Intelligent lightweight design method for rudder and wing component based on multi-objective collaborative optimization

Abstract

The invention relates to the technical field of aerospace structure design, and provides a multi-objective collaborative optimization-based rudder wing component intelligent lightweight design method, which comprises the steps of parametric modeling, pneumatic simulation automation process establishment, pneumatic performance data set construction, CNN pneumatic proxy model construction and NSGA-II algorithm multi-objective optimization. By constructing a parameter-driven automatic modeling framework, key geometric features of a component are defined as design parameters, and parameterized representation and automatic generation of a geometric model are realized; the aerodynamic performance of the aerospace rudder wing component can be quickly and accurately predicted according to the input geometric parameters of the component through the built CNN aerodynamic agent model, and the intelligent lightweight design of the aerospace rudder wing component is realized through building the multi-target collaborative optimization method of fusion of the CNN aerodynamic agent model and the NSGA-II algorithm. The rigid contradiction between ultimate light weight and high aerodynamic efficiency in traditional design is solved, and the design requirement for rapid iteration of novel aerospace equipment is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aerospace structure design, and particularly relates to an intelligent lightweight design method for aileron wing components based on multi-objective collaborative optimization. BACKGROUND

[0002] Extreme lightweight is an important driving force for aerospace components to achieve cost control and efficient transportation. As a key functional component in the field of aerospace, the lightweight design of aileron wings can improve the effective payload of equipment, reduce launch and operation costs, and enhance the mobility and mission adaptability of equipment, thereby promoting the development of aerospace equipment towards low cost and large-scale application. The core challenge of aileron wing lightweight design lies in balancing the inherent contradiction between structure weight reduction and aerodynamic performance. Aerodynamic performance is a key indicator for determining the realization of component functions, but traditional lightweight design methods often affect key aerodynamic parameters due to the reduction of lift area while pursuing structure weight reduction. The contradiction between the two seriously restricts the optimization space of aileron wing performance. In addition, the development of new aerospace equipment is fast and short in cycle, so the original trial-and-error development mode cannot meet the current development needs.

[0003] In the current technology, the design concept of aileron wing components is limited to single-dimensional optimization of aerodynamic performance, and the key design dimension of structure lightweight is generally ignored. Existing optimization strategies mainly use single-objective optimization methods, focusing only on aerodynamic characteristic optimization, and fail to establish a multi-objective collaborative optimization framework that balances the inherent contradiction between aileron wing lightweight (minimum structure weight) and high aerodynamic performance (maximum lift-drag ratio, optimization of control efficiency, etc.), resulting in a whole design that is difficult to achieve optimal comprehensive performance. In addition, to solve the problem of low computational efficiency in the traditional optimization process, existing technologies generally use surrogate models to replace computational fluid dynamics (CFD) simulations, and approximate mathematical models are constructed to predict the response of the optimization target. Although this method can accelerate the design iteration process, there is still a bottleneck in efficiency. The accuracy of the surrogate model is highly dependent on a large number of training samples, and the acquisition of each sample requires complete execution of CFD standard processes such as geometry modeling, mesh generation, and numerical calculation. The current simulation process lacks automation processing capability, making it difficult to achieve efficient computation of a large number of samples.

[0004] Therefore, in order to solve the above problems, those skilled in the art need to design an intelligent lightweight design method for aileron wing components based on multi-objective collaborative optimization. SUMMARY

[0005] To solve the above technical problems, the present application provides an intelligent lightweight design method for aileron wing components based on multi-objective collaborative optimization, which comprises:

[0006] Step 1: Parametric modeling

[0007] Based on the CAD secondary development technology (such as SolidWorks API, NX Open) to build a parameter-driven automatic modeling framework, the key geometric features of the component are defined as design parameters, and the parametric representation of the geometric model is realized through programmed modeling. According to the change of design variables, the corresponding component geometric structure and CFD calculation model are automatically generated;

[0008] Step 2: Establishing the automatic process of aerodynamic simulation

[0009] The grid automatic generation algorithm constructed by ANSYS Fluent Meshing script tool generates mixed structure-unstructured grid. Through Fluent Journal script, the automatic configuration of solution parameters (including convergence criteria, relaxation factor, etc.) is realized. Finally, based on the task scheduling system of Windows batch processing, the automatic submission of multi-condition simulation tasks is realized, so as to establish a complete automatic process of aerodynamic simulation.

[0010] Step 3: Aerodynamic performance dataset construction

[0011] The Latin Hypercube Sampling (LHS) method is used to uniformly stratify sample the key geometric parameters of the component in the design space, ensuring the optimal space filling of each dimension parameter, and generating parameter combinations with good spatial distribution characteristics. Then the established automatic process of aerodynamic simulation is called to solve the aerodynamic performance indicators corresponding to each sample point, forming a complete "design parameter-aerodynamic performance" mapping dataset and performing normalization processing. Stratified Random Sampling method is used to divide the mapping dataset into training set and test set, and ensure the consistency of the probability distribution of the two subsets in the parameter space, laying a data foundation for subsequent proxy model construction. The Latin Hypercube Sampling method is used to sample the key geometric parameters of the component, ensuring that the parameter combinations have good spatial distribution characteristics in the design space, providing a comprehensive and representative data foundation for subsequent model training. By stratified random sampling method, the dataset is divided into training set and test set, and the probability distribution consistency of the two subsets in the parameter space is ensured, which further improves the reliability of the dataset and helps to improve the generalization ability and prediction accuracy of the subsequent proxy model.

[0012] Step 4: Construction of CNN aerodynamic proxy model

[0013] A convolutional neural network model with feature extraction and regression prediction ability is built based on a Matlab environment, aerodynamic performance (lift coefficient CL, drag coefficient CD and lift-drag ratio CL / CD) of a component is predicted according to input key geometric parameters of the component, and the root mean square error and the determination coefficient are used to evaluate the prediction accuracy of the model, and a CNN aerodynamic surrogate model is built.

[0014] Step 5: Multi-objective optimization of NSGA-II algorithm

[0015] The calculation parameters of the non-dominated sorting genetic algorithm (NSGA-II) are set, an initial population is randomly generated, each individual represents a set of structural design parameters, the CNN surrogate model is called to calculate the objective function value, non-dominated sorting and crowded distance calculation are performed, individuals with low non-dominated sorting level and large crowded distance are selected into the parent population, and the child population is generated through crossover and mutation, the parent and child populations are merged, and environmental selection is performed according to the non-dominated sorting and crowded distance, the elite individuals are reserved to form a new generation population, and the iteration is terminated until the termination condition is met, and the first non-dominated front is output as the Pareto optimal solution set, which simultaneously satisfies the collaborative optimization of the objective functions f1 and f2.

[0016] Further, the key geometric features of the component include independent variables such as chord length, grid spacing, inner grid thickness, outer grid thickness and local sweep angle.

[0017] Further, a design vector of corresponding dimension is established according to the number of design parameters, and the design space of the corresponding dimension is uniformly layered and sampled.

[0018] Further, the convolutional neural network model architecture includes an input layer, a convolutional layer, a maximum pooling layer, a fully connected layer and a regression layer, the design parameters are used as the input layer, the output layer has the same number of neurons as the aerodynamic performance to be predicted, the input is feature extracted through the convolutional layer, and the Dropout is introduced to reduce the risk of overfitting, and finally the global information is integrated through the fully connected layer, and the continuous prediction value is output by the regression layer.

[0019] Further, the aerodynamic performance includes lift coefficient CL, drag coefficient CD and lift-drag ratio CL / CD.

[0020] Further, the calculation parameters of the non-dominated sorting genetic algorithm include objective function F=(f1, f2), initial population size N, maximum iteration number T max , crossover probability P c , mutation probability P m .

[0021] The present application has the following advantages:

[0022] (1) The application realizes parameterized representation and automatic generation of a geometric model by constructing a parameter-driven automated modeling framework, defining key geometric features of a component as design parameters, getting rid of the tediousness of traditional manual modeling, and greatly shortening the modeling time;

[0023] (2) The application realizes automation of aerodynamic simulation through a script tool, the whole aerodynamic simulation process does not need manual intervention, reduces errors caused by manual operation, improves the processing efficiency of simulation tasks, and can complete a large amount of simulation calculation of different working conditions in a short time;

[0024] (3) The application can quickly and accurately predict the aerodynamic performance of a component according to the input geometric parameters of the component through the built CNN aerodynamic proxy model, compared with the traditional aerodynamic simulation method, the proxy model greatly shortens the performance prediction time, and uses root mean square error, determination coefficient and other evaluation indexes to evaluate and optimize the model, ensuring the accuracy of the prediction result and providing a reliable basis for subsequent optimization design;

[0025] (4) The application adopts NSGA-Ⅱ algorithm for multi-objective optimization, through reasonable setting of calculation parameters, through the processes of population generation, selection, crossover, mutation and environment selection, finally outputs a Pareto optimal solution set, the solution set can simultaneously satisfy the collaborative optimization of multiple objective functions, provides a series of optimal design schemes for designers, on the basis of batch generation of high-precision samples and rapid prediction of aerodynamic performance, realizes the collaborative optimization of structure weight reduction and aerodynamic efficiency, provides key technical support for the improvement of comprehensive performance of new generation of aerospace equipment;

[0026] (5) The application realizes intelligent lightweight design of aerospace rudder wing components through the multi-objective collaborative optimization method of combining the CNN aerodynamic proxy model and the NSGA-Ⅱ algorithm, solves the rigid contradiction between "extreme lightweight and high aerodynamic efficiency" in traditional design, meets the design requirements of rapid iteration of new type of aerospace equipment, and promotes the intelligent and efficient upgrading of aerospace structure design. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of lightweight design in the application.

[0028] Figure 2 is a schematic diagram of key geometric features of a grid rudder in the application.

[0029] Figure 3 is a result schematic diagram of the lift coefficient predicted by the CNN aerodynamic proxy model in the application.

[0030] Figure 4 is a result schematic diagram of the drag coefficient predicted by the CNN aerodynamic proxy model in the application.

[0031] Figure 5 Figure 1 is a schematic diagram of the grid fin optimization result of the NSGA-II algorithm based on the CNN agent model in the application. DETAILED DESCRIPTION

[0032] The technical solutions of the application are described in further detail below in combination with specific embodiments, but the embodiments are not used to limit the application, and any similar structure and similar changes thereof that adopt the application shall be included in the protection scope of the application. The dots in the application represent the relationship of sum, and the English letters in the application are case-sensitive.

[0033] As shown in Figures 1-2 , a lightweight design is performed on a grid fin of a certain space vehicle, and the specific design method is as follows:

[0034] Step 1, parameterized modeling

[0035] Based on the secondary development technology of SolidWorks, a parameterized modeling system of the grid fin is constructed by using the VBA macro programming technology, the grid spacing d, chord length c, inner grid thickness t1, outer grid thickness t2, and local sweepback angle a of the grid fin are taken as the design parameters, the numerical input function of the design parameters is realized by developing the Visual Basic user interactive form, after the user inputs the parameter values in the form interface, the system automatically calls the program, creates and updates the three-dimensional model and the external flow field model of the grid fin according to the input design parameters, as shown in Figure 2 , thereby establishing a parameterized modeling framework to provide a basis for subsequent performance prediction and optimization design.

[0036] Step 2, automatic process of aerodynamic simulation

[0037] The grid automatic generation algorithm constructed by using the ANSYS Fluent Meshing script tool generates structure-unstructured polyhedral hybrid grids, for the complex geometric features of the grid fin, the boundary layer structured grid is encrypted on the surface of the arrow and the grid fin, the unstructured polyhedral grid is applied in the external flow field area, and the programmatic driving of the key parameters such as grid size, layer number and growth rate is realized through the script tool, in terms of solving, based on the Reynolds average Navier-Stokes (RANS) equation, the k-ω SST turbulence model is adopted for closure, the second-order upwind format is set for spatial discretization, and the automatic solution of the lift coefficient and the drag coefficient is realized through the Fluent Journal script; finally, based on the task scheduling system of Windows batch processing, the parameterized configuration and batch submission of multi-condition simulation tasks are realized through the workflow script, forming the automatic process of aerodynamic simulation from geometric modeling, grid generation to numerical solution.

[0038] Step 3, aerodynamic performance dataset construction

[0039] The geometry of the lattice rudder can be represented as a 5-dimensional design vector, i.e. X = [d c t1 t2 a]. According to the engineering practical constraints of each design variable, determine its value range, use Latin hypercube sampling (LHS) method to perform uniform stratified sampling of the 5-dimensional design space, in order to verify the feasibility and calculation efficiency of the method, 100 samples are extracted in the preliminary research stage, i.e. Wherein, X i represents the i-th sample, d i represents the lattice spacing of the lattice rudder in the i-th sample, c i represents the chord length of the lattice rudder in the i-th sample, t1 i represents the inner lattice thickness of the lattice rudder in the i-th sample, t2 i represents the outer lattice thickness of the lattice rudder in the i-th sample, a i represents the local sweep angle of the lattice rudder in the i-th sample; based on the automatic process of aerodynamic simulation established in step 2, the aerodynamic coefficients of each sample point at 0.8Ma and 10° angle of attack are calculated, and a training dataset containing 100 sample points is generated Wherein, Y i represents the output aerodynamic performance of the i-th sample, CL i represents the lift coefficient of the i-th sample, CD i represents the drag coefficient of the i-th sample; and the dataset is normalized and divided into training set and test set according to a certain proportion.

[0040] Step 4, CNN aerodynamic proxy model construction

[0041] A one-dimensional convolutional neural network (1D-CNN) for regression task is established by Matlab, the 5 design parameters of the lattice rudder are taken as the input layer, and the output layer has 2 neurons, i.e. lift coefficient CL and drag coefficient CD, two convolutional layers are designed, the local features are extracted by the first convolutional layer, after ReLU activation and maximum pooling dimension reduction, the second convolutional layer is stacked to further abstract high-order features, and 10% Dropout is introduced to reduce the risk of overfitting; finally, the global information is integrated through the fully connected layer, and the continuous prediction value is output by the regression layer, the prediction result is as shown in Figures 3-4 The training and testing of the constructed CNN aerodynamic proxy model are carried out through the training set and the test set.

[0042] Step 5, NSGA-Ⅱ algorithm multi-objective optimization

[0043] Based on the structure weight reduction and aerodynamic performance optimization problem of the grid rudder, the above 5 key geometric characteristics (d, c, t1, t2, alpha) are taken as design variables, the minimum volume (weight reduction) and the maximum lift-drag ratio (aerodynamic efficiency) are taken as multi-objective optimization objectives, the constraint condition (CL is greater than or equal to 0.25) is set, and the feasible region of the design variable is limited, and the optimization problem is converted into a minimum problem, and the optimization objective is:

[0044]

[0045] And meet:

[0046]

[0047] In the formula, f1, f2 are objective functions; V is the volume of the grid rudder;

[0048] The NSGA-II algorithm based on the CNN agent model is used to solve the above optimization problem, the calculation of the optimization objective function f2 in all populations calls the CNN agent model, and the parameter setting of the algorithm is: the initial population is 80; the maximum iteration number is 100; the crossover probability is 0.7; the mutation probability is 0.02, and the optimization result is as shown in Figure 5 The Pareto optimal solution set contains 40 Pareto points, the optimal solution is selected, the optimal geometric parameters of the grid rudder can be obtained, and the geometric parameters are input into the parameterized modeling window interface to output the optimal grid rudder geometric structure.

[0049] The application realizes the intelligent lightweight design of the aerospace rudder wing component by constructing the multi-objective collaborative optimization method of the CNN aerodynamic agent model and the NSGA-II algorithm, solves the rigid contradiction of "extreme lightweight and high aerodynamic efficiency" in the traditional design, meets the design requirements of the rapid iteration of new type aerospace equipment, and promotes the intelligent and efficient upgrading of the aerospace structure design.

[0050] Although the preferred embodiments of the application have been described, those skilled in the art can make further changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all changes and modifications falling within the scope of the application.

Claims

1. A smart lightweight design method for rudder components based on multi-objective collaborative optimization, characterized in that, The design method includes: Step 1: Parametric Modeling A parameter-driven automated modeling framework is built based on CAD secondary development technology. The key geometric features of the components are defined as design parameters. The parametric representation of the geometric model is realized through programmatic modeling. Based on the changes in design variables, the corresponding component geometric structure and CFD calculation model are automatically generated. Step 2: Establish an automated process for pneumatic simulation An automatic mesh generation algorithm built using script tools generates a structured-unstructured hybrid mesh, and the solution parameters are automatically configured using scripts. Finally, a Windows batch processing task scheduling system is used to automatically submit multi-condition simulation tasks, thereby establishing a complete automated aerodynamic simulation process. Step 3: Construction of Aerodynamic Performance Dataset The Latin hypercube sampling method is used to uniformly stratify and sample the key geometric parameters of the components in the design space to generate parameter combinations. Then, the established automated aerodynamic simulation process is called to solve the aerodynamic performance index corresponding to each sample point, forming a complete "design parameter-aerodynamic performance" mapping dataset, which is then normalized. The mapping dataset is divided into training and test sets using a stratified random sampling method. Step 4: Construction of CNN aerodynamic surrogate model A convolutional neural network model with feature extraction and regression prediction capabilities is built to predict the aerodynamic performance of the input component based on its key geometric parameters. The accuracy of the prediction is evaluated, and a CNN aerodynamic surrogate model is constructed. Step 5: Multi-objective optimization using the NSGA-II algorithm The computational parameters of the non-dominated sorting genetic algorithm are set, an initial population is randomly generated, and each individual represents a set of structural design parameters. The constructed CNN surrogate model is called to calculate the objective function value. Non-dominated sorting and crowding distance calculations are performed, and individuals with low non-dominated sorting levels and large crowding distances are selected to enter the parent population. After crossover and mutation, offspring populations are generated. The parent and offspring populations are merged, and environmental selection is performed based on non-dominated sorting and crowding distance. Elite individuals are retained to form a new generation population. The iteration continues until the termination condition is met, and the first non-dominated frontier is output as the Pareto optimal solution set.

2. The intelligent lightweight design method for rudder components based on multi-objective collaborative optimization according to claim 1, characterized in that, The key geometric features of the component include chord length, grid spacing, inner grid thickness, outer grid thickness, and local sweep angle.

3. The intelligent lightweight design method for rudder components based on multi-objective collaborative optimization according to claim 1, characterized in that, In step 3, design vectors of corresponding dimensions are established according to the number of design parameters, and the design space of the corresponding dimensions is uniformly stratified and sampled.

4. The intelligent lightweight design method for rudder components based on multi-objective collaborative optimization according to claim 1, characterized in that, The convolutional neural network model architecture includes an input layer, a convolutional layer, a max pooling layer, a fully connected layer, and a regression layer. The design parameters are used as the input layer, and the output layer consists of the same number of neurons whose aerodynamic performance needs to be predicted. The convolutional layer extracts features from the input and introduces Dropout to reduce the risk of overfitting. Finally, the fully connected layer integrates global information, and the regression layer outputs continuous predicted values.

5. The intelligent lightweight design method for rudder components based on multi-objective collaborative optimization according to claim 1, characterized in that, The aerodynamic performance includes lift coefficient, drag coefficient, and lift-to-drag ratio.

6. The intelligent lightweight design method for rudder components based on multi-objective collaborative optimization according to claim 1, characterized in that, The computational parameters of the non-dominated sorting genetic algorithm include the objective function, initial population size, maximum number of iterations, crossover probability, and mutation probability.

Images