3D printing self-adaptive wing design method based on generative artificial intelligence

Abstract

The invention discloses a 3D printing self-adaptive wing design method based on generative artificial intelligence, and relates to the crossing field of 3D printing and artificial intelligence. In order to solve the technical problems of complex manufacturing, low reliability and large weight of a traditional process in the prior art, the technical scheme provided by the invention comprises the following steps: pre-training a large language model to obtain a trained basic model; receiving self-adaptive wing structure design parameters input by a user and generating a design scheme script; calling three-dimensional modeling software, automatically establishing a wing three-dimensional model based on the design scheme script, and outputting a model file; calling simulation analysis software, performing wing performance simulation analysis based on the model file, and outputting analysis result data; and calling the basic model to carry out design scheme evaluation and iterative optimization based on the analysis result data until an optimized model file meeting the design requirement is generated. The method can be applied to intelligent design and automatic manufacturing of self-adaptive wing structures of aerospace crafts.

Description

Technical Field

[0001] It relates to the cross - field of 3D printing and artificial intelligence, specifically to the 3D printing adaptive wing design based on generative artificial intelligence. Background Art

[0002] In recent years, with the rapid development of the aerospace field, the requirements for the performance of aircraft have gradually increased. Adaptive wing technology has attracted much attention because it can dynamically change the wing shape according to flight conditions during flight. Currently, adaptive wings usually use complex mechanical structures to achieve the deformation function, and each component needs to be manufactured separately and then assembled into the overall wing. This traditional manufacturing method has the following deficiencies:

[0003] After each component is manufactured separately and then assembled, the production cycle is long, it is difficult to guarantee the assembly accuracy, and the labor cost is high;

[0004] The connection structure between components is complex, and it is easy to cause wear under long - term dynamic deformation loads, resulting in structural fatigue, reducing the service life and reliability;

[0005] The complex mechanical drive structure increases the weight of the wing, which is not conducive to the lightweight design of the aircraft.

[0006] To solve the above problems, some 3D printing solutions based on lattice structures have emerged in recent years. By filling lattice structures with different densities and strengths in different parts of the wing, a gradient distribution of strength and flexibility is achieved, thus taking into account both structural strength and deformation performance. However, the introduction of lattice structures significantly increases the design complexity, the prediction of the mechanical properties of different lattice structures is difficult, and the traditional "design - simulation - iterative optimization - production" process is long and costly. At the same time, it has high requirements for the professional skills of designers and is difficult to be quickly popularized and applied.

[0007] Although artificial intelligence, especially machine learning methods, have been initially applied in the field of structural design in recent years, there are still problems such as large data requirements, poor model generalization ability, and disconnection between model output and simulation toolchains, which seriously limit the application scenarios and design efficiency of automated design methods.

[0008] In summary, there is an urgent need for a more efficient and intelligent adaptive wing design and manufacturing method to solve the technical bottlenecks of complex manufacturing, low reliability, and large weight existing in traditional processes, and to promote the rapid development and engineering application of adaptive wing technology. Summary of the Invention

[0009] To solve the technical problem that in the prior art, there is an urgent need for a more efficient and intelligent adaptive wing design and manufacturing method to solve the technical bottlenecks of complex manufacturing, low reliability, and large weight existing in traditional processes, the technical solution provided by the present invention is as follows:

[0010] A 3D printing adaptive wing design method based on generative artificial intelligence, comprising:

[0011] The step of pre-training a large language model to obtain a trained basic model;

[0012] The step of deploying the basic model to a data integration management platform, receiving the adaptive wing structure design parameters input by the user and generating a design scheme script;

[0013] The step of calling 3D modeling software to automatically establish a 3D model of the wing based on the design scheme script and output a model file;

[0014] The step of calling simulation analysis software to perform wing performance simulation analysis based on the model file and output analysis result data;

[0015] The step of calling the basic model to perform design scheme evaluation and iterative optimization based on the analysis result data until an optimized model file meeting the design requirements is generated.

[0016] Furthermore, a preferred implementation is provided. The pre-training of the large language model is carried out in a manner combining Few-Shot Learning and reinforcement learning with human feedback, and training prompt words are respectively set for the wing main structure, deformation components, wing tips, and shape memory alloy wires.

[0017] Furthermore, a preferred implementation is provided. The data integration management platform centrally classifies and stores the generated model files, simulation analysis result data, and design scheme scripts, and provides a data visualization function to facilitate users to view, edit, and call the data.

[0018] Furthermore, a preferred implementation is provided. The 3D modeling software uses SolidWorks software to automatically generate 3D models of the wing main structure, deformation components, wing tips, and internal lattice filling structures by running the design scheme script, and outputs them in the.formats of.step files and.stl files.

[0019] Based on the same inventive concept, the present invention also provides a 3D printing adaptive wing design device based on generative artificial intelligence, comprising:

[0020] A module for pre-training a large language model to obtain a trained basic model;

[0021] A module for deploying the basic model to a data integration management platform, receiving the adaptive wing structure design parameters input by the user and generating a design scheme script;

[0022] A module that calls 3D modeling software to automatically establish a 3D model of the wing based on the design scheme script and outputs a model file;

[0023] A module that calls simulation analysis software to perform wing performance simulation analysis based on the model file and outputs analysis result data;

[0024] A module that calls the basic model to evaluate and iteratively optimize the design scheme based on the analysis result data until an optimized model file meeting the design requirements is generated.

[0025] Based on the same inventive concept, the present invention also provides a 3D printing adaptive wing printing method based on generative artificial intelligence, including:

[0026] A step of calling an automated 3D printing module to automatically set 3D printing parameters based on the optimized model file and output a slice file for 3D printing.

[0027] Based on the same inventive concept, the present invention also provides a 3D printing adaptive wing printing device based on generative artificial intelligence, including:

[0028] A module that calls an automated 3D printing module to automatically set 3D printing parameters based on the optimized model file and output a slice file for 3D printing.

[0029] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computing program, and when the computer program is read by a computer, the computer executes the method.

[0030] Based on the same inventive concept, the present invention also provides a computer including a processor and a storage medium, and when the processor reads the computer program stored in the storage medium, the computer executes the method.

[0031] Based on the same inventive concept, the present invention also provides a computer program product, which, as a computer program, realizes the method when the computer program is executed.

[0032] Compared with the prior art, the beneficial effects of the technical solution provided by the present invention are as follows:

[0033] Through the solution generation module, the present invention uses a large language model to automatically generate Python code and scripts, realizing the whole process from user parameter input to automatic generation of the structural scheme, effectively shortening the design cycle in the traditional "design - simulation - iterative optimization - production" mode. In the prior art, based on manual design and traditional optimization algorithms, a large amount of manual intervention and repeated modification are required, and the automation degree of this solution significantly improves the design efficiency.

[0034] The present invention realizes the centralized management of model data, scripts, and simulation results by constructing an integrated management platform, ensuring the stability and accuracy of data transmission between different functional modules. The traditional technical solutions lack a unified data management platform, and the data transfer between modules relies on manual intervention, which is prone to data inconsistency and loss problems. Therefore, this solution improves the reliability of data interaction.

[0035] The present invention realizes an automated iterative optimization process through the simulation and scheme verification modules. The simulation analysis results are visualized and then handed over to the large language model for autonomous evaluation and feedback of optimization suggestions, thereby realizing the precise and rapid iteration of design schemes. Compared with the traditional simulation verification and manual analysis mode, this solution significantly improves the accuracy and iteration speed of design optimization and reduces the dependence on the experience of designers.

[0036] The present invention calls the large language model through the automated 3D printing module to automatically complete the optimization settings of printing parameters, including material selection, printing speed, and layer thickness setting, reducing the uncertainty and process fluctuation risks brought by manual parameter adjustment in the traditional process. Compared with the existing method of manually setting parameters, this solution significantly improves the stability and consistency of 3D printing quality.

[0037] The present invention optimizes the large language model by using a hybrid training method of "Few-Shot Learning + RLHF", realizing the improvement of the model's fast migration and generalization capabilities through few-shot data, and overcoming the bottleneck of the large data demand of traditional machine learning models. Compared with the traditional machine learning methods based on large amounts of data training, this solution significantly reduces the difficulty and cost of model deployment and improves the flexibility and economy of model applications.

[0038] The present invention uses a modular and distributed deployment method to organically integrate modeling software, finite element analysis software, computational fluid dynamics analysis software, and 3D printing software into a unified platform, realizing the collaborative and efficient operation between platforms. In the existing technology, each module usually runs independently and it is difficult to achieve seamless connection. This solution effectively improves the efficiency of multi-platform collaborative work and the overall production speed.

[0039] During the design process of the present invention, through refined modular design prompts such as the wing body, deformation components, wing tips, and shape memory alloy wire selection, the refined control and rapid generation of complex wing structures are realized, solving problems such as the difficult prediction and large design difficulty of complex lattice structures in existing adaptive wing designs, and effectively improving the reliability and feasibility of wing structure designs.

[0040] It can be applied to the intelligent design and automated manufacturing of adaptive wing structures of aerospace aircraft. Brief Description of the Drawings

[0041] Figure 1 Flowchart of the adaptive wing design method for 3D printing based on generative artificial intelligence;

[0042] Figure 2 Schematic diagram of the main interface of the integrated management platform system;

[0043] Figure 3 Schematic diagram of the interface for generating solutions display and iterative optimization;

[0044] Figure 4 Schematic diagram of the solution visualization interface;

[0045] Figure 5 Schematic diagram of the 3D printing parameter and model generation interface. Specific implementation mode

[0046] To make the advantages and benefits of the technical solution provided by the present invention more clearly reflected, the technical solution provided by the present invention will now be further described in detail with reference to the accompanying drawings. Specifically:

[0047] Embodiment 1. This embodiment provides an adaptive wing design method for 3D printing based on generative artificial intelligence, including:

[0048] The step of pre-training the large language model to obtain a trained basic model;

[0049] The step of deploying the basic model to the data integration management platform, receiving the adaptive wing structure design parameters input by the user, and generating a design scheme script;

[0050] The step of calling 3D modeling software to automatically establish a three-dimensional wing model based on the design scheme script and output a model file;

[0051] The step of calling simulation analysis software to perform wing performance simulation analysis based on the model file and output analysis result data;

[0052] The step of calling the basic model to perform design scheme evaluation and iterative optimization based on the analysis result data until an optimized model file meeting the design requirements is generated.

[0053] The pre-training of the large language model is carried out in a way that combines Few-Shot Learning and reinforcement learning with human feedback, and training prompt words are set for the wing main structure, deformation components, wing tips, and shape memory alloy wires respectively.

[0054] The data integration management platform centrally classifies and stores the generated model files, simulation analysis result data, and design scheme scripts, and provides a data visualization function to facilitate users to view, edit, and call the data.

[0055] The 3D modeling software uses SolidWorks software to automatically generate 3D models of the wing main structure, deformation components, wing tips, and internal lattice filling structures by running the design scheme script, and outputs them in the formats of.step files and.stl files.

[0056] A 3D printing adaptive wing printing method based on generative artificial intelligence is also provided, including:

[0057] Calling an automated 3D printing module to automatically set 3D printing parameters based on the optimized model file and output a sliced file for 3D printing.

[0058] Embodiment 2. This embodiment is a detailed description of the technical solution provided in Embodiment 1. Specifically:

[0060] The present invention proposes a 3D printing adaptive wing design method and system based on generative artificial intelligence. The specific implementation process is as follows:

[0061] First, briefly describe as follows:

[0062] (1) Select and pre-train a large language model suitable for adaptive wing design;

[0063] (2) Build a data integration management platform to achieve data integration, management, and invocation of the model and software;

[0064] (3) Construct functional units including a scheme design, simulation analysis, and 3D printing module, and integrate them into a unified platform;

[0065] (4) Use the pre-trained large language model to automatically generate a design scheme based on the design requirements input by the user;

[0066] (5) Perform simulation and iterative optimization on the generated design scheme;

[0067] (6) Finally, determine the optimized model structure, automatically set 3D printing parameters, and complete 3D printing manufacturing.

[0068] The specific detailed implementation method is as follows:

[0069] Step 1, select and pre-train a large language model suitable for adaptive wing design

[0070] Select large language models suitable for local deployment, such as DeepSeek-R1, Qwen2, or ChatGLM, as the base model for solution generation. Select a model version with an appropriate parameter scale according to the hardware conditions, and use a hybrid training method of "Few-Shot Learning + RLHF" for pre-training. The specific approach is as follows: First, refine the adaptive wing structure into four main components: "wing main structure", "deformable components", "wing tip", and "selection of shape memory alloy wires". Design and generate prompt words separately for different components, and construct corresponding small amounts of example data as the training set. Then, guide the model to generate the target content by designing prompt words, and at the same time construct a reward model. During the training process, continuously guide the large language model to iterate and optimize, so that the model can quickly master the key points of wing design with fewer samples and have generalization ability.

[0071] Step 2: Build a data integration management platform to achieve data integration, management, and invocation between the model and software.

[0072] Build a unified data integration management platform to centrally manage the "model files", "Python script files", and "simulation result files" during the design process, and provide a standardized interface for data invocation. The main interface of the platform has a file management function area, supporting users to view, edit, and invoke the above data files at any time. The platform can achieve integrated invocation of Python compilers, SolidWorks modeling software, ABAQUS finite element analysis software, ANSYS Fluent computational fluid dynamics analysis software, and 3D printing slicing software, facilitating rapid data transmission and processing between modules.

[0073] Step 3: Construct functional units including solution design, simulation analysis, and 3D printing modules, and integrate them into a unified platform.

[0074] Integrate the design system, simulation system, and automated 3D printing system in the form of functional modules through the integrated management platform and perform distributed deployment. Specifically, it includes: The solution generation module is used for the invocation of the large language model and the generation of design scripts; the simulation and solution verification module is used for automatic simulation analysis and solution evaluation of the model; the automated 3D printing module is used for model repair, printing parameter setting, and model slicing. All functional units are interconnected through the platform data interface to ensure data sharing and collaborative work.

[0075] Step 4: Use the pre-trained large language model to automatically generate a design solution based on the design requirements input by the user.

[0076] The user inputs the basic parameters and design requirements of the wing body, deformable components, wing tips, and shape memory alloy wires in the design interface of the integrated management platform, and adds additional design objective constraints in the input box. The platform converts the user's input into design prompts, calls the pre-trained large language model, automatically generates a structural solution that meets the design requirements, and outputs it to the integrated management platform in the form of a Python script to form an executable modeling script (such as Python_Model.py) for the next simulation analysis.

[0077] Step Five: Conduct simulation and iterative optimization on the generated design solution

[0078] In the simulation and solution verification module, the specific implementation is as follows:

[0079] First, use SolidWorks software to perform 3D modeling by calling the Python_Model.py script generated in the previous step, and generate a.step file and a.stl file containing detailed dimension information, and save them to the integrated management platform;

[0080] Next, the platform calls the ABAQUS finite element analysis software, and uses the pre-generated Python_simulation.py script to automatically import the.step file to complete the mechanical simulation analysis. The analysis content includes the bending resistance performance of the wing body structure, the deformation ability of the deformed part, the variable angle of the wing tip, and the driving performance of the shape memory alloy wire, and generates the corresponding simulation result file;

[0081] At the same time, the platform calls the ANSYS Fluent software, and uses a dedicated Python script to automatically perform the aerodynamic performance analysis of the wing, including the simulation of lift, drag, and aerodynamic moment, and generates the aerodynamic analysis result file;

[0082] The simulation results are output to the large language model in the form of visual data, and the large language model independently analyzes and evaluates whether the design solution meets the initial design requirements. If not, the large language model automatically proposes optimization suggestions and conducts the next round of iterative optimization until the simulation results meet the preset design criteria or reach the user's satisfaction, forming the final optimized solution model.

[0083] Step Six: Finally determine the optimized model structure, automatically set the 3D printing parameters, and complete the 3D printing manufacturing

[0084] Import the optimized.stl model file after simulation into the automated 3D printing module, and use the large language model to automatically check and repair the model file to avoid defects such as overhangs, voids, and deformations; then automatically set key parameters including printing material, printing layer thickness, printing speed, etc., call the professional 3D printing slicing software to complete the model slicing and generate the 3D printing file; finally, send the 3D printing file to the online automatic 3D printing device through the integrated management platform to automatically complete the 3D printing manufacturing of the wing structure.

[0085] In summary, the technical solution provided by the present invention can realize the automated intelligent design and manufacturing of the adaptive wing structure, greatly shortening the R & D cycle, improving the design accuracy and manufacturing reliability, and having significant engineering application value.

[0086] Embodiment 3. Combine Figures 1-5 To illustrate this embodiment, this embodiment further describes the above-provided technical solution in detail through specific examples. Specifically:

[0087] The 3D printing adaptive wing design method based on generative artificial intelligence includes:

[0088] Step S1: Select a large language model as the basic model for generating the design scheme and pre-train the model;

[0089] Step S2: Build a data integration management platform to integrate, manage, and call the generated data and scripts as well as the software to be used;

[0090] Step S3: Integrate the modeling software, finite element simulation software, and 3D printing slicing software into the ability units of the management platform, and the corresponding design system, modeling system, simulation system, and 3D printing system are deployed and run in a distributed manner; among them, the design system refers to the system that executes the 3D printing adaptive wing design scheme based on generative artificial intelligence;

[0091] Step S4: Call the large language model through local deployment to generate the 3D printer wing scheme, including the selection of prompt words for different wing parts and the corresponding model generation;

[0092] Step S5: Convert the generated model into a Python script, and use the simulation system to call the generated Python script to achieve automatic calculation. Output the obtained result file to the integrated management platform. The integrated management platform performs data visualization and uploads the data to the large language model for analysis to determine whether the design requirements are met. If the design requirements are not met, iterate and optimize again and output again, and start from the beginning to execute Step S5. This process can be manually intervened to perform manual annotation on the data. Even if the design requirements are met, if the user is not satisfied with the results, the user can still choose to continue iterating and optimizing. If satisfied with the results, choose to generate a 3D printing file.

[0093] Step S6: Import the generated model into the 3D printing system, and call the large language model to complete the setting of 3D printing parameters, including layer thickness, printing speed, material selection, etc. After completing the parameter setting, send the sliced file to the online 3D printer to complete the printing.

[0094] The 3D printing adaptive wing design system based on generative artificial intelligence includes:

[0095] A solution generation module that selects a large language model as the basic model to generate solution designs and optimize the solutions.

[0096] An integrated management platform that completes data integration, management, and invocation by constructing an integrated management platform to improve design efficiency and achieve automation.

[0097] A simulation and solution verification module that encapsulates the simulation software and modeling software through the integrated management platform, and completes the simulation and verification of the solution in the software by calling the generated Python script between them, and outputs the result data to the integrated management platform. The large language model analyzes these result data to confirm whether the design requirements are met and performs iterative optimization.

[0098] (4) An automated 3D printing module that outputs the generated 3D printing model to the integrated management platform. The integrated management platform calls the large language model to complete the setting of 3D printing parameters and material selection, and inputs the generated result file into the online printer to complete the printing.

[0099] The system architecture and implementation method of the present invention

[0100] 1. System architecture:

[0101] The core of the present invention is to generate Python code and scripts through generative artificial intelligence, use the integrated management platform to run and call these codes and scripts to complete the solution design, and achieve modular design. The modules are connected together through data output and invocation.

[0102] Based on the integrated management platform, modular design is carried out, specifically including:

[0103] (1) Solution Generation Module: First, deploy the large language model locally. Optional models include multiple parameter versions such as deepseek-r1, Qwen2, ChatGLM, etc. Use the method of "Few-Shot Learning + RLHF" (that is, embed a small number of examples in the prompt, guide the model to generate the target content by designing the prompt (Prompt), train the reward model, and guide the iterative optimization of the large language model) to pre-train the model. During training, divide the adaptive wing structure into three main components: "wing main structure", "deformable component", and "wing tip", and set different prompts for different structures. When running this module, users can add other constraint conditions on the basis of the prompt to optimize the adaptive wing structure. After this module runs, a Python script named Python_Model is generated, and the modeling can be completed by running the script.

[0104] (2) Integrated Management Platform: This module mainly integrates, manages, and calls the generated data including "model files", "script files", and "result files", and provides corresponding interfaces to call other modules to achieve data transmission between different modules. And the design and optimization of the solution can be completed by running the corresponding modeling software, finite element analysis software, and model slicing software distributedly on the main interface of the integrated management platform.

[0105] (3) Simulation and Solution Verification Module: This module mainly conducts simulation to verify the generated experimental solution. First, the large language model outputs a Python script named Python_simulation, and the simulation is completed by calling the finite element analysis software to run the Python_simulation script. After the simulation, a result file is output. The result file can be visualized through the data visualization function to facilitate users to optimize the solution. The integrated management platform inputs the generated result file into the large language model, and the large language model judges whether it meets the design requirements and conducts iterative optimization, and outputs the optimized solution.

[0106] (4) Automated 3D Printing Module: The integrated management platform inputs the generated model file into this module. This module completes the setting of printing parameters and material selection by calling the large language model, and calls the model slicing software to slice the model. After completion, the user selects to export the model and printing parameters, and sends the printing file to the online 3D printing device to complete the printing.

[0107] 2. Implementation Method:

[0108] (1) Model Deployment and Localized Operation: Based on open-source large language models for local deployment, both running and calling are based on virtual environments to ensure data controllability during operation and stable calling. Ensure that the dependencies between functional modules and the integrated management platform are independent of other systems, facilitating maintenance and management among modules.

[0109] (2) Model Training and Prompt Generation: Use the method of "Few-Shot Learning + RLHF" (i.e., embed a small number of examples in the prompt, guide the model to generate target content by designing the prompt (Prompt), train the reward model, and guide the iterative optimization of the large language model) to pre-train the model. During training, the adaptive wing structure is divided into four main components: "wing main structure", "deformable component", "wing tip", and "shape memory alloy wire selection", and different prompts are set for different structures. Users can input parameters according to the preset prompts and can directly input the design requirements of the scheme, the background of the scheme use, and additional design goals through the dialog box to help the large language model better understand the specific content of the scheme. In addition to the necessary design parameters, users can also customize the wing structure and lattice structure in the modeling requirements dialog box on the main interface, and support users to upload marked scheme files for model training or scheme optimization.

[0110] (3) Scheme Generation: The integrated management platform outputs the design parameters and design requirements input by the user to the large language model. The large language model completes the scheme design based on these parameters and user requirements. After completing the scheme design, it converts the scheme into a Python script that can be directly called by modeling software and simulation software. The integrated management platform saves the generated Python script as an executable file that can be directly called, and then runs the modeling software and simulation software in a distributed manner for result calculation. The modeling software completes automatic modeling by calling and running the Python_Model.py script file, outputs the size and structural features of the model, and outputs the result as a model file; the simulation software completes the simulation of the model by calling and running the Python_simulation.py script file to verify whether the deformation ability and mechanical strength of the adaptive wing meet the strength requirements, and verify whether the lattice structure introduced in the design scheme achieves lightweight design while ensuring performance, and saves the simulation results as result files.

[0111] (4) Data integration and solution evaluation: The generated model files and result files are centrally stored and classified and managed according to the solution version, supporting users to view, edit, and upload data at any time on the main interface of the integration management platform, and allowing different functional modules to call and modify the data. Through the data visualization function, the stored model files and result files are converted into text files and graphic files that can be recognized by the large language model and displayed in the interface. The recognizable files are uploaded to the large language model for analysis and evaluation. If the design requirements are not met, iterative optimization is performed to improve the solution design. After completing the iterative optimization, if the user is still not satisfied with the optimization result, they can choose to continue the optimization iteration.

[0112] (5) 3D printing parameter setting and automatic printing: Call the automated 3D printing module to run the optimized model file to generate the corresponding 3D printing model, call the large language model to complete the setting of 3D printing parameters, use professional slicing software to complete slicing, and output the 3D printing file. After exporting the model and 3D printing file, send them to the online 3D printer to complete the printing.

[0113] Specifically,

[0114] The overall design idea of the present invention: Automatically complete parameter design, modeling, solution verification, iterative optimization, and 3D printing parameter setting according to the design requirements input by the user on the main interface of the integration management platform, help the user complete the solution design of the 3D printed adaptive wing, and perform the solution design according to the user's requirements, which speeds up the R & D cycle and production speed of the 3D printed adaptive wing. The system mainly includes four main parts: the solution generation module, the integration management platform, the simulation and solution verification module, and the automated 3D printing module.

[0115] The solution production module is implemented by calling the locally deployed large language model through a Python script, and converting the text file input by the user into a prompt word for overall solution design. Before this, the model needs to be pre-trained using the method of "Few-Shot Learning + RLHF". The adaptive wing structure is divided into four main components: "wing main structure", "deformation component", "wing tip", and "shape memory alloy wire selection", and different prompt words and training sets are set for different structures. Constraint words such as "output a complete executable code file", "output a directly callable Python script file", and "optimize the code to ensure the robustness of the code" are added to the Python script to ensure the accuracy of the code and script output by the large language model, facilitating the user to directly run the code and script without additional debugging.

[0116] The integrated management platform is responsible for the centralized management, invocation, and editing of data, and serves as the center for data invocation and conversion between different modules, supporting users to run the functional components of each module distributively through the centralized management platform.

[0117] The simulation verification module can automatically run the corresponding tool software, and call the script files and data stored in the integrated management platform to conduct model analysis, mechanical property analysis, lattice parameter analysis, weight analysis, deformation ability analysis, and aerodynamic performance analysis, and send the calculation results back to the integrated management platform for storage as result files. Model analysis is to use the solidwork software to call the Python_Model.py script file generated by the large language model to complete the modeling, and store the model size parameters in the result file of the integrated management platform after parameterization. Mechanical property analysis, lattice parameter analysis, weight analysis, and deformation ability analysis are all completed using the commercial finite element analysis software ABAQUAS. The simulation simulation is completed by calling the Python_simulation.py script file. The main verification contents include (1) the bending resistance performance, structural strength and stiffness of the wing main structure, the anti-deformation ability and weight reduction effect of the filled lattice, (2) the deformation ability and bending resistance strength of the deformed part, the deformation ability and weight reduction effect of the filled lattice (3) the variable angle range, deformation strength, and bending resistance section coefficient of the wing tip part, (4) the driving force, driving strain, and driving current of the shape memory alloy wire. The calculation results are submitted to the result file of the integrated management platform. The aerodynamic performance of the aircraft wing is analyzed using ANSYS Fluent to obtain the aerodynamic parameters of the wing under different flight states, such as lift, drag, aerodynamic moment, etc., and the calculated data obtained is saved as a result file. After the calculation and analysis, the result file is converted into a text file and a graphic file that can be recognized by the large language model for data visualization, and submitted to the large language model for analysis and verification to see if the initial design requirements are met. If the design requirements are not met, iterative optimization is carried out. After meeting the design requirements, if the user wants to continue optimizing the scheme, they can choose to continue iterative optimization.

[0118] After completing the above steps, the user can set the parameters for 3D printing and slice the model, call and run the automated 3D printing module to call the optimized model file to generate the corresponding 3D printing model. The large language model inputs the set printing parameters into this module to complete the parameter setting, uses professional slicing software to complete the model slicing, outputs the 3D printing file to the integrated management platform, and the user sends the printing file to the online automatic 3D printer through the platform to complete the printing.

[0119] As Figure 1 shown: The present invention provides a 3D printing adaptive wing design method based on generative artificial intelligence, including:

[0120] Step S1: Select a large language model as the base model for generating design solutions and pre-train the model;

[0121] Step S2: Build a data integration management platform to integrate, manage, and call the generated data and scripts;

[0122] Step S3: Integrate modeling software, finite element simulation software, and 3D printing slicing software into the capability units of the management platform, corresponding design systems, modeling systems, simulation systems, and 3D printing systems, and deploy and run them in a distributed manner; among them, the design system refers to the system that executes the 3D printing adaptive wing design solution based on generative artificial intelligence;

[0123] Step S4: Call the large language model through local deployment to generate a 3D printer wing solution, including the selection of prompt words for different wing parts and the corresponding model generation, and convert the generated model into a Python script for convenient calling in subsequent steps;

[0124] Step S5: Use the simulation system to call the generated Python script to achieve automatic calculation, output the obtained result file to the integration management platform, the integration management platform performs data visualization and uploads the data to the large language model for analysis to determine whether the design requirements are met. If the design requirements are not met, re-iterate and optimize and re-output, and start from scratch to execute Step S5. This process can be manually intervened to perform manual annotation on the data. Although the design requirements are met, if the user is not satisfied with the result, the user can still choose to continue iterating and optimizing. If satisfied with the result, choose to generate a 3D printing file.

[0125] Step S6: Import the generated model into the 3D printing system, call the large language model to complete the setting of 3D printing parameters, including layer thickness, printing speed, material selection, etc. After completing the parameter setting, send the sliced file to an online 3D printer to complete the printing.

[0126] In Step S1, select a suitable large language model as the system base model and pre-train the model, including:

[0127] Step S1.1: Select models such as deepseek-r1, Qwen2, ChatGLM, etc. as the system base model:

[0128] In terms of the number of model parameters, deepseek-r1, Qwen2, and ChatGLM provide multiple versions of large and small parameters for local deployment. Considering the hardware condition limitations, provide usage versions with smaller and larger numbers of parameters. The diverse model selection can fully adapt to the system operation requirements under different hardware conditions.

[0129] Step S1.2: Pre-train the model. The training method uses "Few-Shot Learning + RLHF" (that is, embed a small number of examples in the prompt, guide the model to generate the target content by designing the prompt, train the reward model, and guide the large language model to iteratively optimize). When training, divide the adaptive wing structure into four main components: "wing main structure", "deformable component", "wing tip", and "shape memory alloy wire selection", and set different prompts for different structures. Users can input parameters according to the preset prompts, and can directly input the scheme design requirements, scheme usage background, and additional design goals through the dialog box to help the large language model better understand the specific content of the scheme.

[0130] Step S2: Build a data integration management platform to integrate, manage, and call the generated data, scripts, and software to be used:

[0131] Step S2.1: Build a data integration management platform: Since the present invention needs to implement the adaptive wing scheme design based on the large language model, which not only includes the overall scheme generation based on the large language model, but also covers the upstream and downstream links such as requirement analysis and scheme evaluation. To ensure the availability and consistency of data, build a data integration management platform as the data integration, management, and call platform of the system. Classify and store the design products in each stage of each design process in the form of structured data. As Figure 2 shown, in the left file management function area, it includes model files, script files, and result files, allowing users to complete the call and editing of data by running different function modules, and supporting users to view and edit model files and script files in the "Generate Code" and "Model Preview" windows in the main interface as shown in Figure 2 shown.

[0132] Step S3: Integrate the Python compiler, 3D modeling software, finite element simulation software, and 3D printing slicing and parameter setting software into the ability units of the management platform

[0133] Step S3.1: As Figure 2As shown, corresponding function buttons are provided to support users in running to different function modules, and the required software is integrated into each function module. The "Generate Model" function button mainly calls a large language model to complete the scheme design; the "Run Code" function button calls a Python compiler to generate a model preview; the "Model Preview" function button calls a Python script by running Solidwork software to perform a model preview, and displays the result in the "Dual-Window Model Preview" interface; the "Export Model and Print Parameters" function button can call an automated 3D printing module, which integrates 3D printing slicing and parameter setting software supporting an online 3D printer; the "Simulation" function button can call a simulation and scheme verification module, which is the main function module and integrates solidwork modeling software, ABAQUAS finite element analysis software, ANSYS Fluent, and a Python compiler. The solidwork software is responsible for 3D modeling, ABAQUAS is responsible for analyzing the wing structure, mechanical properties, and deformation ability, ANSYS Fluent is responsible for analyzing the aerodynamics of the wing, and the Python compiler is responsible for visualizing the data in the result file.

[0134] In step S4, the large language model is called in a local deployment manner to generate a 3D printer wing scheme, including the selection of prompt words for different wing parts and the generation of corresponding models.

[0135] Step S4.1: Call the large language model selected in step S1.1, and the specific implementation is as follows:

[0136] The model is called by embedding a Python script in the independently developed integrated management platform and integrated into a scheme generation module. When the user enters the basic parameters and design requirements of different structural parts of the wing in the input box, by clicking Figure 2 the "Generate Model" button on the main interface of the integrated management platform as shown, the prompt words of different structural components are spliced. After the prompt word splicing is completed, the large language model is called to complete the scheme design, and it is converted into a relevant script file for subsequent step calls.

[0137] Step S4.2: After the scheme design is completed, the design scheme is output through the data interface of the integrated management platform. The integrated management platform stores the design schemes generated by the large language model separately according to types, which is convenient for subsequent data management and calls. Users can operate on the generated files through the file management interface on the left side of the main interface.

[0138] In step S5, the user clicks the simulation verification button on the main interface to run the simulation verification module to complete the evaluation and optimization of the design scheme. It mainly includes:

[0139] Step S5.1: Run the Python_Model.py script through Solidwork software to build a 3D model. Specifically, it includes:

[0140] Use Python to interact with the SolidWorks API to create part sketches, features, and insert text. Configure the pywin32 library to interact with SolidWorks, and use the generated Python_Model.py script to achieve 3D modeling in Solidwork software. Mark the dimensions of each component of the wing, such as the wing main structure, deformable components, wing tips, and internal filled lattice, and output the results as a.step file for easy editing and calling during subsequent simulation, and at the same time output the.stl file of the model for subsequent 3D printing. The generated files are all stored in the result file.

[0141] Step S5.2: Analyze and verify the mechanical properties and deformation ability of the wing model through ABAQUAS finite element analysis software. The specific implementation is as follows:

[0142] Through the Python script, the modeling, solution, and analysis processes of ABAQUS can be automated. Use the Python_simulation.py script generated by the large language model to import the.step file generated in Step S5.1 into ABAQUAS. Preset the boundary conditions, load settings, mesh generation, and material properties for the simulation in the script to achieve the automation of the modeling, solution, and analysis processes, obtain the final calculation results, and save them as result files. The parameters verified in this step mainly include (1) the bending resistance performance, structural strength and stiffness of the wing main structure, the anti-deformation ability and weight reduction effect of the filled lattice; (2) the deformation ability and bending resistance strength of the deformable part, the deformation ability and weight reduction effect of the filled lattice; (3) the variable angle range, deformation strength, and bending resistance section coefficient of the wing tip part; (4) the driving force, driving strain, and driving current of the shape memory alloy wire.

[0143] Step S5.3: Analyze the aerodynamic performance of the aircraft wing using ANSYS Fluent:

[0144] Using PyFluent as the Python interface of Ansys Fluent to achieve the call of ANSYS Fluent to Python scripts. Use the Python_Fluent.py script to import the.step file generated in step S5.1 into ANSYS Fluent to complete the model establishment, and then complete the mesh generation and boundary condition setting in the Python script (mainly set the boundary conditions for fluid flow calculation, including inlet velocity, outlet pressure, wing surface wall conditions, etc., as well as the required turbulence model and discrete phase model, etc.), perform computational simulation (use the ANSYS Fluent software to simulate the aerodynamic performance of the set wing model to obtain aerodynamic parameters of the wing under different flight conditions, such as lift, drag, aerodynamic moment, etc.), and save the obtained calculation data as a result file.

[0145] Step S5.4: Visualize the result file and iteratively optimize the results:

[0146] Convert the result file output in the above steps into text files and graphic files that can be recognized by the large language model, then perform data visualization, and submit the result file to the large language model. The large language model will autonomously judge whether the design parameters meet the standards. If not, mark the unqualified data and continue iterative optimization. After each optimization iteration, perform simulation to verify whether the design scheme meets the requirements, and visually display the results of each simulation. Figure 3 Generate a scheme display and iterative optimization interface. When the user clicks the "Data Visualization" button, these data will be visually displayed on the interface (such as Figure 4 ), which is convenient for the user to view. When the large language model determines that the design scheme meets the standards, the user can still click the "Continue Iterative Optimization" button to jump to the main interface and re-enter the design requirements in the main interface to continue optimization. If satisfied with the generated scheme, click the "Generate 3D Printing File" button to enter the next step.

[0147] Step S6: Import the generated model into the 3D printing system, call the large language model to complete the setting of 3D printing parameters, and send the sliced file to an online 3D printer to complete the printing.

[0148] Step S6.1: Import the.stl file generated in step S5.1 into the 3D printing system ( Figure 5 The figure shows the interface of the 3D printing system. The upper part is the model preview, and the lower part is the printing parameter design). Call the large language model to first repair the 3D printed model to prevent defects such as overhangs, holes, and model deformation during printing, and then design the relevant 3D printing parameters, including material selection, layer thickness, printing speed, etc.

[0149] Step S6.2: Import the repaired model and printing parameters in Step S6.1 into professional slicing software to complete model slicing, output the generated 3D printing file to the integrated management platform, and the user sends the printing file to the online automatic 3D printer through the platform to complete printing.

[0150] The advantages of the technical solution provided by this embodiment are as follows:

[0151] Efficient and automated design: Utilize large language models to automatically generate design solutions, code, and scripts, realizing the full-process automation from parameter input to 3D printing file generation, and greatly shortening the design cycle.

[0152] Precise iteration and optimization: Through the automatic simulation verification feedback mechanism, combined with the intelligent judgment of design defects by large language models, support rapid iteration and optimization to improve the accuracy and reliability of the solution.

[0153] Data integration and modular management: Build a unified data integration management platform to realize the centralized management of model files, scripts, and simulation results and the efficient call between modules, ensuring the stable transmission and real-time update of data in each link.

[0154] Multi-platform collaborative operation: Integrate multiple software such as CAD modeling, finite element simulation, computational fluid dynamics analysis, and 3D printing slicing to realize collaborative work between multiple platforms and ensure the seamless connection of the entire design-to-manufacturing process.

[0155] Flexibly adapt to different hardware and application scenarios: Support the deployment of multiple large language models and parameter versions, can select appropriate solutions according to hardware conditions, and at the same time has good scalability and is applicable to other adaptive structure design fields.

[0156] The above further describes the technical solution provided by the present invention in several specific embodiments to highlight the advantages and beneficial effects of the technical solution provided by the present invention. However, the above several specific embodiments are not used as limitations to the present invention. Any reasonable modifications and improvements, combinations of implementation manners, and equivalent replacements based on the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for designing an adaptive wing for 3D printing based on generative artificial intelligence, characterized in that, Including: The step of pre-training a large language model to obtain a trained basic model; The step of deploying the basic model to a data integration management platform, receiving the adaptive wing structure design parameters input by the user and generating a design scheme script; The step of calling 3D modeling software to automatically establish a 3D wing model based on the design scheme script and output a model file; The step of calling simulation analysis software to perform wing performance simulation analysis based on the model file and output analysis result data; The step of calling the basic model to perform design scheme evaluation and iterative optimization based on the analysis result data until an optimized model file meeting the design requirements is generated.

2. The 3D printing adaptive wing design method based on generative artificial intelligence according to claim 1, wherein The pre-training of the large language model is carried out in a way that combines Few-Shot Learning and reinforcement learning with human feedback, and training prompt words are respectively set for the wing main structure, deformation components, wing tips, and shape memory alloy wires.

3. The 3D printing adaptive wing design method based on generative artificial intelligence according to claim 1, wherein The data integration management platform centrally classifies and stores the generated model files, simulation analysis result data, and design scheme scripts, and provides a data visualization function to facilitate users to view, edit, and call the data.

4. The 3D printing adaptive wing design method based on generative artificial intelligence according to claim 1, wherein The 3D modeling software uses SolidWorks software to automatically generate 3D models of the wing main structure, deformation components, wing tips, and internal lattice filling structures by running the design scheme script, and outputs them in the formats of.step files and.stl files.

5. 3D printing adaptive wing design device based on generative artificial intelligence, characterized in that, Including: A module for pre-training a large language model to obtain a trained basic model; A module for deploying the basic model to a data integration management platform, receiving the adaptive wing structure design parameters input by the user and generating a design scheme script; A module for calling 3D modeling software to automatically establish a 3D wing model based on the design scheme script and output a model file; A module for calling simulation analysis software to perform wing performance simulation analysis based on the model file and output analysis result data; A module for calling the basic model to perform design scheme evaluation and iterative optimization based on the analysis result data until an optimized model file meeting the design requirements is generated.

6. A 3D printing adaptive wing printing method based on generative artificial intelligence, characterized in that, Including: The step of calling an automated 3D printing module to automatically set 3D printing parameters based on the optimized model file described in claim 1 and output a slice file for 3D printing.

7. 3D printing adaptive wing printing device based on generative artificial intelligence, characterized in that, Including: A module for calling an automated 3D printing module to automatically set 3D printing parameters based on the optimized model file described in claim 1 and output a slice file for 3D printing.

8. A computer storage medium for storing a computing program, characterized in that, When the computer program is read by a computer, the computer executes the method described in claim 1.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method described in claim 1.

10. A computer program product, as a computer program, characterized in that, When the computer program is executed, the method described in claim 1 is implemented.

Images