A method, system, equipment, and medium for global optimization design of the aerodynamic shape of a vehicle.

Abstract

This invention belongs to the field of vehicle shape optimization design and discloses a method, system, device, and medium for global optimization design of vehicle aerodynamic shape. The method includes: constructing a parametric model of the vehicle based on free deformation technology; dividing the parametric model into regions and arranging control points in each candidate deformation region; analyzing the drag reduction rate of each candidate deformation region under typical deformation modes to obtain an initial design space; reducing the dimensionality of the design variables in the initial design space using principal component analysis to obtain a dimensionality-reduced design space; constructing a Kriging surrogate model and performing global optimization on the Kriging surrogate model using a genetic algorithm to obtain a preferred region; and applying adjoint optimization to the preferred region to obtain the globally optimal shape design scheme. The technical solution of this invention achieves intelligent and efficient global optimization of complex shapes under given constraints.

Description

Technical Field

[0001] This invention belongs to the field of vehicle shape optimization design, and in particular relates to a method, system, equipment and medium for global optimization design of vehicle aerodynamic shape. Background Technology

[0002] The external shape design of transportation vehicles (such as aircraft, high-speed trains and automobiles, and submarines) plays a crucial role in their overall design, determining their aerodynamic performance and significantly impacting their maneuverability, economy, and safety. Aerodynamic shape optimization refers to obtaining a geometric shape with optimal aerodynamic performance under certain constraints. Optimizing the aerodynamic shape of transportation vehicles can effectively reduce drag during operation, control the lift-to-drag ratio, and improve volume, structural strength, and stability. It is an important method for enhancing the overall performance of transportation vehicles and reducing operating costs.

[0003] Virtual Engineering (VR) is a product development platform that integrates geometric models, engineering analysis, simulation, optimization, and decision-making tools within a computer environment to enable multi-domain collaboration. The rapid development of artificial intelligence in recent years has provided a novel approach to solving various optimization problems in industrial applications, propelling virtual engineering technology to new heights. Global optimization of the aerodynamic shape of a vehicle is a highly complex aerodynamic problem. Traditional computational fluid dynamics (CFD) and virtual engineering techniques cannot achieve global optimization of aerodynamic shapes. In existing technologies, to achieve global optimization design of vehicle aerodynamic shapes, artificial intelligence technology is organically combined with mature virtual engineering to obtain a more advanced and powerful intelligent virtual engineering global optimization design method. However, the following problems still exist: Low optimization efficiency: Traditional shape optimization methods require a large number of samples, have high computational costs, and are difficult to handle optimization problems with more than 10 variables.

[0004] Local Optimality Trap: Gradient-based methods (such as adjoint optimization methods) are prone to getting trapped in local optima and cannot guarantee global optima.

[0005] The contradiction between accuracy and cost: In problems such as vehicle flow around a target, simulation errors are large and computational resource requirements are huge.

[0006] Poor versatility: Existing methods are mostly specialized tools, making it difficult to adapt to various types of vehicles.

[0007] To address the above problems, this invention provides a global optimization design scheme for the aerodynamic shape of a vehicle. Summary of the Invention

[0008] The purpose of this invention is to provide a method, system, device, and medium for global optimization design of the aerodynamic shape of a vehicle, so as to solve the problems existing in the prior art.

[0009] In a first aspect, to achieve the above objectives, the present invention provides a global optimization design method for the aerodynamic shape of a vehicle, comprising: S1: Constructing a parametric model of the vehicle based on free deformation technology; S2: Divide the parameterized model into regions to obtain candidate deformation regions corresponding to different parts. Arrange control points in each candidate deformation region. Analyze the drag reduction rate of each candidate deformation region under typical deformation mode to screen effective design variables and corresponding deformation ranges. Construct an initial design space based on the screened effective design variables and corresponding deformation ranges. S3: Based on principal component analysis, the design variables of the initial design space are reduced in dimension to obtain the dimension-reduced design space; S4: Construct a kriging surrogate model based on the dimensionality-reduced design space and response, and perform global optimization of the kriging surrogate model based on a genetic algorithm to obtain the preferred region; S5: Enable accompanying optimization for the preferred region to obtain the globally optimal shape design scheme.

[0010] Optionally, the process of constructing the parameterized model in step S1 specifically includes: Based on the free deformation technique of non-uniform rational splines, the geometric surface of the vehicle is embedded in the control box, and a mathematical mapping from physical space to parameter space is established through NURBS basis functions to complete the construction of the parameterized model.

[0011] Optionally, step S2 specifically includes: The head / tail surfaces of the parameterized model are divided into multiple candidate deformation regions, and control points are arranged in each candidate deformation region. Examine the inward and outward deformation of each candidate deformation region: if either deformation mode increases the drag of the candidate deformation region, remove the corresponding control point of the candidate deformation region; if outward deformation increases the drag, set the variable range of the candidate deformation region to inward deformation; if inward deformation increases the drag, set the variable range of the candidate deformation region to outward deformation; by analyzing the drag reduction rate of each candidate deformation region under typical deformation modes, determine the effective deformation region and its corresponding deformation range with the goal of maintaining the streamlined shape after deformation; construct the initial design space based on the selected effective design variables and corresponding deformation ranges.

[0012] Optionally, step S3 specifically includes: Principal component analysis is used to reduce the dimensionality of the design variables in the initial design space, and a predetermined number of principal components are used to characterize the geometric features of the initial design space.

[0013] Optionally, the process of obtaining the preferred region in step S4 specifically includes: Initial sample points are generated in the dimension-reduced design space using Latin hypercube sampling or adaptive sampling, and the response corresponding to each initial sample point is calculated. Construct the Kriging proxy model; A random combination of design variables is generated as the initial population. The individuals in the initial population are updated multiple times using a non-dominated sorting genetic algorithm. During each iteration, the target value corresponding to each combination of design variables is calculated based on the Kriging surrogate model. The initial population is updated based on the target value corresponding to each combination of design variables until the number of iterations reaches a preset value or the surrogate model outputs the optimal region with approximate accuracy.

[0014] Optionally, the process of obtaining the globally optimal shape design scheme in step S5 specifically includes: Based on the high-precision gradient information of the adjoint optimization method, a local fine-grained search is performed on the preferred region to output the globally optimal shape design scheme.

[0015] Secondly, to achieve the above objectives, the present invention provides a global optimization design system for the aerodynamic shape of a vehicle, comprising: The shape parameterization module is used to construct a parameterized model of the vehicle based on free deformation technology; The design space pre-optimization module is used to divide the parameterized model into regions to obtain candidate deformation regions corresponding to different parts. Control points are arranged in each candidate deformation region. By analyzing the drag reduction rate of each candidate deformation region under typical deformation mode, effective design variables and corresponding deformation ranges are screened. Based on the screened effective design variables and corresponding deformation ranges, the initial design space is constructed. The dimension reduction module is used to reduce the dimension of the design variables in the initial design space using principal component analysis to obtain the dimension-reduced design space. The global optimization module is used to construct a kriging surrogate model based on the dimensionality-reduced design space and response, and to perform global optimization on the kriging surrogate model based on a genetic algorithm to obtain the preferred region. The accompanying optimization module is used to enable accompanying optimization for the preferred area to obtain the globally optimal shape design scheme.

[0016] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform a global optimization design method for the aerodynamic shape of a vehicle according to the first aspect.

[0017] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a global optimization design method for the aerodynamic shape of a vehicle as described in the first aspect.

[0018] The technical effects of this invention are as follows: This invention constructs a complete and modular virtual engineering optimization process, which integrates a shape parameterization module (using free deformation technology), a design space pre-optimization module (screening effective design variables based on regional sensitivity analysis), a reduction model module (realizing the mapping of design variables from high-dimensional to low-dimensional), a surrogate model optimization module (using an improved Kriging surrogate model combined with a multi-objective point-addition strategy for global optimization), and an adjoint optimization module (for local fine-tuning). These modules are sequentially connected and share data, forming an automated and intelligent end-to-end optimization system from coarse screening to fine-tuning, achieving significant control effects.

[0019] The parameterization method and optimization process proposed in this invention do not depend on the geometric features of a specific vehicle. By adjusting the arrangement of the control box and the deformation area, it can be seamlessly applied to the aerodynamic or hydrodynamic shape optimization of various vehicles such as aircraft, high-speed trains, automobiles, and underwater vehicles, demonstrating the versatility and strong scalability of the method. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is an example of an artificial intelligence-based virtual engineering method for optimizing the aerodynamic shape of a vehicle in this invention. Figure 2 Control point P in this embodiment of the invention i,j,k A schematic diagram of the model mesh point Q; Figure 3This is a schematic diagram of the eight candidate deformation regions of the front / rear vehicles in an embodiment of the present invention; Figure 4 The cumulative contribution rate and mean square error of the principal components after PCA order reduction in this embodiment of the invention; Figure 5 This is a schematic diagram of the optimization process based on the agent model in an embodiment of the present invention; Figure 6 This is a schematic diagram of the accompanying optimization process for high-speed rail head shape drag reduction in an embodiment of the present invention; Figure 7 This refers to the three-car train model and computational domain in this embodiment of the invention. Figure 8 This is a scatter plot of the high-speed train shape optimization process with drag reduction as the objective in an embodiment of the present invention; Figure 9 This is a comparison diagram of the model before and after optimization in an embodiment of the present invention. Detailed Implementation

[0022] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as a limitation of the present invention, but rather as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0023] It should be understood that the terminology used in this invention is merely for describing particular embodiments and is not intended to limit the invention. Furthermore, with respect to numerical ranges in this invention, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Every smaller range between any stated value or intermediate value within a stated range, and any other stated value or intermediate value within said range, is also included in this invention. The upper and lower limits of these smaller ranges may be independently included or excluded from the range.

[0024] Various modifications and variations can be made to the specific embodiments described in this specification without departing from the scope or spirit of the invention, as will be apparent to those skilled in the art. Other embodiments derived from this specification will also be obvious to those skilled in the art. This application specification and embodiments are merely exemplary.

[0025] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments. Example

[0027] like Figure 1 - Figure 9As shown, this embodiment provides a global optimization design method for the aerodynamic / hydrodynamic shape of transportation vehicles (high-speed trains, aircraft, automobiles, underwater vehicles, etc.). This method integrates advanced technologies such as virtual engineering, surrogate models, machine learning, adjoint methods, parameter dimensionality reduction, and high-performance computing, aiming to achieve intelligent and efficient global optimization of complex shapes under given constraints. It mainly includes the following parts: Integrated Intelligent Optimization Process Architecture: This embodiment constructs a complete and modular virtual engineering optimization process. This process integrates a shape parameterization module (using free deformation technology), a design space pre-optimization module (screening effective design variables based on regional sensitivity analysis), a reduction-order model module (realizing the mapping of design variables from high-dimensional to low-dimensional), a surrogate model optimization module (using an improved Kriging surrogate model combined with a multi-objective point-addition strategy for global optimization), and an adjoint optimization module (for local fine-tuning). These modules are sequentially connected and share data, forming an automated and intelligent end-to-end optimization system from coarse screening to fine-tuning.

[0028] A collaborative dimensionality reduction method for design variables based on sensitivity analysis and Principal Component Analysis (PCA) is proposed: a two-stage strategy of "screening first, then compressing." First, through pre-optimization of the design space, ineffective or negatively impactful design variables are eliminated based on sensitivity analysis of local deformation on aerodynamic / hydraulic properties (such as drag), constructing a high-quality initial design space. Then, Principal Component Analysis is applied to further reduce the dimensionality of the remaining variables. Only a small number of principal components (e.g., 15) are needed to characterize most of the geometric features of the original high-dimensional design space (e.g., 40 variables) (cumulative contribution rate > 99.8%), thus significantly reducing the dimensionality and computational cost of the optimization problem while ensuring optimization accuracy.

[0029] The hybrid optimization strategy of "global optimization using a surrogate model + local refinement using an adjoint optimization method" innovatively combines machine learning-based surrogate model optimization with gradient-based adjoint optimization. First, an improved Kriging surrogate model and a multi-objective optimization algorithm are used to perform an efficient global search in the dimensionality-reduced design space, quickly approximating the globally optimal region. Then, the adjoint optimization method is initiated in the optimal region, utilizing its high-precision gradient information for local refinement search, ultimately obtaining the globally optimal shape design. This strategy effectively avoids the problems of traditional gradient methods easily getting trapped in local optima, and the insufficient accuracy of pure surrogate model methods near extreme points.

[0030] A cross-platform, universal parameterization and optimization framework: The parameterization method and optimization process proposed in this embodiment do not depend on the geometric features of a specific vehicle. By adjusting the control box arrangement and deformation area, it can be seamlessly applied to the aerodynamic or hydrodynamic shape optimization of various vehicles such as aircraft, high-speed trains, automobiles, and underwater vehicles, demonstrating the versatility and strong scalability of the method.

[0031] The technical solution of this embodiment includes the following core modules and processes ( Figure 1 ): Shape parameterization: Employing a free deformation (FFD)-based parameterization method, flexible control of complex 3D shapes is achieved, supporting smooth deformation of thousands of control points. Design space pre-optimization: Through regional sensitivity analysis, control variables that significantly affect the objective function are screened to construct a high-quality initial design space. Order reduction model: A nonlinear dimensionality reduction method based on PCA reduces design variables from high to low dimensions, retaining most geometric features. Machine learning-based surrogate model optimization: An improved Kriging model is constructed, combined with the NSGA-II optimization algorithm using multi-objective point addition strategies (such as EI, MSE, MSP, etc.) to achieve efficient global optimization. Adjoint optimization: Based on surrogate optimization, the adjoint method is used for local fine-tuning optimization to further improve drag reduction.

[0032] 1. Shape Parameterization: Shape parameterization employs Free Deformation (FFD) technology based on non-uniform rational B-splines, embedding the geometric surface into a control box. A mathematical mapping from physical space to parameter space is established using NURBS basis functions. NURBS basis functions are defined recursively. The process begins with a piecewise constant zero-order basis function; based on this, any higher-order basis function can be constructed through a linear combination of two lower-order basis functions. In the local 3D coordinate system o′·x′y′z′, p′, q′, and r′ represent the order of the basis functions in the three directions, respectively, with the number of control points being l′, m′, and n′, and the node vector lengths being L′, M′, and N′, respectively. The FFD governing equations are: (1) in These are the coordinates of the mesh points on the object's surface in the global coordinate system. For the global coordinates of the control points, The control points are weighted. The local coordinates of each grid point are solved using Newton's iteration method, and then the smooth deformation of the object is achieved by moving the control points. Figure 2 ).

[0033] 2. Design Space Pre-optimization: To obtain the largest possible design space, surface deformation of all areas at the train's head and tail is considered. However, aerodynamic drag is differently sensitive to deformation at different locations. If the design space includes variables that are insensitive or increase drag, optimization efficiency will be reduced and final performance will be affected. Therefore, this embodiment obtains an effective design space while meeting industry geometric constraints through design space pre-optimization. The specific method is as follows: the head / tail surfaces are divided into multiple regions, and control points are arranged in each region (…). Figure 3 Numerical simulations using a parametric train model were conducted to examine the impact of deformation in each region on drag changes. Both inward and outward deformation were checked for each region; if a certain deformation led to an increase in drag, the corresponding control point for that region was removed; if outward deformation resulted in a significant increase in drag, the variable range for that region was set to inward deformation, and vice versa. Finally, the drag reduction rate (DR=(C)) under typical deformation modes in each region was analyzed. d,0 -C d ) / C d,0 In the formula, DR is the drag reduction ratio, and C is the drag reduction ratio. d,0 It is the aerodynamic drag coefficient before control, C d This is the controlled aerodynamic drag coefficient. The effective deformation region and its deformation range are determined to ensure the streamlined shape is maintained after deformation. After checking all regions, a design space containing 40 variables and their respective ranges is obtained, involving 1500 control points.

[0034] 3. Reduced-Order Model: To reduce the dimensionality of the design space and the number of initial samples requiring numerical simulation evaluation, a reduced-order model based on Principal Component Analysis (PCA) was constructed. This process can be viewed as an encoding / decoding process: shape modification in the high-dimensional design space is determined by variables... Define variables that are encoded into a low-dimensional latent space using PCA. The decoding process reconstructs the original shape modification by minimizing the shape approximation error (such as the normalized mean square error). Specifically, the original shape is discretized into N nodes, and the shape modification vector... Include Each entry (each node deforms in three directions) is implemented by uniformly sampling to obtain M variables, whose shapes are then modified and organized into a matrix. Calculate the covariance matrix. Solve for its eigenvalues orthogonal eigenvectors (Principal Components). Selecting the first r principal components with the largest variance to form matrix Ψ, the dimensionless representation of the shape modification vector is: The normalized mean square error is: (2) In the formula Represents the least square error approximation; Let k be the k-th eigenvalue of the matrix covariance matrix C; Let be the deformation vector corresponding to the k-th sample. Applying this ROM to 40 design variables, it was found that the first 15 principal components are sufficient to make ε < 0.2%, with a cumulative contribution rate. The accuracy reaches 99.8%, thus maintaining high precision even when the design space dimension is reduced from 40 to 15. Figure 4 ).

[0035] 4. Machine Learning-Based Proxy Model Optimization: This module uses machine learning techniques to learn the knowledge of the objective function or constraint function about the design variables from a small number of samples, and constructs a proxy model ( Figure 5 This embodiment uses the Kriging model, which is suitable for complex nonlinear engineering problems. The general Kriging model represents the objective function y(x) as the sum of the regression term and the error term: (3) In the formula As basis functions, Let ϵ(x) be the parameter to be estimated, with zero mean and zero variance. The normal distributed random error. Kriging's predicted value is: (4) In the formula For the correlation function vector, a Gaussian exponential correlation model is used: (5) Estimating hyperparameters by maximizing the likelihood function , and , The sampling strategy includes static sampling and improved adaptive sampling. Static sampling uses Latin hypercube sampling (LHS) to generate a small number of initial samples to train the Kriging model; adaptive sampling determines the new sample positions based on four filling sampling criteria in sequence: maximizing expected improvement (EI), minimizing surrogate prediction (MSP), maximizing mean squared error (MSE), and a combination of MSP and MSE. The multi-objective filling sampling strategy balances exploration and utilization, and solves the sub-optimization problem using a non-dominated sorting genetic algorithm (NSGA-II) to obtain four new sampling positions. The optimization termination conditions include the maximum number of optimization iterations and the approximate accuracy of the surrogate model (e.g., ).

[0036] 5. Adjoint optimization: Adjoint optimization is placed as a refinement module after machine learning optimization to further reduce drag. Figure 6 Let the control parameter be... Surface mesh is The state variable is objective function Find the total derivative of J using the chain rule: (6) The Jacobian matrix Solving column-wise using the RBF deformation algorithm. Both the Primal and adjoint solutions are obtained using STAR-CCM+. ® (v15.08) Calculation. To maintain consistency in head and tail shapes, only the head shape is optimized. After each mesh deformation, it is copied to the tail shape using custom code. In the adjoint optimization, nearly 1800 control points are symmetrically arranged around the entire head surface. The optimization process includes: solving the steady-state flow field of the RANS equations; calculating the direct sensitivity of the objective function to the solution and the mesh; calculating the total derivative of the objective function with respect to the mesh by solving the adjoint equations; executing the adjoint differential mesh deformation algorithm to calculate the total derivative of the objective function with respect to the control variables; updating the control variables and deforming the mesh; copying the head shape to the tail shape and iterating repeatedly. A suitable deformation step size is determined through trial and error. .

[0037] The specific implementation example of this embodiment is as follows: This embodiment uses a 1:1 three-car train model, combining the SSTk-ω turbulence model and the RANS equations, to analyze the incoming flow velocity U. ∞ Numerical simulations were performed on the flow around the vehicle model under the operating condition of 400 km / h. Figure 7 ).

[0038] By combining the samples generated during the design space pre-optimization, surrogate model optimization, and adjoint optimization processes, an optimization curve is obtained, such as... Figure 8 As shown. The original head shape has a length L, width W, and height H of 8.70m, 3.36m, and 3.90m, respectively. For a three-car train (smooth car body, including bogies, no pantograph, and pressure difference drag at the front and rear accounting for 38.6% of the total drag), the optimal head shape is obtained at 400km / h. Figure 9 The total drag reduction rate is 10.9%, and the piezoresistive resistance at the beginning and end is reduced by 28.2% (L=9.70m, W=3.36m, H=3.97m).

[0039] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for global optimization design of the aerodynamic shape of a vehicle, characterized in that, include: S1: Constructing a parametric model of the vehicle based on free deformation technology; S2: Divide the parameterized model into regions to obtain candidate deformation regions corresponding to different parts. Arrange control points in each candidate deformation region. Analyze the drag reduction rate of each candidate deformation region under typical deformation mode to screen effective design variables and corresponding deformation ranges. Construct an initial design space based on the screened effective design variables and corresponding deformation ranges. S3: Based on principal component analysis, the design variables of the initial design space are reduced in dimension to obtain the reduced design space. S4: Construct a kriging surrogate model based on the dimensionality-reduced design space and response, and perform global optimization of the kriging surrogate model based on a genetic algorithm to obtain the preferred region; S5: Enable accompanying optimization for the preferred region to obtain the globally optimal shape design scheme.

2. The global optimization design method for the aerodynamic shape of a vehicle according to claim 1, characterized in that, The process of constructing the parameterized model in step S1 specifically includes: Based on the free deformation technique of non-uniform rational splines, the geometric surface of the vehicle is embedded in the control box, and a mathematical mapping from physical space to parameter space is established through NURBS basis functions to complete the construction of the parameterized model.

3. The global optimization design method for the aerodynamic shape of a vehicle according to claim 1, characterized in that, Step S2 specifically includes: The head / tail surfaces of the parameterized model are divided into multiple candidate deformation regions, and control points are arranged in each candidate deformation region. Examine the inward and outward deformation of each candidate deformation region: if either deformation mode increases the drag of the candidate deformation region, remove the corresponding control point of the candidate deformation region; if outward deformation increases the drag, set the variable range of the candidate deformation region to inward deformation; if inward deformation increases the drag, set the variable range of the candidate deformation region to outward deformation; by analyzing the drag reduction rate of each candidate deformation region under typical deformation modes, determine the effective deformation region and its corresponding deformation range with the goal of maintaining the streamlined shape after deformation; construct the initial design space based on the selected effective design variables and corresponding deformation ranges.

4. The global optimization design method for the aerodynamic shape of a vehicle according to claim 1, characterized in that, Step S3 specifically includes: Principal component analysis is used to reduce the dimensionality of the design variables in the initial design space, and a predetermined number of principal components are used to characterize the geometric features of the initial design space.

5. The global optimization design method for the aerodynamic shape of a vehicle according to claim 1, characterized in that, The process of obtaining the preferred region in step S4 specifically includes: Initial sample points are generated in the dimension-reduced design space using Latin hypercube sampling or adaptive sampling, and the response corresponding to each initial sample point is calculated. Construct the Kriging proxy model; A random combination of design variables is generated as the initial population. The individuals in the initial population are updated multiple times using a non-dominated sorting genetic algorithm. During each iteration, the target value corresponding to each combination of design variables is calculated based on the Kriging surrogate model. The initial population is updated based on the target value corresponding to each combination of design variables until the number of iterations reaches a preset value or the surrogate model outputs the optimal region with approximate accuracy.

6. The global optimization design method for the aerodynamic shape of a vehicle according to claim 1, characterized in that, The process of obtaining the globally optimal shape design scheme in step S5 specifically includes: Based on the high-precision gradient information of the adjoint optimization method, a local fine-grained search is performed on the preferred region to output the globally optimal shape design scheme.

7. A global optimization design system for the aerodynamic shape of a vehicle, characterized in that, include: The shape parameterization module is used to construct a parameterized model of the vehicle based on free deformation technology; The design space pre-optimization module is used to divide the parameterized model into regions to obtain candidate deformation regions corresponding to different parts. Control points are arranged in each candidate deformation region. By analyzing the drag reduction rate of each candidate deformation region under typical deformation mode, effective design variables and corresponding deformation ranges are screened. Based on the screened effective design variables and corresponding deformation ranges, the initial design space is constructed. The dimension reduction module is used to reduce the dimension of the design variables in the initial design space using principal component analysis to obtain the dimension-reduced design space. The global optimization module is used to construct a kriging surrogate model based on the dimensionality-reduced design space and response, and to perform global optimization on the kriging surrogate model based on a genetic algorithm to obtain the preferred region. The accompanying optimization module is used to enable accompanying optimization for the preferred area to obtain the globally optimal shape design scheme.

8. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program and the processor runs the computer program to enable the electronic device to perform a global optimization design method for the aerodynamic shape of a vehicle according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements a global optimization design method for the aerodynamic shape of a vehicle as described in any one of claims 1-6.

Images