Indoor environment-friendly decoration three-dimensional computer aided design system

Abstract

The application discloses an indoor environment-friendly decoration three-dimensional computer aided design system, and belongs to the field of indoor decoration three-dimensional computer aided design. The system comprises a three-dimensional modeling and attribute management module, a multi-source pollutant three-dimensional space coupling release real-time simulation module, an environment constraint driven generative material collaborative optimization module and a visual interactive rendering module. The simulation module performs forward inference through a physical information neural network agent model embedded with a convection diffusion equation physical residual error, realizes real-time prediction of multi-material superimposed pollutant concentration in the design stage, and the optimization module takes indoor air quality standard limit value as a constraint, and solves a material replacement scheme among pollutant exceeding, aesthetic deviation and budget deviation targets by using a multi-objective genetic algorithm. The application realizes real-time simulation of pollutant concentration and automatic optimization of materials driven by environmental protection constraints in the design stage.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional computer-aided design and building information modeling for interior decoration, specifically to a three-dimensional computer-aided design system for environmentally friendly interior decoration that combines a physical information neural network proxy model with a multi-objective genetic optimization algorithm, enabling real-time simulation of indoor pollutant concentration and automatic optimization of materials driven by environmental constraints during the design phase. Background Technology

[0002] Interior decoration 3D computer-aided design software typically includes geometric modeling, material rendering, attribute management, and drawing output modules. It is used to construct 3D geometries such as walls, floors, and furniture, and assign visual material attributes such as color and texture. In the field of pollutant detection, current standards mainly use the environmental testing chamber method to determine the release rate of a single decorative material sample, obtaining data on the release rates of formaldehyde and volatile organic compounds under specific temperature and humidity conditions. In the field of fluid simulation, computational fluid dynamics software can obtain the distribution results of pollutant concentration fields within a space by meshing the target space and numerically iterating and solving the mass, momentum, and concentration transport equations. However, the computation time is usually measured in minutes to hours. The closest implementation to this invention is computer-aided design modeling combined with offline computational fluid dynamics. The combined workflow of physical dynamics simulation involves designers completing modeling and material selection in 3D design software, exporting the geometric model, and importing it into computational fluid dynamics software for mesh generation, boundary condition setting, and iterative solution to obtain pollutant concentration distribution results. If the concentration exceeds the standard, the designer returns to the design software to manually adjust the scheme and repeat the above process. In this combined workflow, the computer-aided design workstation and the computational fluid dynamics simulation workstation exchange data unidirectionally via file import and export. The design and simulation processes are disconnected, and there is a lack of a mechanism to use indoor air quality standards as constraints to drive automatic material optimization and replacement. This existing technical solution cannot meet the need for real-time adjustment of the scheme and simultaneous environmental performance feedback during the design phase, nor can it effectively predict the cumulative release effects of multiple materials before completion. Summary of the Invention

[0003] The technical problem this invention aims to solve is that existing 3D computer-aided design software for interior decoration cannot perform real-time simulation and evaluation of the concentration distribution of pollutants released by multiple decorative materials superimposed in the same space during the design stage. There are technical barriers such as data fragmentation and high computational latency between the existing design process and professional computational fluid dynamics simulation software. Furthermore, existing design software lacks a mechanism to use indoor air quality standards as constraints to drive automatic optimization and replacement of materials, resulting in pollution problems being passively discovered only after completion, leading to high error correction costs.

[0004] To address the aforementioned technical problems, this invention provides a three-dimensional computer-aided design system for indoor environmentally friendly decoration, comprising a three-dimensional modeling and attribute management module, a real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants, an environmentally constrained generative material collaborative optimization module, and a visualization and interactive rendering module.

[0005] The 3D modeling and attribute management module is used to receive modeling operations to construct a 3D geometric model of the interior space, and associate the types of decorative materials, geometric exposure areas, and spatial coordinate information of the geometric primitives of walls, floors, and furniture in the geometric model. At the same time, it connects to the material environmental attribute database to obtain data on the release rates of formaldehyde and volatile organic compounds of the materials.

[0006] The real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants is connected to the three-dimensional modeling and attribute management module. It receives geometric exposure area, spatial coordinates, and environmental parameters such as temperature, humidity, and ventilation path. It performs forward inference through a pre-trained physical information neural network surrogate model and outputs three-dimensional distribution field data of the superimposed concentrations of formaldehyde and volatile organic compounds in the indoor space. During the training process of the surrogate model, the convection-diffusion equation is embedded into the training loss function in the form of physical residual loss terms, so that the surrogate model can complete forward inference without meshing after training.

[0007] The environmentally constrained generative material co-optimization module is connected to the simulation module. It compares the three-dimensional distribution field data with the preset indoor air quality standard limits. When there are evaluation units with concentrations exceeding the limits, it calls the material environmental property database and uses a multi-objective genetic algorithm to calculate and output material replacement schemes under the conditions of satisfying preset aesthetic color constraints and engineering budget constraints.

[0008] The visualization and interactive rendering module is connected to the simulation and optimization modules. It overlays and renders the three-dimensional distribution field data in the form of a three-dimensional cloud map onto the three-dimensional spatial geometric model, and presents the material replacement scheme in the form of an interactive interface. At the same time, it receives the designer's confirmation results and feeds them back to the three-dimensional modeling and attribute management module, forming a closed-loop data flow structure.

[0009] The beneficial effects achieved by this invention include:

[0010] 1. By replacing the traditional computational fluid dynamics mesh generation and iterative solution process with a physical information neural network proxy model, millisecond-level real-time simulation of pollutant concentration distribution in multi-material superposition during the 3D design stage is achieved, enabling designers to obtain environmental performance feedback immediately after adjusting materials or layout.

[0011] 2. The convection-diffusion equation is embedded into the training process as a physical residual loss term, so that the surrogate model can have prediction accuracy under the constraints of physical laws while ensuring inference speed, and reduce the prediction bias generated by the pure data-driven model outside the coverage of training data.

[0012] 3. Using indoor air quality standard limits as constraints for generative design, a linkage mechanism for exceeding standards judgment and multi-objective optimization is constructed. Material replacement schemes are solved among the three objectives of pollutant exceeding standards, aesthetic color deviation, and engineering budget deviation. The environmental governance process is moved from passive remediation after completion to proactive prevention in the design stage, reducing governance costs and improving the reliability of the solution implementation. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the overall system architecture of the present invention.

[0014] Figure 2 This is a flowchart of the multi-objective genetic algorithm optimization process of the present invention;

[0015] Figure 3 This is a flowchart illustrating the working sequence of the system of the present invention. Detailed Implementation

[0016] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0017] I. System Overall Architecture

[0018] The indoor environmentally friendly decoration three-dimensional computer-aided design system provided in this embodiment includes a three-dimensional modeling and attribute management module, a real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants, an environmentally constrained generative material collaborative optimization module, and a visualization and interactive rendering module. The system also connects to an external material environmental property database.

[0019] The 3D modeling and attribute management module includes a geometric modeling unit and a material attribute association unit. The geometric modeling unit receives modeling operations from designers and constructs a 3D geometric model of the interior space, including geometric elements such as walls, floors, ceilings, and furniture. The material attribute association unit associates the selected decorative materials with the geometric elements, including their types, geometric exposure areas, and spatial coordinate information. It also connects to a material environmental attribute database to obtain data on the formaldehyde, volatile organic compound release rate per unit area, price, and color characteristics of the corresponding materials. The release rate data is stored in the form of release rate curves with temperature and humidity as independent variables, and is derived from the material's environmental product declaration or third-party testing report.

[0020] The real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants includes a physical information neural network surrogate model inference unit, which is connected to the three-dimensional modeling and attribute management module. It receives geometric exposure area, spatial coordinates, and environmental parameters such as temperature, humidity, and ventilation path as inputs. Through the pre-trained surrogate model, it performs forward inference and outputs three-dimensional distribution field data of the superimposed concentrations of formaldehyde and volatile organic compounds in the indoor space.

[0021] The environmentally constrained generative material co-optimization module includes an exceedance judgment unit and a multi-objective optimization calculation unit, which are connected to the simulation module. It receives three-dimensional distribution field data and compares it with indoor air quality standard limits. When an exceedance area is detected, it calls the material environmental property database and uses a multi-objective genetic algorithm to calculate and output a material replacement scheme under the premise of meeting the designer's preset aesthetic color scheme and engineering budget constraints.

[0022] The visualization and interactive rendering module is connected to the simulation and optimization modules. It overlays and renders the pollutant concentration distribution field as a 3D cloud map in the design scene. The material replacement scheme recommended by the optimization module is presented to the designer in the form of an interactive interface for confirmation or adoption. The adjustment results confirmed by the designer are fed back to the 3D modeling and attribute management module through this module, which re-triggers the cyclical process of modeling data extraction, proxy model reasoning, concentration cloud map rendering, exceedance judgment, material optimization calculation and scheme recommendation, forming a closed-loop workflow of design, simulation, generation and feedback, until the superimposed concentration of pollutants in all assessment units in the indoor space meets the indoor air quality standard limit requirements.

[0023] II. A Lightweight Real-Time Simulation Method for Three-Dimensional Spatial Coupling Release of Multi-Source Pollutants

[0024] The spatiotemporal distribution of pollutant concentrations in indoor air is described using the convection-diffusion equation:

[0025] ;

[0026] In the formula, Representing spatial coordinates At any moment The concentration of pollutants, This represents the indoor air velocity vector, determined by the ventilation path parameters. This represents the molecular diffusion coefficient of pollutants in the air. Represents the gradient operator, Represents the Laplace operator. Indicates the first A type of decorative material in coordinates place, time The intensity of pollutant release sources This represents the total number of decorative materials involved in the coupling calculation.

[0027] The intensity of the material release source is expressed as:

[0028] ;

[0029] In the formula, Indicates the first The geometric exposed area of ​​the material is automatically extracted by the 3D modeling module. Indicates the first This material at temperature ,humidity The release rate per unit area under the given conditions is derived from measured release rate curves in the material environmental property database. The Dirac function is used to confine the material release source to its coordinate position in space. .

[0030] Training a physical information neural network surrogate model includes the following steps:

[0031] We will obtain the true distribution data of the concentration field from computational fluid dynamics simulations under various standard indoor spatial geometric layouts, corresponding to multiple combinations of material types, exposed areas, release rates, and temperature and humidity parameters, as a training dataset. ,in Indicates the first The input feature vector of each sampling point This represents the corresponding true concentration. Indicates the total number of sampling points;

[0032] Constructing a proxy model based on a multi-layer fully connected neural network ,in This represents the set of trainable parameters for the network. The network input is a feature vector composed of spatial coordinates, material geometric exposure area, release rate, and environmental parameters. The output is the predicted concentration value at that point.

[0033] The convection-diffusion equation is embedded as a soft constraint in the training process. The partial derivatives of the network output with respect to coordinates and time are calculated by automatic differentiation to construct the physical residual loss term.

[0034] ;

[0035] In the formula, This indicates the number of configuration points used to calculate the physical residual. Indicates the first The coordinates and time of each configuration point are input, and the meanings of the other symbols are consistent with the aforementioned convection-diffusion equation;

[0036] The mean squared error between the network's predicted values ​​and the true values ​​at the ground truth sampling points is used as the data fitting loss term:

[0037] ;

[0038] The total loss function is obtained by weighting the data fitting loss term, physical residual loss term, and boundary condition loss term according to preset weight coefficients and summing them.

[0039] ;

[0040] In the formula, , These represent the weighting coefficients of the physical constraint loss term and the boundary condition loss term, respectively. The boundary condition loss term is used to constrain concentration or flux conditions at boundaries such as walls and vents. A gradient descent optimizer is used to optimize the network parameters. The process is iteratively updated until the total loss function converges to a preset threshold, resulting in a trained surrogate model. Its parameters are then exported and embedded into the simulation module for inference.

[0041] During the real-time inference phase, after the designer completes at least one operation—material selection adjustment or spatial layout adjustment—the simulation module automatically extracts the geometric exposure area, spatial coordinates, and environmental temperature, humidity, and ventilation path parameters of each decorative material in the current scene. This data is then assembled into a feature vector set covering the three-dimensional mesh sampling points of the interior space and input into the surrogate model. Forward reasoning is performed to obtain the predicted concentrations of formaldehyde and volatile organic compounds at each sampling point. The predicted values ​​are then organized into three-dimensional concentration distribution field data and transmitted to the visualization interactive rendering module and the environmental constraint-driven generative material co-optimization module.

[0042] III. Environmentally Constrained Generative Material Collaborative Optimization Method

[0043] The exceedance determination unit divides the three-dimensional concentration distribution field into several assessment units according to a spatial grid, and compares the predicted concentration value of each assessment unit with the indoor air quality standard limit. If the concentration of a certain evaluation unit exceeds the limit, the unit and its associated decorative materials will be marked as objects to be optimized.

[0044] The multi-objective optimization computational unit employs a multi-objective genetic algorithm to solve for material replacement schemes, specifically including:

[0045] Replacement schemes for each decorative material within the space to be optimized are coded as chromosome individuals. Each chromosome consists of several gene loci, and each gene locus corresponds to a candidate replacement material number for a specific decorative material. The candidate replacement material set is derived from a set of materials in the material environmental attribute database that share the same functional category as the raw materials. Several chromosome individuals are randomly generated within the candidate material set, with a scale defined by a preset value. initial population .

[0046] Calculate the following three target function values ​​for each individual chromosome in the population: (Target function for pollutant exceedance)

[0047] ;

[0048] In the formula, Indicates the total number of evaluation units. This indicates the result obtained after substituting the material scheme corresponding to the current chromosome into the surrogate model and re-inferring. Predicted concentration values ​​for each assessment unit.

[0049] Aesthetic color deviation objective function:

[0050] ;

[0051] In the formula, This indicates the total number of materials involved in the replacement. Indicates the first The weighting coefficient of each material in the designer's aesthetic scheme. , These represent the color feature vectors of the replaced material and the original design material, respectively. This represents the Euclidean distance between vectors.

[0052] Project budget deviation objective function:

[0053] ;

[0054] In the formula, Indicates the first The unit price of the replacement material This indicates the area where the material is used. This indicates the total project budget preset by the designer.

[0055] According to all individuals with chromosomes in the population , , The three objective function values ​​are non-dominated and sorted, dividing the population into several non-dominated levels. Crowding distance is calculated between individuals within the same level. A binary tournament selection method is used to select parent individuals from the current population. Genetic crossover and mutation operations are performed on the parent individuals according to preset crossover probabilities, generating the offspring population. , the parent population With offspring population After merging, the non-dominated sorting and congestion calculation are re-performed. The top levels are selected in descending order of congestion distance within the same level, from front to back of the non-dominated hierarchy. Individuals constitute the next generation population. Repeat the above steps of objective function calculation, non-dominated sorting, selection crossover mutation and population merging until the preset termination condition is reached, output Pareto solution set, and sort the solutions in the Pareto solution set according to the target preference weights set by the designer in advance. Select the top-ranked material replacement schemes and recommend them to the designer for confirmation through the visual interactive rendering module.

[0056] IV. System Working Sequence Example

[0057] After the designer completes the preliminary design, the system executes the following cyclical process: modeling data extraction, proxy model inference, concentration cloud map rendering, exceedance judgment, material optimization calculation, scheme recommendation, and designer confirmation feedback. The designer first completes spatial geometric modeling and material selection in the 3D modeling and attribute management module. Then, the simulation module performs forward inference based on the current modeling data and outputs the concentration distribution field. The visualization interactive rendering module overlays and renders the concentration distribution field as a cloud map on the 3D scene. The exceedance judgment unit determines whether there are any exceedance assessment units. If not, the current design scheme is maintained. If so, the multi-objective optimization calculation unit calculates the material replacement scheme and presents it through the visualization interactive rendering module. After the designer confirms or makes adjustments, the adjustment results are fed back to the modeling module. The system then re-executes the above cycle until the concentration of all assessment units meets the indoor air quality standard limit requirements.

[0058] V. Alternative Implementation Methods

[0059] In the simulation of pollutant three-dimensional spatial coupling and release, as an alternative implementation method, instead of using a physical information neural network proxy model, a data-driven deep learning regression model trained based on historical computational fluid dynamics simulation data can be used. Without explicitly introducing physical equation residual constraints, the mapping relationship between spatial geometric features and concentration distribution can be learned solely by using labeled data, which can also achieve rapid concentration prediction in the design stage.

[0060] In the material co-optimization stage, as an alternative implementation method, instead of using the non-dominated sorting genetic algorithm with elitist strategy, a multi-objective particle swarm optimization algorithm based on Pareto front approximation can be used. The non-dominated solution set is found by iteratively updating the position and velocity of the particle swarm in the multi-dimensional objective space. Similarly, a set of candidate material replacement schemes can be obtained among the objectives of pollutant exceedance, aesthetic deviation, and budget deviation.

[0061] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A three-dimensional computer-aided design system for environmentally friendly interior decoration, characterized in that, include: The 3D modeling and attribute management module is used to receive modeling operations to construct a 3D geometric model of the interior space, and associate the types of decorative materials, geometric exposed areas, and spatial coordinate information of geometric primitives of walls, floors, and furniture in the geometric model. At the same time, it connects to the material environmental property database to obtain data on the release rates of formaldehyde and volatile organic compounds of the materials. A real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants is communicatively connected to the three-dimensional modeling and attribute management module. It is used to receive the geometric exposure area, spatial coordinates, and environmental parameters such as temperature, humidity, and ventilation path. It performs forward inference through a pre-trained physical information neural network proxy model and outputs three-dimensional distribution field data of the superimposed concentrations of formaldehyde and volatile organic compounds in the indoor space. During the training process of the physical information neural network proxy model, the convection-diffusion equation is embedded into the training loss function in the form of physical residual loss terms, so that the proxy model can complete forward inference without mesh division after training. An environmentally constrained generative material co-optimization module is communicatively connected to the real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants. It is used to compare the three-dimensional distribution field data with the preset indoor air quality standard limit. When there is an evaluation unit with a concentration exceeding the indoor air quality standard limit, it calls the material environmental protection attribute database and uses a multi-objective genetic algorithm to calculate and output a material replacement scheme under the conditions of satisfying the preset aesthetic color constraints and engineering budget constraints. The visualization and interactive rendering module is communicatively connected to the real-time simulation module for the three-dimensional spatial coupling and release of multi-source pollutants and the environmental constraint-driven generative material collaborative optimization module. It is used to overlay and render the three-dimensional distribution field data in the form of a three-dimensional cloud map onto the three-dimensional spatial geometric model and to present the material replacement scheme in the form of an interactive interface.

2. The system according to claim 1, characterized in that, The material environmental property database stores data on material types, release rates per unit area, prices, and color characteristics. The release rate per unit area data is stored in the form of release rate curves with temperature and humidity as independent variables.

3. The system according to claim 1, characterized in that, The physical information neural network surrogate model is a multi-layer fully connected neural network. Its input is a feature vector composed of spatial coordinates, material geometric exposure area, release rate, and environmental parameters. The output is the predicted value of pollutant concentration at the corresponding location. The training loss function is composed of a data fitting loss term, a physical residual loss term, and a boundary condition loss term, which are weighted and summed according to preset weight coefficients. The data fitting loss term represents the mean square error between the predicted value and the true value at the sampling point of the computational fluid dynamics simulation by the surrogate model. The physical residual loss term represents the residual obtained by substituting the partial derivatives of the surrogate model output with respect to coordinates and time into the convection-diffusion equation.

4. The system according to claim 1, characterized in that, The real-time simulation module for the three-dimensional spatial coupling release of multi-source pollutants is also used to: after detecting that the designer has completed at least one of the following operations, namely material selection adjustment or spatial layout adjustment, automatically extract the geometric exposure area, spatial coordinates and environmental parameters of the adjusted scene, assemble them into a feature vector set covering the three-dimensional grid sampling points of the indoor space, and input the feature vector set into the physical information neural network proxy model for forward inference to obtain the updated three-dimensional distribution field data.

5. The system according to claim 1, characterized in that, The multi-objective genetic algorithm in the environmentally constrained generative material collaborative optimization module encodes the replacement schemes of each decorative material in the space to be optimized as chromosome individuals. Each chromosome individual corresponds to three objective function values: pollutant exceedance objective function value, aesthetic color deviation objective function value, and engineering budget deviation objective function value. The pollutant exceedance objective function value represents the sum of the parts of the predicted concentration values ​​of each evaluation unit that exceed the indoor air quality standard limit. The aesthetic color deviation objective function value represents the sum of the weighted distances between the color feature vectors of the replaced materials and the original design materials. The engineering budget deviation objective function value represents the absolute value of the deviation between the total cost of the replaced materials and the preset total engineering budget.

6. The system according to claim 5, characterized in that, The multi-objective genetic algorithm classifies and selects individuals in the population according to non-dominated sorting and crowding distance. It generates offspring populations through crossover and mutation operations, merges the parent and offspring populations, re-sorts and calculates crowding, and selects individuals to form the next generation population according to a preset population size. When the population iteration reaches a preset maximum iteration count and / or the Pareto solution set remains unchanged within a preset number of generations, the Pareto solution set is output, and a material replacement scheme is recommended from the Pareto solution set based on preset target preference weights.

7. The system according to claim 1, characterized in that, The visualization and interactive rendering module is also used to receive the designer's confirmation of the material replacement scheme and to feed back the confirmed adjustment results to the 3D modeling and attribute management module, forming a closed-loop data flow structure of design, simulation, generation and feedback.

8. A three-dimensional computer-aided design method for environmentally friendly interior decoration, characterized in that, include: Construct a three-dimensional geometric model of the interior space, associate the types of decorative materials, geometric exposure areas, and spatial coordinate information of the geometric primitives of walls, floors, and furniture in the geometric model, and obtain data on the release rates of formaldehyde and volatile organic compounds from the materials; The geometric exposure area, spatial coordinates, and environmental parameters such as temperature, humidity, and ventilation path are obtained. Forward reasoning is performed through a pre-trained physical information neural network proxy model to obtain three-dimensional distribution field data of the superimposed concentration of formaldehyde and volatile organic compounds in the indoor space. During the training process of the physical information neural network proxy model, the convection-diffusion equation is embedded into the training loss function in the form of physical residual loss term. The three-dimensional distribution field data is compared with the preset indoor air quality standard limit. When there is an evaluation unit where the concentration exceeds the indoor air quality standard limit, a multi-objective genetic algorithm is used to calculate the material replacement scheme under the conditions of satisfying the preset aesthetic color constraints and engineering budget constraints. The three-dimensional distribution field data is superimposed and rendered in the form of a three-dimensional cloud map on a three-dimensional spatial geometric model, and the material replacement scheme is presented.

9. The method according to claim 8, characterized in that, The training of the physical information neural network proxy model includes: The true distribution data of the concentration field obtained by computational fluid dynamics simulation under multiple sets of standard indoor space geometric layout conditions were used as training datasets. Construct a multi-layer fully connected neural network as a proxy model; Using the convection-diffusion equation as a soft constraint, the partial derivatives of the surrogate model output with respect to coordinates and time are calculated by automatic differentiation to construct the physical residual loss term; The mean squared error between the predicted values ​​and the true values ​​of the proxy model at the sampling points of the training dataset is calculated as the data fitting loss term. The total loss function is obtained by weighting and summing the data fitting loss term, physical residual loss term, and boundary condition loss term according to preset weight coefficients, and then using a gradient descent optimizer to iteratively update the network parameters until the total loss function converges.

10. The method according to claim 8, characterized in that, The steps for calculating material replacement options include: The replacement schemes for each decorative material in the space to be optimized are encoded as chromosome individuals and an initial population is generated. Calculate the objective function values ​​for pollutant exceedance, aesthetic color deviation, and engineering budget deviation for each individual chromosome. Individuals in the population are classified according to non-dominated ordering and crowding distance, and offspring populations are generated through selection, crossover, and mutation operations. The parent and offspring populations are merged, reordered, and individuals are selected to form the next generation population. When the number of population iterations reaches the preset maximum number of iterations and / or the Pareto solution set remains unchanged within a preset number of consecutive generations, the Pareto solution set is output, and a material replacement scheme is selected from the Pareto solution set according to the preset target preference weight.

Images