Material ratio and optimization method and system for 3D printing

Abstract

This invention discloses a method and system for material formulation and optimization in 3D printing, belonging to the field of 3D printing technology. Specifically, it includes: constructing an intelligent material information database; utilizing machine learning to achieve dynamic material classification and automatic identification of low-dimensional feature regions in the 3D model of the product; converting project requirement text into structured data, combining it with a deep learning prediction model to recommend material categories and formulations, and verifying this using simulation technology; printing simulated samples of low-dimensional feature regions for testing, and evaluating sample matching using image recognition technology; based on the comprehensive simulation results, iteratively optimizing the material formulation using an adaptive multi-objective optimization algorithm to obtain the optimal solution; performing overall 3D printing based on the optimized formulation, adjusting printing parameters to adapt to low-dimensional feature requirements, and performing post-processing and performance evaluation after product completion, thereby improving the accuracy and efficiency of material selection and printed products.

Description

Technical Field

[0001] The present invention belongs to the technical field of 3D printing, and in particular to a material proportioning and optimization method and system for 3D printing. Background Art

[0002] 3D printing technology, also known as additive manufacturing technology, is a manufacturing technology that builds three-dimensional objects by stacking or adding materials layer by layer. With the rapid development of science and technology, 3D printing technology, as a cutting-edge manufacturing technology, has demonstrated its unique advantages and application potential in many fields, from aerospace, automobile manufacturing to medical, construction and daily consumer goods production. Compared with traditional manufacturing methods of cutting or removing materials, 3D printing has higher design freedom and material utilization. However, the efficiency of 3D printing technology and the quality of finished products depend to a large extent on the materials used and their ratio and optimization methods.

[0003] The selection of materials needs to consider physical properties, chemical properties, printability, and cost. Different application scenarios have different requirements for materials. Therefore, choosing suitable materials and optimizing their ratios are crucial to improving the performance of 3D printed products.

[0004] For example, the patent with publication number CN116766593A discloses an in-situ powder proportioning and mixing system for multi-material 3D printing, including: an actuator, a vibration plate, an angle adjustment device, etc. The actuator is fastened to the vibration plate, and an inner gear ring is provided inside the vibration plate, a pipe clamp I is provided at the bottom, and an adjustment gear is provided inside the inner gear ring; the angle adjustment device includes an angle adjustment plate, a motor and a pipe clamp II, and the central angle of the angle adjustment plate is connected to the central axis of the inner gear ring; the motor output shaft on the angle adjustment plate passes through the angle adjustment plate and is keyed to the axis of the adjustment gear, and the three-way micro-pipeline includes branch pipes A, branch pipes B and branch pipes C, branch pipes A and B are feed ports, branch pipe C is a discharge port, and is connected to the top feed port of the dropper; the discharge port at the bottom of the dropper is directly opposite to the XY workbench. This technical solution realizes the in-situ proportioning and mixed transportation of dual powders, and the powder proportioning is convenient to adjust, the response speed is fast, the transition powder is small, and the uniformity of powder mixing is improved.

[0005] The above existing technologies all have the following problems: the degree of automation and intelligence in the entire 3D printing process is relatively low, and the flexibility in material types and ratios is poor, which reduces the accuracy and efficiency of printed products. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes a material ratio and optimization method and system for 3D printing, constructs an intelligent material information database, and uses machine learning to realize dynamic classification of materials and automatic identification of low-dimensional feature areas of product three-dimensional models; by converting project requirement text into structured data, combining deep learning prediction models to recommend material categories and ratios, and using simulation technology to simulate and verify; printing low-dimensional feature area simulation samples for testing, and using image recognition technology to evaluate the sample matching degree; based on the comprehensive simulation results, using an adaptive multi-objective optimization algorithm to iteratively optimize the material ratio to obtain the best solution; performing overall 3D printing according to the optimized ratio, and adjusting printing parameters to adapt to low-dimensional feature requirements, and performing post-processing and performance evaluation after the product is completed, thereby improving the accuracy and efficiency of material selection and printed products.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] Material ratio and optimization methods for 3D printing, including:

[0009] Step S1: construct an intelligent material information library and use machine learning algorithms to perform dynamic classification. At the same time, construct a three-dimensional model of the product and automatically identify low-dimensional feature areas in the three-dimensional model;

[0010] Step S2: convert the project requirement text into structured data, and load the pre-trained deep learning prediction model. By inputting the structured data, the recommended material categories and candidate material ratio combinations are obtained. At the same time, the candidate material ratio combinations are simulated using simulation software. According to the simulation results, simulated samples of low-dimensional feature areas are printed, and simulated samples under different ratios are tested.

[0011] Step S3: After printing is completed, the printed simulation sample is image recognized by a camera and compared with a preset low-dimensional feature area standard image to evaluate the matching degree of the printed simulation sample;

[0012] Step S4: Based on the comprehensive analysis of the simulation results, an adaptive multi-objective optimization algorithm is used to iteratively optimize the material ratio to obtain the best ratio scheme;

[0013] Step S5: According to the optimal ratio scheme, different materials are mixed in proportion, and sent to the 3D printer for overall printing, and the preset printing parameters are adjusted according to the requirements of the low-dimensional feature area. At the same time, the printed product is post-processed, and the performance of the post-processed product is evaluated;

[0014] The specific steps of step S1 include:

[0015] S1.1: Collect multi-source material data, establish an intelligent material information database using a relational database based on the multi-source material data, and pre-process the data in the intelligent material information database;

[0016] S1.2: Load the pre-trained support vector machine model, use the support vector machine model to dynamically classify the pre-processed smart material information library data, and divide the materials into N categories;

[0017] S1.3: Set the design parameters and constraints of the product, and conduct preliminary screening in the classified material information database to determine the product material;

[0018] S1.4: Use computer-aided design software to construct a three-dimensional model of the product based on design parameters, constraints and product materials, and check and modify the three-dimensional model;

[0019] The specific steps of step S1 also include:

[0020] S1.5: Setting low-dimensional feature region features and slice parameters , h, d, use the slicing algorithm to slice the three-dimensional model and generate M two-dimensional slice images. The slicing algorithm formula is:

[0021] ;

[0022] Among them, M represents the number of two-dimensional slice images, h represents the slice thickness, d represents the slice spacing, and l represents the slice direction of the three-dimensional model. The maximum dimension length on Indicates the starting offset of the slice. Indicates the end offset of the slice, represents the slice overlap, Express Round down, represents the maximum value function;

[0023] S1.6: Perform grayscale transformation on each two-dimensional slice image to generate a slice grayscale image ;

[0024] The specific steps of step S1 also include:

[0025] S1.7: Identify edge information in the slice grayscale image and perform Perform image segmentation to obtain regional slice grayscale image , where r represents the number of segmented regions;

[0026] S1.8: Collect known low-dimensional feature region templates and build a template library. At the same time, extract features from each region slice grayscale image obtained in S1.7 to obtain the region slice grayscale image features. ,in, represents the nth feature in the grayscale image of the rth regional slice, and n represents the number of grayscale image features of different regional slices;

[0027] The specific steps of step S1 also include:

[0028] S1.9: In Each template in the mobile template library is calculated by Similarity between ,according to Value, get each template in The best matching position in , where e represents the number of templates in the template library, Indicates the e-th template in the template library and The similarity is:

[0029] ;

[0030] in, represents the weight factor of the nth feature in the grayscale image of the region slice, Represents the characteristics of the e-th template in the template library, express The model, , represents the adjustment factor;

[0031] S1.10: Set the threshold of the low-dimensional feature region template to ,

[0032] like , then the area is a low-dimensional feature area, and the identified low-dimensional feature area is highlighted.

[0033] Specifically, the specific steps of the adaptive multi-objective optimization algorithm in step S4 include:

[0034] S4.1: Set the size of the wolf pack to L, the number of iterations to R, and the search space, and randomly generate an initial wolf pack in the search space, where each wolf represents a candidate solution for the material ratio;

[0035] S4.2: According to step S3, the evaluation results of each wolf in the initial wolf pack are obtained. According to the simulation results and the evaluation results, the multi-objective function values ​​of the initial wolf pack are calculated. In combination with the current state of the wolf pack and the characteristics of the objective function, the weights of each objective function are adaptively adjusted. The multi-objective function values The formula is:

[0036] f ( x i ) = F ( p ( x i ), C ( x i ), Na ( x i ), Nr ( x i )), i ∈ [ 1 , B ] ;

[0037] in, represents the candidate solution of the i-th material ratio, represents a multi-objective function, represents the evaluation value of the candidate solution of the i-th material ratio in terms of matching degree, represents the cost evaluation value of the candidate solution of the i-th material ratio, represents the durability evaluation value of the candidate solution of the i-th material ratio, represents the evaluation value of the candidate solution of the i-th material ratio in terms of heat resistance, and B represents the number of candidate solutions of the material ratio.

[0038] Specifically, the specific steps of the adaptive multi-objective optimization algorithm in step S4 also include:

[0039] S4.3: Use the improved non-dominated sorting algorithm to sort the wolves in the wolf pack in descending order and assign the wolves to different Pareto fronts;

[0040] S4.4: Select the first wolf from the Pareto frontier as the leader wolf, and predict the optimal step size and search direction based on the machine learning model;

[0041] S4.5: The last L-1 following wolves in the Pareto front search according to the predicted optimal step length and search direction based on the position of the leader wolf and their own experience, and update their own positions while searching;

[0042] S4.6: Perform simulation evaluation on the updated wolf pack and calculate the new multi-objective function value , and reconstruct the non-dominated sorting and Pareto frontier;

[0043] S4.7: Steps S4.4 to S4.6 are repeated until the number of iterations R is reached, and the Pareto optimal solution set is output to obtain the best matching scheme.

[0044] Specifically, the specific steps of S4.3 include:

[0045] S4.31: Create an empty list, set, and frontier,

[0046] S4.32: Divide the wolf pack into q subsets, and for each pair of wolves in each wolf pack subset and , to determine whether there is a dominance relationship,

[0047] If satisfied ,but Dominate , and Add to in the dominance list;

[0048] S4.33: Find the wolf that is not dominated , forming the first Pareto frontier, and setting ,in, represents the tth non-dominated wolf, t represents the number of non-dominated wolves, and level represents The Pareto rank of the medium wolf.

[0049] Specifically, the specific steps of S4.3 also include:

[0050] S4.34: For each wolf in the current frontier , iterate over all the wolves in its dominance list , and for each , check whether it has been assigned a Pareto rank. If not, set its Pareto rank to the rank of the current frontier plus 1 and add it to the new frontier set;

[0051] S4.35: Repeat step S4.34 until no wolves can be added to the new frontier, and add the newly constructed frontier to the list of all frontiers;

[0052] S4.36: Repeat steps S4.34-S4.35 until all wolves are assigned to frontiers, and output all frontiers and the wolves they contain.

[0053] Material ratio and optimization system for 3D printing, including: model building module, material recommendation module, evaluation module, adaptive optimization module and post-processing module;

[0054] The model building module is used for dynamic classification of materials and building of three-dimensional models;

[0055] The material recommendation module is used to recommend material categories and candidate material ratio combinations using a deep learning prediction model;

[0056] The evaluation module is used to capture the simulated sample image after printing through a camera, compare it with the preset low-dimensional feature area standard image, and evaluate the matching degree of the simulated sample;

[0057] The adaptive optimization module is used to iteratively optimize the material ratio using an adaptive multi-objective optimization algorithm according to the simulation results to obtain the best ratio scheme;

[0058] The post-processing module is used to perform 3D printing according to the optimal ratio scheme, and to perform post-processing and performance evaluation on the printed products.

[0059] Specifically, the model building module includes a three-dimensional model building unit, and the three-dimensional model building unit is used to build a three-dimensional model of the product and identify low-dimensional feature areas from the three-dimensional model.

[0060] Specifically, the evaluation module includes a matching evaluation unit, which is equipped with an image recognition algorithm. The image recognition algorithm is used to perform image recognition on the printed simulation sample, and compare the recognized features with a preset low-dimensional feature area standard image to evaluate the matching degree.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. The present invention proposes a material ratio and optimization system for 3D printing, and optimizes and improves the architecture, operation steps and processes. The system has the advantages of simple process, low investment and operation costs, and low production work costs.

[0063] 2. The present invention proposes a material ratio and optimization method for 3D printing. By constructing an intelligent material information database and a deep learning prediction model, it achieves accurate recommendation and optimization of material categories and ratios, and improves the accuracy and efficiency of material selection and ratios.

[0064] 3. The present invention proposes a material ratio and optimization method for 3D printing, which combines simulation, image recognition and multi-objective optimization algorithm to ensure precise control of the printing process and improve the matching, uniformity and overall performance of 3D printed products. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of the material ratio and optimization method for 3D printing of the present invention;

[0066] Figure 2 This is a flow chart showing the principle of the material ratio and optimization method for 3D printing of the present invention;

[0067] Figure 3 A flow chart of identifying low-dimensional feature regions in a three-dimensional model of a material ratio and optimization method for 3D printing according to the present invention;

[0068] Figure 4 A flow chart of the best material ratio scheme for 3D printing and the optimization method of the present invention;

[0069] Figure 5 This is a diagram of the material ratio and optimization system architecture for 3D printing according to the present invention. DETAILED DESCRIPTION

[0070] Example 1

[0071] See also Figure 1 and Figure 2 , an embodiment provided by the present invention: a material ratio and optimization method for 3D printing, applicable to the field of 3D printing of jewelry, comprising the following steps:

[0072] Step S1: construct an intelligent material information library and use machine learning algorithms to perform dynamic classification. At the same time, construct a three-dimensional model of the product and automatically identify low-dimensional feature areas in the three-dimensional model;

[0073] Step S2: convert the project requirement text into structured data, and load the pre-trained deep learning prediction model. By inputting the structured data, the recommended material categories and candidate material ratio combinations are obtained. At the same time, the candidate material ratio combinations are simulated using simulation software. According to the simulation results, simulated samples of low-dimensional feature areas are printed, and simulated samples under different ratios are tested.

[0074] Furthermore, the specific steps of step S2 include:

[0075] S2.1: Encode the information in the project requirement text into XML structured data format and preprocess the structured data;

[0076] S2.2: Input the preprocessed structured data into the deep learning model to perform a prediction operation. At the same time, the deep learning model will output a recommended material category and a candidate material ratio combination. The prediction process of the deep learning model predicting the output result based on the input data is the prior art content in this field, and is not the inventive solution of the present application, and will not be described in detail here.

[0077] S2.3: Configure the simulation environment according to project requirements, input the candidate material ratio combinations obtained in S2.2 into the simulation software, conduct simulation experiments, and record the parameters and result data during the simulation process;

[0078] S2.4: Based on the simulation experiment results, use a 3D printer to print out simulated samples of low-dimensional feature areas, and perform performance tests on the printed simulated samples to verify whether their performance in actual applications meets project requirements. Performance tests include strength tests and corrosion resistance tests.

[0079] Step S3: After printing is completed, the printed simulation sample is image recognized by a camera and compared with a preset low-dimensional feature area standard image to evaluate the matching degree of the printed simulation sample;

[0080] Furthermore, the specific steps of step S3 include:

[0081] S3.1: Use a camera to shoot the printed simulation sample. During the shooting process, adjust the shooting distance and angle according to the pixel resolution and focal length of the camera to ensure that the image can cover the entire simulation sample;

[0082] S3.2: Preprocessing the collected image, including denoising, enhancement, cropping, and scaling operations, and extracting feature information for comparison from the preprocessed image, wherein the feature may be color, texture, or shape, and the feature extraction is based on a traditional image processing algorithm. The feature extraction process based on the traditional image processing algorithm is the prior art content in this field, and is not the inventive solution of the present application, and is not described in detail here;

[0083] S3.3: prepare a preset low-dimensional feature region standard image, and similarly, perform feature extraction with reference to step S3.2;

[0084] S3.4: Compare the features of the image to be identified with the features of the standard image, and calculate the difference by Euclidean distance , where v represents the number of features of the image to be identified, and the calculation formula of the Euclidean distance is the prior art content in this field, which is not the inventive solution of the present application and will not be described in detail here;

[0085] S3.5: Setting thresholds ,

[0086] like , then the matching degree is considered low;

[0087] like , then the matching degree is considered high.

[0088] Step S4: Based on the comprehensive analysis of the simulation results, an adaptive multi-objective optimization algorithm is used to iteratively optimize the material ratio to obtain the best ratio scheme;

[0089] Step S5: According to the optimal ratio scheme, different materials are mixed in proportion and sent to the 3D printer for overall printing. According to the requirements of the low-dimensional feature area, the preset printing parameters are adjusted. At the same time, the printed products are post-processed and the performance of the post-processed products is evaluated.

[0090] Example 2

[0091] See also Figure 3 and Figure 4 In this embodiment, the specific steps of step S1 include:

[0092] S1.1: Collect multi-source material data, establish an intelligent material information database using a relational database based on the multi-source material data, and pre-process the data in the intelligent material information database;

[0093] Among them, the methods of obtaining multi-source material data include suppliers, scientific research institutions, industry databases, research literature and patent databases; material data include: physical properties, chemical composition, mechanical properties, cost information, processing performance and environmental friendliness of materials; preprocessing includes: data cleaning, data integration, data conversion and data normalization / standardization.

[0094] Furthermore, the process of establishing an intelligent material information database using a relational database includes:

[0095] (1) Selecting a database system: In the present invention, the MySQL relational database system is selected based on factors such as data volume, concurrent access volume, data security, and maintainability;

[0096] (2) Database model: Design the database table structure, field types, indexes, and constraints based on data requirements and the structure of integrated material data to ensure that the database model can accurately reflect the relationships and hierarchical structures between data;

[0097] (3) Create database and table: Create database and table in the selected database system and fill in initial data according to the database schema;

[0098] (4) Optimize database performance: Optimize database performance and query efficiency by designing indexes, partitioning tables, and adjusting query statements.

[0099] S1.2: Load the pre-trained support vector machine model, use the support vector machine model to dynamically classify the pre-processed smart material information library data, and divide the materials into N categories, wherein the training process of the support vector machine model and the data classification process of the support vector machine model are the prior art content in this field, and are not the inventive solution of the present application, and are not described in detail here;

[0100] It should be noted that material categories include but are not limited to: metals, plastics, ceramics, and composite materials.

[0101] S1.3: Set the design parameters and constraints of the product, and conduct preliminary screening in the classified material information database to determine the product material;

[0102] It should be noted that before building a three-dimensional model of a product, it is necessary to clarify the design parameters of the product, such as size, weight, strength, corrosion resistance, conductivity, and design constraints, such as cost constraints, processing difficulty, and environmental protection requirements.

[0103] S1.4: Use computer-aided design software to construct a three-dimensional model of the product based on design parameters, constraints and product materials, and check and modify the three-dimensional model;

[0104] Furthermore, the specific process of constructing the three-dimensional model of the product includes:

[0105] (1) Start the software and create a new project: Open the computer-aided design software and create a new project file;

[0106] (2) Set up the working environment: adjust the software's view, unit, and precision settings as needed;

[0107] (3) Constructing the basic shape: Use the cube provided by the software to construct the basic shape of the product according to the design parameters;

[0108] (4) Add features and details: Add features and details to the basic shape through stretching, rotation, mirroring, and chamfering operations to meet design requirements;

[0109] (5) Specify material: Specify the selected product material in the model;

[0110] (6) Use the software’s inspection tools to conduct a comprehensive inspection of the 3D model to find and repair possible errors, such as geometric conflicts, dimensional errors, and unclosed holes;

[0111] (7) Simulation analysis: According to the design requirements and constraints, the model is simulated and analyzed to evaluate the performance of the model;

[0112] (8) Correction and optimization: Based on the results of simulation analysis, the 3D model is corrected and optimized to ensure that it meets the design requirements.

[0113] S1.5: Setting low-dimensional feature region features and slice parameters , h, d, use the slicing algorithm to slice the three-dimensional model and generate M two-dimensional slice images. The slicing algorithm formula is:

[0114] ;

[0115] Among them, M represents the number of two-dimensional slice images, h represents the slice thickness, d represents the slice spacing, and l represents the slice direction of the three-dimensional model. The maximum length of the dimension on the z-axis is selected in the invention, Indicates the starting offset of the slice. Indicates the end offset of the slice, represents the slice overlap, Express Round down, represents the maximum value function;

[0116] It should be noted that, based on the prior art, the present invention clarifies the slicing direction, takes into account the starting offset and the ending offset, and introduces the concept of slice overlap, which can more accurately reflect the distance between adjacent slices in the actual slicing process, and can calculate the actual number of valid slices when considering slice overlap, thereby increasing the flexibility and applicability of the formula to adapt to different slicing scenarios and requirements.

[0117] S1.6: Perform grayscale transformation on each two-dimensional slice image to generate a slice grayscale image ;

[0118] S1.7: Identify edge information in the slice grayscale image and perform Perform image segmentation to obtain regional slice grayscale image , where r represents the number of segmented regions;

[0119] In the present invention, an edge detection algorithm of a first-order differential operator is used to identify edge information in a slice grayscale image. The first-order differential operator includes a Roberts operator, a Prewitt operator, and a Sobel operator. The edge detection algorithm of the first-order differential operator is the prior art content in the field, is not an inventive solution of the present application, and is not described here.

[0120] S1.8: Collect known low-dimensional feature region templates and build a template library. At the same time, extract features from each region slice grayscale image obtained in S1.7 to obtain the region slice grayscale image features. ,in, represents the nth feature in the grayscale image of the rth regional slice, and n represents the number of grayscale image features of different regional slices. In the present invention, a feature extraction method of principal component analysis is adopted, and the feature extraction process of principal component analysis is the prior art content in this field, which is not the inventive solution of the present application and will not be described in detail here;

[0121] It should be noted that in the present invention, low-dimensional feature regions and high-dimensional feature regions are distinguished based on the shape, texture, color, and position characteristics of the product. If it is a low-dimensional feature region, it means that the shape is single, the texture is simple, the color is relatively uniform, or it is at the edge position.

[0122] S1.9: In Each template in the mobile template library is calculated by Similarity between ,according to Value, get each template in The best matching position in , where e represents the number of templates in the template library, Indicates the e-th template in the template library and The similarity is:

[0123] ;

[0124] in, represents the weight factor of the nth feature in the grayscale image of the region slice, Represents the characteristics of the e-th template in the template library, express The model, , represents the adjustment factor;

[0125] It should be noted that, based on the prior art, the present invention introduces weights , adjustment factor , , which makes the similarity calculation more flexible and can adapt to different application scenarios. The use of makes it possible to take into account the importance of features in different dimensions when calculating similarity, which helps to highlight the impact of key features on similarity.

[0126] S1.10: Set the threshold of the low-dimensional feature region template to ,

[0127] like , then the area is a low-dimensional feature area, and the identified low-dimensional feature area is highlighted.

[0128] The specific steps of the adaptive multi-objective optimization algorithm in step S4 include:

[0129] S4.1: Set the size of the wolf pack to L, the number of iterations to R, and the search space, and randomly generate an initial wolf pack in the search space, where each wolf represents a candidate solution for the material ratio;

[0130] S4.2: According to step S3, the evaluation results of each wolf in the initial wolf pack are obtained. According to the simulation results and the evaluation results, the multi-objective function values ​​of the initial wolf pack are calculated. In combination with the current state of the wolf pack and the characteristics of the objective function, the weights of each objective function are adaptively adjusted. The multi-objective function values The formula is:

[0131] f ( x i ) = F ( p ( x i ), C ( x i ), Na ( x i ), Nr ( x i )), i ∈ [ 1 , B ] ;

[0132] in, represents the candidate solution of the i-th material ratio, represents a multi-objective function, represents the evaluation value of the candidate solution of the i-th material ratio in terms of matching degree, represents the cost evaluation value of the candidate solution of the i-th material ratio, represents the durability evaluation value of the candidate solution of the i-th material ratio, represents the evaluation value of the candidate solution of the i-th material ratio in terms of heat resistance, and B represents the number of candidate solutions of the material ratio;

[0133] S4.3: Use the improved non-dominated sorting algorithm to sort the wolves in the wolf pack in descending order and assign the wolves to different Pareto fronts;

[0134] S4.4: Select the first wolf from the Pareto frontier as the leader wolf, and predict the optimal step length and search direction according to the machine learning model, wherein the implementation process of the machine learning model for prediction after inputting the data is the prior art content in this field, and is not the inventive solution of the present application, and will not be described in detail here;

[0135] S4.5: The last L-1 following wolves in the Pareto front search according to the predicted optimal step length and search direction based on the position of the leader wolf and their own experience, and update their own positions while searching. The position update formula is:

[0136] ;

[0137] in, represents the updated position, and u represents the position before the update. represents the optimal step length, Indicates the position of the leader wolf, Indicates the current position of the wolf. represents random disturbance;

[0138] S4.6: Perform simulation evaluation on the updated wolf pack and calculate the new multi-objective function value , and reconstruct the non-dominated sorting and Pareto frontier;

[0139] S4.7: Steps S4.4 to S4.6 are repeated until the number of iterations R is reached, and the Pareto optimal solution set is output to obtain the best matching scheme.

[0140] The specific steps of S4.3 include:

[0141] S4.31: Create an empty list, set, and frontier,

[0142] S4.32: Divide the wolf pack into q subsets, and for each pair of wolves in each wolf pack subset and , to determine whether there is a dominance relationship,

[0143] If satisfied ,but Dominate , and Add to in the dominance list;

[0144] S4.33: Find the wolf that is not dominated , forming the first Pareto frontier, and setting ,in, represents the tth non-dominated wolf, t represents the number of non-dominated wolves, and level represents Pareto rank of the medium wolf;

[0145] S4.34: For each wolf in the current frontier , iterate over all the wolves in its dominance list , for each , check whether it has been assigned a Pareto rank. If not, set its Pareto rank to the rank of the current frontier plus 1 and add it to the new frontier set;

[0146] S4.35: Repeat step S4.34 until no wolves can be added to the new frontier, and add the newly constructed frontier to the list of all frontiers;

[0147] S4.36: Repeat steps S4.34-S4.35 until all wolves are assigned to frontiers, and output all frontiers and the wolves they contain.

[0148] Example 3

[0149] See also Figure 5 Another embodiment provided by the present invention is a material ratio and optimization system for 3D printing, comprising:

[0150] Model building module, material recommendation module, evaluation module, adaptive optimization module and post-processing module;

[0151] Model building module, which is used to dynamically classify materials and build three-dimensional models using machine learning algorithms;

[0152] The material recommendation module is used to recommend material categories and candidate material ratio combinations using a deep learning prediction model;

[0153] An evaluation module is used to capture the simulated sample image after printing through a camera, compare it with a preset low-dimensional feature area standard image, and evaluate the matching degree of the simulated sample;

[0154] The adaptive optimization module is used to iteratively optimize the material ratio using an adaptive multi-objective optimization algorithm based on the simulation results to obtain the best ratio solution;

[0155] The post-processing module is used to perform 3D printing according to the optimal ratio scheme, adjust the printing parameters to meet the needs of the low-dimensional feature area, and perform post-processing and performance evaluation on the printed products.

[0156] The model building module includes: an information base management unit, a classification unit and a three-dimensional model building unit;

[0157] Information library management unit, used for inputting, updating and maintaining smart material information;

[0158] Classification unit, which is used to classify materials using machine learning algorithms to improve the efficiency and accuracy of material selection;

[0159] The three-dimensional model building unit is used to automatically build a three-dimensional model according to the product design using computer-aided design software and identify low-dimensional feature areas therein.

[0160] The material recommendation module includes: demand analysis unit, material recommendation unit and simulation verification unit;

[0161] Requirements parsing unit, used to convert unstructured project requirements text into structured data that can be processed by computers;

[0162] The material recommendation unit is used to load the pre-trained model and recommend appropriate material categories and ratio combinations based on structured data;

[0163] The simulation verification unit is used to simulate the candidate material ratio using simulation software, evaluate its performance, and print simulated samples of low-dimensional feature areas.

[0164] The evaluation module includes: an image acquisition unit, an image recognition unit and a matching degree evaluation unit;

[0165] An image acquisition unit, used for capturing an image of a simulated sample using a camera;

[0166] An image recognition unit, used for processing and feature extraction of the collected images;

[0167] The matching evaluation unit is used to compare the identified features with the preset standard image to evaluate the matching degree.

[0168] The adaptive optimization module includes: a data integration unit, an optimization algorithm unit and an optimal ratio determination unit;

[0169] Data integration unit, used to integrate simulation results and related data to provide input for optimization algorithms;

[0170] Optimization algorithm unit, used to iteratively optimize material ratio using adaptive multi-objective optimization algorithm;

[0171] The optimal ratio determination unit is used to determine the optimal material ratio scheme according to the optimization result.

[0172] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. A material ratio and optimization method for 3D printing, characterized in that: include: Step S1: construct an intelligent material information library and use machine learning algorithms to perform dynamic classification. At the same time, construct a three-dimensional model of the product and automatically identify low-dimensional feature areas in the three-dimensional model; Step S2: convert the project requirement text into structured data, and load the pre-trained deep learning prediction model. By inputting the structured data, the recommended material categories and candidate material ratio combinations are obtained. At the same time, the candidate material ratio combinations are simulated using simulation software. According to the simulation results, simulated samples of low-dimensional feature areas are printed, and simulated samples under different ratios are tested. Step S3: After printing is completed, the printed simulation sample is image recognized by a camera and compared with a preset low-dimensional feature area standard image to evaluate the matching degree of the printed simulation sample; Step S4: Based on the comprehensive analysis of the simulation results, an adaptive multi-objective optimization algorithm is used to iteratively optimize the material ratio to obtain the best ratio scheme; Step S5: According to the optimal ratio scheme, different materials are mixed in proportion, and sent to the 3D printer for overall printing, and the preset printing parameters are adjusted according to the requirements of the low-dimensional feature area. At the same time, the printed product is post-processed, and the performance of the post-processed product is evaluated; The specific steps of step S1 include: S1.1: Collect multi-source material data, establish an intelligent material information database using a relational database based on the multi-source material data, and pre-process the data in the intelligent material information database; S1.2: Load the pre-trained support vector machine model, use the support vector machine model to dynamically classify the pre-processed smart material information library data, and divide the materials into N categories; S1.3: Set the design parameters and constraints of the product, and conduct preliminary screening in the classified material information database to determine the product material; S1.4: Use computer-aided design software to construct a three-dimensional model of the product based on design parameters, constraints and product materials, and check and modify the three-dimensional model; The specific steps of step S1 also include: S1.5: Setting low-dimensional feature region features and slice parameters , h, d, use the slicing algorithm to slice the three-dimensional model and generate M two-dimensional slice images. The slicing algorithm formula is: ； Among them, M represents the number of two-dimensional slice images, h represents the slice thickness, d represents the slice spacing, and l represents the slice direction of the three-dimensional model. The maximum dimension length on Indicates the starting offset of the slice. Indicates the end offset of the slice, represents the slice overlap, Express Round down, represents the maximum value function; S1.6: Perform grayscale transformation on each two-dimensional slice image to generate a slice grayscale image ; The specific steps of step S1 also include: S1.7: Identify edge information in the slice grayscale image and perform Perform image segmentation to obtain regional slice grayscale image , where r represents the number of segmented regions; S1.8: Collect known low-dimensional feature region templates and build a template library. At the same time, extract features from each region slice grayscale image obtained in S1.7 to obtain the region slice grayscale image features. ,in, represents the nth feature in the grayscale image of the rth regional slice, and n represents the number of grayscale image features of different regional slices; The specific steps of step S1 also include: S1.9: In Each template in the mobile template library is calculated by The similarity between ,according to Value, get each template in The best matching position in , where e represents the number of templates in the template library, Indicates the eth template in the template library and The similarity is: ； in, represents the weight factor of the nth feature in the grayscale image of the region slice, Represents the characteristics of the e-th template in the template library, express The model, , represents the adjustment factor; S1.10: Set the threshold of the low-dimensional feature region template to , like , then the area is a low-dimensional feature area, and the identified low-dimensional feature area is highlighted.

2. The material ratio and optimization method for 3D printing according to claim 1, characterized in that: The specific steps of the adaptive multi-objective optimization algorithm in step S4 include: S4.1: Set the size of the wolf pack to L, the number of iterations to R, and the search space, and randomly generate an initial wolf pack in the search space, where each wolf represents a candidate solution for the material ratio; S4.2: According to step S3, the evaluation results of each wolf in the initial wolf pack are obtained. According to the simulation results and the evaluation results, the multi-objective function values ​​of the initial wolf pack are calculated. In combination with the current state of the wolf pack and the characteristics of the objective function, the weights of each objective function are adaptively adjusted. The multi-objective function values The formula is: ； in, represents the candidate solution of the i-th material ratio, represents a multi-objective function, represents the evaluation value of the candidate solution of the i-th material ratio in terms of matching degree, represents the cost evaluation value of the candidate solution of the i-th material ratio, represents the durability evaluation value of the candidate solution of the i-th material ratio, represents the evaluation value of the candidate solution of the i-th material ratio in terms of heat resistance, and B represents the number of candidate solutions of the material ratio.

3. The material ratio and optimization method for 3D printing according to claim 2, characterized in that: The specific steps of the adaptive multi-objective optimization algorithm in step S4 also include: S4.3: Use the improved non-dominated sorting algorithm to sort the wolves in the wolf pack in descending order and assign the wolves to different Pareto fronts; S4.4: Select the first wolf from the Pareto frontier as the leader wolf, and predict the optimal step size and search direction based on the machine learning model; S4.5: The last L-1 following wolves in the Pareto front search according to the predicted optimal step length and search direction based on the position of the leader wolf and their own experience, and update their own positions while searching; S4.6: Perform simulation evaluation on the updated wolf pack and calculate the new multi-objective function value , and reconstruct the non-dominated sorting and Pareto frontier; S4.7: Steps S4.4 to S4.6 are repeated until the number of iterations R is reached, and the Pareto optimal solution set is output to obtain the best matching scheme.

4. The material ratio and optimization method for 3D printing according to claim 3, characterized in that: The specific steps of S4.3 include: S4.31: Create an empty list, set, and frontier, S4.32: Divide the wolf pack into q subsets, and for each pair of wolves in each wolf pack subset and , to determine whether there is a dominance relationship, If satisfied ,but Dominate , and Add to in the dominance list; S4.33: Find the wolf that is not dominated , forming the first Pareto frontier, and setting ,in, represents the tth non-dominated wolf, t represents the number of non-dominated wolves, and level represents The Pareto rank of the medium wolf.

5. The material ratio and optimization method for 3D printing according to claim 4, characterized in that: The specific steps of S4.3 also include: S4.34: For each wolf in the current frontier , iterate over all the wolves in its dominance list , for each , check whether it has been assigned a Pareto rank. If not, set its Pareto rank to the rank of the current frontier plus 1 and add it to the new frontier set; S4.35: Repeat step S4.34 until no wolves can be added to the new frontier, and add the newly constructed frontier to the list of all frontiers; S4.36: Repeat steps S4.34-S4.35 until all wolves are assigned to frontiers, and output all frontiers and the wolves they contain.

6. A material proportioning and optimization system for 3D printing, which is used to implement the material proportioning and optimization method for 3D printing according to any one of claims 1 to 5, characterized in that: include: Model building module, material recommendation module, evaluation module, adaptive optimization module and post-processing module; The model building module is used for dynamic classification of materials and building of three-dimensional models; The material recommendation module is used to recommend material categories and candidate material ratio combinations using a deep learning prediction model; The evaluation module is used to capture the simulated sample image after printing through a camera, compare it with the preset low-dimensional feature area standard image, and evaluate the matching degree of the simulated sample; The adaptive optimization module is used to iteratively optimize the material ratio using an adaptive multi-objective optimization algorithm according to the simulation results to obtain the best ratio scheme; The post-processing module is used to perform 3D printing according to the optimal ratio scheme, and to perform post-processing and performance evaluation on the printed products.

7. The material ratio and optimization system for 3D printing according to claim 6, characterized in that: The model building module includes a three-dimensional model building unit, which is used to build a three-dimensional model of the product and identify low-dimensional feature areas from the three-dimensional model.

8. The material proportioning and optimization system for 3D printing according to claim 7, characterized in that: The evaluation module includes a matching evaluation unit, in which an image recognition algorithm is configured. The image recognition algorithm is used to perform image recognition on the printed simulation sample, and compare the recognized features with a preset low-dimensional feature area standard image to evaluate the matching degree.

Images