Aircraft structure optimization method, system, equipment and medium

Abstract

The invention provides an aircraft structure optimization method, system and device and a storage medium, and belongs to the technical field of aircraft falling adaptability design, and the method comprises the following steps: building a finite element model of an aircraft fuselage frame section, analyzing a load transmission path in an aircraft falling impact process, building an energy absorption performance decomposition model, and calculating an energy absorption performance of the aircraft according to the energy absorption performance decomposition model. According to the energy absorption performance decomposition model, constructing an evaluation system of the layered falling adaptability level of the fuselage frame section; constructing an initial data set, and performing data enhancement on the initial data set by using an adversarial neural network algorithm to obtain an enhanced data set; optimizing an initial weight and a threshold parameter of the prediction neural network model through a parameter optimization algorithm to obtain an improved prediction neural network model, and inputting the data in the enhanced data set into the prediction neural network model to obtain an optimal design parameter combination; the aircraft structure is optimized according to the optimal design parameter combination; by means of the method, the aircraft structure with the excellent falling performance can be obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft crashworthiness design, and particularly relates to an aircraft structure optimization method, system, device and storage medium. Background Technique

[0002] Studying the crash performance of aircraft is an important topic for protecting the safety of passengers and is a concentrated embodiment of the passive safety of aircraft. For the performance of the aircraft fuselage structure under crash impact, it involves the mutual hybridization of material nonlinearity, geometric nonlinearity, and boundary nonlinearity problems, and at the same time involves the coupling relationship between multiple structural design parameters, making it difficult to predict the structural performance independently.

[0003] In existing research on the impact process of aircraft fuselage structures, most of the predictions of impact energy dissipation methods rely too much on commercial finite element simulation programs. However, since collecting the energy absorption characteristics of aircraft fuselage structures is a time-consuming cycle, including experiments and simulation calculations, the calculation cost is relatively high. Therefore, some research uses the ANN neural network algorithm to predict the crash performance of aircraft. However, when using this method, due to the insufficient amount of original sample data, it is impossible to obtain better aircraft structural design parameters, resulting in a low prediction accuracy of aircraft crash performance. Summary of the Invention

[0004] In order to overcome the above-mentioned deficiencies of the existing technology, the present invention provides an aircraft structure optimization method, which includes the following steps: Build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo area and the lower passenger cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model; Construct an initial data set of key variables for aircraft structure design, and use the generative adversarial network algorithm GAN to perform data enhancement operations on the initial data set to obtain an enhanced data set; Optimize the initial weights and threshold parameters of the prediction neural network model BP through a parameter optimization algorithm to obtain an improved prediction neural network model. Input the data in the enhanced data set into the improved prediction neural network model, and take the maximization of the evaluation index in the evaluation system as the goal, and iterate the improved prediction neural network model to obtain an optimal design parameter combination; optimize the aircraft structure according to the optimal design parameter combination.

[0005] Preferably, the construction of the evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model includes the following steps: Decompose the crash kinetic energy dissipation process of the aircraft fuselage frame section in the lower cargo area and the lower passenger cabin area, distribute the energy absorption, deformation stiffness allowed by the structural design, and remaining load-bearing strength to each component, and perform simulation analysis on the finite element model; Extract the load-displacement response curve from the simulation analysis results, collect the peak load, load efficiency, and specific energy absorption rate according to the load-displacement response curve, and quantify the peak load, load efficiency, and specific energy absorption rate into comprehensive evaluation indicators to obtain an evaluation system for the hierarchical crashworthiness level of the fuselage frame section.

[0006] Preferably, the initial data set of the key variables for the aircraft structural design is constructed as follows: Based on non-linear correlation analysis, select the key variables in the structural design process, establish the sample space of the key variables for structural design, and form the initial data set; the key variables include the structural thickness and ply angle of the aircraft components.

[0007] Preferably, use the Generative Adversarial Network (GAN) algorithm to perform data augmentation on the initial data set to obtain an augmented data set, including the following steps: Input the data in the initial data set into the GAN algorithm. Through the iterative processing of the input data by the generator model and discriminator model of the GAN, the data in the initial data set is expanded. When the discrimination probability of the generator model and the discriminator model reaches the 0.5 threshold, the augmented data set is obtained.

[0008] Preferably, the finite element model of the aircraft fuselage frame section is built as follows: Establish a numerical model of the aircraft fuselage frame section, simplify the numerical model, simplify the upper structure of the aircraft fuselage frame section that is not the main load-bearing part into mass nodes and beam elements, and build the finite element model of the aircraft fuselage frame section.

[0009] Preferably, before using the GAN algorithm to perform data augmentation on the initial data set, the data in the initial data set is also normalized.

[0010] The present invention also provides an aircraft structure optimization system, including: An evaluation system acquisition module, used to build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo area and the lower passenger cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model; A data set augmentation module, used to construct an initial data set of the key variables for the aircraft structural design, and use the GAN algorithm to perform data augmentation on the initial data set to obtain an augmented data set; A structure optimization module is used to optimize the initial weights and threshold parameters of the prediction neural network model BP through a parameter optimization algorithm, obtain an improved prediction neural network model, input the data in the enhanced dataset into the improved prediction neural network model, and iterate the improved prediction neural network model with the goal of maximizing the evaluation index in the evaluation system to obtain an optimal combination of design parameters; optimize the aircraft structure according to the optimal combination of design parameters.

[0011] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the aircraft structure optimization method.

[0012] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute the aircraft structure optimization method.

[0013] The aircraft structure optimization method provided by the present invention has the following beneficial effects: By enhancing the initial dataset to obtain an enhanced dataset, the present invention can make up for the deficiency of the original sample data volume, enhance the generalization ability of the prediction model, effectively maximize the energy absorption of the fuselage structure, thereby effectively improving the crash energy absorption characteristics of the aircraft fuselage structure and providing support for the anti-crash structure design of the aircraft; by constructing an energy absorption performance decomposition model for the lower part of the cargo hold and the lower part of the passenger cabin of the fuselage frame section, and constructing an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model, the present invention can comprehensively consider the crashworthiness performance indicators such as the structural strength, stiffness, and energy absorption of each sub-region of the aircraft fuselage structure; by constructing an improved prediction neural network model, inputting the data in the enhanced dataset into the improved prediction neural network model, and iterating the improved prediction neural network model with the goal of maximizing the evaluation index in the evaluation system to obtain an optimal combination of design parameters, the present invention can apply the improved prediction neural network model to the prediction of the aircraft crash process, obtain an optimal combination of design parameters for optimizing the aircraft structure, and realize the optimization of the aircraft structure.

[0014] By combining the finite element method with the neural network model to carry out hybrid prediction, the present invention uses less finite element method calculations, realizes a large reduction in calculation time and calculation cost, makes it possible to carry out multi-parameter optimization of aircraft crashworthiness design in a large design space, and realizes a multi-level decomposition, rapid prediction, and optimization method for aircraft crashworthiness performance. Description of the Drawings

[0015] To more clearly illustrate the embodiments of the present invention and their design solutions, the accompanying drawings required for this embodiment will be briefly introduced below. The accompanying drawings in the following description are only partial embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0016] Figure 1 It is a schematic flow chart of the aircraft structure optimization method according to the embodiment of the present invention; Figure 2 It is a finite element simplified model of the aircraft fuselage frame section; Figure 3 It is a design parameter optimization flow chart; Figure 4 It is a total error comparison chart.

[0017] Explanation of reference numerals: 1 - connecting truss, 2 - aviation seat, 3 - cabin floor cross beam, 4 - cabin floor support column, 5 - cargo hold floor, 6 - cargo hold floor support column, 7 - aircraft skin. Detailed implementation manners

[0018] In order to enable those skilled in the art to better understand the technical solutions of the present invention and implement them, the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0019] Embodiment The present invention provides an aircraft structure optimization method, specifically as Figure 1 shown, including the following steps: Step 1: Build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo hold area and the lower cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model.

[0020] Specifically, the energy absorption performance decomposition model is constructed through the following steps: (1) Structure partitioning: Divide the aircraft fuselage frame section into a lower cargo hold sub - area (including the cargo hold floor support column structure, cargo hold floor structure, etc.) and a lower cabin sub - area (including the cabin floor support column structure, cabin floor cross beam structure, etc.).

[0021] (2) Energy distribution: Based on the load transfer path, distribute the total energy absorbed by the aircraft during the crash to the components of each sub - area according to the functional weight, and form the component - level energy absorption target value E i .

[0022] (3) Constraint setting: Set the allowable deformation stiffness threshold (δ max,i ) and the remaining bearing strength threshold (σ min,i ) for each component.

[0023] (4) Verification mechanism: Verify whether the energy absorbed by the component meets E i ≥E 目标 , and at the same time meet δ i ≤δ max,i and σ i ≥σ min,i .

[0024] First, establish a numerical model of the aircraft fuselage frame section, simplify the numerical model, simplify the upper structure of the aircraft fuselage frame section that does not mainly bear the load into mass nodes and beam elements, and build a finite element model of the aircraft fuselage frame section. Specifically, as Figure 2 shown, it includes the connecting truss 1, the aviation seat 2, the cabin floor cross beam 3, the cabin floor support column 4, the cargo hold floor 5, the cargo hold floor support column 6, and the aircraft skin 7.

[0025] Second, decompose the crash kinetic energy dissipation process of the aircraft fuselage frame section in the lower cargo hold area and the lower cabin area, distribute the energy absorption and the allowable deformation stiffness and remaining bearing strength of the structure design to each component, obtain an energy absorption performance decomposition model, and perform a simulation analysis on the energy absorption performance decomposition model.

[0026] Extract the load-displacement response curve from the simulation analysis results, collect the peak load, load efficiency, and specific energy absorption rate according to the load-displacement response curve, and quantify the peak load, load efficiency, and specific energy absorption rate into comprehensive evaluation indicators to obtain an evaluation system for the hierarchical crashworthiness level of the fuselage frame section. The comprehensive evaluation indicators specifically include the energy absorption rate, specific energy absorption rate, peak load, mean load, load efficiency, deformation stiffness, and structural displacement of the fuselage frame section.

[0027] Step 2: Construct an initial data set of the key variables for aircraft structural design, and use the generative adversarial network algorithm GAN to perform data augmentation operations on the initial data set to obtain an augmented data set.

[0028] First, select the key variables in the structural design process, reasonably distribute the crash impact kinetic energy based on the Complex Proportional Assessment (COPRAS), establish a sample space of the key variables for structural design, and form an initial data set.

[0029] Select the key variables for the fuselage frame section design, and based on the Design of Experiments (DOE) method, conduct a simulation matrix design for multiple working conditions to obtain the original input matrix.

[0030] Conduct a non - linear correlation analysis and parameter sensitivity analysis on the original input matrix. Through the non - linear correlation analysis, explore the relationships between features, and select the features with obvious correlation degrees and the original input matrix to form an initial data set. Normalize the data in the initial data set. The normalization formula is as follows: ; In the formula, X normal is the normalized input matrix; X is the original input matrix; X min is the minimum value in the original input matrix; X max is the maximum value in the original input matrix; The original input matrix includes the ply angles of composite material components, structural thicknesses, and some material parameters.

[0031] Secondly, under the condition of a small sample size, input the normalized input matrix into the Generative Adversarial Network (GAN) algorithm. Through the generative model and discriminative model of GAN, iteratively process the input data to fully expand the normalized input matrix. When the discrimination probabilities of the generative model and the discriminative model reach the 0.5 threshold, generate a large number of reliable data from the small data set to obtain an enhanced data set.

[0032] In addition, in the GAN model, random noise data is also introduced to prevent the model from overfitting and enhance the generalization ability of the model.

[0033] Step 3: Optimize the initial weights and threshold parameters of the prediction neural network model BP through a parameter optimization algorithm to obtain an improved prediction neural network model. Input the data in the enhanced data set into the improved prediction neural network model, and take the maximization of the evaluation index in the evaluation system as the goal, and iterate the improved prediction neural network model to obtain the optimal design parameter combination; Optimize the aircraft structure according to the optimal design parameter combination.

[0034] The parameter optimization algorithm can be a global optimization algorithm such as Particle Swarm Optimization (PSO), Simulated Annealing (SA), or Greedy algorithm. In this embodiment, the Genetic Algorithm (GA) is specifically selected. The initialization parameters of the prediction neural network model (Backpropagation Neural Networks, BP) are optimized by the Genetic Algorithm GA. The optimal initial weight parameters are screened out based on the fitness function and passed to the prediction neural network model BP to obtain the improved prediction neural network model GA-BP. The data in the enhanced dataset is input into GA-BP, and the optimal solution within the feasible region is searched through a large number of iterations of the BP neural network model to search for the optimal result and design parameters, ultimately realizing the lightweight layout and controllable energy absorption design of the lower structure of the fuselage frame section.

[0035] The specific process is as Figure 3 shown. The input vector in the enhanced dataset is input into the prediction neural network model BP after data processing for BP network initialization and initial weights are assigned. After initialization coding by the GA genetic algorithm, chromosome selection, crossover, and mutation operations are performed based on the fitness function, and the population is iteratively updated. The optimal fitness value and average fitness value of each iteration are collected. When the number of runs reaches the maximum number of iterations set by the genetic algorithm, it is judged whether the termination condition is reached. If the termination condition is not met, return to reselect the chromosome and perform iteration again according to the previous steps until the BP network initialization weight value corresponding to the optimal fitness value (i.e., the optimal weight) is screened out. The obtained optimal weight is input into the prediction neural network model BP for iterative calculation and error analysis, and the weight is updated. Then it is judged whether the training target is terminated. If so, retain the output vector of the prediction neural network model BP and use the prediction neural network model BP to realize data prediction within the prediction space. If the termination condition is not met, return to reselect the optimal weight and perform iteration again according to the previous steps until the preset number of iterations is reached or the error threshold is passed.

[0036] In addition, in this embodiment, the activation function of GA-BP is the SELU algorithm improved based on the RELU algorithm, and the error function of GA-BP is a combined form of the mean absolute error (MAE), root mean square error (RMSE), and mean absolute error rate (MAPE).

[0037] The present invention also provides an aircraft structure optimization system, including an evaluation system acquisition module, a data set enhancement module, a parameter optimization module, and a structure optimization module. The evaluation system acquisition module is used to build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo area and the lower passenger cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model; the data set enhancement module is used to construct an initial data set of key variables for aircraft structure design, and perform data enhancement operations on the initial data set using the generative adversarial network algorithm GAN to obtain an enhanced data set; the structure optimization module is used to optimize the initial weights and threshold parameters of the prediction neural network model BP through a parameter optimization algorithm to obtain an improved prediction neural network model, input the data in the enhanced data set into the improved prediction neural network model, and iterate the improved prediction neural network model with the goal of maximizing the evaluation index in the evaluation system to obtain an optimal design parameter combination; optimize the aircraft structure according to the optimal design parameter combination.

[0038] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the aircraft structure optimization method.

[0039] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute the aircraft structure optimization method. Compared with the traditional single BP algorithm prediction, in the iterative process of the GA-BP algorithm, the GA algorithm effectively reduces the initial error of the prediction network by optimizing the initial network weights of BP, improves the error descent rate, and is more likely to reach the convergence state. Moreover, the final error of the GA-BP algorithm is smaller and the prediction effect is better, as Figure 4 shown.

[0040] Compared with the traditional aircraft structure optimization method, which only relies on the finite element method for iterative calculation, the workload is huge, and the time cost and computing power consumption are excessive. The optimization method of the present invention combines the traditional finite element method with multiple neural network models for hybrid optimization, uses less finite element method calculations, realizes a large reduction in calculation time and calculation cost, and makes it possible to carry out multi-parameter optimization of aircraft crashworthiness design in a large design space.

[0041] The above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of the technical solutions that can be obviously obtained by those skilled in the art within the technical scope disclosed by the present invention all belong to the protection scope of the present invention.

Claims

1. An aircraft structure optimization method, characterized in that, It includes the following steps: Build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo area and lower passenger cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model; Construct an initial data set of key variables for aircraft structural design, and use the Generative Adversarial Network (GAN) algorithm to perform data augmentation operations on the initial data set to obtain an augmented data set; Optimize the initial weights and threshold parameters of the Prediction Neural Network (BP) model through a parameter optimization algorithm to obtain an improved prediction neural network model. Input the data in the augmented data set into the improved prediction neural network model, and iterate the improved prediction neural network model with the goal of maximizing the evaluation index in the evaluation system to obtain the optimal design parameter combination; Optimize the aircraft structure according to the optimal design parameter combination.

2. The aircraft structure optimization method according to claim 1, characterized in that, The construction of the evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model includes the following steps: Decompose the crash kinetic energy dissipation process of the aircraft fuselage frame section in the lower cargo area and lower passenger cabin area, allocate the energy absorption, deformation stiffness allowed by the structural design, and remaining load-bearing strength to each component, and perform simulation analysis on the finite element model; Extract the load-displacement response curve from the simulation analysis results, collect the peak load, load efficiency, and specific energy absorption rate according to the load-displacement response curve, and quantify the peak load, load efficiency, and specific energy absorption rate into comprehensive evaluation indicators to obtain the evaluation system for the hierarchical crashworthiness level of the fuselage frame section.

3. The aircraft structure optimization method according to claim 1, wherein The construction of the initial data set of key variables for aircraft structural design is specifically as follows: Based on non-linear correlation analysis, select key variables in the structural design process, establish a sample space of structural design key variables, and form an initial data set; The key variables include the structural thickness and ply angle of aircraft components.

4. The aircraft structure optimization method according to claim 3, wherein Using the Generative Adversarial Network (GAN) algorithm to perform data augmentation operations on the initial data set to obtain an augmented data set includes the following steps: Input the data in the initial data set into the Generative Adversarial Network (GAN) algorithm, and perform iterative processing on the input data through the generative model and discriminant model of GAN to expand the data in the initial data set. When the discrimination probability of the generative model and the discriminant model reaches the 0.5 threshold, an augmented data set is obtained.

5. The aircraft structure optimization method according to claim 1, characterized in that The construction of the finite element model of the aircraft fuselage frame section is specifically as follows: Establish a numerical model of the aircraft fuselage frame section, simplify the numerical model, simplify the non-primary load-bearing upper structure of the aircraft fuselage frame section into mass nodes and beam elements, and build a finite element model of the aircraft fuselage frame section.

6. The aircraft structure optimization method according to claim 1, wherein Before using the Generative Adversarial Network (GAN) algorithm to perform data augmentation operations on the initial data set, it also includes normalizing the data in the initial data set.

7. An aircraft structure optimization system, characterized in that, It includes: An evaluation system acquisition module is used to build a finite element model of the aircraft fuselage frame section, analyze the load transfer path during the aircraft crash impact process, establish an energy absorption performance decomposition model for the lower cargo area and the lower passenger cabin area of the fuselage frame section, and construct an evaluation system for the hierarchical crashworthiness level of the fuselage frame section according to the energy absorption performance decomposition model; A data set enhancement module is used to construct an initial data set of key variables for aircraft structural design, and perform data enhancement operations on the initial data set using the generative adversarial network algorithm GAN to obtain an enhanced data set; A structure optimization module is used to optimize the initial weights and threshold parameters of the prediction neural network model BP through a parameter optimization algorithm to obtain an improved prediction neural network model. Input the data in the enhanced data set into the improved prediction neural network model, and iterate the improved prediction neural network model with the goal of maximizing the evaluation index in the evaluation system to obtain an optimal combination of design parameters; Optimize the aircraft structure according to the optimal combination of design parameters.

8. A computer device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the aircraft structure optimization method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the aircraft structure optimization method according to any one of claims 1 to 6.

Images