A three-dimensional aerodynamic shape design method, device, equipment and medium for an aircraft

Abstract

The present application discloses a three-dimensional aerodynamic shape design method, device, equipment and medium for an aircraft, which relates to the field of fluid mechanics and includes: training an initial aerodynamic shape geometric basic model by using the three-dimensional point cloud shape data of the aircraft to obtain a target aerodynamic shape geometric basic model; training an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model by using the low-dimensional latent space feature vectors and aerodynamic performance data extracted from the target aerodynamic shape geometric basic model to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model; inputting the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate a plurality of three-dimensional point cloud shape data, and using the target aerodynamic performance prediction model to screen out the qualified target three-dimensional point cloud shape data from the plurality of three-dimensional point cloud shape data to complete the three-dimensional aerodynamic shape design of the target aircraft. The present application can realize the three-dimensional aerodynamic shape design of aircraft with different configurations.

Description

Technical Field

[0001] The present invention relates to the field of fluid mechanics, and particularly to a method, device, equipment and medium for designing the three-dimensional aerodynamic shape of an aircraft. Background Art

[0002] The aerodynamic shape design is a key link in aircraft design, which has an important impact on flight performance and flight safety. Especially the aerodynamic shape design in the concept stage is an important bridge between requirements and solutions. It is the starting point of the entire shape design cycle and the basis for subsequent multidisciplinary design, and determines from the "gene" the degree to which the design scheme meets the usage requirements. In the aerodynamic shape concept design stage, the least known information and the highest design freedom exist in the entire design process. And the concept design scheme will directly affect the quality of the final design result, which poses quite high requirements for the designer's demand understanding ability, design experience and knowledge reserve.

[0003] Currently, the aerodynamic shape concept design process mainly includes key links such as "layout selection - parameterization - physical modeling - simulation (ground test) - performance analysis - feedback optimization". This method mainly uses parametric design technology and relies on the designer's knowledge reserve and design experience. There are mainly the following three problems: First, it strongly depends on expert experience and it is difficult to propose an initial scheme. Concept design has the least known information and the highest design freedom in the entire design process, and the concept design scheme will directly affect the quality of the final design result, which poses quite high requirements for the designer's demand understanding ability, design experience and knowledge reserve. Second, parametric design restricts the topological structure and the design exploration space is limited. Traditional parametric methods mainly include splines, shape transformation, parametric sections, etc., which map a set of parameters to a smooth curve or surface through a function for subsequent optimization. The aerodynamic shape topology of the parametric method is fixed, the range of parameter values is limited, the design space exploration is insufficient, and it is difficult to ensure the diversity of the scheme. Third, the design process involves repeated trial and error and the design efficiency is still low. During the design process, it is necessary to calculate and analyze the performance of the design scheme to give a candidate scheme, which is a process of repeated trial and error and has low efficiency, becoming a bottleneck in the entire aircraft development process. Therefore, how to design the aerodynamic shape efficiently and accurately is an urgent problem to be solved at present. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for designing the three-dimensional aerodynamic shape of an aircraft, which can accurately and efficiently complete the design of the three-dimensional aerodynamic shape of the aircraft according to requirements. The specific scheme is as follows:

[0005] In the first aspect, the present application discloses a method for designing the three-dimensional aerodynamic shape of an aircraft, including:

[0006] Construct an initial geometric basic model of the aerodynamic shape, and use the three-dimensional point cloud shape data of a preset number of aircraft to train the initial geometric basic model of the aerodynamic shape to obtain a corresponding target geometric basic model of the aerodynamic shape;

[0007] Construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract the low-dimensional latent space feature vectors corresponding to the three-dimensional point cloud shape data through the target geometric basic model of the aerodynamic shape;

[0008] Use the low-dimensional latent space feature vectors and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model;

[0009] When there is a design request for the three-dimensional aerodynamic shape of a target aircraft, input the aerodynamic performance indicators in the design request into the target aerodynamic shape generation model to generate three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators, and use the target aerodynamic performance prediction model to screen out the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data;

[0010] Among them, the construction of the initial geometric basic model of the aerodynamic shape includes:

[0011] Use the Transformer architecture, encoder, and decoder to construct an initial geometric basic model of the aerodynamic shape; among them, the encoder and the decoder both include a number of Transformer modules; the Transformer module includes a self-attention mechanism and a multi-layer perceptron;

[0012] The construction of the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model includes:

[0013] Construct an initial aerodynamic shape generation model according to the Transformer architecture through the generation model and the decoder, and construct an initial aerodynamic performance prediction model based on the encoder and the mapping neural network; the generation model is a diffusion model.

[0014] Optionally, before using the three-dimensional point cloud shape data of a preset number of aircraft to train the initial geometric basic model of the aerodynamic shape, it further includes:

[0015] Generate three-dimensional point cloud shape data of a preset number of aircraft through a parameterization method, and use a Cartesian grid solver to obtain the aerodynamic performance data corresponding to the three-dimensional point cloud shape data of the aircraft.

[0016] Optionally, inputting the aerodynamic performance indicators in the design request into the target aerodynamic shape generation model to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indicators includes:

[0017] Encoding the aerodynamic performance indicators and time steps in the design request by using a multi-layer perceptron corresponding to the target aerodynamic shape generation model to obtain corresponding encoded information;

[0018] Generating a feature vector based on the encoded information and the diffusion model, and decoding the feature vector by using the decoder to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indicators.

[0019] Optionally, screening out qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators by using the target aerodynamic performance prediction model includes:

[0020] Inputting the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators and the working conditions of the target aircraft into the target aerodynamic performance prediction model to obtain respective target aerodynamic performance data corresponding to the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators;

[0021] Screening each of the target aerodynamic performance data to determine target aerodynamic performance data meeting preset conditions;

[0022] Screening out qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators according to the target aerodynamic performance data meeting the preset conditions.

[0023] Optionally, the method further includes:

[0024] Generating a target number of three-dimensional point cloud shape data based on the target aerodynamic shape generation model and the aerodynamic performance indicators in the design request, so that the staff can directly complete the three-dimensional aerodynamic shape design of the target aircraft according to the target number of three-dimensional point cloud shape data.

[0025] In a second aspect, the present application discloses a device for three-dimensional aerodynamic shape design of an aircraft, including:

[0026] A first model acquisition module, configured to construct an initial aerodynamic shape geometric basic model, and train the initial aerodynamic shape geometric basic model by using three-dimensional point cloud shape data of a preset number of aircraft to obtain a corresponding target aerodynamic shape geometric basic model;

[0027] A feature vector extraction module, configured to construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract a low-dimensional latent space feature vector corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model;

[0028] A second model acquisition module, configured to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model by using the low-dimensional latent space feature vector and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data, so as to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model;

[0029] A point cloud shape data determination module, configured to, when there is a design request for the three-dimensional aerodynamic shape of a target aircraft, input the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate several pieces of three-dimensional point cloud shape data corresponding to the aerodynamic performance index, and use the target aerodynamic performance prediction model to screen out qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data;

[0030] Wherein, the first model acquisition module includes:

[0031] A first model construction unit, configured to construct an initial aerodynamic shape geometric basic model by using a Transformer architecture, an encoder, and a decoder; wherein, both the encoder and the decoder include a plurality of Transformer modules; the Transformer module includes a self-attention mechanism and a multi-layer perceptron;

[0032] The feature vector extraction module includes:

[0033] A second model construction unit, configured to construct an initial aerodynamic shape generation model according to the Transformer architecture through a generation model and the decoder, and construct an initial aerodynamic performance prediction model based on the encoder and a mapping neural network; the generation model is a diffusion model.

[0034] In a third aspect, the present application discloses an electronic device, including:

[0035] A memory, configured to store a computer program;

[0036] A processor, configured to execute the computer program to implement the method for designing the three-dimensional aerodynamic shape of an aircraft as described above.

[0037] In a fourth aspect, the present application discloses a computer-readable storage medium, configured to store a computer program, wherein the computer program, when executed by a processor, implements the method for designing the three-dimensional aerodynamic shape of an aircraft as described above.

[0038] The present application first constructs an initial aerodynamic shape geometric basic model, and uses the three-dimensional point cloud shape data of a preset number of aircraft to train the initial aerodynamic shape geometric basic model to obtain the corresponding target aerodynamic shape geometric basic model; then constructs an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extracts the low-dimensional latent space feature vectors corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model; then uses the low-dimensional latent space feature vectors and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model; finally, when there is a design request for the three-dimensional aerodynamic shape of the target aircraft, the aerodynamic performance index in the design request is input into the target aerodynamic shape generation model to generate a plurality of three-dimensional point cloud shape data corresponding to the aerodynamic performance index, and the target aerodynamic performance prediction model is used to screen out the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data. It can be seen that the present application constructs a three-dimensional point cloud aerodynamic shape data set to calculate the aerodynamic performance of the aircraft; by building a three-dimensional aerodynamic shape geometric basic model and training this model, the low-dimensional feature extraction and shape reconstruction of the three-dimensional aerodynamic shape are realized. Build a three-dimensional aerodynamic shape generation model and train this model to realize the rapid generation of the three-dimensional point cloud aerodynamic shape. Build an aerodynamic performance prediction model and train this model to realize the rapid prediction of the aerodynamic performance of the three-dimensional point cloud aerodynamic shape, and can quickly screen the three-dimensional aerodynamic shape obtained by the generation model to obtain a point cloud aerodynamic shape that meets the input index requirements. In this way, the present application uses the point cloud form to uniformly geometrically represent the three-dimensional aerodynamic shape, can quickly generate a large number of aerodynamic shapes in the form of point clouds, can realize the end-to-end intelligent generation from the aerodynamic performance index to the aerodynamic shape, and reduce the number of iterative optimizations. At the same time, the generated three-dimensional aerodynamic shape is relatively smooth, and the aerodynamic performance meets the input index requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0040] Figure 1 It is a flowchart of a method for designing a three-dimensional aerodynamic shape of an aircraft disclosed in the present application;

[0041] Figure 2A schematic diagram of a model training process disclosed in this application;

[0042] Figure 3 A schematic diagram of an aerodynamic shape scheme in the form of point cloud disclosed in this application;

[0043] Figure 4 (a) A schematic diagram of aerodynamic performance comparison disclosed in this application; Figure 4 (b) Another schematic diagram of aerodynamic performance comparison disclosed in this application;

[0044] Figure 5 A schematic diagram of the structure of a three-dimensional aerodynamic shape design device for an aircraft disclosed in this application;

[0045] Figure 6 A structural diagram of an electronic device disclosed in this application. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] Currently, there are mainly the following three problems in the conceptual design of aerodynamic shapes: First, it strongly relies on expert experience, and it is difficult to propose an initial scheme. In the entire design process, the conceptual design has the least known information and the highest design freedom, and the conceptual design scheme will directly affect the quality of the final design result, which poses quite high requirements for the designer's demand understanding ability, design experience, and knowledge reserve. Second, the parametric design restricts the topological structure, and the design exploration space is limited. The aerodynamic shape topology of the parametric method is fixed, the parameter value range is limited, the design space exploration is insufficient, and it is difficult to ensure the diversity of the scheme. Third, the design process involves repeated trial and error, and the design efficiency is still low. In the design process, it is necessary to calculate and analyze the performance of the design scheme and then give a candidate scheme, which is a process of repeated trial and error with low efficiency and has become a bottleneck in the entire aircraft development process. To solve the above technical problems, this application discloses a three-dimensional aerodynamic shape design method, device, equipment, and medium for an aircraft, which can accurately and efficiently complete the design of the three-dimensional aerodynamic shape of the aircraft according to requirements.

[0048] See Figure 1 As shown, the embodiments of the present invention disclose a three-dimensional aerodynamic shape design method for an aircraft, including:

[0049] Step S11: Construct an initial geometric basis model of the aerodynamic shape. Use the three-dimensional point cloud shape data of a preset number of aircraft to train the initial geometric basis model of the aerodynamic shape, and obtain the corresponding target geometric basis model of the aerodynamic shape.

[0050] In this embodiment, before constructing the initial geometric basis model of the aerodynamic shape, the three-dimensional point cloud shape data of a preset number of aircraft is generated by a parameterization method, and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data of the aircraft is obtained by using a Cartesian grid solver. That is to say, first, the dataset needs to be prepared, and then an initial geometric basis model of the aerodynamic shape is constructed using the Transformer architecture, an encoder, and a decoder. Among them, both the encoder and the decoder include a number of Transformer modules. The Transformer module includes a self-attention mechanism and a multi-layer perceptron. Specifically, the three-dimensional geometric basis model of the aerodynamic shape is constructed using the Transformer architecture and the "encoder-decoder" structure. The three-dimensional point cloud shape of the aircraft is used as the model input, and the reconstructed point cloud is used as the model output. The hidden space feature vector connects the encoder and the decoder. The model encoder and decoder are composed of a number of Transformer modules. Each Transformer module contains sub-modules such as a self-attention mechanism (self-atten-tion) and a multi-layer perceptron (MLP, Multilayer Perceptron). The function of the encoder is to encode the three-dimensional shape represented by the point cloud into a low-dimensional hidden space feature. The function of the decoder is the opposite of the encoder, which is to reconstruct the hidden space feature into a three-dimensional shape represented by the corresponding point cloud. Specifically, the encoder is used to reduce the dimension of the three-dimensional aerodynamic shape data so that the generation model can be trained and generated in the low-dimensional space, and then the decoder can reconstruct these low-dimensional data into a three-dimensional aerodynamic shape. Then, use the three-dimensional point cloud shape data of a preset number of aircraft to train the initial geometric basis model of the aerodynamic shape, and obtain the corresponding target geometric basis model of the aerodynamic shape. In this way, the trained target geometric basis model of the aerodynamic shape can perform dimensional compression and feature extraction of the three-dimensional aerodynamic shape.

[0051] Step S12: Construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract the low-dimensional hidden space feature vector corresponding to the three-dimensional point cloud shape data through the target geometric basis model of the aerodynamic shape.

[0052] In this embodiment, the main structure of the aerodynamic shape generation model includes a generation model and a feature vector decoder (the Decoder module of the geometric basic model), which takes aerodynamic indicators as input and point cloud shapes as output. Specifically, it includes an aerodynamic indicator input layer, a multi-layer perceptron encoding network, a generation model, a latent space feature vector layer, a feature vector decoder), an aerodynamic shape output layer, etc. The generation model uses a diffusion model, and the model architecture uses the same Transformer architecture as the feature vector decoder. In practical applications, the Unet (U-shaped network, a deep learning architecture) architecture can also be used according to requirements. The main structure of the aerodynamic performance prediction model includes a feature vector encoder (the Encoder module of the geometric basic model) and a mapping neural network. The model input is the point cloud shape and working conditions, and the output is the aerodynamic coefficient. The mapping model can use a residual neural network, a multi-layer fully connected neural network, etc. At the same time, this application can extract the low-dimensional latent space feature vector corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model.

[0053] Step S13: Use the low-dimensional latent space feature vector and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model, so as to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model.

[0054] In this embodiment, after the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model are constructed, this application will use the low-dimensional latent space feature vector and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model, so as to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model. That is, the training process of the model is completed, so as to realize the rapid generation of the three-dimensional point cloud aerodynamic shape according to the trained model.

[0055] Step S14: When there is a design request for the three-dimensional aerodynamic shape of the target aircraft, input the aerodynamic performance indicators in the design request into the target aerodynamic shape generation model to generate several three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators, and use the target aerodynamic performance prediction model to screen out the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data.

[0056] In this embodiment, when there is a design request for the three-dimensional aerodynamic shape of the target aircraft, the present application can input the aerodynamic performance indicators in the design request into the target aerodynamic shape generation model to generate three-dimensional point cloud shape data corresponding to a number of aerodynamic performance indicators. In this process, the inference process of the model is as follows: First, the multi-layer perceptron encoding network encodes the input aerodynamic design indicators and time steps, and the encoded information is used as a condition to guide the diffusion model to generate feature vectors. Then, the decoder decodes the feature vectors to obtain a point cloud aerodynamic shape that meets the requirements of the input indicators. Generally speaking, the multi-layer perceptron corresponding to the target aerodynamic shape generation model can be used to encode the aerodynamic performance indicators and time steps in the design request to obtain corresponding encoded information; then, based on the encoded information and the diffusion model, feature vectors are generated, and the decoder is used to decode the feature vectors to generate three-dimensional point cloud shape data corresponding to a number of aerodynamic performance indicators. Through the input aerodynamic index requirements and the trained target aerodynamic shape generation model, a large amount of three-dimensional point cloud shape data can be output.

[0057] Then, the three-dimensional point cloud shape data corresponding to a number of aerodynamic performance indicators and the operating conditions of the target aircraft are input into the target aerodynamic performance prediction model to obtain the target aerodynamic performance data corresponding to the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators; then, the target aerodynamic performance data are screened to determine the target aerodynamic performance data that meet the preset conditions; finally, according to the target aerodynamic performance data that meet the preset conditions, the target three-dimensional point cloud shape data that meet the conditions are screened out from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators. That is to say, the trained target aerodynamic performance prediction model can be used to quickly predict the aerodynamic performance of the three-dimensional point cloud aerodynamic shape, quickly screen the three-dimensional aerodynamic shape obtained by the generation model, obtain the final point cloud aerodynamic shape data, and then complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data. In addition, during the process of completing the three-dimensional aerodynamic shape design of the target aircraft, a target number of three-dimensional point cloud shape data can also be generated based on the target aerodynamic shape generation model and the aerodynamic performance indicators in the design request, so that the staff can directly complete the three-dimensional aerodynamic shape design of the target aircraft according to the target number of three-dimensional point cloud shape data. In other words, according to the performance of the aerodynamic shape generation model, a small amount of point cloud shapes can also be generated and directly used as alternative solutions for designers to complete the three-dimensional aerodynamic shape design of the target aircraft.

[0058] It can be seen that in this application, a three-dimensional point cloud aerodynamic shape dataset is constructed to calculate the aerodynamic performance of the aircraft. By building a three-dimensional aerodynamic shape geometric basic model and training this model, low-dimensional feature extraction and shape reconstruction of the three-dimensional aerodynamic shape are realized. A three-dimensional aerodynamic shape generation model is built and trained to achieve the rapid generation of the three-dimensional point cloud aerodynamic shape. An aerodynamic performance prediction model is built and trained to achieve the rapid prediction of the aerodynamic performance of the three-dimensional point cloud aerodynamic shape, and the three-dimensional aerodynamic shapes obtained from the generation model can be quickly screened to obtain the point cloud aerodynamic shapes that meet the input index requirements. In this way, this application uses the point cloud form to uniformly geometrically represent the three-dimensional aerodynamic shape, can quickly generate a large number of aerodynamic shapes in the form of point clouds, can realize the end-to-end intelligent generation from aerodynamic performance indicators to aerodynamic shapes, and reduce the number of iterative optimizations. At the same time, the generated three-dimensional aerodynamic shapes are relatively smooth, and the aerodynamic performance meets the input index requirements.

[0059] As can be seen from the previous embodiment, this application mainly includes three parts: a three-dimensional shape geometric basic model, an aerodynamic performance prediction model, and an aerodynamic shape generation model. Next, taking the intelligent design of a wing-body fusion layout airliner as an example, the specific process of the three-dimensional aerodynamic shape design method of the aircraft will be specifically described.

[0060] First of all, it should be known that the three-dimensional shape geometric basic model is used to extract the low-dimensional latent space feature vector of the three-dimensional aerodynamic shape and has the ability to decode the three-dimensional aerodynamic shape from the latent space feature vector; the aerodynamic shape generation model is a generative model that embeds aerodynamic performance guidance on the decoder of the basic model for the aerodynamic shape generation task and is trained in the latent space to achieve the controllable generation of the three-dimensional aerodynamic shape latent space feature vector, and the latent space feature vector is decoded into a point cloud form through the decoder; the shape aerodynamic prediction model is a mapping network added on the encoder of the basic model for the aerodynamic coefficient prediction task to achieve rapid inference and prediction of the aerodynamic performance.

[0061] When designing the three-dimensional aerodynamic shape of the aircraft, first prepare the dataset. The reference shape of the wing-body fusion layout airliner consists of four parts: the central body, the fusion section, the outer wing section, and the vertical tail. 10,000 three-dimensional shapes in the form of point clouds are generated as the training dataset by the parameterization method, and the number of points for each shape is 50,000. The Cartesian grid solver is used to solve its aerodynamic performance. Then model training is carried out, and the training process is as Figure 2As shown in the figure. The construction, training, and inference of the model are all implemented based on the open-source deep learning framework MindSpore. The point cloud reconstruction loss function of the geometric basic model is calculated using the chamfer distance. The mean square error (MSE) is selected as the loss function for the aerodynamic performance prediction model, and the L2 loss is used as the loss function for the aerodynamic shape generation model. First, the three-dimensional point cloud shape data of the wing-body fusion layout airliner prepared is used to train the geometric basic model; then, the generation model and the performance prediction model are trained with the aerodynamic performance data and the low-dimensional latent space feature vectors obtained from the geometric basic model.

[0062] After the model training is completed, as Figure 2 shown in the figure. Using the aerodynamic performance index as the input, a large number of aerodynamic shape schemes in the form of point clouds as Figure 3 shown in the figure are generated by the generation model, and then the aerodynamic coefficients of the schemes are quickly predicted by the aerodynamic performance prediction model. The better schemes are selected as the alternative schemes for the designer according to the satisfaction of the input values (the aerodynamic performance indexes input into the model), and then the three-dimensional aerodynamic shape design of the aircraft is completed. In addition, a small number of point cloud shapes can be generated directly as alternative schemes for the designer according to the performance of the aerodynamic shape generation model. The comparison of the aerodynamic performance calculated by the aerodynamic performance prediction model for a certain shape in the aerodynamic shape scheme is as Figure 4 shown in the figure. The calculated value is very close to the expected value (input value), which indicates that the alternative schemes generated by the model of the present application can meet the requirements of the input indexes and can be provided to the designer for subsequent design.

[0063] In this way, the present application uses the point cloud form to uniformly geometrically represent the three-dimensional aerodynamic shape, can quickly generate a large number of aerodynamic shapes in the form of point clouds, can realize the end-to-end intelligent generation from the aerodynamic performance index to the aerodynamic shape, and reduce the number of iterative optimizations. At the same time, the generated three-dimensional aerodynamic shape is relatively smooth, and the aerodynamic performance meets the requirements of the input indexes.

[0064] See Figure 5 shown in the figure. An embodiment of the present invention discloses a device for designing a three-dimensional aerodynamic shape of an aircraft, including:

[0065] The first model acquisition module 11 is used to construct an initial aerodynamic shape geometric basic model, and train the initial aerodynamic shape geometric basic model with the three-dimensional point cloud shape data of a preset number of aircraft to obtain the corresponding target aerodynamic shape geometric basic model;

[0066] The feature vector extraction module 12 is used to construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract the low-dimensional latent space feature vectors corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model;

[0067] The second model acquisition module 13 is configured to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model by using the low-dimensional latent space feature vector and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data, so as to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model;

[0068] The point cloud shape data determination module 14 is configured to, when there is a design request for the three-dimensional aerodynamic shape of a target aircraft, input the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate a plurality of three-dimensional point cloud shape data corresponding to the aerodynamic performance index, and use the target aerodynamic performance prediction model to screen out the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data.

[0069] It can be seen that in this application, a three-dimensional point cloud aerodynamic shape data set is constructed to calculate the aerodynamic performance of the aircraft; by building a three-dimensional aerodynamic shape geometric basic model and training this model, low-dimensional feature extraction and shape reconstruction of the three-dimensional aerodynamic shape are realized. A three-dimensional aerodynamic shape generation model is built and trained to realize the rapid generation of the three-dimensional point cloud aerodynamic shape. An aerodynamic performance prediction model is built and trained to realize the rapid prediction of the aerodynamic performance of the three-dimensional point cloud aerodynamic shape, and the three-dimensional aerodynamic shapes obtained by the generation model can be quickly screened to obtain the point cloud aerodynamic shapes that meet the input index requirements. In this way, this application uses the point cloud form to uniformly geometrically represent the three-dimensional aerodynamic shape, can quickly generate a large number of aerodynamic shapes in the form of point clouds, can realize the end-to-end intelligent generation from the aerodynamic performance index to the aerodynamic shape, and reduce the number of iterative optimizations. At the same time, the generated three-dimensional aerodynamic shape is relatively smooth, and the aerodynamic performance meets the input index requirements.

[0070] In some specific embodiments, the device may further include:

[0071] The aerodynamic performance data acquisition module is configured to generate three-dimensional point cloud shape data of a preset number of aircrafts by a parameterization method, and use a Cartesian grid solver to obtain the aerodynamic performance data corresponding to the three-dimensional point cloud shape data of the aircrafts.

[0072] In some specific embodiments, the first model acquisition module 11 may specifically include:

[0073] The first model construction unit is configured to build an initial aerodynamic shape geometric basic model by using a Transformer architecture, an encoder, and a decoder; wherein, both the encoder and the decoder include a plurality of Transformer modules; the Transformer module includes a self-attention mechanism and a multi-layer perceptron.

[0074] In some specific embodiments, the feature vector extraction module 12 may specifically include:

[0075] A second model construction unit, configured to construct an initial aerodynamic shape generation model based on the Transformer architecture through a generative model and the decoder, and construct an initial aerodynamic performance prediction model based on the encoder and a mapping neural network; the generative model is a diffusion model.

[0076] In some specific embodiments, the point cloud shape data determination module 14 may specifically include:

[0077] An encoding information acquisition unit, configured to encode the aerodynamic performance indicators and time steps in the design request by using a multi-layer perceptron corresponding to the target aerodynamic shape generation model, and obtain corresponding encoding information;

[0078] A decoding unit, configured to generate a feature vector based on the encoding information and the diffusion model, and decode the feature vector by using the decoder to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indicators.

[0079] In some specific embodiments, the point cloud shape data determination module 14 may specifically include:

[0080] A target aerodynamic performance data acquisition unit, configured to input the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators and the operating conditions of the target aircraft into the target aerodynamic performance prediction model to obtain respective target aerodynamic performance data corresponding to the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators;

[0081] A target aerodynamic performance data determination unit, configured to screen the respective target aerodynamic performance data to determine target aerodynamic performance data that meets preset conditions;

[0082] A target three-dimensional point cloud shape data screening unit, configured to screen out qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indicators according to the target aerodynamic performance data that meets the preset conditions.

[0083] In some specific embodiments, the device may further include:

[0084] A three-dimensional point cloud shape data generation module, configured to generate a target number of three-dimensional point cloud shape data based on the target aerodynamic shape generation model and the aerodynamic performance indicators in the design request, so that a staff member can directly complete the three-dimensional aerodynamic shape design of the target aircraft according to the target number of three-dimensional point cloud shape data.

[0085] Furthermore, an embodiment of the present application also discloses an electronic deviceFigure 6 It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment. The content in the figure should not be regarded as any limitation on the scope of use of this application.

[0086] Figure 6 It is a schematic structural diagram of an electronic device 20 provided by an embodiment of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the three-dimensional aerodynamic shape design method of the aircraft disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0087] In this embodiment, the power supply 23 is used to provide operating voltages for the various hardware devices on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows is any communication protocol applicable to the technical solution of this application, and specific limitations are not imposed on it here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application requirements, and no specific limitations are imposed here.

[0088] In addition, as a carrier for resource storage, the memory 22 may be a read-only memory, a random access memory, a magnetic disk, or an optical disc, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.

[0089] Among them, the operating system 221 is used to manage and control the various hardware devices and the computer program 222 on the electronic device 20, and it may be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the three-dimensional aerodynamic shape design method of the aircraft executed by the electronic device 20 disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs that can be used to complete other specific tasks.

[0090] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the three-dimensional aerodynamic shape design method disclosed above. For the specific steps of this method, reference may be made to the corresponding content disclosed in the foregoing embodiments, and details are not repeated here.

[0091] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0092] Those skilled in the art can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0093] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be directly implemented by hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0094] Finally, it should also be noted that in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0095] The above has introduced the technical solution provided by this application in detail. Specific examples are used herein to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A three-dimensional aerodynamic shape design method for an aircraft, characterized in that, Including: Construct an initial aerodynamic shape geometric basic model, and train the initial aerodynamic shape geometric basic model using the three-dimensional point cloud shape data of a preset number of aircraft to obtain the corresponding target aerodynamic shape geometric basic model; Construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract the low-dimensional latent space feature vectors corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model; Train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model using the low-dimensional latent space feature vectors and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model; When there is a design request for the three-dimensional aerodynamic shape of a target aircraft, input the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indexes, and use the target aerodynamic performance prediction model to screen out the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance indexes, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data; Among them, the constructing of the initial aerodynamic shape geometric basic model includes: Construct an initial aerodynamic shape geometric basic model using the Transformer architecture, an encoder, and a decoder; wherein, both the encoder and the decoder include a plurality of Transformer modules; the Transformer module includes a self-attention mechanism and a multi-layer perceptron; wherein, the encoder is used to reduce the dimension of the three-dimensional aerodynamic shape data, so that the generation model is trained and generated in the low-dimensional space, and then the decoder can reconstruct these low-dimensional data into a three-dimensional aerodynamic shape; The constructing of the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model includes: Construct an initial aerodynamic shape generation model according to the Transformer architecture through the generation model and the decoder, and construct an initial aerodynamic performance prediction model based on the encoder and a mapping neural network; the generation model is a diffusion model based on the Transformer architecture, and generates feature vectors that meet the aerodynamic performance indexes in the latent space; The inputting of the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indexes includes: Use the multi-layer perceptron corresponding to the target aerodynamic shape generation model to encode the aerodynamic performance index and the time step in the design request to obtain the corresponding encoded information; Generate feature vectors based on the encoded information and the diffusion model, and use the decoder to decode the feature vectors to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indexes.

2. The three-dimensional aerodynamic shape design method of the aircraft according to claim 1, wherein Before training the initial aerodynamic shape geometric basic model using the three-dimensional point cloud shape data of a preset number of aircraft, it further includes: Generate the three-dimensional point cloud shape data of a preset number of aircraft through a parametric method, and use a Cartesian grid solver to obtain the aerodynamic performance data corresponding to the three-dimensional point cloud shape data of the aircraft.

3. The three-dimensional aerodynamic shape design method of an aircraft according to claim 1 or 2, characterized in that The screening of the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index by using the target aerodynamic performance prediction model includes: Input the three-dimensional point cloud shape data corresponding to the aerodynamic performance index and the working conditions of the target aircraft into the target aerodynamic performance prediction model to obtain the respective target aerodynamic performance data corresponding to the three-dimensional point cloud shape data corresponding to the aerodynamic performance index; Screen the respective target aerodynamic performance data to determine the target aerodynamic performance data that meet the preset conditions; Screen the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index according to the target aerodynamic performance data that meet the preset conditions.

4. The three-dimensional aerodynamic shape design method of an aircraft according to claim 1, characterized in that, It further includes: Generate a target number of three-dimensional point cloud shape data based on the target aerodynamic shape generation model and the aerodynamic performance index in the design request, so that the staff can directly complete the three-dimensional aerodynamic shape design of the target aircraft according to the target number of three-dimensional point cloud shape data.

5. A three-dimensional aerodynamic shape design device for an aircraft, characterized in that, It includes: A first model acquisition module, configured to construct an initial aerodynamic shape geometric basic model, and use the three-dimensional point cloud shape data of a preset number of aircraft to train the initial aerodynamic shape geometric basic model to obtain the corresponding target aerodynamic shape geometric basic model; A feature vector extraction module, configured to construct an initial aerodynamic shape generation model and an initial aerodynamic performance prediction model, and extract the low-dimensional latent space feature vector corresponding to the three-dimensional point cloud shape data through the target aerodynamic shape geometric basic model; A second model acquisition module, configured to train the initial aerodynamic shape generation model and the initial aerodynamic performance prediction model by using the low-dimensional latent space feature vector and the aerodynamic performance data corresponding to the three-dimensional point cloud shape data to obtain a target aerodynamic shape generation model and a target aerodynamic performance prediction model; A point cloud shape data determination module, configured to, when there is a design request for the three-dimensional aerodynamic shape of a target aircraft, input the aerodynamic performance index in the design request into the target aerodynamic shape generation model to generate a plurality of three-dimensional point cloud shape data corresponding to the aerodynamic performance index, and use the target aerodynamic performance prediction model to screen the qualified target three-dimensional point cloud shape data from the three-dimensional point cloud shape data corresponding to the aerodynamic performance index, so as to complete the three-dimensional aerodynamic shape design of the target aircraft based on the target three-dimensional point cloud shape data; Among them, the first model acquisition module includes: The first model construction unit is used to construct an initial aerodynamic shape geometric basic model by using the Transformer architecture, an encoder, and a decoder; wherein, both the encoder and the decoder include a plurality of Transformer modules; the Transformer module includes a self-attention mechanism and a multi-layer perceptron; wherein, the encoder is used to reduce the dimension of the three-dimensional aerodynamic shape data, so that the generation model is trained and generated in the low-dimensional space, and then the decoder can reconstruct these low-dimensional data into a three-dimensional aerodynamic shape; The feature vector extraction module includes: The second model construction unit is used to construct an initial aerodynamic shape generation model according to the Transformer architecture through a generation model and the decoder, and construct an initial aerodynamic performance prediction model based on the encoder and a mapping neural network; the generation model is a diffusion model based on the Transformer architecture, and generates feature vectors that meet the aerodynamic performance indicators in the latent space; The point cloud shape data determination module includes: The encoding information acquisition unit is used to encode the aerodynamic performance indicators and time steps in the design request by using the multi-layer perceptron corresponding to the target aerodynamic shape generation model, and acquire the corresponding encoding information; The decoding unit is used to generate feature vectors based on the encoding information and the diffusion model, and decode the feature vectors by using the decoder to generate three-dimensional point cloud shape data corresponding to a plurality of the aerodynamic performance indicators.

6. An electronic device, characterized in that, It includes: A memory for storing a computer program; A processor for executing the computer program to implement the three-dimensional aerodynamic shape design method for an aircraft according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, For storing a computer program, wherein the computer program, when executed by the processor, implements the three-dimensional aerodynamic shape design method for an aircraft according to any one of claims 1 to 4.

Images