Three-dimensional aircraft aerodynamic configuration feature construction and aerodynamic force prediction method based on cross section area

Through a three-dimensional aircraft aerodynamic profile feature construction method based on cross-sectional area, combined with principal component analysis and machine learning model, the problems of low aerodynamic prediction efficiency and inconsistent feature expression are solved, and efficient and accurate aerodynamic prediction is achieved.

CN118886360APending Publication Date: 2024-11-01AVIC SHENYANG AERODYNAMICS RES INST
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202411012426.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-01

AI Technical Summary

Technical Problem

The aerodynamic prediction methods in the prior art are inefficient and cannot uniformly express the geometric characteristics of different topological profiles, resulting in wasted aerodynamic prediction accuracy and computing resources.

Method used

Using a three-dimensional aircraft aerodynamic profile characteristic construction method based on cross-sectional area, geometric digital model is generated through the aircraft aerodynamic concept design software, numerical simulation and aerodynamic calculation are performed, and the dimensionality reduction cross-sectional area characteristics are analyzed using principal components, and a machine learning model is constructed for gas. Dynamic prediction.

Benefits of technology

It realizes unified expression and aerodynamic prediction of different topological appearances, reduces the demand for computing resources, and improves the efficiency and accuracy of aerodynamic prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118886360A_ABST
    Figure CN118886360A_ABST
Patent Text Reader

Abstract

The invention discloses a three-dimensional aircraft aerodynamic configuration feature construction and aerodynamic force prediction method based on a cross section area, and belongs to the technical field of aircraft aerodynamic force prediction. The problems that in the prior art, a traditional aerodynamic force prediction method is low in obtaining efficiency, and different aerodynamic shape geometric features cannot be expressed in a unified mode are solved. The method comprises the following steps: carrying out numerical simulation on aerodynamic configuration geometric mathematical models of aircrafts with different aerodynamic configurations to obtain flight condition characteristics of the aircrafts and aerodynamic coefficients of all aerodynamic configurations and the aircrafts under flight conditions; creating a plane to intercept the outer contour of the aerodynamic configuration of the aircraft, obtaining the cross section area of the aircraft, and carrying out feature dimension reduction on the cross section area by using principal component analysis to obtain the aerodynamic configuration features of the aircraft after dimension reduction; and selecting a machine learning algorithm for modeling according to the aerodynamic force training data set to obtain an aerodynamic force prediction model, and training the aerodynamic force prediction model to obtain a predicted aerodynamic force coefficient. The aerodynamic force prediction efficiency is effectively improved, and the method can be applied to wind tunnel tests.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a three-dimensional aircraft aerodynamic shape feature structure and an aerodynamic force prediction method, and in particular to a three-dimensional aircraft aerodynamic shape feature structure and an aerodynamic force prediction method based on cross-sectional area, belonging to the technical field of aircraft aerodynamic force prediction. Background Art

[0002] The acquisition of aerodynamic force in aircraft design is the focus of design. There are usually three means to obtain aerodynamic force in the aircraft design stage, including theoretical analysis, wind tunnel test and CFD (numerical simulation) calculation. Among the three means, the theoretical analysis method is suitable for simple shapes and uses a linear superposition and correction model. The accuracy of its aerodynamic force calculation results is low; the wind tunnel test uses a scaled model in a special artificial airflow facility, namely a wind tunnel, and obtains aerodynamic force through simulation tests according to the similarity law. The wind tunnel test has high accuracy, but the model processing and test cycle is long and expensive; CFD calculation discretizes the continuous flow field around the aircraft into a finite number of calculation points, uses the NS equation to describe the physical law that the aircraft turbulence conforms to, and applies the partial differential equation solution method to calculate the flow field and aerodynamic force. The obtained aerodynamic force has high accuracy, does not require model processing, and the acquisition cycle and cost are relatively less than wind tunnel tests, but it still requires supercomputing for time-consuming calculations, and the cost and time consumed for high-precision and high-fidelity calculations are no less than those for wind tunnel tests; therefore, efficiently and high-quality acquisition of aerodynamic force of a given shape aircraft is the goal that is always pursued in aircraft design.

[0003] With the rise of artificial intelligence technology, the use of data-driven methods to conduct aerodynamic research has provided another means for efficient and high-quality prediction of aerodynamic forces. For example, in optimization design, the aerodynamic layout parameterization method is used to describe the aerodynamic shape of the aircraft, the Kriging method is used to model the aerodynamic force of the aircraft given the aerodynamic layout parameters, and the optimization algorithm is used to optimize the aerodynamic shape parameters. However, this aerodynamic layout parameterization method can only describe the aerodynamic shape with unchanged topology, and cannot provide a unified geometric feature description for the aerodynamic layouts of different topological shapes. Therefore, its scope of use is limited.

[0004] In order to describe the geometric features of different topological shapes in a unified way, geometric deep learning methods have been successfully introduced, such as symbolic example functions, voxels, point clouds, etc. These methods can describe a wide variety of geometric shapes in a unified way. The data is usually represented by a large matrix, and then a convolutional neural network is used to extract its deep features. However, the data size is large and feature extraction requires a professional graphics processor to perform long-term calculations.

[0005] In summary, a method for constructing aerodynamic shape features and predicting aerodynamic forces is needed that can uniformly express different topological shapes, has a small scale of expression data, has a small amount of dimensionality reduction feature extraction calculation, has low hardware requirements, and has a uniform geometric expression. Summary of the invention

[0006] A brief overview of the present invention is provided below in order to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify key or important parts of the present invention, nor is it intended to limit the scope of the present invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to a more detailed description discussed later.

[0007] In view of this, in order to solve the problem that traditional aerodynamic force prediction methods in the prior art have low acquisition efficiency and different aerodynamic shape geometric characteristics cannot be expressed uniformly, the present invention provides a three-dimensional aircraft aerodynamic shape feature construction and aerodynamic force prediction method based on cross-sectional area.

[0008] The technical solution is as follows: A method for constructing aerodynamic shape characteristics of a three-dimensional aircraft based on cross-sectional area and predicting aerodynamic force, comprising the following steps:

[0009] S1. Generate geometric digital models of aircraft aerodynamic shapes with different aerodynamic shapes using aircraft aerodynamic concept design software;

[0010] S2. numerically simulate the geometric numerical models of the aerodynamic shapes of aircraft with different aerodynamic shapes through an aerodynamic solver to obtain different flight conditions of the aircraft, obtain the flight condition characteristics of the aircraft and the aerodynamic coefficients of all aerodynamic shapes and aircraft flight conditions;

[0011] S3. Create a plane to intercept the outer contour of the aerodynamic shape of the aircraft, obtain the cross-sectional area of ​​the aircraft, use principal component analysis to reduce the dimension of the cross-sectional area, and obtain the aerodynamic shape characteristics of the aircraft after the dimension reduction;

[0012] S4. Using the reduced-dimensional aircraft aerodynamic shape features and aircraft flight condition features as input features, and using the corresponding true aerodynamic coefficients as output labels, to construct an aerodynamic training data set;

[0013] S5. Select a machine learning algorithm for modeling based on the aerodynamic training data set to obtain an aerodynamic prediction model, train the aerodynamic prediction model, and input the reduced-dimensional aircraft aerodynamic shape characteristics and aircraft flight condition characteristics to obtain predicted aerodynamic coefficients.

[0014] Furthermore, the S1 specifically includes the following steps:

[0015] S11. Determine the layout of the selected aircraft aerodynamic shape;

[0016] S12. Use aircraft aerodynamic concept design software to construct a basic shape corresponding to the layout form;

[0017] S13. Design basic parameters of the aircraft components according to the basic shape to obtain unified design parameters;

[0018] Specifically: The basic parameters include the span of the wing, the position of the wing inflection point, the root chord length, the chord length and the twist angle at both ends of each wing section, the dihedral angle of each wing section, the airfoil parameters, etc.

[0019] S14. Determine the design parameters of the selected aircraft that are different from the unified design parameters, calculate the range of change of the design parameters into the unified design parameters, and generate a design parameter sample using Latin hypercube sampling;

[0020] S15. Based on the design parameter samples, generate the aircraft aerodynamic shape geometric digital models of different aerodynamic shapes corresponding to all the design parameter samples in the aircraft aerodynamic concept design software.

[0021] Furthermore, the step S2 specifically includes the following steps:

[0022] S21. Establish an automatic grid generation script to perform aerodynamic calculation grid division on the geometric numerical model of the aerodynamic shape of the aircraft with different aerodynamic shapes, and generate aerodynamic calculation grids for all aerodynamic shapes;

[0023] S22. Select an aerodynamic solver to read the aerodynamic calculation grids of all aerodynamic shapes, set different aircraft flight conditions, solve the aerodynamic coefficients through the aerodynamic solver, traverse the aerodynamic calculation grids of all aerodynamic shapes and aircraft flight conditions, and generate the aerodynamic coefficients under all aerodynamic shapes and aircraft flight conditions.

[0024] Furthermore, the step S3 specifically includes the following steps:

[0025] S31. Construct a three-dimensional space coordinate system, and create N planes perpendicular to the X-axis, Y-axis and Z-axis, respectively, N ≥ 100, the starting position of the plane is the minimum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, and the end position is the maximum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, that is, 3N planes are obtained;

[0026] S32. Use the created 3N planes to intercept the outer contour of the aircraft's aerodynamic shape in the X-axis, Y-axis and Z-axis directions to obtain all cross-sectional areas of the aircraft, and use the row vector data composed of the 3N cross-sectional area data in the X-axis, Y-axis and Z-axis directions as the cross-sectional area characteristics of the aircraft's aerodynamic geometry;

[0027] S33. Using principal component analysis to perform feature dimensionality reduction on the cross-sectional area characteristics of the aerodynamic geometry of the aircraft, and obtaining new features of the aerodynamic shape of the aircraft after dimensionality reduction;

[0028] Specifically: 3N cross-sectional area data in the X-axis, Y-axis, and Z-axis directions were standardized to ensure that each variable had a similar scale;

[0029] The standardization process is expressed as:

[0030]

[0031] Among them, X 标准化 is the cross-sectional area after normalization, X is the cross-sectional area, is the mean of all cross-sectional areas, S is the standard deviation of the cross-sectional areas;

[0032] According to the standardized cross-sectional area X 标准化 , integrate to obtain the standardized cross-sectional area data matrix X, and calculate the covariance matrix C between the cross-sectional area data;

[0033] The covariance matrix C is expressed as:

[0034]

[0035] Where m is the number of cross sections and T is the matrix transpose sign;

[0036] Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and corresponding eigenvectors. Sort the eigenvectors according to the size of the eigenvalues, and select the first K eigenvectors as principal components. The corresponding K eigenvalues ​​are the aerodynamic shape characteristics of the aircraft after dimensionality reduction.

[0037] Further, in S5, the aerodynamic training data set is randomly divided, wherein 64% is a training set, 16% is a validation set, and 20% is a test set;

[0038] Select a machine learning algorithm, that is, construct an artificial neural network. The artificial neural network includes 1 input layer, 2 hidden layers and 1 output layer. Each hidden layer contains 200 neurons, and the output layer has 100 neurons. The hidden layer uses the ReLU activation function, establishes a Gaussian process, selects the radial basis function as the kernel function, and selects the gradient descent method as the optimization method;

[0039] Using machine learning algorithms, training is carried out on the training set, verification is carried out on the verification set, an aerodynamic prediction model is generated, and testing is carried out on the test set. The reduced-dimensional aircraft aerodynamic shape characteristics and aircraft flight condition characteristics of the samples in the test set are input into the aerodynamic prediction model, and the corresponding predicted aerodynamic coefficients are output.

[0040] The beneficial effects of the present invention are as follows: based on the geometric digital model of the aerodynamic shape of the aircraft, the present invention extracts the cross-sectional area in three directions, and performs PCA (principal component analysis) dimensionality reduction on the cross-sectional area characteristics in three directions, and then uses the reduced dimensionality features as the geometric features of the aircraft, which solves the problem of inconsistent geometric feature expression methods and parameters of aerodynamic shapes with different topologies, and effectively associates aerodynamic shapes with different topologies; the present invention uses the CFD method to carry out aerodynamic simulation calculations on aircraft with different shapes to obtain aerodynamic data labels corresponding to the aircraft, adopts a machine learning model, takes the cross-sectional area characteristics in three directions after dimensionality reduction and the flight condition characteristics of the aircraft as the input of the machine learning model, takes the corresponding aerodynamic coefficients as the output labels, establishes an aerodynamic prediction model, and realizes aerodynamic prediction; compared with other general geometric feature expression methods, such as SDF (signed distance field), point cloud and other methods, the present invention has the advantages of clear physical meaning, concise expression, and small amount of calculation. The present invention can effectively reduce the resource requirements for modeling and training of aerodynamic prediction models, and improve the efficiency of aerodynamic prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0042] Figure 1 It is a flow chart of a method for constructing aerodynamic shape characteristics of a three-dimensional aircraft and predicting aerodynamic force based on cross-sectional area;

[0043] Figure 2 The present invention is a schematic flow chart of an embodiment of a method for constructing aerodynamic shape characteristics of a three-dimensional aircraft based on cross-sectional area and predicting aerodynamic force;

[0044] Figure 3 The schematic diagram of an embodiment of a process for extracting features of a cross-sectional area of ​​an aircraft;

[0045] Figure 4 Schematic diagram of an embodiment of the comparison between the actual aerodynamic coefficient and the predicted aerodynamic coefficient. DETAILED DESCRIPTION

[0046] In order to make the technical solutions and advantages of the embodiments of the present invention more clearly understood, the exemplary embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0047] refer to Figure 1-Figure 4The present embodiment is described in detail. A method for constructing aerodynamic shape characteristics of a three-dimensional aircraft based on cross-sectional area and predicting aerodynamic force specifically includes the following steps:

[0048] S1. Generate geometric digital models of aircraft aerodynamic shapes with different aerodynamic shapes using aircraft aerodynamic concept design software;

[0049] S2. numerically simulate the geometric numerical models of the aerodynamic shapes of aircraft with different aerodynamic shapes through an aerodynamic solver to obtain different flight conditions of the aircraft, obtain the flight condition characteristics of the aircraft and the aerodynamic coefficients of all aerodynamic shapes and aircraft flight conditions;

[0050] S3. Create a plane to intercept the outer contour of the aerodynamic shape of the aircraft, obtain the cross-sectional area of ​​the aircraft, use principal component analysis to reduce the dimension of the cross-sectional area, and obtain the aerodynamic shape characteristics of the aircraft after the dimension reduction;

[0051] S4. Using the reduced-dimensional aircraft aerodynamic shape features and aircraft flight condition features as input features, and using the corresponding true aerodynamic coefficients as output labels, to construct an aerodynamic training data set;

[0052] S5. Select a machine learning algorithm for modeling based on the aerodynamic training data set to obtain an aerodynamic prediction model, train the aerodynamic prediction model, and input the reduced-dimensional aircraft aerodynamic shape characteristics and aircraft flight condition characteristics to obtain predicted aerodynamic coefficients.

[0053] Furthermore, the S1 specifically includes the following steps:

[0054] S11. Determine the layout of the selected aircraft aerodynamic shape;

[0055] S12. Use aircraft aerodynamic concept design software to construct a basic shape corresponding to the layout form;

[0056] S13. Design basic parameters of the aircraft components according to the basic shape to obtain unified design parameters;

[0057] Specifically: The basic parameters include the span of the wing, the position of the wing inflection point, the root chord length, the chord length and the twist angle at both ends of each wing section, the dihedral angle of each wing section, the airfoil parameters, etc.

[0058] S14. Determine the design parameters of the selected aircraft that are different from the unified design parameters, calculate the range of change of the design parameters into the unified design parameters, and generate a design parameter sample using Latin hypercube sampling;

[0059] S15. Generate aerodynamic geometry models of different aerodynamic shapes of the aircraft corresponding to all design parameter samples in the aircraft aerodynamic concept design software according to the design parameter samples;

[0060] Specifically, conceptual design software includes OpenVSP, CEASIOM, etc. The layout form, for example, the three-wing V-tail layout, divides the aerodynamic shape into basic components such as the fuselage, wing, and air intake.

[0061] Furthermore, the step S2 specifically includes the following steps:

[0062] S21. Establish an automatic grid generation script to perform aerodynamic calculation grid division on the geometric numerical model of the aerodynamic shape of the aircraft with different aerodynamic shapes, and generate aerodynamic calculation grids for all aerodynamic shapes;

[0063] S22. Select an aerodynamic solver to read the aerodynamic calculation grids of all aerodynamic shapes, set different aircraft flight conditions, solve the aerodynamic coefficients through the aerodynamic solver, traverse the aerodynamic calculation grids of all aerodynamic shapes and aircraft flight conditions, and generate aerodynamic coefficients under all aerodynamic shapes and aircraft flight conditions;

[0064] Specifically, pneumatic solvers such as Fluent, StarCCM++ and other software or other self-developed fluid pneumatic calculation tools can all be used for reading.

[0065] Furthermore, the step S3 specifically includes the following steps:

[0066] S31. Construct a three-dimensional space coordinate system, and create N planes perpendicular to the X-axis, Y-axis and Z-axis, respectively, N ≥ 100, the starting position of the plane is the minimum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, and the end position is the maximum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, that is, 3N planes are obtained;

[0067] S32. Use the created 3N planes to intercept the outer contour of the aircraft's aerodynamic shape in the X-axis, Y-axis and Z-axis directions to obtain all cross-sectional areas of the aircraft, and use the row vector data composed of the 3N cross-sectional area data in the X-axis, Y-axis and Z-axis directions as the cross-sectional area characteristics of the aircraft's aerodynamic geometry;

[0068] S33. Using principal component analysis to perform feature dimensionality reduction on the cross-sectional area characteristics of the aerodynamic geometry of the aircraft, and obtaining new features of the aerodynamic shape of the aircraft after dimensionality reduction;

[0069] Specifically: 3N cross-sectional area data in the X-axis, Y-axis, and Z-axis directions were standardized to ensure that each variable had a similar scale;

[0070] The standardization process is expressed as:

[0071]

[0072] Among them, X 标准化 is the cross-sectional area after normalization, X is the cross-sectional area, is the mean of all cross-sectional areas, S is the standard deviation of the cross-sectional areas;

[0073] According to the standardized cross-sectional area X 标准化 , integrate to obtain the standardized cross-sectional area data matrix X, and calculate the covariance matrix C between the cross-sectional area data;

[0074] The covariance matrix C is expressed as:

[0075]

[0076] Where m is the number of cross sections and T is the matrix transpose sign;

[0077] Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and corresponding eigenvectors, sort the eigenvectors according to the size of the eigenvalues, select the first K eigenvectors as the principal components, and the corresponding K eigenvalues ​​are the aerodynamic shape characteristics of the aircraft after dimensionality reduction;

[0078] Specifically, in this embodiment, the value of K is 7;

[0079] In S4, the cross-sectional area after feature dimensionality reduction, i.e., the seven principal component parameters, is used as the cross-sectional area feature of the aerodynamic geometry of the aircraft, the aircraft flight condition, i.e., the Mach number, angle of attack, sideslip angle, and Reynolds number obtained in step S2, is used as the aircraft flight condition feature, and the corresponding aerodynamic coefficient is used as the output label to construct an aerodynamic training data set.

[0080] Further, in S5, the aerodynamic training data set is randomly divided, wherein 64% is a training set, 16% is a validation set, and 20% is a test set;

[0081] Select a machine learning algorithm, that is, construct an artificial neural network. The artificial neural network includes 1 input layer, 2 hidden layers and 1 output layer. Each hidden layer contains 200 neurons, and the output layer has 100 neurons. The hidden layer uses the ReLU activation function, establishes a Gaussian process, selects the radial basis function (RBF) as the kernel function, and selects the gradient descent method as the optimization method;

[0082] Using machine learning algorithms, training is carried out on the training set, verification is carried out on the verification set, an aerodynamic prediction model is generated, and testing is carried out on the test set. The aerodynamic shape characteristics and flight condition characteristics of the aircraft after dimensionality reduction of the samples in the test set are input into the aerodynamic prediction model, and the corresponding predicted aerodynamic coefficients are output;

[0083] Specifically, the machine learning algorithm includes methods such as artificial neural network and Gaussian process. In this embodiment, for the artificial neural network, the learning rate is 0.0001, the optimizer is Adam, and after the predicted aerodynamic coefficient, i.e., the predicted CL, is output, it can be compared with the actual aerodynamic coefficient, i.e., the sample true value CL in step S4 to verify the prediction effect, and the data can be discarded or stored;

[0084] When using the aerodynamic prediction model to carry out aerodynamic prediction, the given reduced-dimensional aircraft aerodynamic shape cross-sectional area characteristics and flight condition characteristics are directly input to predict the predicted aerodynamic coefficients of the aircraft under given flight conditions.

[0085] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and teaching purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is illustrative, not restrictive, with respect to the scope of the present invention, which is defined by the appended claims.

Claims

1. A method for constructing aerodynamic shape characteristics of a three-dimensional aircraft based on cross-sectional area and predicting aerodynamic force, characterized in that: The following steps are involved: S1. Generate geometric digital models of aircraft aerodynamic shapes with different aerodynamic shapes using aircraft aerodynamic concept design software; S2. numerically simulate the geometric numerical models of the aerodynamic shapes of aircraft with different aerodynamic shapes through an aerodynamic solver to obtain different flight conditions of the aircraft, obtain the flight condition characteristics of the aircraft and the aerodynamic coefficients of all aerodynamic shapes and aircraft flight conditions; S3. Create a plane to intercept the outer contour of the aerodynamic shape of the aircraft, obtain the cross-sectional area of ​​the aircraft, use principal component analysis to reduce the dimension of the cross-sectional area, and obtain the aerodynamic shape characteristics of the aircraft after the dimension reduction; S4. Using the reduced-dimensional aircraft aerodynamic shape features and aircraft flight condition features as input features, and using the corresponding true aerodynamic coefficients as output labels, to construct an aerodynamic training data set; S5. Select a machine learning algorithm for modeling based on the aerodynamic training data set to obtain an aerodynamic prediction model, train the aerodynamic prediction model, and input the reduced-dimensional aircraft aerodynamic shape characteristics and aircraft flight condition characteristics to obtain predicted aerodynamic coefficients.

2. A method for constructing aerodynamic shape characteristics of a three-dimensional aircraft and predicting aerodynamic force based on cross-sectional area according to claim 1, characterized in that: The S1 specifically includes the following steps: S11. Determine the layout of the selected aircraft aerodynamic shape; S12. Use aircraft aerodynamic concept design software to construct a basic shape corresponding to the layout form; S13. Design basic parameters of the aircraft components according to the basic shape to obtain unified design parameters; Specifically: The basic parameters include the span of the wing, the position of the wing inflection point, the root chord length, the chord length and the twist angle at both ends of each wing section, the dihedral angle of each wing section, the airfoil parameters, etc. S14. Determine the design parameters of the selected aircraft that are different from the unified design parameters, calculate the range of change of the design parameters into the unified design parameters, and generate a design parameter sample using Latin hypercube sampling; S15. Based on the design parameter samples, generate aircraft aerodynamic shape geometric digital models of different aerodynamic shapes corresponding to all design parameter samples in the aircraft aerodynamic concept design software.

3. A method for constructing aerodynamic shape characteristics and predicting aerodynamic force of a three-dimensional aircraft based on cross-sectional area according to claim 2, characterized in that: The S2 specifically includes the following steps: S21. Establish an automatic grid generation script to perform aerodynamic calculation grid division on the geometric numerical model of the aerodynamic shape of the aircraft with different aerodynamic shapes, and generate aerodynamic calculation grids for all aerodynamic shapes; S22. Select an aerodynamic solver to read the aerodynamic calculation grids of all aerodynamic shapes, set and calculate different aircraft flight conditions, solve the aerodynamic coefficients through the aerodynamic solver, traverse the aerodynamic calculation grids of all aerodynamic shapes and aircraft flight conditions, and generate the aerodynamic coefficients under all aerodynamic shapes and aircraft flight conditions.

4. A method for constructing aerodynamic shape characteristics of a three-dimensional aircraft and predicting aerodynamic force based on cross-sectional area according to claim 3, characterized in that: The S3 specifically includes the following steps: S31. Construct a three-dimensional space coordinate system, and create N planes perpendicular to the X-axis, Y-axis and Z-axis, respectively, N ≥ 100, the starting position of the plane is the minimum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, and the end position is the maximum value of the X-axis, Y-axis and Z-axis coordinates of the aircraft aerodynamic shape geometry model, that is, 3N planes are obtained; S32. Use the created 3N planes to intercept the outer contour of the aircraft's aerodynamic shape in the X-axis, Y-axis and Z-axis directions to obtain all cross-sectional areas of the aircraft, and use the row vector data composed of the 3N cross-sectional area data in the X-axis, Y-axis and Z-axis directions as the cross-sectional area characteristics of the aircraft's aerodynamic geometry; S33. Using principal component analysis to perform feature dimensionality reduction on the cross-sectional area characteristics of the aerodynamic geometry of the aircraft, and obtaining new features of the aerodynamic shape of the aircraft after dimensionality reduction; Specifically: 3N cross-sectional area data in the X-axis, Y-axis, and Z-axis directions were standardized to ensure that each variable had a similar scale; The standardization process is expressed as: Among them, X 标准化 is the cross-sectional area after normalization, X is the cross-sectional area, is the mean of all cross-sectional areas, S is the standard deviation of the cross-sectional areas; According to the standardized cross-sectional area X 标准化 , integrate to obtain the standardized cross-sectional area data matrix X, and calculate the covariance matrix C between the cross-sectional area data; The covariance matrix C is expressed as: Where m is the number of cross sections and T is the matrix transpose sign; Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and corresponding eigenvectors. Sort the eigenvectors according to the size of the eigenvalues, and select the first K eigenvectors as principal components. The corresponding K eigenvalues ​​are the aerodynamic shape characteristics of the aircraft after dimensionality reduction.

5. A method for constructing aerodynamic shape characteristics and predicting aerodynamic force of a three-dimensional aircraft based on cross-sectional area according to claim 4, characterized in that: In S5, the aerodynamic training data set is randomly divided, wherein 64% is a training set, 16% is a validation set, and 20% is a test set; Select a machine learning algorithm, that is, construct an artificial neural network. The artificial neural network includes 1 input layer, 2 hidden layers and 1 output layer. Each hidden layer contains 200 neurons, and the output layer has 100 neurons. The hidden layer uses the ReLU activation function, establishes a Gaussian process, selects the radial basis function as the kernel function, and selects the gradient descent method as the optimization method; Using machine learning algorithms, training is carried out on the training set, verification is carried out on the verification set, an aerodynamic prediction model is generated, and testing is carried out on the test set. The reduced-dimensional aircraft aerodynamic shape characteristics and aircraft flight condition characteristics of the samples in the test set are input into the aerodynamic prediction model, and the corresponding predicted aerodynamic coefficients are output.

Citation Information

Cited By

  • Aircraft aerodynamic characteristic intelligent reconstruction method based on multi-source data fusion

    CN119337510A

  • Training method and device for aerodynamic force prediction model of aircraft

    CN119808634A

  • Training method and device for aircraft aerodynamic prediction model

    CN119808634B

  • Aircraft three-dimensional aerodynamic configuration design method, device, equipment and medium

    CN120086985A

  • Intelligent aerodynamic force prediction model for three-dimensional aircraft and transfer learning method

    CN120197511A