Three-dimensional model design optimization method and system based on deep learning

Abstract

The invention relates to the technical field of building design, in particular to a three-dimensional model design optimization method and system based on deep learning. The method comprises the following steps: acquiring original three-dimensional model data; performing data preprocessing on the acquired three-dimensional model data, constructing an intelligent analysis engine based on the preprocessed three-dimensional model data, and performing parameter optimization according to the constructed intelligent analysis engine, including performing multi-target optimization on an output result of the intelligent analysis engine through an optimization algorithm; carrying out prediction simulation by utilizing the optimized intelligent analysis engine, and setting simulation parameters to carry out performance prediction; according to the method, the deep learning module of the intelligent analysis engine can adapt to different types of three-dimensional model data, performance simulation and evaluation can be carried out on three-dimensional designs in different fields, and the universality and adaptability of the design method are improved.

Description

Technical Field

[0001] The present invention relates to the field of architectural design technology, and in particular to a three-dimensional model design optimization method and system based on deep learning. Background Art

[0002] In the field of architectural design, with the development of the times, the scale and complexity of projects are constantly increasing, and traditional architectural 3D design methods have gradually exposed many drawbacks. Traditional architectural 3D design mainly relies on the personal experience and manual operation of designers. Designers often need to spend a lot of time and energy to build architectural 3D models. In this process, they evaluate the feasibility and performance of the design through manual calculations and analysis. For example, for the stability analysis of building structures, designers may need to perform complex calculations based on empirical formulas, which is not only inefficient, but also prone to errors due to human factors.

[0003] When it comes to displaying the exterior and interior structure of a building, traditional methods have limitations. It is difficult to fully and accurately present the details of a building, and it is impossible for relevant personnel to discover design problems such as unreasonable spatial layout and insufficient structural stability in advance. For example, in traditional three-dimensional model displays, the spatial relationship between different floors may not be clearly displayed, resulting in the discovery of unreasonable staircase locations during the actual construction process, which in turn leads to frequent changes and adjustments, greatly increasing project costs and time cycles.

[0004] In addition, traditional design methods are not accurate enough in building energy consumption analysis and environmental impact assessment. The evaluation of building lighting, ventilation, insulation and other performance often relies on empirical judgment or simple calculation tools, which is difficult to meet the requirements of sustainable development of modern buildings. For example, when evaluating the energy consumption of a building, it may not be possible to accurately consider the impact of light and temperature changes in different seasons and time periods on energy demand. At this stage, a three-dimensional model design optimization method and system based on deep learning is needed. Summary of the invention

[0005] In order to solve the obvious deficiencies in efficiency, accuracy, coordination and adaptability of traditional three-dimensional design methods, the present invention provides a three-dimensional model design optimization method and system based on deep learning.

[0006] In the first aspect, the present invention provides a three-dimensional model design optimization method based on deep learning, which adopts the following technical solution: A three-dimensional model design optimization method based on deep learning, comprising: Obtain original 3D model data; Perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; Build an intelligent analysis engine based on preprocessed 3D model data, including deep learning module, machine learning module and data mining module; Optimize parameters based on the constructed intelligent analysis engine, including multi-objective optimization of the output results of the intelligent analysis engine through optimization algorithms; Use the optimized intelligent analysis engine to perform prediction simulation, including integrated structural analysis software, energy consumption simulation software and lighting simulation software, model performance simulation, and set simulation parameters for performance prediction; Share prediction simulation results in real time and integrate intelligent analysis.

[0007] Furthermore, the data is cleaned, and the format conversion and feature extraction of the cleaned three-dimensional model data are performed, including using a Gaussian filtering algorithm to denoise the three-dimensional model data, then using linear interpolation to complete missing or incomplete data, and converting the three-dimensional model data from different sources into a unified internal format, and finally using a computer graphics algorithm to extract the geometric features of the model, and combining the material library and physical property calculation tools to extract the physical features of the model.

[0008] Furthermore, the construction of the deep learning module, the machine learning module and the data mining module includes constructing a hybrid model of a convolutional neural network and a recurrent neural network, wherein the convolutional neural network is used to extract the spatial features of the three-dimensional model, and the recurrent neural network is used to process the temporal information of the model, collect and analyze the results to construct a training data set, and use transfer learning and reinforcement learning techniques to improve the generalization ability of the model.

[0009] Furthermore, the construction of the deep learning module, the machine learning module and the data mining module also includes using support vector machines and random forest learning classification algorithms to classify and evaluate the performance of the model, and after the classification is completed, using a regression algorithm to predict the performance indicators of the model.

[0010] Furthermore, the construction of the deep learning module, the machine learning module and the data mining module also includes calculating the association rules between the model parameters through the Apriori algorithm, and finally performing K-means clustering analysis based on the design cases to classify potential design patterns and optimization strategies.

[0011] Furthermore, the output results of the intelligent analysis engine are subjected to multi-objective optimization by the optimization algorithm, including determining the optimization direction and objectives according to the output results of the intelligent analysis engine, iterating the optimization algorithm using a genetic algorithm, making the optimization algorithm balance between multiple objectives by setting appropriate objective functions and constraints, calculating the fitness value of each solution according to the objective function, and determining the final number of iterations according to the fitness value.

[0012] Furthermore, the predictive simulation is performed using the optimized intelligent analysis engine, including importing the optimized three-dimensional model data into the structural analysis software, determining the range and values ​​of the simulation parameters according to the optimization objectives, performing multi-parameter combination simulation according to the values ​​of the simulation parameters, and performing predictive analysis according to the set simulation parameters using the analysis software.

[0013] In the second aspect, a three-dimensional model design optimization system based on deep learning includes: The data acquisition module is configured to: acquire original three-dimensional model data; The preprocessing module is configured to: perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; The conversion module is configured to: construct an intelligent analysis engine based on the preprocessed three-dimensional model data, including constructing a deep learning module, a machine learning module and a data mining module; The optimization module is configured to: optimize parameters according to the constructed intelligent analysis engine, including performing multi-objective optimization on the output results of the intelligent analysis engine through an optimization algorithm; The prediction module is configured to: use the optimized intelligent analysis engine to perform prediction simulation, including integrating structural analysis software, energy consumption simulation software and illumination simulation software to simulate model performance, and set simulation parameters to perform performance prediction; The transformation module is configured to: share the prediction simulation results in real time and integrate intelligent analysis.

[0014] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for the three-dimensional model design optimization method based on deep learning.

[0015] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the three-dimensional model design optimization method based on deep learning.

[0016] In summary, the present invention has the following beneficial technical effects: 1. The present invention can adapt to different types of three-dimensional model data through the deep learning module of the intelligent analysis engine, continuously learn and adapt to new design tasks and scenarios through transfer learning and reinforcement learning techniques, and flexibly adjust the objective function and constraints according to different design requirements and optimization goals in the multi-objective optimization process, so that the design method can adapt to various complex design requirements, and integrate a variety of professional software to perform performance simulation and evaluation on three-dimensional designs in different fields, thereby improving the versatility and adaptability of the design method.

[0017] 2. The present invention imports the optimized three-dimensional model data into the structural analysis software, energy consumption simulation software and lighting simulation software to achieve multi-faceted performance simulation, ensure the reliability and rationality of the design, reduce manual operation errors, determine the range and value of the simulation parameters according to the optimization objectives, and perform multi-parameter combination simulation, which can explore the performance of different design schemes, help designers find the optimal combination of design parameters, and improve design quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the overall process of a three-dimensional model design optimization method based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0020] Example 1 Reference Figure 1 , a three-dimensional model design optimization method based on deep learning in this embodiment includes: Obtain original 3D model data; Perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; Build an intelligent analysis engine based on preprocessed 3D model data, including deep learning module, machine learning module and data mining module; Optimize parameters based on the constructed intelligent analysis engine, including multi-objective optimization of the output results of the intelligent analysis engine through optimization algorithms; Use the optimized intelligent analysis engine to perform prediction simulation, including integrated structural analysis software, energy consumption simulation software and lighting simulation software, model performance simulation, and set simulation parameters for performance prediction; Share prediction simulation results in real time and integrate intelligent analysis.

[0021] Specifically, a 3D model design optimization method based on deep learning includes the following steps: like Figure 1As shown, S1, obtaining original three-dimensional model data; First, a large amount of raw data related to a specific field is collected as input. These data include various parameters of the three-dimensional design model, performance indicators, historical design cases, etc. In the field of architectural design, the data include the building's dimensions, material information, energy consumption data, lighting conditions, etc.

[0022] S2. performing data preprocessing on the acquired 3D model data, including data cleaning, and performing format conversion and feature extraction on the cleaned 3D model data; First, the acquired data is denoised by using the Gaussian filtering algorithm to denoise the 3D model data. The basic principle is to determine the weight of the filter according to the shape of the Gaussian function. For a 2D image (the 3D model data can be regarded as a combination of a series of 2D slices), the pixel value at the coordinate is The value after Gaussian filtering can be calculated by the following formula: , in, Expressed as pixels, Expressed as the standard deviation, it determines the smoothness of the filter. It means that after Gaussian filtering, the coordinates are The pixel value at Indicates that the original image is at coordinates The pixel value at Represents the normalization coefficient, which is then used to complete missing or incomplete data using linear interpolation. Assume that two data points are known and , the value to be interpolated at, the linear interpolation formula is: , in, Represents the horizontal coordinate of the position to be interpolated. and Represented as the horizontal coordinates of two known data points, and Respectively expressed as corresponding to and The vertical coordinate of the data point, It is expressed as the value obtained by interpolation at the horizontal axis x. Finally, the format conversion tool is used to convert the 3D model data from different sources (such as RVT, STEP, IGES, etc.) into a unified internal format. The computer graphics algorithm is used to extract the geometric features of the model, such as volume, surface area, and shape parameters (aspect ratio, circularity, etc.). The volume can be obtained by integral calculation, the surface area can be obtained by calculating the sum of the areas of triangles on the surface of the model, and the shape parameters are calculated by the corresponding calculation formula. For example: for a 3D model, its maximum and minimum dimensions in the directions of the three coordinate axes can be determined first. It is assumed that the maximum dimension in the x-axis direction is maxX and the minimum dimension is minX; the maximum dimension in the y-axis direction is maxY and the minimum dimension is minY; the maximum dimension in the z-axis direction is maxZ and the minimum dimension is minZ. The aspect ratio can be defined as max (length) / min (length), where length can be the ratio of the maximum value to the minimum value among maxX - minX, maxY - minY or maxZ - minZ. Combined with the material library and physical property calculation tool, the physical characteristics of the model are extracted, such as material density, thermal conductivity, structural strength, etc. The material density can be obtained from the material library, and the thermal conductivity and structural strength can be calculated by the corresponding physical calculation formulas.

[0023] S3. Building an intelligent analysis engine based on the preprocessed 3D model data, including building a deep learning module, a machine learning module, and a data mining module; The acquired raw data is fed into the deep learning module to build a hybrid model of convolutional neural network (CNN) and recurrent neural network (RNN). CNN is used to extract the spatial features of the 3D model, such as the shape and structure of the object; RNN is used to process the temporal information of the model, such as the changing trend in the design process. The model parameters are adjusted using the back-propagation algorithm, and the model is trained using a training data set constructed from a large number of design cases and analysis results. At the same time, transfer learning and reinforcement learning techniques are used to improve the generalization and adaptability of the model. CNN automatically learns the spatial features of the image (the 2D slice of the 3D model) through multiple convolutional layers and pooling layers. The operation of the convolutional layer can be regarded as a convolution operation between the input image and a set of learnable convolution kernels. For a 2D image, the output value O(x,y) after convolution at the coordinate (x,y) can be calculated by the following formula: , Among them, I(x,y) is the pixel value of the input image, K(i,j) is the value of the convolution kernel, and the pooling layer usually uses maximum pooling or average pooling to reduce the resolution of the feature map while retaining the main features. For example, maximum pooling takes the maximum value in a small area as the output. Then a recurrent neural network (RNN) is constructed to process time series information. RNN is suitable for processing sequence data. In a three-dimensional model, it can be used to process information that changes over time. The basic unit of RNN is a neuron, and its output depends not only on the current input, but also on the hidden state of the previous moment. The formula is: , in, is the input at the current moment, is the hidden state at the current moment, and is the weight matrix, is a bias term. Transfer learning uses the knowledge learned in other related tasks to accelerate the learning of the current task. For example, the weights of the CNN model pre-trained on a large-scale image dataset can be used as the initial value and applied to the feature extraction task of the three-dimensional model. The feature vector output by the deep learning stage is then used as the input of the machine learning module to the machine learning module. The specific processing process is as follows: classification algorithms such as support vector machine (SVM) and random forest (RF) are used to classify and evaluate the model performance to determine the pros and cons of the model, applicable scenarios, etc. At the same time, regression algorithms such as linear regression and ridge regression are used to predict the performance indicators of the model, such as energy consumption, illumination, etc. SVM classifies data by finding an optimal hyperplane. For the binary classification problem, assuming that the training data is ,in, is the input feature vector, is the category label. The goal of SVM is to find a hyperplane that maximizes the distance between the two types of data points and the hyperplane. RF is an ensemble learning algorithm consisting of multiple decision trees. For an input sample, each decision tree will give a classification result. The final classification result is determined by majority voting. The decision tree is constructed by partitioning the features. At each node, an optimal feature and partition point are selected so that the purity of the sub-nodes after partitioning is the highest. Purity can be measured by indicators such as information entropy and Gini index. Linear regression is used to predict continuous numerical target variables. The regression coefficient is estimated by minimizing the mean square error (MSE). Ridge regression adds a regularization term to linear regression to prevent overfitting. The original 3D model data, the intermediate results of deep learning and machine learning are used as inputs to mine the association rules between design parameters through the Apriori algorithm or FP-Growth algorithm to discover potential design patterns and relationships. K-means cluster analysis or hierarchical cluster analysis is performed on similar design cases to better understand the design space and discover optimization strategies.

[0024] Finally, the data mining module uses the Apriori algorithm to mine frequent item sets and association rules. Frequent item sets refer to the collection of items that appear frequently in a data set. The algorithm iterates continuously to first find the frequent 1-item set, then generate the frequent 2-item set based on the frequent 1-item set, and so on, until no larger frequent item set is found. The form of the association rule is , where A and B are item sets, support and confidence are two indicators to measure the strength of association rules. Support refers to the proportion of transactions containing A and B in the total transactions; confidence refers to the proportion of transactions containing A and B in transactions that also contain B.

[0025] Then, K initial cluster centers are randomly selected, and each data point is assigned to the cluster with the nearest cluster center. Next, the center of each cluster is recalculated, and this process is repeated until the cluster center no longer changes or the maximum number of iterations is reached.

[0026] S4. Optimizing parameters of the constructed intelligent analysis engine, including multi-objective optimization of the output results of the intelligent analysis engine through an optimization algorithm; According to the output results of the intelligent analysis engine, the optimization direction and goal are determined. For example, for material selection optimization, the appropriate material combination can be selected according to the performance requirements and cost constraints of the model; for structural layout optimization, the structural characteristics of the model can be analyzed to find a more reasonable layout solution to improve strength and stability; for lighting design optimization, the light source parameters and model surface materials can be adjusted to improve the lighting effect, and the genetic algorithm (GA) can be used to simulate natural selection, crossover and mutation processes to find the optimal solution in the search space, considering the conflicts between different design goals, such as cost, energy consumption, aesthetics, etc., and the optimization algorithm can make trade-offs between multiple goals through the set objective functions and constraints. The objective function is expressed as: , Among them, C represents the target cost, E represents energy consumption, and B represents aesthetics. Expressed as a weight coefficient, find a design with the best comprehensive performance. Encode the design parameters. For genetic algorithms, the parameters can be encoded as chromosomes. Through selection, crossover, and mutation, the optimization process is continuously iterated to gradually approach the optimal solution. In each iteration, an excellent design solution is selected according to the fitness value, and genetic operations are performed to generate a new design solution. A maximum number of iterations T is pre-set. When the current number of iterations t reaches T, the optimization process is stopped. When the stopping condition is met, the optimization process is stopped and the optimal design parameter combination and performance indicators are output.

[0027] S5. Use the optimized intelligent analysis engine to perform prediction simulation, including integrating structural analysis software, energy consumption simulation software and illumination simulation software to simulate model performance, and set simulation parameters to perform performance prediction; 1. Integrated structural analysis software: such as ANSYS, etc., import the optimized 3D model data into the structural analysis software to evaluate the structural strength, stability and deformation of the model. By setting different load conditions and boundary constraints, the stress conditions of the model in actual use can be simulated, possible problems can be predicted, and a basis for structural optimization can be provided.

[0028] 2. Integrated energy consumption simulation software: such as EnergyPlus, which is used to analyze the energy consumption performance of the model. Based on the model's geometry, material properties, lighting conditions and other factors, the energy consumption simulation software can calculate the model's energy consumption under different operating conditions, such as electricity and heat. By optimizing the design parameters, the model's energy consumption can be reduced and energy efficiency can be improved.

[0029] 3. Integrated lighting simulation software: such as Radiance, etc., to simulate the lighting effect of the model. You can set different light source positions, intensities and colors, as well as the surface material and reflection characteristics of the model to evaluate indicators such as lighting uniformity, brightness and color reproduction. The lighting simulation results can provide a reference for lighting design and improve the visual effect and comfort of the model.

[0030] The specific integration method is to develop a data interface to ensure data exchange and interaction between different simulation software. The data interface is to import the geometric data of the 3D model from the design software to the structural analysis software, and then pass the results to the energy consumption simulation software and the lighting simulation software after the structural analysis. For the analysis model, the performance indicators can be obtained by calling its API or reading the output file to perform comprehensive performance simulation. Since different simulation software may support different model formats, model conversion is required to ensure that the 3D model can be correctly loaded and run in each simulation software.

[0031] Determine the range and value of simulation parameters according to design requirements and optimization goals. For example, in structural analysis, you can set different load sizes, directions, and points of action; in energy consumption simulation, you can adjust factors such as indoor temperature, humidity, and human activity; in lighting simulation, you can change parameters such as the position, intensity, and color of the light source. Consider the mutual influence between different parameters and perform multi-parameter combination simulation. By adjusting the values ​​of multiple parameters, you can explore the performance of different design solutions and find the optimal parameter combination.

[0032] Run the integrated professional software to perform performance prediction. According to the set simulation parameters, the software will calculate and analyze the model and generate corresponding performance index results, such as structural stress, deformation, energy consumption value, light intensity, etc. Analyze the performance prediction results to evaluate the pros and cons of the design scheme. You can compare the performance of different design schemes on various performance indicators to determine the optimal design scheme. At the same time, the analysis results can also reveal the problems and potential improvement directions in the design, and put forward further optimization suggestions based on the performance prediction results.

[0033] S6. Share the prediction simulation results in real time and integrate intelligent analysis.

[0034] Integrate version control mechanisms, such as Git, to manage design changes and history. Git records the version information of each submission, supports branch management and merge operations, and integrates the intelligent analysis engine into the collaborative work platform. By calling the interface of the intelligent analysis engine, real-time intelligent analysis services are realized.

[0035] Example 2 The difference between this embodiment and embodiment 1 is that this embodiment provides a three-dimensional model design optimization system based on deep learning, including: The data acquisition module is configured to: acquire original three-dimensional model data; The preprocessing module is configured to: perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; The conversion module is configured to: construct an intelligent analysis engine based on the preprocessed three-dimensional model data, including constructing a deep learning module, a machine learning module and a data mining module; The optimization module is configured to: optimize parameters according to the constructed intelligent analysis engine, including performing multi-objective optimization on the output results of the intelligent analysis engine through an optimization algorithm; The prediction module is configured to: use the optimized intelligent analysis engine to perform prediction simulation, including integrating structural analysis software, energy consumption simulation software and illumination simulation software to simulate model performance, and set simulation parameters to perform performance prediction; The transformation module is configured to: share the prediction simulation results in real time and integrate intelligent analysis.

[0036] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for a three-dimensional model design optimization method based on deep learning.

[0037] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the three-dimensional model design optimization method based on deep learning.

[0038] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A three-dimensional model design optimization method based on deep learning, characterized in that: include: Obtain original 3D model data; Perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; Build an intelligent analysis engine based on preprocessed 3D model data, including deep learning module, machine learning module and data mining module; Optimize parameters based on the constructed intelligent analysis engine, including multi-objective optimization of the output results of the intelligent analysis engine through optimization algorithms; Use the optimized intelligent analysis engine to perform prediction simulation, including integrated structural analysis software, energy consumption simulation software and lighting simulation software, model performance simulation, and set simulation parameters for performance prediction; Share prediction simulation results in real time and integrate intelligent analysis.

2. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The data is cleaned, and format conversion and feature extraction are performed on the cleaned three-dimensional model data, including denoising the three-dimensional model data using a Gaussian filtering algorithm, then completing missing or incomplete data using linear interpolation, and converting the three-dimensional model data from different sources into a unified internal format, and finally extracting the geometric features of the model using a computer graphics algorithm, and extracting the physical features of the model in combination with a material library and a physical property calculation tool.

3. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The construction of the deep learning module, the machine learning module and the data mining module includes constructing a hybrid model of a convolutional neural network and a recurrent neural network, wherein the convolutional neural network is used to extract the spatial features of the three-dimensional model, the recurrent neural network is used to process the temporal information of the model, the analysis results are collected to construct a training data set, and transfer learning and reinforcement learning techniques are used to improve the generalization ability of the model.

4. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The construction of the deep learning module, the machine learning module and the data mining module also includes using support vector machines and random forest learning classification algorithms to classify and evaluate the performance of the model, and after the classification is completed, using a regression algorithm to predict the performance indicators of the model.

5. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The construction of the deep learning module, the machine learning module and the data mining module also includes calculating the association rules between the model parameters through the Apriori algorithm, and finally performing K-means clustering analysis based on the design cases to classify potential design patterns and optimization strategies.

6. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The output results of the intelligent analysis engine are optimized by the optimization algorithm, including determining the optimization direction and objectives according to the output results of the intelligent analysis engine, iterating the optimization algorithm by using a genetic algorithm, making the optimization algorithm balance between multiple objectives by setting appropriate objective functions and constraints, calculating the fitness value of each solution according to the objective function, and determining the final number of iterations according to the fitness value.

7. The three-dimensional model design optimization method based on deep learning according to claim 1, characterized in that: The predictive simulation is performed using the optimized intelligent analysis engine, including importing the optimized three-dimensional model data into the structural analysis software, determining the range and value of the simulation parameters according to the optimization target, performing multi-parameter combination simulation according to the value of the simulation parameters, and performing predictive analysis according to the set simulation parameters using the analysis software.

8. A three-dimensional model design optimization system based on deep learning, characterized in that: include: The data acquisition module is configured to: acquire original three-dimensional model data; The preprocessing module is configured to: perform data preprocessing on the acquired 3D model data, including data cleaning, and perform format conversion and feature extraction on the cleaned 3D model data; The conversion module is configured to: construct an intelligent analysis engine based on the preprocessed three-dimensional model data, including constructing a deep learning module, a machine learning module and a data mining module; The optimization module is configured to: optimize parameters according to the constructed intelligent analysis engine, including performing multi-objective optimization on the output results of the intelligent analysis engine through an optimization algorithm; The prediction module is configured to: use the optimized intelligent analysis engine to perform prediction simulation, including integrating structural analysis software, energy consumption simulation software and illumination simulation software to simulate model performance, and set simulation parameters to perform performance prediction; The transformation module is configured to: share the prediction simulation results in real time and integrate intelligent analysis.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing a three-dimensional model design optimization method based on deep learning as described in claim 1.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing a three-dimensional model design optimization method based on deep learning as described in claim 1.

Images