A method, device, equipment and medium for aerodynamic shape optimization of a complex object

Abstract

The application provides a method and device for aerodynamic shape optimization of a complex object, equipment and a medium, comprising: performing displacement parameter transformation on a plurality of model points of an initial three-dimensional geometric model of the object, and then performing deformation processing on the initial three-dimensional geometric model by combining a direct free deformation technique to generate a new three-dimensional geometric model; performing aerodynamic performance prediction processing on point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and outputting aerodynamic performance parameters of each new three-dimensional geometric model; if the aerodynamic performance parameters of the new three-dimensional geometric model are multiple, searching and selecting a plurality of non-dominated optimal solutions from the plurality of new three-dimensional geometric models based on a non-dominated sorting and a reference point guiding mechanism, dynamically adjusting the displacement parameters of the model points based on a clustering analysis result of the plurality of non-dominated optimal solutions, and outputting an optimal aerodynamic shape design scheme when a preset convergence condition is met. The optimal aerodynamic shape design scheme of the object can be quickly and accurately generated.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerodynamics and object shape optimization technology, in particular to a complex object aerodynamic shape optimization method, device, equipment and medium. BACKGROUND

[0002] With the continuous improvement of industrial design and engineering optimization demand, the shape optimization of complex objects (such as cars, airplanes, ships, etc.) has become a key link to improve performance and energy efficiency. The traditional shape optimization method depends on manual iteration, parameterized modeling and finite element, finite volume and other numerical simulation techniques, although it has certain reliability and maturity, but the process highly depends on artificial experience, the overall efficiency is low, the cycle is long, and it is difficult to support high frequency, high dimension parameter deformation analysis and rapid iteration optimization. In addition, when facing complex nonlinear design space or multi-objective trade-off, the traditional method has obvious limitations in searching for global optimal solution, which is difficult to meet the demand of modern engineering for efficient and intelligent optimization. Therefore, how to improve the accuracy of the aerodynamic shape optimization of complex objects has become a technical problem that cannot be underestimated. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a complex object aerodynamic shape optimization method, device, equipment and medium, which can quickly and accurately generate the optimal aerodynamic shape design scheme of the object through direct free deformation technology, aerodynamic performance prediction model and intelligent optimization process for multiple aerodynamic performance parameters.

[0004] The embodiment of the present application provides a complex object aerodynamic shape optimization method, which comprises:

[0005] Displacement parameter transformation is performed on a plurality of model points of an initial three-dimensional geometric model of an object, and a plurality of new three-dimensional geometric models are generated by deforming the initial three-dimensional geometric model in combination with direct free deformation technology;

[0006] Point cloud data of the plurality of new three-dimensional geometric models is subjected to aerodynamic performance prediction processing based on an aerodynamic performance prediction model, and aerodynamic performance parameters of each new three-dimensional geometric model are outputted;

[0007] If the aerodynamic performance parameters of the new three-dimensional geometric models are multiple, a plurality of non-dominated optimal solutions are selected from the plurality of new three-dimensional geometric models based on non-dominated sorting and reference point guiding mechanism, the displacement parameters of the model points are dynamically adjusted based on the clustering analysis results of the plurality of non-dominated optimal solutions, and when the preset convergence condition is met, the optimal aerodynamic shape design scheme of the object is outputted.

[0008] In a possible implementation, after outputting the aerodynamic performance parameter of each new three-dimensional geometric model, the aerodynamic shape optimization method further comprises:

[0009] If the aerodynamic performance parameter of the new three-dimensional geometric model is one, the displacement parameters of the model points corresponding to the plurality of new three-dimensional geometric models are iteratively optimized based on a genetic algorithm, and when a preset convergence condition is met, the optimal aerodynamic shape design scheme of the object is output.

[0010] In a possible implementation, for any new three-dimensional geometric model, the aerodynamic performance prediction model is used to perform aerodynamic performance prediction processing on the point cloud data of the plurality of new three-dimensional geometric models, and the aerodynamic performance parameter of each new three-dimensional geometric model is output, comprising:

[0011] The new three-dimensional geometric model is subjected to feature extraction and encoding processing in the aerodynamic performance prediction model, and the voxel features corresponding to each point cloud data are determined;

[0012] The plurality of voxel features are processed based on a convolution layer, a linear layer, a normalization layer, and an activation function, and a plurality of embedding features are output;

[0013] The plurality of embedding features are sequentially processed based on a local attention layer, a feedforward network layer, a residual connection layer, and a normalization layer in a multi-layer encoder, and a plurality of multi-scale features are output;

[0014] The plurality of multi-scale features are subjected to global pooling processing to output a global feature, the global feature is subjected to mapping processing based on a regressor, and the aerodynamic performance parameter is output.

[0015] In a possible implementation, the non-dominated sorting and reference point guided mechanism is used to search and select a plurality of non-dominated optimal solutions from the plurality of new three-dimensional geometric models, comprising:

[0016] A population is formed based on the plurality of new three-dimensional geometric models and corresponding aerodynamic performance parameters, and all individuals in the population are sorted according to a non-dominated relationship to generate a non-dominated level; wherein the first level in the non-dominated level is a completely non-dominated solution set, and the second level is a non-dominated solution set remaining after the first level is removed;

[0017] A set of reference points is uniformly arranged in the target optimization space, when selecting the next generation individuals in the population, the individuals with high non-dominated level close to the reference points are preferentially selected, and the individuals with high non-dominated level are used as the non-dominated optimal solutions; wherein the reference points are optimization targets corresponding to the plurality of aerodynamic performance parameters.

[0018] In one possible implementation, the clustering analysis result based on the plurality of non-dominated optimal solutions is used to dynamically adjust the displacement parameters of the model points, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, including:

[0019] The plurality of non-dominated optimal solutions are subjected to clustering analysis to identify solution clusters with different performance tendencies.

[0020] The spatial range of the displacement parameters of the model points of each solution cluster is directionally contracted, and in the next generation of evolution, the distribution proportion of the population among the clusters is dynamically adjusted according to the distribution density of each solution cluster in the target space, by continuing to use the reference point mechanism and the adaptive population allocation strategy.

[0021] When the aerodynamic performance parameters of a solution cluster meet a preset convergence condition, the solution cluster is taken as the optimal aerodynamic shape design scheme of the object.

[0022] In one possible implementation, the genetic algorithm is used to iteratively optimize the displacement parameters of the model points corresponding to the plurality of new three-dimensional geometric models, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, including:

[0023] The displacement parameters of the plurality of model points are taken as the initial population in the genetic algorithm.

[0024] The aerodynamic performance parameters are taken as the fitness function, the population is subjected to multi-generation iterative optimization by using the genetic algorithm, a new generation of individuals is generated each generation, and individuals with high fitness are reserved as optimal solutions.

[0025] According to the distribution of the displacement parameters of the model points of the optimal solutions in the current generation, a search interval of the displacement parameters is determined, and a preset proportion of buffer boundaries is extended on the basis of the search interval.

[0026] When the aerodynamic performance parameters of the individuals corresponding to the displacement parameters of the model points in the search interval meet a preset convergence condition, the individuals, the corresponding three-dimensional geometric models, and the aerodynamic performance parameters are taken as the optimal aerodynamic shape design scheme.

[0027] In one possible implementation, the aerodynamic performance prediction model is determined by the following steps:

[0028] Sample point cloud data of a plurality of sample three-dimensional geometric models and corresponding sample aerodynamic performance parameters are input into a deep learning network model to learn the mapping relationship between the sample point cloud data and the sample aerodynamic performance parameters, and the predicted aerodynamic performance parameters of the sample point cloud data are predicted.

[0029] The deep learning network model is iteratively trained based on a loss value between the predicted aerodynamic performance parameter and the sample aerodynamic performance parameter, and the aerodynamic performance prediction model is determined.

[0030] The embodiment of the present application further provides a device for aerodynamic shape optimization of a complex object, which comprises:

[0031] The free-form deformation processing module is configured to perform displacement parameter transformation on a plurality of model points of an initial three-dimensional geometric model of the object, and perform deformation processing on the initial three-dimensional geometric model by combining a direct free-form deformation technique to generate a plurality of new three-dimensional geometric models;

[0032] The aerodynamic performance prediction module is configured to perform aerodynamic performance prediction processing on point cloud data of the plurality of new three-dimensional geometric models based on the aerodynamic performance prediction model, and output aerodynamic performance parameters of each of the new three-dimensional geometric models.

[0033] The intelligent optimization module is configured to, if the aerodynamic performance parameters of the new three-dimensional geometric models are a plurality of aerodynamic performance parameters, search and select a plurality of non-dominated optimal solutions from the plurality of new three-dimensional geometric models based on a non-dominated sorting and a reference point guiding mechanism, dynamically adjust the displacement parameters of the model points based on clustering analysis results of the plurality of non-dominated optimal solutions, and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met.

[0034] The embodiment of the present application further provides an electronic device, which comprises a processor, a memory and a bus, the memory stores machine readable instructions executable by the processor, the processor and the memory communicate through the bus when the electronic device is running, and the machine readable instructions are executed by the processor to perform the steps of the aerodynamic shape optimization method of the complex object as described above.

[0035] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to perform the steps of the aerodynamic shape optimization method of the complex object as described above.

[0036] The embodiment of the present application provides a kind of complex object aerodynamic shape optimization method, device, equipment and medium, the aerodynamic shape optimization method includes: after the displacement parameter transformation of the multiple model points of the initial three-dimensional geometric model of object, the initial three-dimensional geometric model is deformed and handled to generate multiple new three-dimensional geometric models in combination with direct free deformation technology;The point cloud data of multiple new three-dimensional geometric models is carried out aerodynamic performance prediction processing based on aerodynamic performance prediction model, and the aerodynamic performance parameter of each new three-dimensional geometric model is output;If the aerodynamic performance parameter of the new three-dimensional geometric model is multiple, then search and select multiple non-dominated optimal solution in multiple new three-dimensional geometric models based on non-dominated sorting and reference point guiding mechanism, and the displacement parameter of model point is dynamically adjusted based on the clustering analysis result of multiple non-dominated optimal solution, and when the preset convergence condition is satisfied, the optimal aerodynamic shape design scheme of the object is output.The direct free deformation technology, aerodynamic performance prediction model and intelligent optimization processing process for multiple aerodynamic performance parameters can quickly and accurately generate the optimal aerodynamic shape design scheme of object.

[0037] In order to make the above objects, features and advantages of the present application more apparent, the following will describe a preferred embodiment in detail, and the accompanying drawings will be described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0039] Figure 1 A flow chart of the aerodynamic shape optimization method of a complex object provided by the embodiment of the present application;

[0040] Figure 2 A schematic diagram of the aerodynamic shape optimization method of a complex object provided by the embodiment of the present application;

[0041] Figure 3 A structural schematic diagram of the aerodynamic shape optimization device of a complex object provided by the embodiment of the present application;

[0042] Figure 4 A structural schematic diagram of the aerodynamic shape optimization device of a complex object provided by the embodiment of the present application;

[0043] Figure 5 A structural schematic diagram of the electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0044] The technical solutions and advantages of the embodiments of the present application will be more clearly understood from the following description of the embodiments of the present application with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, every other embodiment obtained by a person skilled in the art without creative work belongs to the scope of protection of the present application.

[0045] Firstly, the application scenarios applicable to the present application are introduced. The present application can be applied to the field of aerodynamics and the field of object shape optimization technology.

[0046] It is found through research that, with the increasing demand for industrial design and engineering optimization, shape optimization of complex objects (such as cars, airplanes, ships, etc.) has become a key link to improve performance and energy efficiency. Traditional shape optimization methods rely on manual iteration, parameterized modeling, and numerical simulation techniques such as finite element and finite volume. Although they have certain reliability and maturity, the process highly depends on human experience, and the overall efficiency is low, the cycle is long, and it is difficult to support high-frequency, high-dimensional parameter deformation analysis and rapid iterative optimization. In addition, when facing complex nonlinear design space or multi-objective trade-off, traditional methods have obvious limitations in searching for global optimal solutions, and it is difficult to meet the demand for efficient and intelligent optimization of modern engineering. Therefore, how to improve the accuracy of aerodynamic shape optimization of complex objects has become a technical problem that cannot be underestimated.

[0047] Based on this, the embodiments of the present application provide an aerodynamic shape optimization method for complex objects, which can quickly and accurately generate the optimal aerodynamic shape design scheme of the object through direct free deformation technology, aerodynamic performance prediction model, and intelligent optimization process for multiple aerodynamic performance parameters.

[0048] Please refer to Figure 1 , Figure 1 The flowchart of the aerodynamic shape optimization method for complex objects provided by the embodiments of the present application. As shown in Figure 1 The aerodynamic shape optimization method provided by the embodiments of the present application includes:

[0049] S101: After performing displacement parameter transformation on multiple model points of an initial three-dimensional geometric model of an object, a direct free deformation technique is used to deform the initial three-dimensional geometric model to generate multiple new three-dimensional geometric models.

[0050] It should be noted that the plurality of control boxes are set on the initial three-dimensional geometric model, the model points in the control boxes are subjected to displacement parameter transformation, and then the initial three-dimensional geometric model subjected to the displacement parameter transformation is subjected to deformation processing by using the direct free deformation technology to generate a new three-dimensional geometric model.

[0051] Specifically, an initial three-dimensional geometric model (such as a DrivAer automobile model, a SUBOFF submarine model, a CHN-T1 aircraft, and the like) is selected, a control box (BoundingBox) capable of completely wrapping the model is automatically calculated, and the size (length x width x height) thereof is determined. A control point grid (such as 11 x 9 x 9) is set in the x, y, and z directions of the control box to form a direct free deformation (DFFD) framework. A deformation region is selected: model points in key regions (such as the front face of an automobile, a tail wing, the bow of a submarine, and the like) are divided on the model surface, and the allowed displacement directions (such as only x and z direction movement) thereof are set. Latin hypercube sampling (LHS) is used to generate 200 or more deformation parameter combinations to ensure uniform coverage of the design space, and the model is driven by DFFD to deform the displacement amount of the model points corresponding to each parameter to generate a new three-dimensional geometric model.

[0052] S102: Perform aerodynamic performance prediction processing on the point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and output the aerodynamic performance parameters of each new three-dimensional geometric model.

[0053] In this step, the point cloud data of the plurality of new three-dimensional geometric models is subjected to aerodynamic performance prediction processing by using the aerodynamic performance prediction model, and the aerodynamic performance parameters of each new three-dimensional geometric model are output.

[0054] It should be noted that the aerodynamic performance prediction model dynamically captures the point cloud distribution characteristics (such as curvature change, pressure gradient, vortex core position, and the like) of the key regions through local geometric feature aggregation and multi-scale transformation, thereby accurately representing the influence of complex flow phenomena (such as separated flow, wake vortex structure) on aerodynamic performance. Compared with traditional gridded CFD data, the aerodynamic performance prediction model directly learns features from unstructured point clouds, which is more suitable for automatic modeling of complex geometries.

[0055] Here, the aerodynamic performance parameters include a drag coefficient, a lift coefficient, and a pressure distribution, and the aerodynamic performance parameters corresponding to different complex objects are different, and this part is not specifically limited.

[0056] In one possible implementation, for any new three-dimensional geometric model, the aerodynamic performance prediction processing on the point cloud data of the plurality of new three-dimensional geometric models based on the aerodynamic performance prediction model and the output of the aerodynamic performance parameters of each new three-dimensional geometric model include:

[0057] (1) Feature extraction and encoding of the new three-dimensional geometric model in the aerodynamic performance prediction model, determining the voxel features corresponding to each point cloud data.

[0058] Here, using PyTorch's Dataset and DataLoader, batch loading and multi-process acceleration are supported. Each batch contains point cloud data of several new three-dimensional geometric models, which are discretized into integer grid coordinates (such as 1 cm per grid). The purpose is to facilitate subsequent spatial ordering and sparse convolution. Spatial serialization (such as z-order, Hilbert curve encoding) is performed on the voxelized point cloud, so that spatially adjacent points are also adjacent in the sequence, facilitating patch grouping and local attention. The point cloud data is converted into a sparse tensor to obtain voxel features.

[0059] (2) Process multiple voxel features based on convolution layers, linear layers, normalization layers, and activation functions, and output multiple embedding features.

[0060] Here, 3D convolution + linear layer is used to map the 3D coordinates of each voxel feature to a higher-dimensional feature space (such as 32 dimensions). Then, after normalization and activation function processing, the hidden layer features are obtained.

[0061] (3) Process multiple embedding features based on local attention layers, feedforward network layers, residual connection layers, and normalization layers in the multi-layer encoder, and output multiple multi-scale features.

[0062] Here, the local attention layer is used for self-attention within the adjacent patch after serialization, capturing spatial local relationships, and the feedforward network layer (MLP) is used to enhance feature expression capability. The residual connection layer and the normalization layer are used to improve the stability of the training. The feature dimension is gradually increased (such as 32→64→128→256) and the point number is gradually reduced after each layer.

[0063] (4) Global pooling processing of multiple multi-scale features outputs global features, and mapping processing of the global features based on a regressor outputs the aerodynamic performance parameters.

[0064] Global pooling (such as average pooling) is performed on all current multi-scale features to obtain global features, and the regressor performs mapping processing on the global features to output the aerodynamic performance parameters.

[0065] In one possible implementation, the aerodynamic performance prediction model is determined by the following steps:

[0066] The sample point cloud data of a plurality of sample three-dimensional geometric models and corresponding sample aerodynamic performance parameters are input into a deep learning network model to learn the mapping relationship between the sample point cloud data and the sample aerodynamic performance parameters, and to predict the predicted aerodynamic performance parameters of the sample point cloud data; the deep learning network model is iteratively trained based on the loss value between the predicted aerodynamic performance parameters and the sample aerodynamic performance parameters, and the aerodynamic performance prediction model is determined.

[0067] Here, if the loss value is greater than a preset loss value, the network parameters of the deep learning network model are updated, and the updated deep learning network model is continuously iteratively trained. If the loss value is less than or equal to the preset loss value, the current deep learning network model is taken as the aerodynamic performance prediction model.

[0068] Before the sample data is input into the deep learning network model, the sample data needs to be preprocessed, including data format processing, data standardization processing, downsampling, and repeated sampling processing.

[0069] It should be noted that the sample aerodynamic performance parameters are determined by the LBM algorithm. The LBM algorithm has strong locality and can greatly improve the calculation efficiency in combination with GPU acceleration, and is particularly suitable for rapid iterative optimization of multiple schemes. In addition, it automatically satisfies the mass conservation, and the boundary processing is flexible (such as dynamic boundary and curved surface boundary), and can accurately capture complex flow characteristics such as separation flow and vortex shedding, thereby providing high-fidelity prediction of aerodynamic forces (such as drag, lift, and lateral force) and flow field details of complex shapes of aircrafts, vehicles, or buildings. Taking a car as an example, the sample aerodynamic performance parameters corresponding to different sample three-dimensional geometric models generated by the DFFD technology can be quickly calculated by the LBM method.

[0070] S103: If the aerodynamic performance parameters of the new three-dimensional geometric model are multiple, a plurality of non-dominated optimal solutions are selected from the plurality of new three-dimensional geometric models based on a non-dominated sorting and a reference point guiding mechanism, the displacement parameters of the model points are dynamically adjusted based on the clustering analysis results of the plurality of non-dominated optimal solutions, and the optimal aerodynamic shape design scheme of the object is output when a preset convergence condition is met.

[0071] In this step, if the aerodynamic performance parameters of the new three-dimensional geometric model are multiple, a plurality of non-dominated optimal solutions are selected from the plurality of new three-dimensional geometric models based on a non-dominated sorting and a reference point guiding mechanism, the displacement parameters of the model points are dynamically adjusted based on the clustering analysis results of the plurality of non-dominated optimal solutions, and the optimal aerodynamic shape design scheme of the object is output when a preset convergence condition is met.

[0072] It should be noted that the optimal aerodynamic shape design scheme includes a three-dimensional geometric model and corresponding aerodynamic performance parameters.

[0073] Here, the convergence condition is whether the aerodynamic performance parameter of the current three-dimensional geometric model meets the preset standard aerodynamic performance parameter. For example, in the automobile wind resistance optimization task, when the wind resistance coefficient is less than 0.001 for 5 generations or reaches the maximum number of iterations (default 500 generations), the optimization is terminated.

[0074] In one possible implementation, the non-dominated sorting and reference point guiding mechanism searches and selects non-dominated optimal solutions from the new three-dimensional geometric models, including:

[0075] A: Based on the new three-dimensional geometric models and corresponding aerodynamic performance parameters, a population is formed, and all individuals in the population are sorted according to the non-dominated relationship to generate a non-dominated level; wherein the first level in the non-dominated level is a completely non-dominated solution set, and the second level is a non-dominated solution set remaining after removing the first level.

[0076] Here, according to the new three-dimensional geometric models and corresponding aerodynamic performance parameters, a population is formed, and all individuals in the population are sorted according to the non-dominated relationship to generate a non-dominated level.

[0077] Wherein the first level in the non-dominated level is a completely non-dominated solution set, and the second level is a non-dominated solution set remaining after removing the first level, and so on.

[0078] It should be noted that a solution is called "non-dominated" if it is not worse than another solution in all optimization objectives and at least one optimization objective is better.

[0079] A set of reference points is uniformly arranged in the target optimization space, and when selecting the next generation of individuals in the population, individuals with high non-dominated levels close to the reference points are preferentially selected, and the individuals with high non-dominated levels are selected as the non-dominated optimal solutions; wherein the reference points are optimization objectives corresponding to a plurality of aerodynamic performance parameters.

[0080] Here, a set of "reference points" is uniformly arranged in the target space (such as wind resistance, lift, wind noise, etc.) in the multi-objective optimization process, and these reference points are used to guide the uniform distribution of the solution set in the high-dimensional target space. When selecting the next generation of individuals in the population, individuals with high non-dominated levels close to the reference points are preferentially selected, and the individuals with high non-dominated levels are selected as the non-dominated optimal solutions.

[0081] Wherein, the determination process of the individual with high non-dominated level close to the reference point is: calculating the Euclidean distance between each non-dominated solution and the reference point, and determining the non-dominated solution with a Euclidean distance less than a preset threshold as the individual with high non-dominated level.

[0082] In one possible implementation, the clustering analysis result based on the plurality of non-dominated optimal solutions is used to dynamically adjust the displacement parameters of the model points, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, including:

[0083] a: clustering analysis is performed on the plurality of non-dominated optimal solutions to identify solution clusters with different performance tendencies.

[0084] Here, the clustering analysis of the plurality of non-dominated optimal solutions is performed on the distribution characteristics of the target space.

[0085] Among them, the "solution cluster" refers to a subset with similar performance characteristics or target preferences identified by clustering operation on non-dominated optimal solutions.

[0086] b: the space range of the displacement parameters of the model points of each solution cluster is directionally contracted, and in the next generation evolution process, the distribution proportion of the population in each cluster is dynamically adjusted according to the distribution density of each solution cluster in the target space, using the reference point mechanism and the adaptive population allocation strategy.

[0087] Here, the position and density of the reference points are dynamically adjusted according to the clustering analysis result, such as deleting "cold points" (reference points rarely occupied), refining "hot points" (solution set intensive area), or moving the reference points as a whole to better cover the actual Pareto front. In the new generation population, solutions close to sparse reference points are preferentially retained to ensure the diversity and uniform distribution of the solution set.

[0088] c: when the aerodynamic performance parameters of the solution cluster meet the preset convergence condition, the solution cluster is taken as the optimal aerodynamic shape design scheme of the object.

[0089] Here, the new generation population is generated by genetic algorithm operations such as crossover and mutation, and the above steps are continued to be executed, and when the aerodynamic performance parameters of the solution cluster meet the preset convergence condition, the solution cluster is taken as the optimal aerodynamic shape design scheme of the object.

[0090] Here, the aerodynamic performance parameters of the solution cluster are also determined by the aerodynamic performance prediction model.

[0091] In this application, multi-objective optimization can use methods such as NSGA-III (the third generation of non-dominated sorting genetic algorithm). Still taking the car as an example, the multi-objective optimization of the aerodynamic shape of the car can be realized by using NSGA-III, which can efficiently balance multiple conflicting objectives (such as reducing the drag coefficient Cd, improving the downforce Cl, and reducing the aerodynamic noise). This method processes the high-dimensional objective space through the reference point mechanism, combines parameterized modeling DFFD to generate diversified design samples, and drives LBM / CFD simulation to evaluate the performance of each scheme. In particular, for the characteristics of multi-objective optimization, this application designs a hierarchical contraction strategy: first, the non-dominated optimal solution is subjected to cluster analysis, and the solution clusters with different performance tendencies are identified, and then the DFFD parameter subspace corresponding to each cluster is subjected to directional contraction, so that the efficient search capability can still be maintained in the 3-dimensional and higher-dimensional objective space. The adaptive population allocation strategy of NSGA-III can effectively maintain the distribution of the Pareto front, combined with the intelligent contraction method of the application, the performance degradation of traditional multi-objective algorithms in more than three objectives can be avoided, and the optimization efficiency is greatly improved.

[0092] In one possible implementation, after outputting the aerodynamic performance parameters of each of the new three-dimensional geometric models, the aerodynamic shape optimization method further comprises:

[0093] If the aerodynamic performance parameters of the new three-dimensional geometric models are one, the displacement parameters of the model points corresponding to a plurality of the new three-dimensional geometric models are iteratively optimized based on a genetic algorithm, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output.

[0094] Here, if the aerodynamic performance parameters are one, the optimization target is one, and the displacement parameters of the model points corresponding to a plurality of the new three-dimensional geometric models are iteratively optimized according to the genetic algorithm, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output.

[0095] In one possible implementation, the displacement parameters of the model points corresponding to a plurality of the new three-dimensional geometric models are iteratively optimized based on a genetic algorithm, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, comprising:

[0096] i: The displacement parameters of a plurality of the model points are taken as initial populations in the genetic algorithm.

[0097] ii: The aerodynamic performance parameters are taken as fitness functions, and a plurality of generations of the populations are iteratively optimized by using the genetic algorithm, a new generation of individuals is generated every generation, and individuals with high fitness are reserved as optimal solutions.

[0098] Here, a genetic algorithm is used to iteratively optimize the population for multiple generations. Each generation generates a new generation of individuals, and the fitness of the new generation of individuals is determined according to the fitness function. Individuals with high fitness are selected as the optimal solution.

[0099] iii: Based on the distribution of displacement parameters of model points of the current optimal solution, determine the search range of displacement parameters, and expand the buffer boundary by a preset ratio based on the search range.

[0100] Here, based on the distribution of displacement parameters of model points in the current generation of the optimal solution, the search interval of displacement parameters is determined, and a buffer boundary with a preset ratio is extended on the basis of the search interval.

[0101] iv: When the aerodynamic performance parameters of the individual corresponding to the displacement parameters of the model points within the search interval meet the preset convergence conditions, the individual, its corresponding three-dimensional geometric model, and aerodynamic performance parameters are taken as the optimal aerodynamic shape design scheme.

[0102] In this specific implementation, the aerodynamic performance parameter is the drag coefficient, and the displacement encoding of the model points is used as the initial population. The drag coefficient is used as the fitness function, and high-performance individuals are gradually selected through multiple generations of iteration, retaining advantageous characteristics as the optimal solution. This scheme introduces an adaptive parameter space contraction strategy: during the optimization process, the search range is dynamically adjusted based on the parameter distribution of the optimal solution. Every 10 generations, the displacement parameters of the DFFD model points are contracted, retaining the parameter range of the top 20% of performers, and providing a 10% buffer boundary. This significantly improves local search efficiency while maintaining population diversity. Compared to gradient-based optimization algorithms, genetic algorithms can avoid getting trapped in local optima, and are particularly suitable for high-dimensional, nonlinear aerodynamic optimization problems with multiple extreme points.

[0103] For further details, please refer to Figure 2 , Figure 2 This is a schematic diagram illustrating a method for optimizing the aerodynamic shape of a complex object, as provided in an embodiment of this application. Figure 2 As shown, a database of complex geometric shapes is formed by DFFD deformation of a complex geometric basic model. The database of complex geometric shapes is then input into the aerodynamic performance prediction model in batches to obtain the aerodynamic performance parameters of the complex geometric shapes. The aerodynamic performance parameters are checked to see if they meet the convergence condition. If not, the model points of different complex geometric shapes are continuously searched for DFFD deformation based on the intelligent optimization algorithm. The aerodynamic performance prediction model is used to predict the aerodynamic performance parameters of the deformed geometry. If the convergence condition is met, the optimal aerodynamic shape design scheme is output.

[0104] The application effectively solves the key bottlenecks of traditional aerodynamic optimization in aspects of design efficiency, global search capability, automation degree and calculation resource dependence, provides an efficient, accurate and low-cost solution for rapid optimization of complex aerodynamic shapes, and significantly improves the practicability and industrial application potential of engineering design.

[0105] The aerodynamic shape optimization method for a complex object provided by the embodiment of the application comprises: performing displacement parameter transformation on a plurality of model points of an initial three-dimensional geometric model of the object, and then performing deformation processing on the initial three-dimensional geometric model by combining a direct free deformation technique to generate a plurality of new three-dimensional geometric models; performing aerodynamic performance prediction processing on point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and outputting an aerodynamic performance parameter of each new three-dimensional geometric model; if the aerodynamic performance parameters of the new three-dimensional geometric models are a plurality of, then searching and selecting a plurality of non-dominated optimal solutions from the plurality of new three-dimensional geometric models based on a non-dominated sorting and reference point guiding mechanism, and dynamically adjusting displacement parameters of the model points based on a clustering analysis result of the plurality of non-dominated optimal solutions, and outputting an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met. Through the direct free deformation technique, the aerodynamic performance prediction model and the intelligent optimization processing process for a plurality of aerodynamic performance parameters, the optimal aerodynamic shape design scheme of the object can be quickly and accurately generated.

[0106] Please refer to Figure 3 、 Figure 4 , Figure 3 Figure 1 is a structural schematic diagram of an aerodynamic shape optimization device for a complex object provided by the embodiment of the application; Figure 4 Figure 2 is another structural schematic diagram of the aerodynamic shape optimization device for the complex object provided by the embodiment of the application. As shown in the figure, the aerodynamic shape optimization device 300 for the complex object comprises: Figure 3

[0107] The free deformation processing module 310 is configured to perform displacement parameter transformation on a plurality of model points of an initial three-dimensional geometric model of the object based on a direct free deformation technique, and then perform deformation processing on the initial three-dimensional geometric model by combining the direct free deformation technique to generate a plurality of new three-dimensional geometric models;

[0108] The aerodynamic performance prediction module 320 is configured to perform aerodynamic performance prediction processing on point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and output an aerodynamic performance parameter of each new three-dimensional geometric model;

[0109] ​The intelligent optimization module 330 is configured to: if the aerodynamic performance parameter of the new three-dimensional geometric model is multiple, search and select multiple non-dominated optimal solutions from the multiple new three-dimensional geometric models based on a non-dominated sorting and a reference point guide mechanism; dynamically adjust the displacement parameter of the model point based on a clustering analysis result of the multiple non-dominated optimal solutions; and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met.

[0110] Further, the intelligent optimization module 330 is further configured to:

[0111] if the aerodynamic performance parameter of the new three-dimensional geometric model is one, iteratively optimize the displacement parameter of the model point corresponding to the multiple new three-dimensional geometric models based on a genetic algorithm; and output the optimal aerodynamic shape design scheme of the object when a preset convergence condition is met.

[0112] Further, the aerodynamic performance prediction module 320 is configured to, for any new three-dimensional geometric model, perform aerodynamic performance prediction processing on the point cloud data of the multiple new three-dimensional geometric models based on an aerodynamic performance prediction model, and output an aerodynamic performance parameter of each new three-dimensional geometric model.

[0113] In the aerodynamic performance prediction model, feature extraction and encoding processing are performed on the new three-dimensional geometric model to determine a voxel feature corresponding to each point cloud data.

[0114] The multiple voxel features are processed based on a convolution layer, a linear layer, a normalization layer and an activation function to output multiple embedding features.

[0115] The multiple embedding features are sequentially processed based on a local attention layer, a feedforward network layer, a residual connection layer and a normalization layer in a multi-layer encoder to output multiple multi-scale features.

[0116] Global pooling processing is performed on the multiple multi-scale features to output a global feature, and the global feature is mapped based on a regressor to output the aerodynamic performance parameter.

[0117] Further, the intelligent optimization module 330 is configured to search and select multiple non-dominated optimal solutions from the multiple new three-dimensional geometric models based on the non-dominated sorting and the reference point guide mechanism, including:

[0118] A population is formed based on the multiple new three-dimensional geometric models and corresponding aerodynamic performance parameters, and all individuals in the population are sorted according to a non-dominated relationship to generate a non-dominated level; wherein the first level in the non-dominated level is a complete non-dominated solution set, and the second level is a non-dominated solution set remaining after the first level is removed.

[0119] A plurality of reference points are uniformly arranged in the target optimization space, and when selecting the next generation of individuals in the population, individuals with high non-dominated levels close to the reference points are preferentially selected as the non-dominated optimal solutions; wherein the reference points are optimization targets corresponding to a plurality of aerodynamic performance parameters.

[0120] Further, the intelligent optimization module 330 is configured to dynamically adjust the displacement parameters of the model points based on the clustering analysis results of the plurality of non-dominated optimal solutions, and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met.

[0121] The plurality of non-dominated optimal solutions are subjected to clustering analysis to identify solution clusters with different performance tendencies.

[0122] The space range of the displacement parameters of the model points of each solution cluster is directionally contracted, and in the next generation evolution process, the distribution proportion of the population among the clusters is dynamically adjusted according to the distribution density of each solution cluster in the target space, by using the reference point mechanism and the adaptive population allocation strategy.

[0123] When the aerodynamic performance parameters of a solution cluster meet the preset convergence condition, the solution cluster is taken as the optimal aerodynamic shape design scheme of the object.

[0124] Further, the intelligent optimization module 330 is configured to iteratively optimize the displacement parameters of the model points corresponding to the plurality of new three-dimensional geometric models based on the genetic algorithm, and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met.

[0125] The displacement parameters of the plurality of model points are taken as initial populations in the genetic algorithm.

[0126] The aerodynamic performance parameters are taken as fitness functions, the population is subjected to multi-generation iterative optimization by using the genetic algorithm, a new generation of individuals is generated each generation, and individuals with high fitness are reserved as optimal solutions.

[0127] According to the distribution of the displacement parameters of the model points of the optimal solutions in the current generation, a search interval of the displacement parameters is determined, and a preset proportion of buffer boundaries is expanded based on the search interval.

[0128] When the aerodynamic performance parameters of the individuals corresponding to the displacement parameters of the model points in the search interval meet the preset convergence condition, the individual, the corresponding three-dimensional geometric model, and the aerodynamic performance parameters are taken as the optimal aerodynamic shape design scheme.

[0129] Further, as Figure 4As shown, the aerodynamic shape optimization device 300 of the complex object further comprises a model training module 340, which determines the aerodynamic performance prediction model by the following steps:

[0130] The sample point cloud data of the plurality of sample three-dimensional geometric models and the corresponding sample aerodynamic performance parameters are input into the deep learning network model to learn the mapping relationship between the sample point cloud data and the sample aerodynamic performance parameters, and the predicted aerodynamic performance parameters of the sample point cloud data are predicted.

[0131] The deep learning network model is iteratively trained based on the loss value between the predicted aerodynamic performance parameters and the sample aerodynamic performance parameters, and the aerodynamic performance prediction model is determined.

[0132] The aerodynamic shape optimization device of the complex object provided by the embodiment of the application comprises: a free deformation processing module, which is configured to perform displacement parameter transformation on a plurality of model points of an initial three-dimensional geometric model of an object based on a direct free deformation technique, and perform deformation processing on the initial three-dimensional geometric model based on the direct free deformation technique to generate a plurality of new three-dimensional geometric models; an aerodynamic performance prediction module, which is configured to perform aerodynamic performance prediction processing on point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and output an aerodynamic performance parameter of each of the new three-dimensional geometric models; and an intelligent optimization module, which is configured to, if the aerodynamic performance parameter of the new three-dimensional geometric model is a plurality, search and select a plurality of non-dominated optimal solutions from the plurality of new three-dimensional geometric models based on a non-dominated sorting and reference point guiding mechanism, dynamically adjust displacement parameters of the model points based on a clustering analysis result of the plurality of non-dominated optimal solutions, and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met. Through the direct free deformation technique, the aerodynamic performance prediction model, and the intelligent optimization processing procedure for the plurality of aerodynamic performance parameters, the optimal aerodynamic shape design scheme of the object can be quickly and accurately generated.

[0133] Please refer to Figure 5 , Figure 5 The structure of an electronic device provided by the embodiment of the application is shown in FIG. 5. Figure 5 As shown in FIG. 5, the electronic device 500 comprises a processor 510, a memory 520, and a bus 530.

[0134] The memory 520 stores machine-readable instructions executable by the processor 510. When the electronic device 500 is running, the processor 510 and the memory 520 communicate through the bus 530. When the machine-readable instructions are executed by the processor 510, the above-mentioned Figure 1 and Figure 2The steps of the aerodynamic shape optimization method of the complex object in the method embodiment are not repeated here.

[0135] The embodiment of the present application further provides a computer readable storage medium, and the computer readable storage medium stores a computer program. Figure 1 and Figure 2 The steps of the aerodynamic shape optimization method of the complex object in the method embodiment are not repeated here.

[0136] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, and is not repeated here.

[0137] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented by other ways. The device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some communication interfaces, devices or units, and can be electrical, mechanical or other forms.

[0138] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on a plurality of network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0139] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0140] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a nonvolatile computer readable storage medium executable by a processor. Based on this understanding, the technical solutions of the present application essentially or the parts of the prior art that make contributions or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0141] Finally, it should be noted that: the above-described embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit them. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily think of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed by the present application, or make equivalent replacements to some of the technical features. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method of aerodynamic shape optimization of a complex object, characterized in that, The aerodynamic shape optimization method comprises: After displacement parameter transformation is performed on a plurality of model points of an initial three-dimensional geometric model of an object, a plurality of new three-dimensional geometric models are generated by performing deformation processing on the initial three-dimensional geometric model in combination with a direct free deformation technique; Aerodynamic performance prediction processing is performed on point cloud data of the plurality of new three-dimensional geometric models based on an aerodynamic performance prediction model, and aerodynamic performance parameters of each of the new three-dimensional geometric models are output; If the aerodynamic performance parameters of the new three-dimensional geometric models are a plurality, a plurality of non-dominated optimal solutions are selected from the plurality of new three-dimensional geometric models based on a non-dominated sorting and reference point guiding mechanism, displacement parameters of the model points are dynamically adjusted based on clustering analysis results of the plurality of non-dominated optimal solutions, and an optimal aerodynamic shape design scheme of the object is output when a preset convergence condition is met; For any one of the new three-dimensional geometric models, the aerodynamic performance prediction processing performed on the point cloud data of the plurality of new three-dimensional geometric models based on the aerodynamic performance prediction model and the output of the aerodynamic performance parameters of each of the new three-dimensional geometric models comprise: Feature extraction and coding processing are performed on the new three-dimensional geometric models in the aerodynamic performance prediction model, and voxel features corresponding to each point cloud data are determined; A plurality of embedded features are output by processing a plurality of the voxel features based on a convolution layer, a linear layer, a normalization layer, and an activation function; A plurality of multi-scale features are output by sequentially processing a plurality of the embedded features based on a local attention layer, a feedforward network layer, a residual connection layer, and a normalization layer in a multi-layer encoder; Global features are output by performing global pooling processing on a plurality of the multi-scale features, and the aerodynamic performance parameters are output by performing mapping processing on the global features based on a regressor; The plurality of non-dominated optimal solutions are selected from the plurality of new three-dimensional geometric models based on the non-dominated sorting and reference point guiding mechanism, which comprises: A population is formed based on the plurality of new three-dimensional geometric models and corresponding aerodynamic performance parameters, and non-dominated ranks are generated by sorting all individuals in the population according to non-dominated relationships; wherein the first rank in the non-dominated ranks is a complete non-dominated solution set, and the second rank is a non-dominated solution set remaining after the first rank is removed; A plurality of reference points are uniformly arranged in a target optimization space, and when next-generation individuals are selected from the population, individuals with high non-dominated ranks close to the reference points are preferentially selected, and the individuals with high non-dominated ranks are taken as the non-dominated optimal solutions; wherein the reference points correspond to optimization targets of a plurality of aerodynamic performance parameters.

2. The aerodynamic shape optimization method of claim 1, wherein, After the aerodynamic performance parameters of each of the new three-dimensional geometric models are output, the aerodynamic shape optimization method further comprises: If the aerodynamic performance parameters of the new three-dimensional geometric models are one, the displacement parameters of the model points corresponding to the plurality of new three-dimensional geometric models are iteratively optimized based on a genetic algorithm, and the optimal aerodynamic shape design scheme of the object is output when a preset convergence condition is met.

3. The aerodynamic shape optimization method of claim 1, wherein, The clustering analysis result based on the multiple non-dominated optimal solutions is used to dynamically adjust the displacement parameters of the model points, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, including: The multiple non-dominated optimal solutions are subjected to clustering analysis, and solution clusters with different performance tendencies are identified; The space range of the displacement parameters of the model points of each solution cluster is directionally contracted, and in the next generation evolution process, the distribution proportion of the population among the clusters is continuously dynamically adjusted according to the distribution density of each solution cluster in the target space; When the aerodynamic performance parameters of a solution cluster meet a preset convergence condition, the solution cluster is taken as the optimal aerodynamic shape design scheme of the object.

4. The aerodynamic shape optimization method of claim 2, wherein, The genetic algorithm is used to iteratively optimize the displacement parameters of the model points corresponding to the multiple new three-dimensional geometric models, and when a preset convergence condition is met, an optimal aerodynamic shape design scheme of the object is output, including: The displacement parameters of the multiple model points are taken as the initial population in the genetic algorithm; The aerodynamic performance parameters are taken as the fitness function, and the population is subjected to multi-generation iterative optimization by using the genetic algorithm, a new generation of individuals are generated each generation, and individuals with high fitness are reserved as optimal solutions; According to the distribution of the displacement parameters of the model points of the optimal solutions in the current generation, the search interval of the displacement parameters is determined, and a preset proportion of buffer boundaries is extended on the basis of the search interval; When the aerodynamic performance parameters of the individuals corresponding to the model points in the search interval meet a preset convergence condition, the individuals, the corresponding three-dimensional geometric models and the aerodynamic performance parameters are taken as the optimal aerodynamic shape design scheme.

5. The aerodynamic shape optimization method of claim 1, wherein, The aerodynamic performance prediction model is determined by the following steps: Sample point cloud data of multiple sample three-dimensional geometric models and corresponding sample aerodynamic performance parameters are input into a deep learning network model to learn the mapping relationship between the sample point cloud data and the sample aerodynamic performance parameters, and the predicted aerodynamic performance parameters of the sample point cloud data are predicted; The deep learning network model is iteratively trained based on the loss value between the predicted aerodynamic performance parameters and the sample aerodynamic performance parameters, and the aerodynamic performance prediction model is determined.

6. An apparatus for aerodynamic shape optimization of complex objects, characterized in that, The aerodynamic shape optimization device includes: A free deformation processing module is configured to transform the displacement parameters of multiple model points of an initial three-dimensional geometric model of an object, and deform the initial three-dimensional geometric model by using a direct free deformation technique to generate multiple new three-dimensional geometric models; An aerodynamic performance prediction module is configured to perform aerodynamic performance prediction processing on the point cloud data of the multiple new three-dimensional geometric models based on an aerodynamic performance prediction model, and output the aerodynamic performance parameters of each new three-dimensional geometric model. The intelligent optimization module is configured to: if the aerodynamic performance parameters of the new three-dimensional geometric models are multiple, search and select multiple non-dominated optimal solutions from the new three-dimensional geometric models based on a non-dominated sorting and a reference point guiding mechanism; dynamically adjust displacement parameters of model points based on clustering analysis results of the multiple non-dominated optimal solutions; and output an optimal aerodynamic shape design scheme of the object when a preset convergence condition is met. The aerodynamic performance prediction module is configured to: for any new three-dimensional geometric model, perform aerodynamic performance prediction processing on point cloud data of the new three-dimensional geometric models based on an aerodynamic performance prediction model, and output aerodynamic performance parameters of each new three-dimensional geometric model. The new three-dimensional geometric models are subjected to feature extraction and encoding processing in the aerodynamic performance prediction model, and voxel features corresponding to each point cloud data are determined. The multiple voxel features are processed based on a convolution layer, a linear layer, a normalization layer, and an activation function, and multiple embedding features are output. The multiple embedding features are sequentially processed based on a local attention layer, a feedforward network layer, a residual connection layer, and a normalization layer in a multi-layer encoder, and multiple multi-scale features are output. Global features are output by performing global pooling processing on the multiple multi-scale features, the global features are mapped based on a regressor, and the aerodynamic performance parameters are output. The intelligent optimization module is configured to search and select multiple non-dominated optimal solutions from the new three-dimensional geometric models based on the non-dominated sorting and the reference point guiding mechanism. A population is formed based on the multiple new three-dimensional geometric models and corresponding aerodynamic performance parameters, all individuals in the population are sorted according to a non-dominated relationship to generate a non-dominated level; the first level in the non-dominated level is a complete non-dominated solution set, and the second level is a non-dominated solution set remaining after the first level is removed. A set of reference points are uniformly arranged in a target optimization space, when selecting next-generation individuals in the population, individuals with high non-dominated levels close to the reference points are preferentially selected, and the individuals with high non-dominated levels are taken as the non-dominated optimal solutions; the reference points correspond to optimization targets of multiple aerodynamic performance parameters.

7. An electronic device, comprising: The processor, the memory, and the bus, the memory stores machine readable instructions executable by the processor, when the electronic device is running, the processor and the memory communicate through the bus, the machine readable instructions are executed by the processor to execute the steps of the aerodynamic shape optimization method of the complex object in any one of claims 1 to 5. The computer readable storage medium stores a computer program, the computer program is executed by the processor to execute the steps of the aerodynamic shape optimization method of the complex object in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, ​

Images