Aviation dispenser airfoil high-dimensional pneumatic optimization method and system based on data-driven support strategy

Abstract

The invention discloses an aviation dispenser airfoil high-dimensional pneumatic optimization method and system based on a data-driven support strategy, relates to and is applicable to the field of aviation dispenser airfoil optimization, and is used for solving the problem of modeling difficulty in existing pneumatic airfoil optimization based on computational fluid dynamics and numerical optimization. According to the method, a high-dimensional CST method is used for parameterizing the missile wing shape of the aviation dispenser, then a geometric domain decomposition algorithm is adopted for obtaining a low-dimensional geometric mode through POD analysis, the missile wing shape of the aviation dispenser is parameterized again, the design dimension is reduced, and therefore optimization and modeling difficulties are reduced. According to the method, a new data-driven support strategy is adopted, the aerodynamic data of the missile wing of the aviation dispenser are fully utilized through a DDSS algorithm, useful information is extracted from the aerodynamic data to construct a low-cost agent model, and meanwhile, the prediction convergence precision of the agent model is effectively improved through a teaching optimization algorithm. Dependence on the accuracy of a real model can be effectively reduced, and the difficulty of high-dimensional modeling is overcome.

Description

Technical Field

[0001] The present invention relates to the field of aviation dispenser airfoil optimization, and in particular to a high-dimensional aerodynamic optimization method and system for aviation dispenser airfoil based on a data-driven support strategy. Background Art

[0002] In the current environment full of uncertainty and complexity, traditional equipment faces unprecedented challenges. Against this background, standoff systems came into being. They can launch attacks at a safe distance away from the enemy's air defense system, significantly reducing the risk to one's own personnel. Among them, the aerial dispenser, as an important standoff system, is highly valued because it can carry a variety of ammunition and disperse it over the target area. At the same time, as an unpowered aircraft, the aerodynamic shape of the aerial dispenser largely determines the combat effectiveness of the aerial dispenser. Optimizing the airfoil can improve the lift-to-drag ratio of the aerial dispenser, thereby enhancing its aerodynamic performance and achieving a longer range and higher maneuverability.

[0003] Aerodynamic shape optimization (ASO) based on computational fluid dynamics (CFD) and numerical optimization is a key research topic in aerodynamic design. So far, a variety of optimization frameworks have been developed to solve different aerodynamic shape optimization problems. However, these optimization frameworks are effective in designing two-dimensional shapes, but face difficulties when applied to complex three-dimensional configurations (such as missile wings, hybrid wing bodies, etc.), which require high-dimensional design variables and make modeling difficult. Summary of the invention

[0004] The present invention is used to solve the problem that high-dimensional design variables are required and modeling is difficult in the existing aerodynamic airfoil optimization based on computational fluid dynamics and numerical optimization.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a high-dimensional aerodynamic optimization method for an aviation dispenser airfoil based on a data-driven support strategy, the method comprising the following steps:

[0007] Step S1: parameterize the airfoil of the aerial dispenser wing using the CST parameterization method and construct the geometric shape of the aerial dispenser wing;

[0008] Step S2: Using the geometric domain decomposition algorithm to reduce the dimension of the constructed geometric shape into a low-dimensional geometric mode through POD analysis, and extract the geometric features of the air dispenser wing. After RBF-based mesh deformation, the CFD initial evaluation results are obtained and put into the historical air dispenser wing aerodynamic parameter database;

[0009] Step S3: Using the DDSS algorithm, a proxy model of the aerial dispenser wing is constructed according to the historical aerodynamic parameter database of the aerial dispenser wing. At the same time, the teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model. By comparing with the actual CFD evaluation results, the shape parameters of the aerial dispenser wing are continuously iteratively optimized until convergence is achieved and the optimal shape parameters are obtained.

[0010] Furthermore, there is a preferred embodiment in which the geometric shape of the aerial dispenser wing is represented by the class function C(ψ) and the shape function S(ψ) of the CST parameterization method.

[0011] The geometry of the air dispenser wing is expressed as:

[0012]

[0013] Among them, ξ represents the dimensionless surface coordinate, ψ (0≤ψ≤1) represents the dimensionless chord coordinate, and ΔZ te / c represents the dimensionless trailing edge thickness.

[0014] Furthermore, there is another preferred embodiment, the above step S2 is specifically as follows:

[0015] Step S21: Generate a series of original sample vectors s of the wings of the aerial dispenser using the high-dimensional CST method and Latin hypercube sampling i (i=1,2,…,N) to construct the sampling result matrix S=s 1 s 2 …s N ,s i Represents the coordinates of the air dispenser wing;

[0016] Step S22: Calculate the average vector based on the sampling result matrix

[0017] Step S23 is based on the average vector The correlation matrix R is constructed by calculating the deviation matrix, and the feature extraction of the correlation matrix R is performed to obtain a series of geometric feature patterns v i and eigenvalue λ i ;

[0018] Step S24: Based on the average vector The geometric shape of the air dispenser wing is obtained by approximating the original sample vector through the geometric characteristic pattern.

[0019] Furthermore, the geometric shape of the air dispenser wing after dimensionality reduction is expressed as:

[0020]

[0021] Among them, α i is the modal coefficient of the i-th order vector.

[0022] Furthermore, the calculation formula for extracting information from the geometric characteristics of the air dispenser wing is as follows:

[0023]

[0024] Where P is the information ratio, M is the number of characteristic functions, and N is the number of high-order modes.

[0025] Furthermore, there is another preferred embodiment, the above step S3 is specifically as follows:

[0026] Step S31: using the DDSS algorithm to extract valid data from the historical aerodynamic data of the aerial dispenser wing, building an aerodynamic parameter database of the aerial dispenser wing in combination with the CFD evaluation results, and building a proxy model of the aerial dispenser wing based on the aerodynamic parameter database;

[0027] Step S32: using a teaching optimization algorithm to obtain the best shape parameters in the proxy model, and obtaining a CFD evaluation result after mesh deformation based on the RBF;

[0028] Step S33: The teaching optimization algorithm continuously optimizes the teaching stage and the student stage according to the CFD evaluation results until the obtained aerial dispenser wing shape parameters converge and the optimal shape parameters are obtained.

[0029] The method for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy described in the present invention can be fully implemented using computer software. Therefore, correspondingly, the present invention also provides a system for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy. The system includes a storage device, and the storage device is used to implement the above-mentioned method for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy.

[0030] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, any one of the above-mentioned high-dimensional aerodynamic optimization methods for an aerial dispenser airfoil based on a data-driven support strategy is executed.

[0031] The present invention also provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes any one of the above-mentioned high-dimensional aerodynamic optimization methods for an aviation dispenser airfoil based on a data-driven support strategy.

[0032] The beneficial effects of the present invention are:

[0033] The present invention proposes a high-dimensional aerodynamic optimization method for the airfoil of an aerial dispenser based on a data-driven support strategy. The airfoil of the aerial dispenser is parameterized by using a high-dimensional CST method, and then a low-dimensional geometric model is obtained through POD analysis using a geometric domain decomposition algorithm to re-parameterize the shape of the aerial dispenser wing, thereby reducing the design dimension and reducing the difficulty of optimization and modeling. In addition, the present invention also adopts a new data-driven support strategy (DDSS), which fully utilizes the aerodynamic data of the aerial dispenser wing through the DDSS algorithm, and extracts useful information from it to build a low-cost proxy model. At the same time, the teaching-learning based optimization algorithm (TLBO) is used to effectively improve the prediction convergence accuracy of the proxy model. It can effectively reduce the dependence on the accuracy of the real model and overcome the difficulties of high-dimensional modeling.

[0034] The invention is suitable for optimizing the high-dimensional aerodynamics of an aviation dispenser airfoil. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 is the unstructured computing grid described in the present invention;

[0037] Figure 2 It is a high-dimensional aerodynamic shape optimization framework structure diagram of the present invention;

[0038] Figure 3 Schematic diagram of the present invention for representing the geometric shape of the wings of an aerial dispenser using a high-dimensional CST method, wherein (a) is a class function and (b) is a shape function;

[0039] Figure 4 The three-dimensional CST parameterization process of the aerial dispenser wing of the present invention;

[0040] Figure 5 It is the teaching optimization algorithm process based on "teaching" and "learning" described in the present invention;

[0041] Figure 6 It is a flow chart of the teaching optimization algorithm described in the present invention;

[0042] Figure 7is a detailed aerodynamic shape optimization (ASO) framework for aerodynamic design as described in the present invention;

[0043] Figure 8 is a flow chart of the direct optimization framework of the present invention;

[0044] Fig. 9 It is the comparative convergence process of the drag coefficient (Cd) described in the present invention. DETAILED DESCRIPTION

[0045] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0046] Implementation method 1, see Figure 2 This embodiment describes a high-dimensional aerodynamic optimization method for an air dispenser airfoil based on a data-driven support strategy, the method comprising the following steps:

[0047] Step S1: parameterize the airfoil of the aerial dispenser wing using the CST parameterization method and construct the geometric shape of the aerial dispenser wing;

[0048] Step S2: Using the geometric domain decomposition algorithm to reduce the dimension of the constructed geometric shape into a low-dimensional geometric mode through POD analysis, and extract the geometric features of the air dispenser wing. After RBF-based mesh deformation, the CFD initial evaluation results are obtained and put into the historical air dispenser wing aerodynamic parameter database;

[0049] Step S3: Using the DDSS algorithm, a proxy model of the aerial dispenser wing is constructed according to the historical aerodynamic parameter database of the aerial dispenser wing. At the same time, the teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model. By comparing with the actual CFD evaluation results, the shape parameters of the aerial dispenser wing are continuously iteratively optimized until convergence is achieved and the optimal shape parameters are obtained.

[0050] In practical application, if Figure 2As shown in the figure, the airfoil of the aerial dispenser wing is first parameterized using the high-dimensional CST method, and then the geometric domain decomposition algorithm (GDD) is used to obtain the low-dimensional geometric pattern through orthogonal decomposition (POD) to re-parameterize the shape of the aerial dispenser wing, thereby reducing the design dimension. In order to reduce the design dimension, the geometric shape is parameterized, which reduces the difficulty of optimization and modeling. At the same time, the gradient-free teaching-learning based optimization (TLBO) optimizer is used to obtain the global optimal solution, and the historical aerodynamic data generated during the optimization process is learned by DDSS to guide and accelerate the optimization process of the aerial dispenser wing shape.

[0051] Embodiment 2: This embodiment specifically describes the various steps of the high-dimensional aerodynamic optimization method for an air dispenser airfoil based on a data-driven support strategy described in the above embodiment 1;

[0052] First, the optimization basis of the airfoil of the aerial dispenser wing is explained:

[0053] For an n-dimensional aerial dispenser wing shape optimization problem, the mathematical optimization model can be expressed as:

[0054]

[0055] Where x is the design variable vector, C d and C l are the drag coefficient and lift coefficient, C l0 is the initial lift coefficient, V and V 0 are the optimized and initial air dispenser wing volumes, respectively. It can be seen that the optimization goal is to minimize the air dispenser drag coefficient under the constraints of lift coefficient and volume.

[0056] Unstructured computational grids such as Figure 1 As shown, the steady-state flow solution is calculated by combining the Reynolds-averaged Navier-Stokes equations with the Spalart-Allmaras turbulence model. This embodiment uses the unit center finite volume method for spatial discretization of the air dispenser wing, uses the symmetric Gauss-Seidel iterative time-marching scheme for pseudo-time step integration, and uses a second-order accurate fully implicit scheme to solve the physical time equation.

[0057] Step S1: parameterize the airfoil of the aerial dispenser wing using the CST parameterization method and construct the geometric shape of the aerial dispenser wing;

[0058] Specifically:

[0059] The high-dimensional CST method is used to represent the geometric shape of the air dispenser wing through the class function C(ψ) and the shape function S(ψ). The geometric shape of the air dispenser wing is expressed as:

[0060]

[0061] Among them, ξ represents the dimensionless surface coordinate, ψ (0≤ψ≤1) represents the dimensionless chord coordinate, and ΔZ te / c represents the dimensionless trailing edge thickness.

[0062] The class function C(ψ) is expressed as:

[0063]

[0064] Among them, N 1 and N 2 The basic geometric features are defined, such as Figure 3 (a) shows that different N 1 and N 2 The values ​​represent different types of geometries, N 1 = 0.5 and N 2 =1.0 represents NACA airfoil, N 1 =1.0 and N 2 =1.0 represents a double-arc airfoil, N 1 = 0.75 and N 2 =0.75 represents the Sears-Haack airfoil.

[0065] The n-order shape function S(ψ) is expressed as:

[0066]

[0067] Among them, b i (i=0,1,...,n) are the Bernstein polynomial control coefficients. The geometry of the fifth-order shape function is as follows: Figure 3 (b) as shown.

[0068] The three-dimensional CST method can be regarded as a set of two-dimensional airfoils distributed in the radial direction. The Bernstein polynomial has an additional radial distribution variable η. The three-dimensional parameterization process of the entire aerial dispenser wing is as follows: Figure 4 As shown, the upper and lower surfaces of the wing can be expressed as:

[0069]

[0070] Where n and m represent the order of the Bernstein polynomial in the chord direction and span direction respectively; Δα t Indicates the twist angle of the aerial dispenser wing.

[0071] From the above formula (5), we can see that by changing b i,j (i=0,1,...,n;j=0,1,...,m) can create a new wing shape, which means that with the increase of spanwise distribution variable η, there are 2×(n+1)·(m+1) design variables on the upper and lower surfaces. These high-dimensional design variables increase the modeling dimension and optimization difficulty of the wing shape of the aerial dispenser. In order to solve this problem, this embodiment also reduces the design dimension of the wing of the aerial dispenser by adopting the geometric domain decomposition method of POD analysis.

[0072] Step S2: Using the geometric domain decomposition algorithm to reduce the dimension of the constructed geometric shape into a low-dimensional geometric mode through POD analysis, and extract the geometric features of the air dispenser wing. After RBF-based mesh deformation, the CFD initial evaluation results are obtained and put into the historical air dispenser wing aerodynamic parameter database;

[0073] Specifically:

[0074] The geometric configuration of the air dispenser wing is extracted by using the POD analysis algorithm, and its mathematical expression is as follows:

[0075] First, a set of sampling results N is generated in the design space, and the sampling matrix S is shown as follows:

[0076] S=s 1 s 2 …s N (6)

[0078] where the vector s i is the wing coordinate of the aerial dispenser in this patent, corresponding to the i-th condition.

[0079] Then, combine the matrix sampling results to calculate the average vector

[0080]

[0081] After that, the deviation matrix is ​​calculated to prevent the absolute value of the first singular value from affecting other values. The deviation matrix X is as follows:

[0082]

[0083] The correlation matrix R is constructed as follows:

[0084] R=X T X (9)

[0086] By analyzing the correlation matrix R, we can obtain the eigenvalue λ under the i-th wing shape condition matrix i and the eigenvector vi , therefore, the geometry of the air dispenser wing can be expressed as:

[0087]

[0088] Among them, α i is the modal coefficient, and the i-th wing shape condition coefficient can be written as

[0089] Combined with the above POD wing shape analysis algorithm, the design process of the aviation dispenser wing using the geometric domain decomposition method is as follows:

[0090] First, a series of aerial dispenser wing sample vectors s are generated using the CST method with high-dimensional design variables and Latin Hypercube Sampling (LHS). i (i=1,2,…,N) to construct the sampling result matrix S=s 1 s 2 …s N , and then use the POD algorithm to extract features and obtain a series of geometric feature patterns v i and eigenvalue λ i (i=1,2,…,N). The original sampling result vector is approximated by the geometric characteristic pattern as follows:

[0091]

[0092] in, is the mean vector, α i is the modal coefficient of the i-th order vector. Usually, the first few modes are sufficient to describe the characteristics of the aerial dispenser wing sample.

[0093] The calculation formula for extracting relevant information from the geometric characteristics of the air dispenser wing is as follows:

[0094]

[0095] Where P is the information ratio, M is the number of characteristic functions, and N is the number of high-order modes.

[0096] The parameter accuracy of geometric domain decomposition using geometric weighted error analysis is:

[0097] z werror =8×10 -4 (13)

[0099] If the generated reduced-dimensional parameterized aerial dispenser wing falls within this specified tolerance range z werror, then it is considered a successful fit, and the initial air dispenser wing condition can be well approximated using only a few geometric modes (between five and ten), and adding ten or so geometric modes is enough to keep all errors within the tolerance range.

[0100] The geometric domain decomposition proposed in this embodiment is a design method that uses the CST algorithm principle to reconstruct a high-dimensional aerodynamic shape design space into a low-order geometric model. This method can effectively improve the shape optimization efficiency and reduce the difficulty of modeling.

[0101] Step S3: Using the DDSS algorithm, a proxy model of the aerial dispenser wing is constructed according to the historical aerodynamic parameter database of the aerial dispenser wing. At the same time, the teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model. By comparing with the actual CFD evaluation results, the shape parameters of the aerial dispenser wing are continuously iteratively optimized until convergence is achieved and the optimal shape parameters are obtained.

[0102] Specifically:

[0103] S31: Use DDSS algorithm to extract valid data from historical aerodynamic data of aerial dispenser wing, build aerodynamic parameter database of aerial dispenser wing in combination with CFD evaluation results, and build a proxy model of aerial dispenser wing based on the aerodynamic parameter database;

[0104] S32: The teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model, and the CFD evaluation results are obtained after the RBF-based mesh deformation;

[0105] S33: The teaching optimization algorithm continuously optimizes the teaching stage and the student stage according to the CFD evaluation results until the obtained aerial dispenser wing shape parameters converge and the optimal shape parameters are obtained.

[0106] Among them, teaching optimization algorithms such as Figure 6 As shown, the specific steps include:

[0107] Step 1: Set parameters: population size N p ; Set termination conditions, such as the maximum number of iterations G max .

[0108] Step 2: Initialize the class (population) and evaluate the knowledge level (fitness value) of the individuals.

[0109] Step 3: Select the best individual as teacher X T , the other individuals are called student X S . Calculate the student mean, that is

[0110] The formula is as follows:

[0111]

[0112] Step 4: Teaching stage. Each student in the class is assigned a specific task according to the teacher's T and the student average Update the difference between

[0113]

[0114]

[0115] where r i is a random factor, and T F is the teaching factor. The calculation formula is as follows:

[0116] r i =rand(1) (17)

[0118] T F =round1+rand(0,1) (18)

[0120] Here rand(1) means creating a one-to-one uniformly distributed random number matrix with elements between 0 and 1; rand[1+rand(0,1)] means rounding the object 1+rand(0,1) to the nearest integer.

[0121] Step 5: Learning phase. Randomly select two students from the class and Learn from each other. They can update themselves according to the following formula:

[0122]

[0123] Step 6: Determine whether the convergence condition is met. If so, the optimization process ends; otherwise, return to step 3 and continue the process.

[0124] The teaching optimization algorithm proposed in this embodiment is a gradient-free optimizer with no parameter adjustment (except for the population parameter N p) and good global search capabilities, it is a gradient-free optimizer based on the "teaching" and "learning" modes. The teaching optimization algorithm defines the concept of "knowledge level" to distinguish between students and teachers, and the overall "knowledge level" is improved in the "teacher stage" and "learner stage". That is, through the "teacher stage" and "learner stage", the overall level of the sample is improved. However, the optimizer using the teaching optimization algorithm does not make full use of historical data. During the optimization process, many individuals with poor performance were eliminated, which wasted valuable aerodynamic information of the aerial dispenser wing. Therefore, this embodiment adopts the DDSS algorithm to achieve re-learning of the historical aerial dispenser wing aerodynamic parameter data to reconstruct the proxy model. These proxy models can improve the optimization efficiency, and their structure is as follows: Figure 5 As shown in the figure, it can be seen that the DDSS algorithm can make full use of the aerodynamic data of the air dispenser wing and extract useful information from it to guide the optimization process. All individuals generated during the optimization process are collected to build a historical aerodynamic database, which is used for secondary learning to build a low-cost proxy model. The air dispenser wing proxy model acts as an auxiliary accelerator in the optimization framework to help the optimizer converge quickly.

[0125] The DDSS iterative design process is as follows:

[0126] First, select the relevant proxy model according to the requirements. Kriging is a well-known proxy model proposed by Krige and widely used in aerodynamic shape optimization. It is used to learn the historical aerodynamic data in DDSS to form a proxy model of aerodynamic shape. The online model learning strategy is adopted as the model management strategy. Since the aerodynamic data is iteratively generated and stored, the proxy model is adaptively updated through incremental data according to the online model learning strategy.

[0127] Then, the model optimal prediction algorithm is used as the main optimization loop algorithm in the teaching optimization algorithm to obtain possible aerodynamic shape "teacher" parameters, thereby realizing the search for the best individual X from the agent model. T,MOP . Evaluate the actual aerodynamic results F(X T,MOP ) and compare it with the current aerodynamic shape "teacher" parameters X T If F(X T,MOP )<F(X T ), the current aerodynamic shape "teacher" parameters are replaced by the teacher obtained by MOP; if not, the current aerodynamic shape "teacher" parameters X T The new aerodynamic shape "teacher" parameters then generate new individuals through the teaching phase.

[0128] Furthermore, this embodiment adopts the model optimal prediction algorithm as the main optimization loop algorithm in the teaching optimization algorithm. The model optimal prediction method (Model Optimal Prediction, MOP) is used to obtain the best prediction individual of the proxy model. Different from the traditional optimization algorithm based on the proxy model, the competition mechanism in the teaching optimization algorithm effectively improves the prediction convergence accuracy of the proxy model. If the proxy model converges successfully, the best prediction individual will become the new teacher in the optimization process. This new and better teacher can educate students more effectively than before, thereby accelerating the search process of the optimizer. If the proxy model fails to converge, it will be collected into the database and the optimizer will continue its normal operation. At the same time, this competition mechanism ensures that the optimizer will not be misled by poor-quality prediction individuals, thereby significantly reducing the dependence of the optimization results on the accuracy of the proxy model.

[0129] Implementation method three, see Figure 7 This embodiment is described. This embodiment is a detailed description of the high-dimensional aerodynamic optimization method of an air dispenser airfoil based on a data-driven support strategy described in the above embodiment.

[0130] like Figure 7 As shown in Figure 1, the first few main wing geometries are parameterized using geometric domain decomposition, and then the RBF-based mesh deformation method is used to update the aerodynamic mesh after the new shape is generated. The aerodynamic performance of the air dispenser wing is evaluated through CFD simulation. The teaching optimization algorithm optimizer drives a large number of aerodynamic evaluations to search for the best aerodynamic shape. DDSS assists the teaching optimization algorithm optimizer to improve its search capability. The detailed pseudo code for reducing the drag of the air dispenser wing is shown in Table 1:

[0131] Table 1

[0132]

[0133]

[0134]

[0135] Embodiment 4: The method for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy described in the above embodiment can be implemented entirely by computer software. Therefore, correspondingly, this embodiment provides a high-dimensional aerodynamic optimization system for an aerial dispenser airfoil based on a data-driven support strategy. The optimization system includes a storage device, and the storage device is used to perform the following steps:

[0136] Step S1: parameterize the airfoil of the aerial dispenser wing using the CST parameterization method and construct the geometric shape of the aerial dispenser wing;

[0137] Step S2: Using the geometric domain decomposition algorithm to reduce the dimension of the constructed geometric shape into a low-dimensional geometric mode through POD analysis, and extract the geometric features of the air dispenser wing. After RBF-based mesh deformation, the CFD initial evaluation results are obtained and put into the historical air dispenser wing aerodynamic parameter database;

[0138] Step S3: Using the DDSS algorithm, a proxy model of the aerial dispenser wing is constructed according to the historical aerodynamic parameter database of the aerial dispenser wing. At the same time, the teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model. By comparing with the actual CFD evaluation results, the shape parameters of the aerial dispenser wing are continuously iteratively optimized until convergence is achieved and the optimal shape parameters are obtained.

[0139] Embodiment 5. This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy as described in any one of the above embodiments is executed.

[0140] Embodiment 6. This embodiment provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes a high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy as described in any one of the above embodiments.

[0141] A computer device provided in this embodiment, the hardware device of this part is a general model and is not shown in the form of a diagram. The system includes a processor and a memory, wherein the processor and the memory can be connected via a bus or other means. The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, and corresponding program instructions / modules. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, so as to realize the high-dimensional aerodynamic optimization method and steps of the aviation dispenser airfoil based on the data-driven support strategy in the above method embodiment.

[0142] Implementation method 7: See Figures 8 to 9 This embodiment is described. This embodiment compares and analyzes the high-dimensional aerodynamic optimization method of the airfoil of an aerial dispenser based on a data-driven support strategy described in the above embodiment with the existing optimization method.

[0143] Specifically:

[0144] Existing direct optimization (DO) frameworks automatically optimize aerodynamic shapes directly using gradient-based or gradient-free optimizers, which search for design solutions that meet specified requirements until the results converge. Figure 8 A flowchart of the direct optimization framework is shown.

[0145] The wing surface of the aerial dispenser missile is parameterized by 48-dimensional CST design variables, and the geometric domain decomposition algorithm is used to extract the first 12 geometric modes. Compared with the high-dimensional aerodynamic optimization of the aerial dispenser airfoil based on the data-driven support strategy and the direct optimization algorithm proposed in the above embodiment, its drag coefficient (C d ) The comparison of the convergence process with the number of CFD simulations is shown in the following figure. Fig. 9 It can be seen that in 1000 CFD simulations, the direct optimization framework reduces the drag coefficient C d Minimize from 0.0436 to 0.0308 (ΔC d =29.4%). At the same time, the optimization algorithm framework proposed in this patent converts the drag coefficient C d Optimized from 0.0436 to 0.0243 (ΔC d =44.3%), which is 14.9% better than the direct optimization framework. The comparison results show that the optimization algorithm framework proposed in this patent performs better in optimization performance, converges faster, and effectively solves this high-dimensional wing optimization problem.

[0146] The above description is only the implementation mode of the present invention and is not limited to the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A high-dimensional aerodynamic optimization method for airfoil of an aerial dispenser based on a data-driven support strategy, characterized in that: The method is: S1: The CST parameterization method is used to parameterize the airfoil of the aerial dispenser missile wing and construct the geometric shape of the aerial dispenser missile wing; S2: The constructed geometric shape is reduced to a low-dimensional geometric pattern through POD analysis using the geometric domain decomposition algorithm, and the geometric features of the air dispenser wing are extracted. After RBF-based mesh deformation, the CFD initial evaluation results are obtained and put into the historical air dispenser wing aerodynamic parameter database; S3: The DDSS algorithm is used to build a proxy model of the aerial dispenser wing based on the historical aerodynamic parameter database of the aerial dispenser wing. At the same time, the teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model. By comparing with the actual CFD evaluation results, the shape parameters of the aerial dispenser wing are continuously iteratively optimized until convergence is achieved and the optimal shape parameters are obtained.

2. The high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 1, characterized in that: The geometry of the air dispenser wing is represented by the class function C(ψ) and shape function S(ψ) of the CST parameterization method.

3. The high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 2, characterized in that: The geometry of the air dispenser wing is expressed as: Among them, ξ represents the dimensionless surface coordinate, ψ (0≤ψ≤1) represents the dimensionless chord coordinate, and ΔZ te / c represents the dimensionless trailing edge thickness.

4. The high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 2, characterized in that: S2 is specifically: S21: Generate a series of original sample vectors s of the wings of aerial dispensers using high-dimensional CST method and Latin hypercube sampling i (i=1,2,…,N) to construct the sampling result matrix S=s1 s2…s N ,s i Represents the coordinates of the air dispenser wing; S22: Calculate the average vector based on the sampling result matrix S23 constructs the correlation matrix R by calculating the deviation matrix according to the average vector s, and extracts features from the correlation matrix R to obtain a series of geometric feature patterns v i and eigenvalue λ i ; S24: Based on the average vector The geometric shape of the air dispenser wing is obtained by approximating the original sample vector through the geometric characteristic pattern.

5. A high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 4, characterized in that: The geometric shape of the air dispenser wing after dimensionality reduction is expressed as: Among them, α i is the modal coefficient of the i-th order vector.

6. A high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 5, characterized in that: The calculation formula for extracting information from the geometric characteristics of the air dispenser wing is as follows: Where P is the information ratio, M is the number of characteristic functions, and N is the number of high-order modes.

7. The high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy according to claim 1, characterized in that: S3 is specifically: S31: Use DDSS algorithm to extract valid data from historical aerodynamic data of aerial dispenser wing, build aerodynamic parameter database of aerial dispenser wing in combination with CFD evaluation results, and build a proxy model of aerial dispenser wing based on the aerodynamic parameter database; S32: The teaching optimization algorithm is used to obtain the optimal shape parameters in the proxy model, and the CFD evaluation results are obtained after the RBF-based mesh deformation; S33: The teaching optimization algorithm continuously optimizes the teaching stage and the student stage according to the CFD evaluation results until the obtained aerial dispenser wing shape parameters converge and the optimal shape parameters are obtained.

8. A high-dimensional aerodynamic optimization system for airfoil of an aerial dispenser based on a data-driven support strategy, characterized in that: The system includes a storage device, which is used to execute the method and steps for high-dimensional aerodynamic optimization of an aerial dispenser airfoil based on a data-driven support strategy as described in claim 1.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes a high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy as described in any one of claims 1 to 7.

10. A computer device, characterized in that: The device includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes a high-dimensional aerodynamic optimization method for an aerial dispenser airfoil based on a data-driven support strategy as described in any one of claims 1 to 7.

Images