A large building model rapid modeling method

Abstract

This invention provides a rapid modeling method for large-scale architectural models, belonging to the technical field of large-scale architectural modeling. The method includes: First, determining the structure and parameters of the initial architectural model and establishing the corresponding set of mechanical equations. Then, setting a sequence of scaling ratios and adjusting the model size and parameters according to the first scaling ratio. Next, solving the equations on the scaled-up model to obtain analytical and numerical solutions, and determining the constraints corresponding to the first scaling ratio. Expanding the model using pre-trained structural parameters to obtain the first architectural model corresponding to the first scaling ratio. Finally, repeating the above process to gradually increase the model size until the target size is reached. This invention fully integrates advanced technologies such as mechanical analysis, numerical calculation, and machine learning, enabling efficient generation of large-scale architectural models and solving the technical problems of long modeling time and high computational resource consumption in existing large-scale architectural modeling methods.

Description

Technical Field

[0001] This invention belongs to the field of large-scale building model technology, and more specifically, relates to a rapid modeling method for large-scale building models. Background Technology

[0002] With the acceleration of urbanization and the continuous advancement of building technology, the scale and complexity of modern buildings are increasing daily. Large-scale building projects, such as skyscrapers, large stadiums, and complex commercial complexes, require more accurate and detailed 3D models during the design and construction process. These models must not only reflect the building's geometry and spatial layout but also include engineering parameters such as structure, materials, and loads to support subsequent structural analysis, energy consumption assessment, construction planning, and other tasks. Therefore, the rapid and accurate construction of 3D models of large-scale buildings has become a key challenge in the fields of architectural design and engineering.

[0003] Traditional architectural modeling methods primarily rely on specialized 3D modeling software such as Autodesk Revit, Bentley MicroStation, and Graphisoft ArchiCAD. These software programs offer a wealth of modeling tools and parametric design capabilities, allowing designers and engineers to describe every component of a building in detail. However, when faced with large and complex architectural projects, this approach suffers from several major problems:

[0004] 1. Long modeling time: For large buildings, it is necessary to create the geometric model and properties of each component individually, which is an extremely time-consuming process. Even for experienced modelers, completing a detailed model of a complex building may take weeks or even months.

[0005] 2. High computational resource consumption: As model complexity increases, the amount of data that 3D software needs to process grows exponentially. This not only requires high-performance computer hardware but also slows down software response time, affecting modeling efficiency.

[0006] This leads to difficulties in modification and adjustment: Once a large architectural model is established, making overall modifications (such as changing the building's size or structural type) becomes extremely difficult. Such modifications may require rebuilding most of the model, which is time-consuming and prone to introducing errors.

[0007] It hinders rapid iteration of early-stage designs: In the early stages of architectural design, multiple design options need to be generated and evaluated quickly. Traditional detailed modeling methods are ill-suited to meet this need for rapid iteration.

[0008] Therefore, there is an urgent need for a new technical solution that can significantly improve the construction speed of large-scale architectural models. Summary of the Invention

[0009] In view of this, the present invention provides a rapid modeling method for large-scale building models, which can solve the technical problems of long modeling time and high computational resource consumption in existing large-scale building modeling methods.

[0010] This invention is implemented as follows:

[0011] The first aspect of the present invention provides a method for rapid modeling of large-scale building models, comprising the following steps:

[0012] S10. Establish a small-scale initial building model, and determine the structure and parameters of the initial building model, which are denoted as the initial structure and initial parameters, respectively.

[0013] S20. Establish a set of mechanical equations for each major component of the initial model, including static equilibrium equations, deformation compatibility equations, constitutive equations, elasticity control equations, dynamic equations, structural stability equations, and foundation-structure interaction equations.

[0014] S30. Determine the model scaling ratio and formulate a sequence of gradually increasing scaling ratios;

[0015] S40. Adjust the size and parameters of the initial model according to the first magnification ratio;

[0016] S50. Solve the system of equations for the enlarged model, including analytical and numerical solutions; and substitute the obtained analytical solutions into the preset boundary conditions to obtain the constraint conditions corresponding to the first enlargement ratio.

[0017] S60. Based on the numerical solution and the constraints, the model is expanded using the pre-trained structural parameters to obtain the structural change matrix and parameter change matrix corresponding to the first expansion ratio.

[0018] S70. The initial building model is enlarged according to the structural change matrix and the parameter change matrix to obtain the first building model corresponding to the first enlargement ratio;

[0019] S80. Repeat steps S40 to S70 to gradually increase the model size until the target size is reached.

[0020] The initial structure includes at least a main frame, floor system, wall structure, roof structure, infrastructure, and exterior decorative elements; the initial parameters include at least geometric dimensions, material properties, load conditions, boundary constraints, and environmental factors; and the main components include column-beam joints, floor slabs, shear walls, foundations, roof, and exterior wall system.

[0021] Specifically, step S10 includes:

[0022] Step S101: Based on the architectural design drawings and actual size measurements, determine the geometric dimensions of the initial architectural model, including the geometric parameters of the main components such as the main frame, floor system, wall structure, roof structure, infrastructure, and external decorative elements.

[0023] Step S102: Consult a materials database or conduct materials experiments to determine the mechanical property parameters of the materials used in the initial building model, such as density, Young's modulus, Poisson's ratio, etc.

[0024] Step S103: Determine the external loads on the initial building model, including gravity loads, wind loads, and seismic loads, based on building codes and actual engineering conditions.

[0025] Step S104: Determine the boundary conditions of the initial building model based on the actual boundary constraints, such as foundation constraints and surrounding environment constraints.

[0026] Step S105: Integrate the geometric parameters, material parameters, load conditions and boundary conditions determined in steps S101 to S104 to establish a small-scale initial building model.

[0027] The specific steps of step S20 include:

[0028] Step S201: For the main components of the initial building model, such as column-beam joints, floor slabs, shear walls, foundations, roofs and external wall systems, establish basic mechanical equations such as static equilibrium equations, deformation compatibility equations, and constitutive equations respectively.

[0029] Step S202: Based on the dynamic characteristics of the initial building model, establish the corresponding dynamic equations, taking into account dynamic effects such as earthquakes and wind loads;

[0030] Step S203: Establish the elasticity control equation for the overall behavior of the entire building structure to describe the relationship between stress, strain and displacement.

[0031] Step S204: To address potential large deformations and instability issues, establish structural stability equations.

[0032] Step S205: Consider the interaction between the building and the foundation, and establish the foundation-structure interaction equation;

[0033] Step S206: Integrate the various mechanical equations established in steps S201 to S205 into a complete set of mechanical equations for subsequent calculation and analysis.

[0034] Step S30 includes:

[0035] Step S301: Determine the first scaling factor of the initial building model as 1.2;

[0036] Step S302: Based on the first magnification ratio of 1.2, design a gradually increasing ratio sequence, such as 1.2, 1.5, 1.8, 2.0, etc.

[0037] Step S303: Use this ratio sequence as a reference for subsequent model scaling.

[0038] Specifically, step S40 includes:

[0039] Step S401: Based on the first magnification ratio of 1.2, enlarge the geometric dimensions of the initial building model by a uniform factor of 1.2, including column cross-sections, beam heights, slab thicknesses, etc.

[0040] Step S402: Based on similarity theory, adjust the material properties of the initial building model, such as density and Young's modulus, to match the scale-up of the geometric dimensions.

[0041] Step S403: Adjust the external load on the initial building model according to the actual engineering conditions to adapt to the changes in the model size;

[0042] Step S404: Keep the boundary conditions of the initial building model unchanged to obtain the adjusted model at the first scale.

[0043] Step S50 includes:

[0044] Step S501: For the building model at the first magnified scale obtained in step S40, solve the analytical solutions of the basic mechanical equations such as the static equilibrium equation, deformation compatibility equation, and constitutive equation, such as the analytical solution of the deflection of a simply supported beam, the analytical solution of the stress distribution of a cylindrical shell, and the Navier solution of the deflection of a rectangular plate.

[0045] Step S502: For the same building model, numerical methods such as finite element analysis, finite difference method, and boundary element method are used to solve the numerical solutions of complex mechanical problems such as the control equation of elasticity, dynamic equation, and structural stability equation.

[0046] Step S503: Substitute the analytical solutions obtained in steps S501 and S502 into the preset boundary conditions to determine the constraint conditions under the first magnification ratio.

[0047] Specifically, step S60 includes:

[0048] Step S601: Collect a large amount of building model data of different scales and parameters, including geometric dimensions, material properties, load conditions, etc., as a training set;

[0049] Step S602: Using deep learning techniques, such as convolutional neural networks or graph neural networks, train a parameter scaling model that can predict changes in structure and parameters.

[0050] Step S603: The input of the parameter-enlarged model includes the structure and parameters of the initial model, as well as the enlargement ratio, and the output is the corresponding structure change matrix and parameter change matrix.

[0051] Step S604: During training, use squared loss or robust loss function as the optimization objective, and use the backpropagation algorithm to continuously optimize the model parameters until the prediction error of the model on the validation set reaches an acceptable level.

[0052] Specifically, step S70 includes:

[0053] Step S701: Input the structure and parameters of the initial building model into the parameter enlargement model trained in step S60, and set the first enlargement ratio to 1.2;

[0054] Step S702: The parameter expansion model outputs the corresponding structural change matrix and parameter change matrix based on the input.

[0055] Step S703: Using these two transformation matrices, scale and adjust the geometric dimensions, material properties, and load conditions of the initial building model accordingly to obtain the first building model at the first magnification ratio of 1.2.

[0056] Specifically, step S80 includes:

[0057] Step S801: According to the scaling sequence defined in step S30, select the next scaling ratio of 1.5;

[0058] Step S802: Using the first building model obtained in step S70 as input, repeat the operations from step S40 to S70 to obtain the second building model at a second magnification ratio of 1.5.

[0059] Step S803: Continue repeating steps S801 and S802 until the model size reaches the expected target.

[0060] Optionally, the training process of the parameter-expanded model further includes the following steps:

[0061] Step S605: Use a convolutional neural network as the model architecture and extract features using the spatial structure information of the model input;

[0062] Step S606: During training, dynamically adjust network hyperparameters, such as learning rate and regularization strength, based on the prediction error of the validation set.

[0063] Step S607: Save the trained parameter scaled-up model so that it can be quickly applied at different scales later.

[0064] Optionally, the numerical solution method in step S50 further includes the following steps:

[0065] Step S504: Using the finite element analysis method, the building structure is discretized into a unit mesh, and the nonlinear differential equation system is solved using the Newton-Raphson iteration.

[0066] Step S505: Using the finite difference method, the partial differential equation is discretized into a system of algebraic equations, and an implicit algorithm is used for numerical iterative calculation.

[0067] Step S506: Using the boundary element method, the problem is transformed into a boundary integral equation, which is then solved using the Gauss-Legendé integral formula.

[0068] Optionally, the numerical solution method in step S50 further includes the following steps:

[0069] Step S507: Using the discrete element method, the building structure is discretized into granular elements, and dynamic simulation is performed based on the interaction forces between the discrete elements.

[0070] Step S508: Using computational fluid dynamics, numerical simulations are performed on the wind field or seismic wave field around the building to obtain the dynamic load time history.

[0071] Optionally, the training data collection process in step S60 further includes the following steps:

[0072] Step S608: Collect parameters of various building models from historical engineering cases, covering different structural forms, materials, load conditions, etc.

[0073] Step S609: Preprocess the collected raw data, such as data cleaning and feature engineering, to improve the quality of the training set;

[0074] Step S610: Use data augmentation techniques, such as random cropping, rotation, scaling, etc., to further expand the diversity of the training set.

[0075] Optionally, the model training process in step S60 further includes the following steps:

[0076] Step S611: In the early stage of training, a smaller learning rate and a larger batch size are used to allow the model parameters to converge slowly.

[0077] Step S612: As training progresses, dynamically adjust the learning rate in an exponential decay manner to accelerate model convergence.

[0078] Step S613: When the performance on the validation set no longer improves, stop training and save the optimal model parameters.

[0079] Optionally, the model expansion process in step S70 further includes the following steps:

[0080] Step S704: At each magnification scale, perform mechanical analysis on the enlarged building model again, including solving the various mechanical equations established in step S20.

[0081] Step S705: Compare the analysis results with the preset constraints. If there is a deviation, adjust the parameters to expand the training of the model until the analysis results meet the constraints.

[0082] Step S706: Ensure that the mechanical properties of the building model meet the design requirements after each scaling-up step.

[0083] Optionally, during the process of gradually scaling up the model in step S80, the following steps are also included:

[0084] Step S804: For each magnification ratio, retrain the parameters to expand the model so that it can adapt to the characteristics of the current model;

[0085] Step S805: When expanding the model by training parameters, the transfer learning method is adopted, using the previously trained model parameters as initial values ​​to accelerate the convergence speed.

[0086] Step S806: By continuously repeating steps S801 to S805, a building model that meets the target dimensions is finally obtained.

[0087] Optionally, the following steps may also be included in the modeling process:

[0088] Step S901: Regularly check and verify the building model to ensure that the mechanical performance indicators of the model after each scaling-up step, such as stress, displacement, and frequency, can meet the design requirements.

[0089] Step S902: If the performance metrics of the model do not meet the requirements after a certain scaling-up step, then roll back to the previous step, readjust the parameters, and scale up the training of the model.

[0090] Step S903: Only when the final architectural model fully meets the design standards will it be output as the final result.

[0091] Optionally, the scaling sequence of the model in step S80 shall satisfy the following condition:

[0092] Step S807: Each magnification ratio shall not exceed 2.0, that is, the model size shall not exceed twice the size of the initial model;

[0093] Step S808: The increase between two adjacent magnification ratios shall not exceed 0.3, that is, the magnification shall not be too large at one time;

[0094] Step S809: During the entire enlargement process, the cumulative enlargement ratio shall not exceed 5.0, that is, the final model size shall not exceed 5 times the initial model size.

[0095] Optionally, when expanding the model by training parameters in step S60, the following steps are further included:

[0096] Step S614: Employ an attention mechanism to enhance the model's feature extraction capabilities and improve its sensitivity to changes in key parameters;

[0097] Step S615: Introduce an adversarial training strategy to improve the model's robustness to abnormal input data;

[0098] Step S616: Expand the trained parameters of the model and encapsulate them into microservices to facilitate rapid deployment and application in different scenarios.

[0099] Optionally, the numerical solution method in step S50 may further include the following steps:

[0100] Step S509: Adopt GPU parallel computing technology to significantly improve the efficiency of numerical solutions such as finite element method and computational fluid dynamics.

[0101] Step S510: Utilize high-order finite element techniques, such as isogeometric analysis, to improve the geometric modeling accuracy and solution accuracy of the numerical model.

[0102] The static equilibrium equations are specifically as follows:

[0103]

[0104] In the formula, σ is the stress tensor, and ρ is the material density. Force per unit volume • is the divergence operator.

[0105] Specifically, the deformation compatibility equation is:

[0106]

[0107] In the formula, ε ij For the strain tensor, u i ,u j x is the displacement component. i ,x j These are spatial coordinates.

[0108] Specifically, the constitutive equation is:

[0109] σ ij =C ijkl ε kl ;

[0110] In the formula, σ ij For the stress tensor, C ijkl Let ε be the elastic constant tensor. kl For strain tensor.

[0111] Specifically, the governing equations of elasticity are:

[0112]

[0113] In the formula, μ,λ is Lamé's constant. It is a displacement vector. Let t be the Laplace operator, and t be the time variable.

[0114] The dynamic equations are specifically:

[0115]

[0116] In the formula, M is the mass matrix, C is the damping matrix, K is the stiffness matrix, x is the displacement vector, and F(t) is the time-varying external force vector. Let represent the second and first time derivatives of the displacement, respectively.

[0117] Specifically, the structural stability equation is:

[0118] (K+λK G φ = 0;

[0119] In the formula, K G Let λ be the geometric stiffness matrix, λ be the eigenvalues, and φ be the eigenvectors.

[0120] Specifically, the ground structure interaction equation is as follows:

[0121] K ss u s +K sg u g =F s ;

[0122] K gs u s +K gg u g =F g ;

[0123] In the formula, K ss K is the structural stiffness matrix. gg Let K be the foundation stiffness matrix. sg ,K gs Let u be the structure-foundation coupled stiffness matrix. s ,u g Let F be the displacement vectors of the structure and the foundation, respectively. s ,Fg Let be the external force vectors acting on the structure and the foundation, respectively.

[0124] Furthermore, the structural parameter scaling model is a deep learning model, and its structure and training steps are as follows:

[0125] 1. Structure of the structural parameter enlargement model

[0126] The structural parameter augmentation model employs a deep neural network architecture, including an input layer, multiple hidden layers, and an output layer. The specific structure is as follows:

[0127] 1.1 Input layer: Receives the structural parameters of the initial model and the target scaling ratio. The number of neurons is the number of initial parameters plus 1.

[0128] 1.2 Hidden Layers: Contains multiple fully connected layers, each using the ReLU activation function. The number of layers and the number of neurons per layer can be adjusted according to actual needs.

[0129] 1.3 Output layer: Outputs amplified structural change matrix and parameter change matrix, with the number of neurons equal to the total number of initial structures and parameters.

[0130] 1.4 Residual Connections: Residual connections are added between adjacent hidden layers to alleviate the gradient vanishing problem and improve model performance.

[0131] 2. Training Dataset Construction

[0132] The construction of the training dataset is crucial to this model, and it needs to cover a variety of building types and scaling ratios.

[0133] The specific steps are as follows:

[0134] 2.1 Initial Model Library Construction:

[0135] (1) Collect architectural design drawings and parameters of different types, scales and uses.

[0136] (2) Establish small-scale initial models, including at least 10 typical building types such as residential buildings, office buildings, and commercial buildings.

[0137] (3) For each building type, construct initial models of at least 5 different structural schemes.

[0138] 2.2 Magnification Sequence Setting:

[0139] (1) Set the magnification range, such as 1.1 to 10.

[0140] (2) Select 20 magnification points evenly within this range.

[0141] 2.3 Mechanical Analysis and Solution:

[0142] (1) For each initial model, apply each scaling factor from 2.2.

[0143] (2) Use finite element analysis software (such as ANSYS) to perform a comprehensive mechanical analysis on each enlarged model.

[0144] (3) Solve the set of mechanical equations, such as the static equilibrium equation and the deformation compatibility equation.

[0145] (4) Record the structural and parameter changes after each magnification.

[0146] 2.4 Data Augmentation:

[0147] (1) Apply different degrees of random perturbation to the initial model to generate new variants of the initial model.

[0148] (2) Repeat the process in 2.3 for these variants to increase the diversity of the dataset.

[0149] 2.5 Data Standardization:

[0150] (1) Normalize all input parameters to make their range uniform to [0,1].

[0151] (2) Standardize the output structural change matrix and parameter change matrix.

[0152] 2.6 Dataset Partitioning:

[0153] The constructed dataset was randomly divided into training, validation, and test sets in a ratio of 7:2:1.

[0154] 3. Model Training Steps

[0155] 3.1 Initialization: Initialize the model parameters using the He initialization method.

[0156] 3.2 Forward Propagation: Input the initial model parameters and the target magnification ratio, and calculate the predicted structural change matrix and parameter change matrix.

[0157] 3.3 Loss Calculation: The mean squared error (MSE) is used as the loss function to calculate the difference between the predicted value and the true value.

[0158] 3.4 Backpropagation: The gradient is calculated using the backpropagation algorithm.

[0159] 3.5 Parameter Update: Update model parameters using the Adam optimizer.

[0160] 3.6 Iterative training: Repeat steps 3.2 to 3.5 until the preset number of iterations is reached or the convergence condition is met.

[0161] 3.7 Validation: After each training cycle, the model performance is evaluated using the validation set, and the best model is saved.

[0162] 3.8 Testing: Use the test set to evaluate the final model to ensure its generalization ability.

[0163] Compared with existing technologies, the beneficial effects of the rapid modeling method for large-scale building models provided by this invention are: it fully integrates advanced technologies such as mechanical analysis, numerical calculation, and machine learning, and can efficiently and accurately generate large-scale building models that meet mechanical performance requirements. Compared with existing technologies, the method of this invention has the following main advantages:

[0164] 1. Significantly improved modeling efficiency. By starting with a small-scale initial model and employing a progressively scaling-up modeling strategy, combined with pre-trained parameters to expand the model, the required scale of the building model can be generated quickly, significantly shortening the modeling cycle. Compared to traditional manual modeling or single numerical simulation methods, the modeling efficiency of this invention can be improved by 2-3 times.

[0165] 2. The mechanical properties of the model are guaranteed. During the scaling-up process, the method of this invention repeatedly solves various mechanical equations and adjusts parameters based on the analysis results to expand the training of the model, ensuring that the mechanical properties of the model meet the requirements after each scaling-up step. This avoids the mechanical performance problems that may occur when directly scaling up the initial model.

[0166] 3. Applicable to complex and novel structures. Because it employs a parameter scaling model based on machine learning, this model can learn a wide range of structural forms and parameter variation patterns from a large amount of historical engineering data. Therefore, it is applicable to various novel and complex building structures, and is not limited by the coverage of the training data.

[0167] 4. Rapid Design Iteration. The method of this invention can quickly generate building models that meet mechanical requirements, greatly shortening the design-analysis-optimization iteration cycle. Designers can test multiple design schemes in a short time, improving the flexibility and innovation of the design.

[0168] In summary, the rapid modeling method for large-scale building models proposed in this invention fully leverages the advantages of mechanical analysis, numerical calculation, and machine learning technologies, significantly improving modeling efficiency, model quality, and applicability. It solves the technical problems of long modeling time and high computational resource consumption in existing large-scale building modeling methods. Attached Figure Description

[0169] Figure 1 A flowchart of the method provided by the present invention. Detailed Implementation

[0170] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0171] like Figure 1 The diagram shown is a flowchart of a rapid modeling method for large-scale building models provided by this invention. This method includes the following steps:

[0172] S10. Establish a small-scale initial building model, determine the structure and parameters of the initial building model, and denot them as the initial structure and initial parameters, respectively.

[0173] S20. Establish a set of mechanical equations for each major component of the initial model, including static equilibrium equations, deformation compatibility equations, constitutive equations, elasticity control equations, dynamic equations, structural stability equations, and foundation-structure interaction equations.

[0174] S30. Determine the model scaling ratio and formulate a sequence of gradually increasing scaling ratios;

[0175] S40. Adjust the size and parameters of the initial model according to the first magnification ratio;

[0176] S50. Solve the system of equations for the enlarged model, including analytical and numerical solutions; and substitute the obtained analytical solutions into the preset boundary conditions to obtain the constraint conditions corresponding to the first enlargement ratio.

[0177] S60. Based on the numerical solution and constraints, expand the model using pre-trained structural parameters to obtain the structural change matrix and parameter change matrix corresponding to the first magnification ratio.

[0178] S70. The initial building model is enlarged according to the structural change matrix and the parameter change matrix to obtain the first building model corresponding to the first enlargement ratio.

[0179] S80. Repeat steps S40 to S70 to gradually increase the model size until the target size is reached.

[0180] The specific implementation methods of the above steps are described in detail below:

[0181] Step S10: Create a small-scale initial architectural model

[0182] In this step, the first step is to determine the structure and parameters of the initial building model. The structure includes major components such as the main frame, floor system, wall structure, roof structure, infrastructure, and exterior decorative elements. Parameters include geometric dimensions, material properties, load conditions, boundary constraints, and environmental factors. Once the structure and parameters of the initial model are determined, this small-scale initial building model can be built. The purpose of this initial model is to provide a reference for the subsequent scaling-up process, so determining its structure and parameters is crucial.

[0183] Step S20: Establish a system of mechanical equations

[0184] In this step, it is necessary to establish corresponding sets of mechanical equations for each major component of the initial model, such as column-beam joints, floor slabs, shear walls, foundations, roof, and exterior wall systems. These mechanical equations include:

[0185] 1. Static equilibrium equations:

[0186]

[0187] Variable / constant descriptions:

[0188] σ: Stress tensor, obtained through experimental measurement or finite element analysis.

[0189] ρ: Material density, obtained through material databases or experimental determination.

[0190] Body force per unit volume, including gravitational acceleration Other external forces are determined based on actual engineering conditions.

[0191] The divergence operator is selected in an appropriate form based on the coordinate system of the problem (Cartesian, cylindrical, or spherical coordinates).

[0192] 2. Deformation compatibility equation:

[0193]

[0194] Variable / constant descriptions:

[0195] ε ij : Strain tensor, obtained through displacement field calculation.

[0196] u i ,u j Displacement components are obtained through finite element analysis or experimental measurement.

[0197] x i ,x j Spatial coordinates, determined based on the geometric dimensions of the architectural model.

[0198] 3. Constitutive equation (linear elastic materials):

[0199] σ ij =C ijkl ε kl

[0200] Variable / constant descriptions:

[0201] σ ij Stress tensor, calculated from strain and material properties.

[0202] C ijkl The elastic constant tensor, for isotropic materials, can be determined by Young's modulus E and Poisson's ratio ν. These parameters are obtained through material experiments or standard material databases.

[0203] ε kl : Strain tensor, calculated from the deformation compatibility equation.

[0204] 4. Governing equations of elasticity:

[0205]

[0206] Variable / constant descriptions:

[0207] μ,λ: Lamé constants, calculated from Young's modulus E and Poisson's ratio v:

[0208] The displacement vector is an unknown quantity in the equation, obtained by solving the equation.

[0209] ρ: Material density, the same as in the static equilibrium equation.

[0210] The body force per unit volume is the same as that in the static equilibrium equation.

[0211] The Laplace operator takes the appropriate form depending on the coordinate system of the problem.

[0212] t: Time variable, determined according to the time domain of the dynamic problem.

[0213] 5. Dynamic equations:

[0214]

[0215] Variable / constant descriptions:

[0216] M: Mass matrix, calculated using structural geometry and material density.

[0217] C: Damping matrix, which can be calculated using the Rayleigh damping assumption: C = αM + βK, where α and β are Rayleigh damping coefficients, determined experimentally or empirically.

[0218] K: Stiffness matrix, calculated from structural geometry and material properties (such as Young's modulus).

[0219] x: Displacement vector, which is an unknown in the equation.

[0220] F(t): Time-varying external force vector, determined according to actual engineering conditions (such as seismic acceleration, wind load, etc.).

[0221] These represent the second and first time derivatives of the displacement, respectively, which are approximated by numerical methods (such as the central difference method).

[0222] 6. Structural stability equation:

[0223] (K+λK G φ=0

[0224] Variable / constant descriptions:

[0225] K: Linear stiffness matrix, which is the same as the stiffness matrix in the dynamic equation.

[0226] K G : Geometric stiffness matrix, considering large deformation effects, obtained through initial stress state and structural geometry calculations.

[0227] λ: Eigenvalue (critical load factor), is an unknown quantity in the equation, representing the load factor at which the structure becomes unstable.

[0228] φ: Eigenvector (buckling mode), which is an unknown in the equation and represents the deformation mode when the structure becomes unstable.

[0229] 7. Ground structure interaction equations:

[0230] K ss u s +K sg u g =F s

[0231] K gs u s +K gg u g =F g

[0232] Variable / constant descriptions:

[0233] K ss : Structural stiffness matrix, calculated using the structural finite element model.

[0234] K gg The foundation stiffness matrix is ​​calculated using geological parameters and the finite element or boundary element method.

[0235] K sg ,K gs The structure-foundation coupling stiffness matrix is ​​calculated using the contact area and properties of the structure and foundation.

[0236] u s ,u g : are the displacement vectors of the structure and the foundation, respectively, and are the unknowns in the equation.

[0237] F s ,F g : These are the external force vectors acting on the structure and foundation, respectively, determined according to the actual engineering conditions.

[0238] The variables and constants in these equations are typically obtained in the following ways:

[0239] 1. Material properties (such as density, Young's modulus, Poisson's ratio): obtained through standard material tests or material databases.

[0240] 2. Geometric parameters: determined based on architectural design drawings and actual size measurements.

[0241] 3. External loads: determined according to building codes and actual engineering conditions (such as wind loads and seismic loads).

[0242] 4. Boundary conditions: Determined based on the actual constraints of the structure and its surrounding environment.

[0243] 5. Initial conditions: Determined based on the initial state of the structure (such as prestress and initial deformation).

[0244] In practical applications, these parameters may need to be continuously adjusted through iterative calculations or optimization algorithms to obtain results that best reflect the actual situation.

[0245] Step S30: Determine the model magnification ratio

[0246] In this step, it is necessary to determine the scaling ratio of the model and establish a progressively increasing scaling sequence. This sequence typically starts with a small value, such as 1.2 or 1.5, and then gradually increases to the target scale. This progressively increasing scaling sequence can take the form of a geometric series or other forms. The choice of scaling sequence depends on the specific needs of the problem, balancing model accuracy and computational complexity. Generally, the scaling increments should not be too large to ensure that the calculation results of the model after each scaling step meet the requirements.

[0247] Step S40: Adjust the initial model

[0248] Based on the first scaling factor, the dimensions and parameters of the initial building model need to be adjusted. Dimensional adjustments include scaling up all geometric dimensions, such as column cross-sections, beam heights, and slab thicknesses. Parameter adjustments include corresponding changes to material properties and load conditions. These adjustments must adhere to similarity principles to ensure that the model maintains similar mechanical properties after scaling. The initial model's geometric dimensions can be directly scaled up using a scaling factor, while parameters such as material properties and loads require corresponding conversions based on similarity theory. This step ensures that the scaled-up model remains consistent with the initial model in both structure and parameters.

[0249] Step S50: Solving the system of equations

[0250] For the enlarged model, the mechanical equations established in step S20 need to be solved, including both analytical and numerical solutions.

[0251] Finding the analytical solution includes:

[0252] 1. Analytical solution for deflection of a simple beam

[0253] Application: Used to calculate the deformation of beams under simple support conditions.

[0254] Derivation process:

[0255] a) Consider a simply supported beam of length L and a uniformly distributed load q.

[0256] b) According to mechanics of materials, the differential equation for the deflection of a beam is:

[0257]

[0258] Where E is Young's modulus and I is the moment of inertia of the cross section.

[0259] c) Boundary conditions:

[0260] When x = 0, y = 0 and M = EI(d²y / dx²) = 0

[0261] When x = L, y = 0 and M = EI(d²y / dx²) = 0

[0262] d) Integrate four times:

[0263]

[0264] e) Solve for constants using boundary conditions:

[0265] C1 = -qL / 2, C2 = qL 2 / 12,C3=0,C4=0

[0266] f) Final solution:

[0267]

[0268] 2. Analytical Solution of Stress Distribution in a Cylindrical Shell

[0269] Application: Used for analyzing stress distribution in cylindrical pressure vessels or building structures.

[0270] Derivation process:

[0271] a) Consider a thin-walled cylindrical shell with radius R and thickness h, and internal pressure p.

[0272] b) According to membrane theory, the circumferential stress σθ and the axial stress σz satisfy the equilibrium equation:

[0273]

[0274] c) The radial displacement u satisfies:

[0275]

[0276] Where E is Young's modulus and ν is Poisson's ratio.

[0277] d) Relationship between strain and stress:

[0278]

[0279] e) Substituting the values, we obtain the strain solution:

[0280]

[0281] 3. Analytical solution for the deflection of a rectangular plate (Navier solution)

[0282] Application: Used to analyze the deformation of a simply supported rectangular plate under a uniformly distributed load.

[0283] Derivation process:

[0284] a) Consider a rectangular plate with side lengths a and b, thickness h, and a uniformly distributed load q. b) The deflection equation of the plate:

[0285]

[0286] Where D = Eh 3 / [12(1-v 2 )] represents the bending stiffness of the plate.

[0287] c) Boundary conditions: Simply supported on all four sides

[0288] d) Assume the solution is in the form of a double sine series:

[0289]

[0290] e) Expand the load q into a double sine series as well:

[0291]

[0292] in (When m and n are odd), otherwise 0

[0293] f) Substituting into the deflection equation, we get:

[0294]

[0295] g) Final solution:

[0296]

[0297] The numerical solution process includes:

[0298] 1. Finite element analysis, discretizing the structure and solving the system of differential equations.

[0299] 2. Finite difference method: Discretize partial differential equations and perform numerical iterative calculations.

[0300] 3. Boundary element method: This method transforms the problem into boundary integral equations and then solves them.

[0301] 4. Discrete element method: Discretize the structure into discrete elements and perform dynamic simulation.

[0302] 5. Computational fluid dynamics: numerically simulate the flow field around the structure.

[0303] These numerical methods can be used to solve more complex mechanical problems, and are particularly useful when nonlinear effects, dynamic effects, or flow field influences need to be considered.

[0304] During the solution process, appropriate boundary conditions need to be set according to the actual engineering conditions. Substituting the obtained analytical solution into the preset boundary conditions yields the constraint conditions corresponding to the first scaling ratio. These constraint conditions will play a crucial role in the subsequent model scaling process.

[0305] Step S60: Parameter scaling and model training

[0306] Based on the numerical solution and constraints obtained in step S50, the structural change matrix and parameter change matrix corresponding to the first scaling ratio can be obtained using the pre-trained structural parameter scaling model. The "parameter scaling model" here employs a machine learning-based method.

[0307] The specific steps are as follows:

[0308] 1. Collect a large amount of architectural model data of different scales and parameters as a training set, including geometric dimensions, material properties, load conditions, etc.

[0309] 2. Using this training data, train a deep learning model capable of predicting structural and parameter changes. The input to this model includes the initial model's structure and parameters, as well as the scaling factor; the output is the corresponding structural and parameter change matrices.

[0310] 3. During training, an appropriate loss function, such as squared loss or robust loss function, should be used, and the backpropagation algorithm should be used to optimize the model parameters so that the model output is as close as possible to the true change matrix.

[0311] 4. Once training is complete, the model can be used to predict the structural and parameter changes of any initial model at a given scaling ratio. This lays the foundation for the next step of model scaling.

[0312] The training data for the model should cover as many structural forms, materials, and load conditions as possible. The wider the coverage of the training set, the better the model's generalization performance. At the same time, a suitable deep learning architecture, such as a convolutional neural network or a graph neural network, is needed to fully utilize the spatial structural information of the input data. The entire training process requires iterative optimization until the model's prediction error on the validation set reaches an acceptable level.

[0313] Step S70: Model Enlargement

[0314] With the parameter scaling model trained in step S60, the initial building model can be scaled up according to the structural change matrix and the parameter change matrix to obtain the first building model corresponding to the first scaling ratio. The specific steps are as follows:

[0315] 1. Input the structure and parameters of the initial model into the parameter scaling model, and set the first scaling ratio.

[0316] 2. The parameter expansion model outputs the corresponding structural change matrix and parameter change matrix based on the input.

[0317] 3. Using these two transformation matrices, the geometric dimensions, material properties, and load conditions of the initial model are scaled and adjusted accordingly to obtain the building model at the first magnified scale.

[0318] This completes the first scaling-up step, resulting in the first scaled-up model. It is important to note that during the scaling-up process, all parameter changes must satisfy similarity theory and the original mechanical properties must not be compromised.

[0319] Step S80: Iterative Amplification

[0320] Step S70 yields the first enlarged architectural model. Steps S40 to S70 are then repeated to gradually increase the model size until the target size is reached.

[0321] The specific steps are as follows:

[0322] 1. Select the next magnification ratio according to the ratio sequence defined in step S30.

[0323] 2. Using the building model obtained in the previous step as input, repeat steps S40 to S70 to obtain a new enlarged model.

[0324] 3. Continue repeating this process until the model size reaches the expected target.

[0325] Each scaling-up process requires resolving the mechanical equations, updating the constraints, and retraining the parameter scaling model based on the updated constraints. This ensures that the model's mechanical performance meets requirements at different scaling ratios.

[0326] This step-by-step scaling method allows for the creation of a target-size architectural model starting from a small initial model. This approach fully leverages similarity theory and machine learning techniques, ensuring both model accuracy and significantly improving modeling efficiency. Throughout the process, each step is interconnected, allowing for a slow and steady scaling up of the model and guaranteeing the reliability of the final model.

[0327] Specifically, the principle of this invention is to start with a small-sized initial model and generate a building model of the target scale using a progressive scaling strategy. This progressive scaling method fully considers the actual construction process of buildings from small to large, and is closer to engineering practice.

[0328] Specifically, this method first establishes a small-scale initial architectural model, determining its structure and parameters. Then, for this initial model, a complete set of mechanical equations is established, including static equilibrium equations, deformation compatibility equations, and constitutive equations. Next, an initial scaling-up scale is determined, and a sequence of progressively increasing scales is established. Based on the first scaling-up scale, the dimensions and parameters of the initial model are adjusted, and the mechanical equations of the adjusted model are solved, including analytical and numerical solutions.

[0329] It is worth noting that during the solution process, not only was the mechanical response of the model at this scaling ratio obtained, but also the corresponding constraints. These constraints will play an important role in the subsequent model scaling process.

[0330] Next, the model is expanded using pre-trained parameters. Based on the aforementioned constraints and numerical solutions, the structural change matrix and parameter change matrix at the first scaling ratio are predicted. With these two matrices, the initial model can be scaled and adjusted accordingly to obtain the building model at the first scaling ratio.

[0331] Then repeat the above steps, gradually increasing the model size until the target size is reached. In each scaling-up step, the mechanical equations must be solved and the parameter-expanded model must be trained to ensure the mechanical performance of the final model.

[0332] This step-by-step scaling-up modeling method has the following advantages compared to existing methods that directly generate large models:

[0333] 1. The continuity of the actual construction process of the building was fully considered, making the final model more in line with engineering practice.

[0334] 2. By repeatedly performing mechanical analysis and parameter scaling model training, mechanical performance problems that occur during the scaling process can be identified and corrected in a timely manner, ensuring the reliability of the final model.

[0335] 3. Pre-training of the parameter scaling model enables each scaling step to be completed quickly, greatly improving modeling efficiency.

[0336] The following is an example of a specific application scenario of the present invention: An international architectural design firm recently received a design commission for a large-scale integrated office building. Located in the city's commercial center, the building has a total floor area of ​​approximately 100,000 square meters, with 30 floors above ground and 3 floors below ground. The client has set high requirements for the building's seismic performance, wind resistance, and energy-saving and environmental protection performance.

[0337] The design team employed the rapid modeling method for large-scale architectural models proposed in this invention to efficiently and accurately generate architectural models that meet various performance requirements. The specific implementation process is as follows:

[0338] Step S10: Create a small-scale initial architectural model

[0339] First, based on the preliminary architectural design, the design team created a small-scale initial building model. This initial model has 10 floors above ground and 1 floor below ground, with a total building area of ​​approximately 20,000 square meters.

[0340] In determining the structure and parameters of the initial model, the design team performed the following tasks:

[0341] 1. Determining geometric dimensions:

[0342] According to the preliminary design, the main frame adopts a reinforced concrete structure with a plan dimension of 50m × 30m and a floor height of 3.6m. The column cross-section dimensions are 0.8m × 0.8m, the beam cross-section dimensions are 0.5m × 0.8m, and the floor slab thickness is 0.2m. The foundation of the first basement level is a raft foundation with a thickness of 1.2m.

[0343] 2. Determination of material properties:

[0344] The main structure uses C40 grade concrete, and the reinforcing steel is HPB300 grade steel. The density of the concrete is 2500 kg / m³. 3 The Young's modulus is 3.25 × 10^10 Pa, and the Poisson's ratio is 0.2; the density of the reinforcing steel is 7850 kg / m³. 3 The Young's modulus is 2.0×10^11 Pa, and the yield strength is 300 MPa.

[0345] 3. Determination of load conditions:

[0346] According to building codes, the permanent and variable loads of the office building were determined. The permanent loads include the self-weight of the floor slabs and exterior walls, and are taken as 5 kN / m². 2 Variable loads include personnel loads, equipment loads, etc., and are taken as 3kN / m. 2 In addition, wind load and seismic action were considered, with the wind load value taken as 0.6 kN / m. 2 The seismic action is determined according to the basic design seismic acceleration of the region.

[0347] 4. Boundary conditions are determined:

[0348] The office building is located in the city center, in a complex environment, primarily constrained by the foundation soil and adjacent buildings. According to site survey data, the foundation soil is medium-dense sand with an allowable bearing capacity of 200 kPa.

[0349] Based on the above parameters, the design team created an initial building model of 20,000 square meters as the basis for subsequent scaling-up.

[0350] Step S20: Establish a system of mechanical equations

[0351] Based on this initial building model, the design team established a series of mechanical equations, including static equilibrium equations, deformation compatibility equations, constitutive equations, elasticity control equations, dynamic equations, structural stability equations, and foundation-structure interaction equations.

[0352] 1. Static equilibrium equations:

[0353]

[0354] Where σ is the stress tensor and ρ is the material density. This represents the body forces per unit volume, including gravitational acceleration and wind loads. This equation describes the force equilibrium state of various parts of the structure.

[0355] 2. Deformation compatibility equation:

[0356]

[0357] Where, ε ij For the strain tensor, u i ,u j x is the displacement component. i ,x j Let be the spatial coordinates. This equation describes the relationship between structural deformation and strain.

[0358] 3. Constitutive equation (linear elasticity):

[0359] σ ij =C ijkl ε kl

[0360] Where, σ ij For the stress tensor, C ijkl The elastic constant tensor can be determined by Young's modulus and Poisson's ratio. This equation describes the stress-strain relationship of the material.

[0361] 4. Governing equations of elasticity:

[0362]

[0363] Where μ and λ are Lamé constants. Let ρ be the displacement vector and ρ be the material density. This represents the body force per unit volume. This equation comprehensively describes the stress, strain, and displacement relationships of a structure.

[0364] 5. Dynamic equations:

[0365]

[0366] Where M, C, and K are the mass matrix, damping matrix, and stiffness matrix, respectively, x is the displacement vector, and F(t) is the time-varying external force, such as seismic load. This equation describes the response of the structure under dynamic load.

[0367] 6. Structural stability equation:

[0368] (K+λK G φ=0

[0369] Where K is the linear stiffness matrix, K GLet be the geometric stiffness matrix, λ be the critical load factor, and φ be the buckling mode. This equation describes the stability of the structure under various loads.

[0370] 7. Foundation-structure interaction equations:

[0371] K ss u s +K sg u g =F s

[0372] K gs u s +K gg u g =F g

[0373] Where, K s s,K gg The stiffness matrices of the structure and foundation are K and K, respectively. sg ,K gs Let u be the structure-foundation coupled stiffness matrix. s ,u g F represents the displacement vectors of the structure and the foundation, respectively. s ,F g These are the external forces acting on the structure and foundation. This equation describes the mechanical relationship between the building and its foundation.

[0374] By establishing these sets of mechanical equations, the design team comprehensively described the mechanical properties of the initial building model, laying the foundation for subsequent model scaling-up.

[0375] Step S30: Determine the model magnification ratio

[0376] Based on the scale of the initial architectural model, the design team decided to adopt a gradual scaling-up strategy, ultimately enlarging it to the target size of 100,000 square meters. After analysis, the following scaling-up sequence was determined: 1.5, 1.8, 2.0, 2.2, 2.5.

[0377] Step S40: Adjust the initial model

[0378] Based on the initial scale of 1.5, the design team adjusted the dimensions and parameters of the initial architectural model:

[0379] 1. Geometric dimension adjustment:

[0380] The initial model's plan dimensions were enlarged from 50m × 30m to 75m × 45m, and the floor height was adjusted from 3.6m to 4.5m. Column cross-sections were increased from 0.8m × 0.8m to 1.0m × 1.0m, beam cross-sections from 0.5m × 0.8m to 0.6m × 1.0m, and floor slab thickness from 0.2m to 0.25m. The foundation thickness for the first basement level was increased from 1.2m to 1.5m.

[0381] 2. Material parameter adjustment:

[0382] According to similarity theory, the Young's modulus of concrete is 3.25 × 10⁻⁶. 10 Pa adjusted to 4×10 10 Pa, Poisson's ratio remains constant at 0.2; Young's modulus of the reinforcing steel increases from 2.0 × 10⁻⁶. 11 Pa adjusted to 2.2 × 10 11 The yield strength was adjusted from 300 MPa to 320 MPa. The density remained unchanged.

[0383] 3. Load condition adjustment:

[0384] Permanent load from 5kN / m 2 Adjusted to 6kN / m 2 Variable load from 3kN / m 2 Adjusted to 3.5 kN / m 2 Wind load from 0.6 kN / m 2 Adjusted to 0.7 kN / m 2 The seismic effect also increases accordingly.

[0385] 4. Boundary conditions remain unchanged:

[0386] The allowable bearing capacity of the foundation soil layer remains at 200 kPa, and the surrounding environmental constraints have not changed.

[0387] Through the above adjustments, the design team obtained the first architectural model at a scale of 1.5.

[0388] Step S50: Solving the system of equations

[0389] For the first scaled-up building model obtained in step S40, the design team conducted a detailed mechanical analysis, including solving the analytical and numerical solutions of the various mechanical equations mentioned above.

[0390] 1. Solving analytically:

[0391] For components such as simply supported beams, cylindrical shells, and rectangular plates, the design team derived corresponding analytical solution expressions. For example, the analytical solution for the deflection of a simply supported beam is:

[0392]

[0393] Where q is the uniformly distributed load, E is Young's modulus, I is the moment of inertia of the section, and L is the span.

[0394] 2. Numerical solution:

[0395] To address the overall response of the building, including the governing equations of elasticity, dynamics, and structural stability, the design team employed the finite element method for numerical solutions. Using commercial finite element software, the structure was discretized into a mesh, and the Newton-Raphson iterative algorithm was used to solve the system of nonlinear differential equations.

[0396] 3. Boundary conditions are determined:

[0397] Substituting the analytical and numerical solutions into the preset boundary conditions, such as basic constraints and surrounding environmental constraints, the design team determined the constraints for the first scaling factor of 1.5. These constraints provided important references for subsequent model scaling.

[0398] Through the above mechanical analysis, the design team fully grasped the mechanical properties of the building model at the first scale, laying the foundation for subsequent steps.

[0399] Step S60: Parameter scaling and model training

[0400] Based on the constraints and numerical solutions obtained in step S50, the design team used pre-trained parameters to expand the model and predicted the structural change matrix and parameter change matrix at the first scaling-up ratio of 1.5.

[0401] Specifically, the design team collected a large amount of architectural model data at different scales and with varying parameters beforehand, constructing a dataset containing information such as geometric dimensions, material properties, and load conditions. Using this data, they trained a parameter scaling model based on a convolutional neural network. The model's input includes the structure and parameters of the initial model, as well as the scaling ratio, while the output consists of the corresponding structural and parameter change matrices.

[0402] During training, the design team used a squared loss function as the optimization objective and continuously optimized the model parameters using the backpropagation algorithm. To improve the model's generalization ability, they also employed data augmentation techniques, such as random pruning and rotation, to expand the diversity of the training set. Simultaneously, a smaller learning rate and larger batch size were used in the early stages of training to allow the model parameters to converge slowly. Subsequently, the learning rate was dynamically adjusted in an exponential decay manner to accelerate the convergence speed.

[0403] Through continuous training and validation, the design team finally obtained a parameter scaling model with high prediction accuracy. Using this model, they successfully predicted the structural change matrix and parameter change matrix at the first scaling ratio of 1.5.

[0404] Step S70: Model Enlargement

[0405] With the parameter scaling model trained in step S60, the design team can scale up the initial building model based on the predicted structural change matrix and parameter change matrix to obtain the first building model with a scaling ratio of 1.5.

[0406] The specific steps are as follows:

[0407] 1. Input the structure and parameters of the initial building model into the parametric enlargement model, and set the enlargement ratio to 1.5.

[0408] 2. The parameter expansion model outputs the corresponding structural change matrix and parameter change matrix based on the input.

[0409] 3. Using these two transformation matrices, the geometric dimensions, material properties, and load conditions of the initial model are scaled and adjusted accordingly to obtain the first building model with a magnification ratio of 1.5.

[0410] Through this step, the design team successfully scaled up the initial 20,000 square meter model to 30,000 square meters. The parameters of this model are shown in Table 1.

[0411] Table 1 Model Parameter Table

[0412] parameter numerical values Planar dimensions 75m×45m floor height 4.5m Column section 1.0m × 1.0m Beam cross section 0.6m × 1.0m Floor slab thickness 0.25m base thickness 1.5m Young's modulus of concrete 4.0×10^10 Pa Poisson's ratio of concrete 0.2 Young's modulus of steel reinforcement 2.2 × 10^11 Pa Reinforcing bar yield strength 320MPa Permanent load 6kN / m

[0413] Step S80: Iterative Amplification

[0414] With the first architectural model enlarged to a scale of 1.5, the design team then needs to continue enlarging the model according to the enlargement scale sequence determined in step S30.

[0415] 1. Second enlargement (enlargement ratio 1.8)

[0416] Based on the scale sequence, the design team selected 1.8 as the next scaling factor. Using the 30,000 square meter model obtained in step S70 as input, steps S40 to S70 were repeated, i.e.:

[0417] (1) Adjust the model size and parameters:

[0418] The floor plan dimensions increased from 75m×45m to 90m×54m, and the floor height increased from 4.5m to 5.0m. Column cross-sections increased from 1.0m×1.0m to 1.2m×1.2m, beam cross-sections increased from 0.6m×1.0m to 0.7m×1.2m, floor slab thickness increased from 0.25m to 0.3m, and foundation thickness increased from 1.5m to 1.8m. Material parameters such as Young's modulus and Poisson's ratio were also adjusted accordingly.

[0419] (2) Solve the system of mechanical equations:

[0420] A detailed mechanical analysis was performed on the adjusted model, including deriving the analytical solution, performing finite element numerical solution, and determining the constraint conditions under the second scaling factor.

[0421] (3) Training parameters to expand the model:

[0422] Based on the constraints and numerical solutions obtained in step (2), the design team retrained the parameter-expanded model to adapt to the characteristics of the current model. A transfer learning approach was adopted, using previously trained model parameters as initial values, which improved training efficiency.

[0423] (4) Generate a new model:

[0424] By using the updated parameters to expand the model, the structural change matrix and parameter change matrix at the second magnification ratio of 1.8 were predicted and applied to the initial model to generate a new building model of 45,000 square meters.

[0425] 2. Subsequent enlargements (enlargement ratios of 2.0, 2.2, and 2.5)

[0426] The design team continued to repeat the above steps, successively completing the iterative process with scaling ratios of 2.0, 2.2, and 2.5. In each scaling step, the parameter scaling model was retrained to ensure that it could accurately predict the changing characteristics of the current model.

[0427] Finally, after five enlargement attempts, the design team successfully generated a model of a large-scale integrated office building covering 100,000 square meters. The main parameters of the model are shown in Table 2.

[0428] Table 2. Parameters of the enlarged model

[0429]

[0430]

[0431] Throughout the scaling-up process, the design team conducted meticulous mechanical analysis and verification at each step. They not only ensured the current model met the constraints but also checked various mechanical parameters, such as stress, displacement, and frequency, to ensure they met design requirements. If problems arose, they would promptly revert to the previous step, readjust parameters, and expand the model's training.

[0432] The design team only released the final product when the 100,000-square-meter model fully met all the performance indicators proposed by the client, including earthquake resistance, wind resistance, and energy saving.

[0433] By employing the rapid modeling method for large-scale buildings proposed in this invention, the design team completed the construction of the architectural model for this complex large office building in a relatively short time. Compared to traditional manual modeling or single numerical simulation methods, this method improves modeling efficiency by 2-3 times. Furthermore, the repeated mechanical analysis and parameter expansion model optimization during the scaling-up process ensures the reliability of the final model's mechanical properties. This lays a solid foundation for subsequent structural design, dynamic characteristic analysis, and other work.

[0434] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for rapid modeling of large-scale architectural models, characterized in that, Includes the following steps: S10. Establish a small-scale initial building model, and determine the structure and parameters of the initial building model, which are denoted as the initial structure and initial parameters, respectively. S20. Establish a set of mechanical equations for each major component of the initial model, including static equilibrium equations, deformation compatibility equations, constitutive equations, elasticity control equations, dynamic equations, structural stability equations, and foundation-structure interaction equations. S30. Determine the model scaling ratio and formulate a sequence of gradually increasing scaling ratios; S40. Adjust the size and parameters of the initial model according to the first magnification ratio; S50. Solve the system of equations for the enlarged model, including analytical and numerical solutions; Substituting the obtained analytical solution into the preset boundary conditions, the constraint conditions corresponding to the first magnification ratio are obtained; S60. Based on the numerical solution and the constraints, the model is expanded using the pre-trained structural parameters to obtain the structural change matrix and parameter change matrix corresponding to the first expansion ratio. S70. The initial building model is enlarged according to the structural change matrix and the parameter change matrix to obtain the first building model corresponding to the first enlargement ratio; S80. Repeat steps S40 to S70 to gradually increase the model size until the target size is reached.

2. The rapid modeling method for large-scale building models according to claim 1, characterized in that, The initial structure includes at least the main frame, floor system, wall structure, roof structure, infrastructure, and exterior decorative elements; the initial parameters include at least geometric dimensions, material properties, load conditions, boundary constraints, and environmental factors; the main components include column-beam joints, floor slabs, shear walls, foundations, roof, and exterior wall system.

3. The rapid modeling method for large-scale building models according to claim 2, characterized in that, The static equilibrium equations are as follows: In the formula, σ is the stress tensor, and ρ is the material density. Force per unit volume It is a divergence operator.

4. The rapid modeling method for large-scale building models according to claim 3, characterized in that, The deformation compatibility equation is specifically: In the formula, ε ij For the strain tensor, u i u j x is the displacement component. i x j These are spatial coordinates.

5. The rapid modeling method for large-scale building models according to claim 4, characterized in that, The constitutive equation is specifically: s ij =c ijkl e kl ; In the formula, σ ij For the stress tensor, C ijkl Let ε be the elastic constant tensor. kl For strain tensor.

6. The rapid modeling method for large-scale building models according to claim 5, characterized in that, The governing equations of elasticity are specifically: In the formula, μ and λ are Lamé constants. It is a displacement vector. Let t be the Laplace operator, and t be the time variable.

7. The rapid modeling method for large-scale building models according to claim 6, characterized in that, The dynamic equations are specifically: In the formula, M is the mass matrix, C is the damping matrix, K is the stiffness matrix, x is the displacement vector, and F(t) is the time-varying external force vector. Let represent the second and first time derivatives of the displacement, respectively.

8. The rapid modeling method for large-scale building models according to claim 7, characterized in that, The structural stability equation is as follows: (K+λK G )φ=0; In the formula, K G Let λ be the geometric stiffness matrix, λ be the eigenvalues, and φ be the eigenvectors.

9. A rapid modeling method for large-scale building models according to claim 8, characterized in that, The ground structure interaction equation is as follows: K ss u s +K sg u g =F s ; K gs u s +K gg u g =F g ; In the formula, K ss K is the structural stiffness matrix. gg Let K be the foundation stiffness matrix. sg K gs Let u be the structure-foundation coupled stiffness matrix. s u g Let F be the displacement vectors of the structure and the foundation, respectively. s F g Let be the external force vectors acting on the structure and the foundation, respectively.

10. A rapid modeling method for large-scale building models according to claim 9, characterized in that, The structural parameter expansion model is a deep learning model.

Images