysis and rigidity parameter calculation, constructing a Timoshenko beam model and combining a graph network to optimize a structure arrangement prediction result, inputting a parameterized modeling program to generate a building structure model, completing multi-working-condition analysis to obtain a structure design scheme, automatically generating design documents and construction information, and establishing a parameter and analysis result corresponding relationship database. Through integrated modeling, structure analysis and design document generation, the application effectively solves the problems of intelligent collaboration and efficient output in the whole structure design process.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an automated design method and system for building structures based on artificial intelligence. Background Technology

[0002] In current building structural design practices, structural engineers typically rely on traditional CAD drawing tools, structural analysis software (such as SAP2000 and PKPM), and manual experience to design and verify building structural schemes. This process often involves extracting building component information from two-dimensional design drawings, manually building a structural model, inputting load cases for internal force analysis, and finally completing component selection and node connection design. Although BIM (Building Information Modeling) has been applied in some projects, it still suffers from disconnects in structural mechanics analysis, load calculation, stiffness control, and construction drawing generation, particularly lacking effective integration with structural dynamic characteristics, automatic component classification, parametric modeling, and construction information generation. This means that design efficiency and accuracy remain largely limited by manual experience and data incompatibility between various software platforms.

[0003] However, the core problem in existing structural design processes lies in the lack of automated modeling and analysis methods for the entire process. Specifically: First, the spatial layout, load combinations, and node connection forms of building components rely on manual input, leading to gaps and redundancies in data transmission. Furthermore, there is a lack of systematic analysis mechanisms for structural stiffness distribution and component connection responses. Second, in the processes of structural layout optimization, modal analysis, and construction drawing generation, a closed-loop system that can automatically achieve the entire process from prediction to modeling and from design to drawing has not yet been established. This results in numerous information silos between the structural model and design documents, hindering rapid iteration and efficient delivery. Especially in large steel structure projects such as shipyard frame structure design, where there are numerous components, complex nodes, and variable working conditions, existing methods struggle to meet the requirements of high efficiency, high precision, and seamless end-to-end workflow. Summary of the Invention

[0004] This application provides an AI-based automated building structure design method and system, which effectively solves the problem of intelligent collaboration and efficient output throughout the entire structural design process by integrating modeling, structural analysis, and design document generation.

[0005] Firstly, this application provides an automated building structure design method based on artificial intelligence. The method includes: collecting data from a building design task to obtain a building information model (BIM) containing collaborative information; performing load analysis and stiffness characteristic parameter calculation on the BIM, and obtaining a structural stiffness distribution scheme by determining the bending stiffness coefficient and shear stiffness ratio; constructing a lumped Timoshenko beam simplified model based on the structural stiffness distribution scheme and integrating a graph network optimization algorithm, obtaining a structural layout prediction result through component relationship analysis and mode shape matching calculation; inputting the structural layout prediction result into a parametric modeling program to generate components, obtaining a building structure model; performing multi-condition analysis and calculation on the building structure model to obtain a structural design scheme; generating design documents and construction information based on the structural design scheme, and establishing a database of correspondence between design parameters and analysis results, thus obtaining the building structure design scheme.

[0006] Secondly, this application provides an AI-based automated building structure design system, the AI-based automated building structure design system comprising: The data acquisition module is used to collect data from architectural design tasks and obtain a building information model containing collaborative information. The calculation module is used to perform load analysis and stiffness characteristic parameter calculation on the building information model, and obtain the structural stiffness distribution scheme by determining the bending stiffness coefficient and shear stiffness ratio. The construction module is used to construct a simplified model of the lumped Timoshenko beam based on the structural stiffness distribution scheme and integrate a graph network optimization algorithm to obtain the structural layout prediction results through component relationship analysis and mode shape matching calculation. The input module is used to input the predicted structural layout results into the parametric modeling program to generate components and obtain a building structure model. The analysis module is used to perform multi-condition analysis and calculation on the building structure model to obtain a structural design scheme; The generation module is used to generate design documents and construction information based on the structural design scheme, and to establish a database of the correspondence between design parameters and analysis results to obtain the building structural design scheme.

[0007] Thirdly, an AI-based automated building structure design device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the AI-based automated building structure design device to execute the aforementioned AI-based automated building structure design method.

[0008] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned artificial intelligence-based automated building structure design method.

[0009] The technical solution provided in this application involves data collection for architectural design tasks to obtain a building information model (BIM) containing collaborative information. This allows information such as building geometry, equipment layout, material parameters, and load application locations to be stored and retrieved in a unified data format, effectively solving the problems of data disconnect and redundant modeling among multiple disciplines in traditional design processes. Furthermore, by performing load analysis and stiffness characteristic parameter calculations on this BIM model, and further determining the bending stiffness coefficient and shear stiffness ratio, a stiffness response distribution dataset across the entire structural domain is constructed, improving the overall structural stress balance and local response identification capabilities. Based on this, a simplified lumped Timoshenko beam model is constructed using a structural stiffness distribution scheme, and a graph network optimization algorithm is integrated. This maps the original spatial component relationships into a graph network with physical properties. Through component relationship analysis and mode shape matching calculations, predictive deduction and adjustment optimization of the structural layout are achieved, significantly improving the rationality of the structural layout and its seismic performance. The predicted structural layout results are then input into a parametric modeling program to generate components. The program intelligently classifies components based on stress type, automatically selects standard or customized sections, and quickly generates calculable and graphically applicable 3D component models, significantly reducing modeling workload and improving consistency. Next, multi-condition analysis and calculations are performed on the generated building structure model, enabling a comprehensive assessment of structural stiffness, displacement, and component internal forces under various load combinations. Based on the response results, component dimensions and connection methods are adjusted to ensure structural safety and economy. Finally, the structural design scheme automatically generates design documents and construction information, establishing a database of correspondence between design parameters and analysis results. This completes the structured archiving and data integration of design outcomes, achieving a closed-loop process from inputting architectural design tasks to outputting a complete structural design scheme. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an embodiment of the automated building structure design method based on artificial intelligence in this application. Figure 2 This is a schematic diagram of one embodiment of the AI-based automated building structure design system in this application. Figure 3 This is a schematic block diagram of the building structure automation design equipment based on artificial intelligence in an embodiment of the present invention. Detailed Implementation

[0012] This application provides an automated design method and system for building structures based on artificial intelligence. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0013] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the automated building structure design method based on artificial intelligence in this application includes: Step S101: Collect data for the architectural design task to obtain a building information model containing collaborative information; Step S102: Perform load analysis and stiffness characteristic parameter calculation on the building information model. By determining the bending stiffness coefficient and shear stiffness ratio, obtain the structural stiffness distribution scheme. Step S103: Based on the structural stiffness distribution scheme, construct a simplified model of the lumped Timoshenko beam and integrate it with the graph network optimization algorithm. Through component relationship analysis and mode shape matching calculation, obtain the structural layout prediction results. Step S104: Input the structural layout prediction results into the parametric modeling program to generate components and obtain the building structure model; Step S105: Perform multi-condition analysis and calculation on the building structure model to obtain the structural design scheme; Step S106: Generate design documents and construction information based on the structural design scheme, and establish a database of the correspondence between design parameters and analysis results to obtain the building structural design scheme.

[0014] It is understood that the executing entity of this application can be an AI-based automated building structure design system, or it can be a terminal or a server; no specific limitation is made here. This application's embodiment uses a server as an example for illustration.

[0015] Specifically, data acquisition uses the architectural design task as the input source and employs multi-source information fusion to construct a Building Information Model (BIM) containing collaborative information. Basic parameters such as the number of floors, span, functional use, and component material types are extracted from the design task. Initial structural parameter values ​​are generated by retrieving project templates from a structural design database. Next, site survey data is used to generate a point cloud dataset using 3D laser scanning technology. The spatial coordinates (x, y, z) of each point in the point cloud are classified into main structural or non-structural parts using boundary recognition algorithms. Then, a semantic segmentation model is used to classify components such as equipment, walls, and floors in the point cloud, mapping them to corresponding component information based on classification labels. Following this, drawings from various disciplines (such as water, electricity, and HVAC plans) are imported, and equipment layout information and technical parameters are extracted using image recognition and annotation parsing algorithms, mapping spatial layout to load application points. All of the above information is encoded and integrated into a multi-dimensional BIM data structure containing fields such as spatial geometry, material properties, load location, and direction of action, forming a complete building information model containing collaborative information.

[0016] The load analysis and stiffness parameter calculation stages require first analyzing the stress state of each component in the building information model. Specifically, the model is structurally identified, extracting the node connection information and relative arrangement relationships of each component to determine whether it is a beam, column, or wall, and then matching the corresponding load type. For example, dead loads come from the structure's self-weight, live loads come from functional requirements, wind loads are graded according to ground roughness and building height, and seismic loads are set according to regional codes for response spectrum data. The above loads are combined according to their direction of action (X / Y / Z axes) and type, forming a load case matrix based on the "Code for Design of Building Structures". On this basis, stiffness analysis is performed, and for each component, based on its size, material, boundary conditions, and connection method, the component section data is used to calculate its deformation resistance in the bending (M) and shear (V) directions, thereby obtaining the bending stiffness (EI) and shear stiffness (GA) values. These stiffness values ​​are spatially distributed and mapped to a structural stiffness distribution map. This map is a three-dimensional array, with each element containing the EI and GA of a spatial element, forming a structural stiffness distribution scheme.

[0017] The structural stiffness data is input into a simplified Timoshenko beam model. Timoshenko beam theory considers shear deformation and rotational inertia, making it more suitable for deep beams or regions with abrupt stiffness changes than conventional Euler-Bernoulli beams. First, based on the stiffness distribution, the structure is discretized into multiple lumped mass and stiffness elements, forming a Timoshenko beam element mesh, with the mass and stiffness matrices of each element defined. Then, these structural elements are used as vertices of a Graph Neural Network (GNN), with the connections between every two components as edges, constructing a structural topology graph. Edge weights are set on the graph, assigning each edge the stiffness transfer capability between two components, and the attribute vector of each node is initialized. Through the message passing mechanism of the GNN, attribute information is passed between adjacent nodes, updating node features. Subsequently, modal analysis is performed to solve for the node vibration modes. Spectral clustering groups structural elements with similar or coupled frequencies into a single module, optimizing the structural layout to ensure the predicted layout most closely matches the dynamic characteristics. The output is the final spatial coordinates of each node and the component connection logic, forming the structural layout prediction result.

[0018] The parametric modeling program receives the structural layout prediction results and first classifies the components into axially dominant, bending moment dominant, and mixed-force components based on the predicted stress information. For each type of component, it queries the standard component database; if a match fails, it calculates the cross-sectional parameters based on the stress data to generate a customized cross-section. Next, it models the component geometry; for example, bending moment dominant components are modeled as I-shaped or box-shaped sections, column components as H-shaped or circular tube sections, and wall components as plate sections. The node connection method is determined based on the component type and the intersection angle to determine whether it is a rigid connection, hinged connection, or flexible transition. Finally, geometric Boolean operations are used to assemble the components and nodes as a whole, resulting in a three-dimensional representation of the building structure model.

[0019] In the multi-condition analysis phase, a load combination list is constructed, including dead loads, live loads, wind loads, seismic loads, and their combinations. For example, "G+0.7Q+0.6W" represents a combination of dead load, live load, and wind load. Internal force analysis of components is performed through static and nonlinear analysis modules to determine whether each component meets design requirements. Internal force data serves as input parameters for the section design optimization module to calculate the minimum section moment of inertia, minimum web thickness, and minimum cross-sectional area. For node connections, finite element analysis is performed based on the bending moment, shear force, and axial force experienced by the nodes to determine the connection method and plate thickness. Deformation data of each component is compared with design limits. When deformation exceeds the limit, section optimization or the addition of lateral support design processes is automatically triggered, ultimately generating an optimized structural design dataset. The design data is input into the document generation module, and the design specification document summarizes the load data, internal force analysis results, component design basis, etc.; construction drawings are exported through parametric modeling, including plan layout, node details, and component list; the material statistics table automatically summarizes the dimensions and quantities of each type of component; and the construction sequence is automatically generated based on the component connection logic. A one-to-one mapping relationship is established between all generated parameters and analysis results, and an indexed design parameter database is constructed for data traceability and updates throughout the entire life cycle.

[0020] In this embodiment, data is collected for the architectural design task to obtain a building information model (BIM) containing collaborative information. This allows information such as building geometry, equipment layout, material parameters, and load application locations to be stored and retrieved in a unified data format, effectively solving the problems of data disconnect and redundant modeling among multiple disciplines in traditional design processes. Furthermore, by performing load analysis and stiffness characteristic parameter calculations on this BIM model, and further determining the bending stiffness coefficient and shear stiffness ratio, a stiffness response distribution dataset across the entire structural domain is constructed, improving the overall structural stress balance and local response identification capabilities. Based on this, a simplified lumped Timoshenko beam model is constructed using the structural stiffness distribution scheme, and a graph network optimization algorithm is integrated. This maps the original spatial component relationships into a graph network with physical properties. Through component relationship analysis and mode shape matching calculations, predictive deduction and adjustment optimization of the structural layout are achieved, significantly improving the rationality of the structural layout and its seismic performance. The predicted structural layout results are then input into a parametric modeling program to generate components. The program intelligently classifies components based on stress type, automatically selects standard or customized sections, and quickly generates calculable and graphically applicable 3D component models, significantly reducing modeling workload and improving consistency. Next, multi-condition analysis and calculations are performed on the generated building structure model, enabling a comprehensive assessment of structural stiffness, displacement, and component internal forces under various load combinations. Based on the response results, component dimensions and connection methods are adjusted to ensure structural safety and economy. Finally, the structural design scheme automatically generates design documents and construction information, establishing a database of correspondence between design parameters and analysis results. This completes the structured archiving and data integration of design outcomes, achieving a closed-loop process from inputting architectural design tasks to outputting a complete structural design scheme.

[0021] In one specific embodiment, the process of performing step S101 may specifically include the following steps: The project requirements analysis of the architectural design task is carried out to obtain the basic parameters of the architectural design. The basic parameters of the building design are processed by on-site condition surveying. By measuring the site topography and surrounding environmental constraints, point cloud data of the building space is obtained. Feature extraction and contour recognition are performed on the point cloud data of the building space. By separating the spatial information of the main building structure and non-structural components, the geometric model of the building space is obtained. Professional data association processing is performed on the architectural space geometric model, and equipment layout drawings and technical parameter tables from various disciplines are imported to obtain multi-disciplinary collaborative data. Based on multi-disciplinary collaborative data, load identification and classification are performed. By calculating the weight of each piece of equipment and determining its position in the structure, a structural load distribution scheme is obtained. The structural load distribution scheme is integrated to obtain a building information model containing collaborative information.

[0022] Specifically, in the project requirements analysis process, the inputs are the design brief, the owner's requirements document, and the functional planning diagram. This text data is parsed using natural language processing tools to extract key structural design parameters, such as load rating, floor height, building use, fire resistance rating, and seismic fortification intensity. These parameters are structured and encoded into an attribute table and matched against reference projects already defined in the project template library to complete missing fields. For example, when the brief specifies "single-story industrial plant with a frame structure, 30 meters span, 10 meters height, and crane operation," the model automatically matches a typical frame structure layout, extracts standard beam-column spacing, rigid connection node types, etc., from existing data, and assigns initial values ​​to the current project.

[0023] On-site surveys were conducted based on basic parameters to collect 3D spatial terrain data and environmental information. The operator used laser scanning equipment to perform area scanning of the building site, generating point cloud data where each point contained spatial coordinates (x, y, z), reflection intensity i, and scanning angle θ, in standard LAS file format. Point cloud preprocessing included denoising (filtering out floating points), registration (unifying coordinate systems across different scanning surfaces), and downsampling. Then, spatial clustering algorithms, such as the density-based DBSCAN method, were introduced to initially partition the terrain changes, automatically classifying ground, elevated areas, and depressions. The output data at this point was a clearly structured surface geometric point cloud, providing a foundation for subsequent contour extraction. When performing contour recognition processing on the point cloud data, the main goal was to effectively separate the main building structure from non-structural components in the point cloud, constructing an accurate architectural spatial geometric model. The processing flow first used a surface fitting algorithm based on normal vectors to detect large planes, vertical surfaces, or continuous edges in the point cloud. Combined with typical component size ranges (e.g., column diameter not less than 300mm, beam span not less than 2m), these feature sets were labeled. A CNN convolutional network is used to classify and recognize sliced ​​point cloud images, enabling the initial segmentation of beams, columns, slabs, walls, etc. Subsequently, geometric reconstruction is performed based on information such as the center of gravity distribution and component spacing to form a complete spatial geometric model, including the three-dimensional bounding box, centerline, and cross-sectional orientation of each component.

[0024] After the spatial geometric model is constructed, drawings and technical parameter tables from various disciplines are imported to achieve multi-disciplinary collaborative data association. Input data consists of CAD or BIM format drawings (such as DWG, IFC, etc.) and equipment parameter tables (such as Excel spreadsheets). Equipment layout information in the drawings is parsed using block recognition and attribute extraction technologies (such as AutoLISP scripts + OCR extraction). For each component in the drawings, such as ducts, cable trays, and transformer boxes, its graphic coordinates, dimensions, and numbers are converted into structured entries in the database and matched with spatial coordinates in the geometric model. Information fusion is achieved through spatial similarity (e.g., component centroid distance < 0.1m), forming a multi-disciplinary collaborative information structure. When identifying and classifying loads based on the above collaborative data, information such as component mass, support method, and direction of action needs to be extracted from the technical parameters. For example, if the equipment parameter table indicates that a transformer weighs 5000kg and is supported under a main beam, the load application point and transmission path are confirmed by comparing the main beam number with the spatial geometric model. All such point loads are categorized into the live load set. For dead loads, such as the self-weight of walls and floor slabs, the volume is directly calculated by multiplying the model's cross-sectional dimensions by the material density. Wind loads and seismic loads are generated according to standards and specifications based on height, openness, and the structure's natural vibration period. All loads are uniformly converted into a structured data format, with each load item containing fields such as load value (N or kN / m²), application location (coordinates), direction of action (vector), and the number of the component acting on it.

[0025] After the load data is prepared, a structural load distribution scheme is formed through load zoning and superposition. In this process, the load path is traced according to the component connection relationships, and the nodes and beams / columns acting on each point load are deduced from the load itself. Multiple types of loads are then superimposed to form a set of structural unit load vectors. This set is ultimately mapped back to the architectural geometric model, achieving a closed-loop transformation from spatial geometry to structural loads.

[0026] In one specific embodiment, the process of performing step S102 may specifically include the following steps: Structural type identification processing is performed on the building information model, and the building structural system data is obtained by analyzing the building floor plan and the distribution characteristics of vertical components. Load sub-item calculations are performed based on building structure system data. By separating dead load, live load, wind load and seismic load and determining their direction of action, multi-type load action data are obtained. Load combination processing is performed on multi-type load data, and critical load case data are obtained by combining load effects under different working conditions according to design specifications. The bending stiffness of the structure is calculated based on the critical load case data. The bending stiffness coefficient is obtained by analyzing the resistance of the structure to bending deformation. Shear deformation analysis is performed on the building information model, and the shear stiffness ratio is obtained by calculating the shear deformation characteristics of the structure under horizontal force. Stiffness distribution optimization is performed based on bending stiffness coefficient and shear stiffness ratio. By balancing the stiffness distribution in each region and considering special treatment in areas of abrupt stiffness change, a structural stiffness distribution scheme is obtained.

[0027] Specifically, the process of structural type identification in a Building Information Model (BIM) begins with the geometric information of each component in the model as input, including its location, length, height, connection relationships, and component type. This data is arranged in a structured table. By analyzing the relative positions, orientations, repetitions, and whether rigid units are closed structures formed by beams and columns in the building plan, combined with the continuity of vertical components in the height direction and the density of connection nodes, the structural system type is determined. If all beams connecting the columns are found to form a regular arrangement on the horizontal plane, and the vertical column components are uninterrupted from the foundation to the roof, and no shear wall units appear, then it can be identified as a steel frame system. This type of identification is achieved through rule matching, that is, by sequentially verifying the structural form in the model according to a series of conformity judgment rules, and querying the database to see if it corresponds to a certain defined structural system type. After the structural system identification is completed, the load component calculation processing stage must be started immediately. At this time, the geometric dimensions, material properties, and positional relationships of each component in the building model are processed. The dead load is derived by multiplying the volume of the component by its material density, including roof steel plates and self-weight components. Live load is determined by the building's purpose, such as considering the periodic loads generated by equipment operation in industrial plants. Wind load is determined by the roof elevation, the structure's windward area, and geographical parameters. However, in this technical solution, the standard wind pressure value matched in the database is uniformly used, and the force is calculated by multiplying it with the structure's projected area. Seismic load is determined by finding the structure's period segment using an existing period correspondence table based on the building height and structural stiffness, and then obtaining the equivalent seismic force according to the standard seismic response spectrum. The key point of this process is that for each load, a specific component is identified as the object of action in the original model, and the direction of the force is clearly marked in the model's spatial coordinate system.

[0028] After all loads are independently determined, the system enters the load combination processing stage. This stage automatically generates different design load cases using a preset load combination matrix. For example, the foundation load case combines dead load and live load; the intermediate load case combines dead load and wind load; and the ultimate load case superimposes dead load, live load, wind load, and seismic force. Each combination corresponds to a specific number and action mechanism. The structural model performs stress analysis under each combination, extracting the maximum internal force response of each member under each load case. Subsequently, the program filters out the load case corresponding to the maximum response of each beam and column from all load case combinations, using it as the control load case for that member. This filtering is achieved by comparing the response values ​​under different load cases row by row, and the critical load case data is recorded in the member attribute table.

[0029] Based on these critical load case data, when calculating the structural bending stiffness, the system first extracts the length, maximum bending value, and deformation data of the corresponding components. After obtaining the deformation using static analysis tools, the system can determine the component's ability to resist bending deformation. The stronger this ability, the greater the component stiffness and the higher the structural stability. The bending stiffness results of all components are then summarized to form a stiffness distribution map of the entire building on a horizontal plane, allowing observation of any areas with significantly insufficient or excessively concentrated stiffness.

[0030] When a building experiences horizontal displacement under earthquake or wind loads, its structural members undergo shear deformation. Therefore, shear deformation analysis is necessary for the building information model (BIM). This process analyzes the horizontal displacement of vertical member nodes, combined with the member's stiffness and connection type, to determine its horizontal shear deformation capacity. The ratio between the horizontal deformation capacity and bearing capacity of each group of members is statistically recorded to describe its shear stiffness ratio. A ratio that is too high or too low is detrimental to the rational distribution of the overall structural stiffness.

[0031] After obtaining the bending and shear stiffness of all components, the overall structure needs to undergo stiffness distribution optimization. This process uses a spatial grid as the unit, adjusting the stiffness values ​​in each small region to smooth stiffness changes between adjacent regions and avoid stress concentration problems caused by abrupt changes in local stiffness. In practice, regions with the largest stiffness differences are identified first, and then the dimensions, cross-sectional forms, or material grades of the components in those regions are adjusted to maintain continuity with the surrounding areas within a reasonable range. All optimization results are synchronously updated back into the building model, forming a structural stiffness distribution scheme with high data consistency and engineering applicability.

[0032] In one specific embodiment, the process of executing step S103 may specifically include the following steps: The structural stiffness distribution scheme is simplified by converting the complex building structure into a Timoshenko beam element with distributed mass and stiffness, thus obtaining the lumped parameter model of the building structure. Based on the lumped parameter model of building structure, a graph network structure is constructed. By treating each structural node as a vertex in the graph and the component connection relationship as an edge in the graph, a structural topology graph network is obtained. The structural topology graph network is processed by assigning edge weights. By calculating the stiffness transfer relationship between adjacent nodes, the bending stiffness coefficient and shear stiffness ratio are mapped to the edge weight values ​​to obtain a weighted structural graph network. Message passing computation is performed based on a weighted structural graph network. By exchanging information between nodes, the structural feature attributes of each node are updated to obtain structural node feature data. The structural node feature data is processed by mode shape calculation. By solving the eigenvalue equation, the vibration period and mode shape of the structure are obtained, and the dynamic characteristic data of the simplified model are obtained. The component layout is optimized based on the simplified model's dynamic characteristic data. By adjusting the node positions and component connection methods, the dynamic characteristics of the simplified model are matched with the target characteristics, and the structural layout prediction results are obtained.

[0033] Specifically, the structural stiffness distribution scheme is simplified by converting the complex three-dimensional structural system into a Timoshenko beam model composed of multiple distributed mass and stiffness elements, while ensuring the fidelity of mechanical properties. This model is characterized by simultaneously considering the structural response under bending and shear deformation, unlike traditional simplified models that only consider bending. During construction, the spatial geometry, material properties, connection nodes, and spatial arrangement of each beam and column are extracted from the Building Information Model (BIM). Then, each continuous beam or column segment is divided into discrete elements according to predefined equivalence rules. Each element is bound to a stiffness value and a mass value, corresponding to the two dominant parameters in the structural stress characteristics. At this point, the entire building structure is transformed into a lumped parameter model composed of discrete component elements. Each component element has a clear physical definition and can be used for subsequent graph model construction. Based on this lumped model, constructing a graph network structure is the next key step. Each node in the graph corresponds to a component connection point, i.e., the connection node between beams and columns or columns and foundations, or between beams in the model. An edge is created between each pair of nodes with a direct physical connection; the existence of an edge indicates a force transmission path between the components. This step requires traversing the entire lumped model, identifying the connection points of each component in space, and recording the node numbers of its start and end points. The resulting topology is a directed graph where the connections between nodes fully reflect the actual connections between components, with no redundant edges. The direction of the edges corresponds to the installation direction of the components, ensuring data consistency during subsequent propagation.

[0034] Subsequently, weights are assigned to each edge in the graph. This operation aims to quantify and map stiffness information from structural mechanics into the graph structure, giving it physical interpretability. The weight of each edge is determined by the bending stiffness and shear stiffness of the components connected to adjacent nodes. Bending stiffness reflects the component's ability to resist deformation, while shear stiffness describes its resistance to deformation under horizontal forces. When calculating edge weights, the cross-sectional dimensions, material modulus, and geometric parameters of the components connected to the two endpoints are extracted, and their stiffness performance under stress is calculated. This result is used as the edge weight. The resulting weighted graph not only possesses a topological structure but also incorporates quantified parameters reflecting physical properties, laying the foundation for subsequent information propagation. Executing message passing computation based on the weighted graph network structure is key to introducing graph neural networks into the structural physics modeling process. In this stage, each node is initialized with multi-dimensional attributes such as the component type, geometric dimensions, material type, load type, and spatial location. These attributes are encoded as vectors. The graph neural network transmits feature information between adjacent nodes according to the set propagation rules, with the edge weights determining the degree and direction of information transmission. After each round of propagation, the node updates its feature vector, reflecting its position in the overall structure and the changing trend of its mechanical behavior. After multiple iterations of information propagation, the feature data of each node not only contains its own information but also incorporates the feature changes transmitted from neighboring nodes, reflecting the mechanical interactions of the overall structural layout.

[0035] After message propagation is complete, mode shape calculations are performed on the node feature data to extract the overall dynamic characteristics of the structure. This stage mainly involves extracting the variation trend of each node's eigenvalues ​​to identify the node's natural vibration period and corresponding mode shape. The vibration period of a node is determined by its stiffness and mass, while the mode shape describes the distribution of the structure's overall vibration patterns at different frequencies. This process, based on the node features in the graph, obtains the global dynamic response mode through frequency domain analysis of the entire graph.

[0036] Based on this vibration mode information, component layout optimization is performed. The optimization objective is to adjust the structural layout to make its vibration characteristics more reasonable and avoid problems such as local vibration mode concentration and abrupt changes in overall vibration modes. This step involves adjusting the position of nodes, i.e., changing the spatial arrangement of components, and optimizing the edge connection method, i.e., modifying the connection sequence or method of components, so that the energy distribution of the overall structure under each principal vibration mode is more uniform and the vibration response is more stable. The optimization algorithm uses a combination of greedy search and constraint condition verification. After each adjustment, the vibration mode characteristics of the graph are recalculated, and the output is the structural layout prediction result when the set target characteristics are met.

[0037] In one specific embodiment, the process of executing step S104 may specifically include the following steps: The structural layout prediction results are processed by component type classification. When the component is mainly subjected to axial force, it is identified as a column component; when the component is mainly subjected to bending moment, it is identified as a beam component; when the component is subjected to both large axial force and bending moment, it is identified as a beam-column composite component. Component classification data is obtained. Based on the component classification data, standard component library matching is performed. If the component's stress characteristics meet the national standard specifications, a standard section is selected. If the component's stress characteristics exceed the standard range, a customized section is generated to obtain the component's parametric description data. Geometric modeling is performed on the parametric description data of the components. When the component is a beam, an I-shaped section or a box section is generated; when the component is a column, an H-shaped section or a circular tube section is generated; and when the component is a wall, a plate section is generated to obtain a detailed model of the component. Node connection processing is performed based on the detailed component model. Rigid connections are formed when beams and columns intersect, hinged connections are formed when supports intersect with the main structure, and transition connections are formed when different structural units connect, thus obtaining node design data. The node design data and component detailed models are assembled as a whole to obtain a preliminary building structure model; Based on the preliminary building structure model, the construction details are improved. When the stress is concentrated at the node, stiffening ribs are added. When the stress at the connection is large, connection reinforcement plates are set. When the foundation is connected to the superstructure, anchoring devices are designed to obtain the building structure model.

[0038] Specifically, the structural layout prediction results serve as input, including the three-dimensional coordinates of each node in space, the start and end nodes of the components, and force vectors and moment data. Logical judgment is applied to these data to extract the axial force and bending moment components of each component on the force coordinate axis. Their magnitudes are normalized, and the ratio of axial force to bending moment is calculated. If the ratio significantly favors axial force, it is labeled as a column component; if bending moment is dominant, it is labeled as a beam component; if both are major influencing factors, it is classified as a beam-column composite component. The classification process uses a rule-based mapping method, automatically assigning component type labels to all components according to this logic and writing them into a component classification table. This table includes fields such as component number, start and end node numbers, main force type, and geometric information source, providing a classification basis for subsequent modeling steps.

[0039] The process then proceeds to the component library matching stage. Based on the component classification table, the components are compared with the national standard component library according to key parameters such as stress properties, dimensions, span, and load rating. The standard component library is a pre-organized database of national design standards, containing various specifications of H-beams, box-sections, circular tube sections, C-beams, etc., and corresponding applicable mechanical ranges and boundary conditions. The comparison logic is as follows: if the current component's span and load rating completely match the data range of a certain standard component, then that section type is directly selected as the matching result; if there are cases where the span is too large or the bending moment exceeds the standard range, then the component is marked as requiring a customized section, and enters the section generation module to generate section dimensions that meet the strength and stiffness requirements based on its maximum internal force and the component's slenderness ratio. All matching results are written into the component parametric description data table, with fields including component type, section form, dimensional parameters, material grade, and connection method constraints. After the component parameter data is determined, the process enters the geometric modeling stage, generating a 3D model based on the parameter data through a rule-driven modeling program. When the component type is a beam, an I-shaped section or a box-shaped section is generated based on the load path and usage function. The I-shaped section is used for medium spans and areas mainly subjected to bending, while the box-shaped section is used for areas with concentrated stress and requiring increased torsional stiffness. When the component is a column, an H-shaped section is generated to withstand combined bidirectional bending moments, and a circular tube section is used for areas with high torsional performance requirements and where the joint connections need a uniform force field. When the component is a wall, a planar plate with thickness parameters is generated, with the section height and thickness derived from a parameter table. The generated model includes complete 3D geometry, local normals, boundary control points, and preset connection point information to ensure direct docking with the joint components.

[0040] After generating the component models, the connection methods need to be designed at the nodes. Matching is performed according to the connection rules recorded in the component classification table. For example, if the intersection of a beam and column is a rigid connection, a rigid constraint feature is generated at the connection location in the component model. If it's a brace connecting to a main beam, a hinged connection is formed, and a single-degree-of-freedom rotation mechanism is set at the corresponding node. If there's a connection between different structural units, such as the intersection of a platform structure and the main plant frame, it's defined as a transition connection. The interface section is designed according to the cross-sectional forms of the two components, and connecting transition components such as transition plates or flexible connectors are inserted. The node design data structure includes connection type, node number, connecting component number, constraint type, structural reinforcement requirements, etc., forming a complete connection information table.

[0041] After the node data and component models are completed, the entire structure is assembled according to its spatial location and connection relationships to form a preliminary model of the overall building structure. In this stage, the node table serves as the control index, and each component model is inserted into the main model structure according to its number. Boolean operations or parametric connection modules are used to establish the connection states between components, forming a complete three-dimensional spatial skeleton model.

[0042] After obtaining the preliminary structural model, detailed structural refinement is performed, primarily involving local modeling and reinforcement of nodes and components at areas of concentrated stress. The criteria for evaluation are as follows: if a node experiences concentrated stress and has a large number of connecting components, stiffening ribs are added to that node to ensure local yielding at the connection; if the connection point is a stress path transition point, such as a beam-column junction or the connection between a roof truss and a roof beam, reinforcing plates are installed to disperse stress; if there is concentrated force transfer at the junction of the foundation component and the superstructure, anchoring devices, such as embedded parts, anchor bolts, or anchor bars, are installed in the connection area to achieve effective force transition between components. These reinforcing components are also modeled parametrically and included in component information tables and node connection tables to maintain data consistency.

[0043] In one specific embodiment, the process of executing step S105 may specifically include the following steps: The building structure model is subjected to load case definition processing. Different design cases are formed by combining dead load, live load, wind load and seismic load. When vertical and horizontal loads exist at the same time, the P-Delta effect is considered to obtain a multi-case load combination scheme. Static analysis is performed based on a multi-condition load combination scheme. Elastic analysis is performed when the structure is in normal service stage, and nonlinear analysis is performed when the structure is close to the limit state to obtain the internal force distribution data of the structure. The structural internal force distribution data is processed for component cross-section design. When the component is subjected to axial force, the cross-sectional area is optimized; when the component is subjected to bending moment, the cross-sectional moment of inertia is optimized; and when the component is subjected to shear force, the web thickness is optimized to obtain the optimized cross-section data of the component. Based on the optimized section data of the components, the node connection verification is performed. When the node is under simple stress, a standard connection is used. When the node is under complex stress, a detailed finite element analysis is performed to obtain the node connection design data. Displacement verification is performed on the node connection design data and component optimized section data. When the horizontal displacement exceeds the limit, the lateral stiffness is increased. When the vertical deflection exceeds the limit, the component section is adjusted. When the local deformation of the component is too large, stiffness reinforcement is added to obtain the structural deformation control scheme. Based on the structural deformation control scheme, the overall performance of the structure is evaluated. When the structural construction is difficult, the node design is simplified to obtain the structural design scheme.

[0044] Specifically, the system defines load cases for the building structure model, with data input from the building geometry model and pre-generated component parameter sets. Based on fields such as component material, dimensions, spatial location, and function in the model, the system automatically constructs sub-item data for dead load, live load, wind load, and seismic load. Dead load is calculated from the component's volume and material density; live load is mapped from typical values ​​retrieved from the building function standard library; wind load is segmented based on building height and wind pressure zone standards; and seismic load is obtained by matching the building period and its seismic intensity level from the response spectrum library. Various loads are applied to component nodes in the form of three-dimensional vectors, and multiple combinations are generated according to load case combination logic, such as dead load plus live load, dead load plus wind load plus seismic load, etc. For combinations with both vertical and horizontal loads, an iterative analysis mode for the P-Delta effect is also enabled, considering the additional force effect caused by structural displacement in the calculation to ensure the completeness of the structural stability analysis. The data output at this stage is the structural input load matrix under multiple load combination scenarios, which provides a basis for subsequent structural response analysis.

[0045] Next, static analysis is performed, employing different analysis models depending on the load combination. When the load is within the normal service stage of the building, elastic analysis is conducted to extract the linear response of components under load, including axial force, bending moment, shear force, and torque. When the load is close to the limit state, such as considering a major earthquake or strong wind ultimate load, nonlinear analysis is initiated, considering nonlinear responses such as material yielding, large component deformation, and connection failure. Elastic analysis uses the matrix displacement method or finite element method to discretize the structure into nodes and elements, solving for the deformation response of each node and the internal force response of each element. Nonlinear analysis introduces factors such as component stiffness reduction, plastic hinge formation, and node rotation, applying iterative corrections to the component elements until the convergence condition is met. The analysis results output as a complete internal force distribution dataset for each component under each load condition, including the location of the maximum internal force, stress value, and deformation. All data are indexed using structural element numbers and load condition numbers to support subsequent optimization.

[0046] When processing the structural internal force distribution data for cross-section design, the maximum internal force data of each member is first read, and then grouped according to its dominant force type. For members dominated by axial force, the minimum cross-sectional area is calculated based on compressive stress and stability; for members dominated by bending, the required moment of inertia is calculated; and for members dominated by shear, the web thickness is optimized based on web shear yield. This process relies on the national steel structure design standard database. By looking up tables or using analytical functions to compare the safety factors of different sections under specified internal forces, the optimal section is selected and written into the member attribute table. If the current section meets the requirements and is a standard member, the original design is maintained; if it does not meet the requirements, the section is upgraded or the section form is adjusted, thus forming a member section optimization dataset.

[0047] Based on the optimized component cross-section data, further verification of node connections is performed, matching connection schemes according to the internal force values ​​at the connections and the component types. For nodes with uniform internal force distribution and simple stress, standard connection templates, such as welded rigid connections and bolted hinged connections, are directly adopted. However, for nodes with complex internal forces, overlapping bending and shear stresses, or multiple component intersections, a detailed finite element analysis module needs to be imported to establish a local node model. The loads, boundary conditions, and material properties of the node components are input to perform local stress-strain analysis, identify weak areas, adjust the connection plate dimensions and weld lengths, and then backfill the node connection design data into the main structural model. Next, the displacement verification stage begins. With the goal of controlling structural deformation, the node displacement results of the overall structural model are read and compared with the displacement limit standards to determine whether they exceed the limits. If the horizontal lateral displacement of the structure under wind load exceeds the allowable value specified in the code, the lateral stiffness of the structure needs to be increased, for example, by adding supporting members or increasing the frame stiffness. If the mid-span deflection of the beam exceeds the limit, the beam section should be optimized to improve stiffness. If excessive local deformation is observed at certain nodes, local reinforcement members should be added, using stiffeners, double-layer members, or variable cross-sections. The output of this stage is a structural deformation control scheme table, recording the numbers of all components that need optimization or reinforcement, optimization measures, and suggestions. Based on all optimization results, the overall structural performance is evaluated. This process comprehensively scores the structural components based on the number of structural components, node complexity, and construction accessibility, and evaluates the feasibility of the structural scheme in conjunction with structural stability, stress rationality, and construction economy. When the construction difficulty score is high, it automatically judges whether the node construction is too complex. If there are multiple irregular connections of components or too many anchoring devices, it is recommended to simplify the node design, adjust the component connection logic, or unify the cross-section standard to achieve a balance between structural performance and construction efficiency. Finally, the structural design scheme data file is output, including all component numbers, cross-section types, connection methods, and control parameters.

[0048] In one specific embodiment, the process of executing step S106 may specifically include the following steps: The structural design scheme is processed by extracting design specifications, and by organizing the design basis, load data and calculation results, a structural design specification document is obtained. Based on the structural design scheme, construction drawings are generated, including plan views, section views, and detailed node views, to obtain a set of structural construction drawings. Material statistics are performed on the structural design scheme. By calculating the quantity and specifications of various components, a material list is obtained. Based on the structural construction drawings, the construction sequence is arranged, and by determining the component installation sequence and construction technical requirements, construction guidance documents are obtained. Data association processing is performed on relevant documents of the structural design scheme. By recording the mapping relationship between design parameters and calculation results, a database of the correspondence between design parameters and analysis results is obtained. The database of correspondences between design parameters and analysis results is integrated and processed to obtain building structure design schemes by integrating various design documents and construction information.

[0049] Specifically, in the design specification extraction stage, based on the component numbers, cross-sectional forms, connection methods, calculated control conditions, and nodal stress data from the structural design scheme database, the automatic document generation module is invoked to automatically fill in the design specification entries using preset templates. In the data processing flow, basic design basis items (such as load codes, steel structure design codes, seismic fortification standards, etc.) are read from the template library and inserted into the document header. Then, the component mechanical data and load case combination table are invoked to provide typical descriptions for each type of component, including the maximum stress case number, control internal force value, corresponding calculated cross-section, and structural form. For example, for a certain frame main beam, the description clearly indicates that its design basis is the combined load "dead + live + wind," the maximum bending moment occurs at the mid-span, the selected cross-section is a box-section steel section, the material grade is Q355B, and the connection type is rigid. All content is derived from the pre-analysis results and parameter tables through automatic filtering and matching, without relying on manual input, ensuring accuracy and consistency. After the design specification document is completed, the construction drawing process immediately begins. The input data for this stage includes the spatial geometric model of the components, a list of connection nodes, a detailed construction model, and a section template library. The output includes construction drawings such as plan views, elevation views, section views, and detailed node drawings. The drawing generation process first involves projecting the main structural frame model to identify the principal orientation of each component and generating a plan view in a two-dimensional coordinate system. Elevations are achieved by projecting components onto the XZ or YZ plane. When drawing sections, the geometric parameters of the component sections are extracted from the model data, and standard I-type, box-type, H-type, or circular tube section outlines are drawn according to component classification. Detailed node drawings are generated by extracting connecting parts, bolt positions, weld symbols, and reinforcing component data from the node connection table, enlarging them, and then generating partial construction drawings. After all drawings are generated, they are uniformly numbered and written into a drawing index table to achieve a one-to-one correspondence between drawings and models.

[0050] After completing the drawings, material statistics are processed for the structural design scheme. Based on the component parametric description table and node connection data, the cross-sectional specifications, length, quantity, and total weight of each type of component are statistically analyzed, and material classification details are output in batches. In the processing flow, all components are grouped by type, and the total quantity of components of the same specification is calculated by multiplying the quantity by the volume and density of the material of a single component. For connecting materials such as bolts and welding rods, the quantity, type, and connection length of the connectors are extracted from the node connection table and calculated by comparing against a standard library. The output is a material list, including steel classification (H-beams, box-sections, round pipes, etc.), connecting material types (high-strength bolts, welding rods), auxiliary materials (anchor plates, stiffening ribs), and their specifications, quantities, and standard codes, used for procurement planning and cost control. A construction sequence arrangement document is generated based on the drawings and material data, serving as the basis for subsequent construction organization. The construction sequence arrangement is based on component numbers, spatial locations, construction node coordinates, and installation height data, sorted according to the installation path from the starting point to the end point, while also considering the structural self-stability conditions to determine whether temporary supports are needed. The process uses a topological sorting algorithm to analyze the component connection diagram, identify the dependencies in the installation sequence of components, assign component partition numbers, and output a component installation batch table and lifting number. Construction technical requirements are automatically generated based on the component installation method and connection node type. For example, when a component is a double-girder box girder with welded rigid connections at the end nodes, the technical requirements automatically include items such as welding sequence, control of welding deformation, and node re-measurement. The final output is a construction guidance document, including component numbers, installation sequence, connection methods, construction techniques, and precautions.

[0051] Data association processing of various documents related to structural design schemes is a crucial step in ensuring data consistency. This process establishes a mapping relationship between structural parameters (such as cross-sectional dimensions, component numbers, and connection methods), calculation results (internal force values, displacements, and stress conditions), design drawings, and material lists through a unified numbering system. During processing, a relational database is built using the component number as the primary key, recording the corresponding design parameters, construction drawing locations, node detail numbers, and structural analysis results for each component. This allows for tracing back to the original design basis and calculation source from any drawing or component information. This database is structured and stored in the model management platform, with interface call functionality to support subsequent construction management and project archiving. By integrating all design parameter data, construction drawings, material details, and construction specifications, a unified building structural design scheme output package is formed. This package includes design specification documents, construction drawing sets, material statistics tables, construction guidance documents, and a parameter analysis database. All content is bidirectionally linked to data tables through unique component codes, ensuring information closure, a clear structure, traceability, reviewability, and executable capabilities.

[0052] The above describes the AI-based automated building structure design method in the embodiments of this application. The following describes the AI-based automated building structure design system in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the AI-based automated building structure design system in this application includes: The data acquisition module 201 is used to collect data from architectural design tasks and obtain a building information model containing collaborative information. The calculation module 202 is used to perform load analysis and stiffness characteristic parameter calculation on the building information model, and obtain the structural stiffness distribution scheme by determining the bending stiffness coefficient and shear stiffness ratio. Module 203 is used to construct a simplified model of a lumped Timoshenko beam based on the structural stiffness distribution scheme and integrate a graph network optimization algorithm to obtain the structural layout prediction results through component relationship analysis and mode shape matching calculation. Input module 204 is used to input the structural layout prediction results into a parametric modeling program to generate components and obtain a building structure model; Analysis module 205 is used to perform multi-condition analysis and calculation on the building structure model to obtain a structural design scheme; The generation module 206 is used to generate design documents and construction information based on the structural design scheme, and to establish a database of the correspondence between design parameters and analysis results to obtain the building structural design scheme.

[0053] Through the collaborative efforts of the aforementioned components, data collection for architectural design tasks yields a Building Information Model (BIM) containing collaborative information. This allows information such as building geometry, equipment layout, material parameters, and load application locations to be stored and retrieved in a unified data format, effectively resolving the issues of data disconnect and redundant modeling across multiple disciplines in traditional design processes. Furthermore, by performing load analysis and stiffness characteristic parameter calculations on this BIM model, and further determining the bending stiffness coefficient and shear stiffness ratio, a stiffness response distribution dataset across the entire structural domain is constructed, enhancing the overall structural stress balance and local response identification capabilities. Based on this, a simplified lumped Timoshenko beam model is constructed using the structural stiffness distribution scheme. Integrating a graph network optimization algorithm, the original spatial component relationships are mapped into a graph network with physical properties. Through component relationship analysis and modal shape matching calculations, predictive deduction and adjustment optimization of the structural layout are achieved, significantly improving the rationality of the structural layout and its seismic performance. The predicted structural layout results are then input into a parametric modeling program to generate components. The program intelligently classifies components based on stress type, automatically selects standard or customized sections, and quickly generates calculable and graphically applicable 3D component models, significantly reducing modeling workload and improving consistency. Next, multi-condition analysis and calculations are performed on the generated building structure model, enabling a comprehensive assessment of structural stiffness, displacement, and component internal forces under various load combinations. Based on the response results, component dimensions and connection methods are adjusted to ensure structural safety and economy. Finally, the structural design scheme automatically generates design documents and construction information, establishing a database of correspondence between design parameters and analysis results. This completes the structured archiving and data integration of design outcomes, achieving a closed-loop process from inputting architectural design tasks to outputting a complete structural design scheme.

[0054] above Figure 2 The AI-based automated building structure design system in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The AI-based automated building structure design equipment in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0055] Figure 3This is a schematic diagram of an AI-based automated building structure design device 300 provided in an embodiment of the present invention. The AI-based automated building structure design device 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the AI-based automated building structure design device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the AI-based automated building structure design device 300 to implement the steps of the aforementioned AI-based automated building structure design method.

[0056] The AI-based automated building structure design device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated structure of the AI-based automated building structure design equipment does not constitute a limitation on the AI-based automated building structure design equipment provided by this invention. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0057] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the artificial intelligence-based automated building structure design method.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an AI-based automated building structure design device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0060] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automated building structure design method based on artificial intelligence, characterized in that, The method includes: Data is collected for architectural design tasks to obtain a building information model containing collaborative information; Load analysis and stiffness characteristic parameter calculation are performed on the building information model. By determining the bending stiffness coefficient and shear stiffness ratio, the structural stiffness distribution scheme is obtained. Based on the aforementioned structural stiffness distribution scheme, a simplified lumped Timoshenko beam model is constructed and integrated with a graph network optimization algorithm. Through component relationship analysis and mode shape matching calculations, structural layout prediction results are obtained. This includes: simplifying the structural stiffness distribution scheme by representing the complex building structure as equivalent to a Timoshenko beam element with distributed mass and stiffness, thus obtaining a lumped parameter model of the building structure; constructing a graph network structure based on the lumped parameter model, using each structural node as a vertex and component connections as edges, thus obtaining a structural topology graph network; and assigning edge weights to the structural topology graph network, and then calculating... The stiffness transfer relationship between adjacent nodes is calculated, mapping the bending stiffness coefficient and shear stiffness ratio to edge weights to obtain a weighted structural graph network. Based on this weighted structural graph network, message passing computation is performed to update the structural characteristic attributes of each node through information exchange between nodes, resulting in structural node characteristic data. The structural node characteristic data is then processed for mode shape calculation, obtaining the vibration period and mode shape of the structure by solving the eigenvalue equations, thus obtaining simplified model dynamic characteristic data. Based on this simplified model dynamic characteristic data, component layout optimization is performed, adjusting node positions and component connection methods to match the simplified model's dynamic characteristics with the target characteristics, resulting in a structural layout prediction result. The predicted structural layout results are input into a parametric modeling program to generate components, resulting in a building structure model. Multi-condition analysis and calculation were performed on the building structure model to obtain the structural design scheme; Based on the structural design scheme, design documents and construction information are generated, and a database of the correspondence between design parameters and analysis results is established to obtain the building structural design scheme.

2. The automated building structure design method based on artificial intelligence according to claim 1, characterized in that, The process of collecting data for architectural design tasks to obtain a building information model containing collaborative information includes: The project requirements analysis of the architectural design task is carried out to obtain the basic parameters of the architectural design. The basic parameters of the building design are processed by on-site condition surveying, and the point cloud data of the building space is obtained by measuring the site topography and surrounding environmental constraints. Feature extraction and contour recognition are performed on the building space point cloud data. By separating the spatial information of the main building structure and non-structural components, a building space geometric model is obtained. The architectural space geometric model is processed by professional data association, and the layout drawings and technical parameter tables of various professional equipment are imported to obtain multi-professional collaborative data. Based on the multi-disciplinary collaborative data, load identification and classification are performed. By calculating the weight of each piece of equipment and determining its position in the structure, a structural load distribution scheme is obtained. The structural load distribution scheme is integrated to obtain a building information model containing collaborative information.

3. The automated building structure design method based on artificial intelligence according to claim 1, characterized in that, The process of performing load analysis and stiffness characteristic parameter calculation on the building information model, and obtaining a structural stiffness distribution scheme by determining the bending stiffness coefficient and shear stiffness ratio, includes: The building information model is subjected to structural type identification processing, and the building structural system data is obtained by analyzing the building plan layout and vertical component distribution characteristics; Based on the data of the building structure system, load sub-item calculations are performed. By separating dead load, live load, wind load and seismic load and determining their direction of action, multi-type load action data are obtained. The load combination processing is performed on the multi-type load data, and the critical load case data is obtained by combining the load effects under different working conditions according to the design specifications. Based on the critical load condition data, the structural bending stiffness is calculated and processed. By analyzing the structure's resistance to bending deformation, the bending stiffness coefficient is obtained. Shear deformation analysis is performed on the building information model, and the shear stiffness ratio is obtained by calculating the shear deformation characteristics of the structure under horizontal force. Stiffness distribution optimization is performed based on the bending stiffness coefficient and the shear stiffness ratio. By balancing the stiffness distribution in each region and considering special treatment in areas of abrupt stiffness change, a structural stiffness distribution scheme is obtained.

4. The automated building structure design method based on artificial intelligence according to claim 1, characterized in that, The step of inputting the structural layout prediction results into a parametric modeling program to generate components and obtain a building structure model includes: The structural layout prediction results are processed by component type classification. When the component is mainly subjected to axial force, it is identified as a column component. When the component is mainly subjected to bending moment, it is identified as a beam component. When the component is subjected to both large axial force and bending moment, it is identified as a beam-column composite component. Component classification data is obtained. Based on the component classification data, standard component library matching processing is performed. If the component's stress characteristics meet the national standard specifications, a standard section is selected. If the component's stress characteristics exceed the standard range, a customized section is generated to obtain the component's parametric description data. Geometric modeling is performed on the parametric description data of the component. When the component is a beam, an I-shaped section or a box section is generated; when the component is a column, an H-shaped section or a circular tube section is generated; when the component is a wall, a plate section is generated, thus obtaining a detailed model of the component. Based on the detailed model of the component, node connection processing is performed. When beams and columns intersect, a rigid connection is formed; when supports intersect with the main structure, a hinged connection is formed; and when different structural units connect, a transition connection is formed, thus obtaining node design data. The node design data and the detailed component model are assembled as a whole to obtain a preliminary building structure model; Based on the preliminary building structure model, the construction details are improved. When the stress is concentrated at the node, stiffening ribs are added. When the stress at the connection is large, connection reinforcement plates are set. When the foundation is connected to the superstructure, anchoring devices are designed to obtain the building structure model.

5. The automated building structure design method based on artificial intelligence according to claim 1, characterized in that, The process of performing multi-condition analysis and calculation on the building structure model to obtain a structural design scheme includes: The building structure model is subjected to load case definition processing. Different design cases are formed by combining dead load, live load, wind load and seismic load. When vertical and horizontal loads exist at the same time, the P-Delta effect is considered to obtain a multi-case load combination scheme. Static analysis is performed based on the aforementioned multi-condition load combination scheme. Elastic analysis is performed when the structure is in normal service stage, and nonlinear analysis is performed when the structure is close to the limit state to obtain the internal force distribution data of the structure. The structural internal force distribution data is processed for component cross-section design. When the component is subjected to axial force, the cross-sectional area is optimized; when the component is subjected to bending moment, the cross-sectional moment of inertia is optimized; and when the component is subjected to shear force, the web thickness is optimized to obtain the optimized cross-section data of the component. Based on the optimized section data of the component, the node connection verification process is carried out. When the node is under simple stress, the standard connection is adopted. When the node is under complex stress, a detailed finite element analysis is performed to obtain the node connection design data. Displacement verification is performed on the node connection design data and the component optimized section data. When the horizontal displacement exceeds the limit, the lateral stiffness is increased. When the vertical deflection exceeds the limit, the component section is adjusted. When the local deformation of the component is too large, stiffness reinforcement is increased to obtain the structural deformation control scheme. Based on the aforementioned structural deformation control scheme, the overall structural performance is evaluated. When the structural construction is difficult, the node design is simplified to obtain the structural design scheme.

6. The automated building structure design method based on artificial intelligence according to claim 1, characterized in that, The process of generating design documents and construction information based on the structural design scheme, and establishing a database of correspondences between design parameters and analysis results to obtain the building structural design scheme includes: The structural design scheme is processed by extracting design specifications, and by organizing the design basis, load data and calculation results, a structural design specification document is obtained. Based on the structural design scheme, construction drawings are generated, including plan views, sectional views, and detailed node views, to obtain a set of structural construction drawings. The material statistics of the structural design scheme are performed, and a material list is obtained by calculating the quantity and specifications of various components. Based on the aforementioned structural construction drawings, the construction sequence is arranged, and by determining the component installation sequence and construction technical requirements, a construction guidance document is obtained. Data association processing is performed on the relevant documents of the structural design scheme. By recording the mapping relationship between design parameters and calculation results, a database of the correspondence between design parameters and analysis results is obtained. The database of the correspondence between the design parameters and the analysis results is integrated and processed to obtain the building structure design scheme by integrating various design documents and construction information.

7. An automated building structure design system based on artificial intelligence, characterized in that, For implementing the AI-based automated building structure design method as described in any one of claims 1-6, the AI-based automated building structure design system comprises: The data acquisition module is used to collect data from architectural design tasks and obtain a building information model containing collaborative information. The calculation module is used to perform load analysis and stiffness characteristic parameter calculation on the building information model, and obtain the structural stiffness distribution scheme by determining the bending stiffness coefficient and shear stiffness ratio. The construction module is used to build a simplified lumped Timoshenko beam model based on the structural stiffness distribution scheme and integrate a graph network optimization algorithm. Through component relationship analysis and mode shape matching calculation, it obtains the structural layout prediction results. This includes: simplifying the structural stiffness distribution scheme by equating the complex building structure to Timoshenko beam elements with distributed mass and stiffness, thus obtaining a lumped parameter model of the building structure; constructing a graph network structure based on the lumped parameter model, by treating each structural node as a vertex and the component connections as edges, thus obtaining a structural topology graph network; and assigning edge weights to the structural topology graph network. By calculating the stiffness transfer relationship between adjacent nodes, the bending stiffness coefficient and shear stiffness ratio are mapped to edge weights, resulting in a weighted structural graph network. Based on this weighted structural graph network, message passing computation is performed to update the structural characteristic attributes of each node through information exchange between nodes, yielding structural node characteristic data. The structural node characteristic data is then processed for mode shape calculation; by solving the eigenvalue equations, the vibration period and mode shape of the structure are obtained, resulting in simplified model dynamic characteristic data. Based on this simplified model dynamic characteristic data, component layout optimization is performed; by adjusting node positions and component connection methods, the dynamic characteristics of the simplified model are matched with the target characteristics, resulting in a structural layout prediction result. The input module is used to input the predicted structural layout results into the parametric modeling program to generate components and obtain a building structure model. The analysis module is used to perform multi-condition analysis and calculation on the building structure model to obtain a structural design scheme; The generation module is used to generate design documents and construction information based on the structural design scheme, and to establish a database of the correspondence between design parameters and analysis results to obtain the building structural design scheme.

8. An automated building structure design device based on artificial intelligence, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the artificial intelligence-based automated design method for building structures as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by a processor, it causes the processor to execute the AI-based automated design method for building structures as described in any one of claims 1 to 6.

Images