A method for predicting aerodynamic parameters, a storage medium and a computer program product

Abstract

Embodiments of the present application provide a method for predicting aerodynamic parameters, a storage medium and a computer program product, the method comprising: obtaining a to-be-predicted object; when the to-be-predicted object is a three-dimensional wing, outputting an aerodynamic parameter prediction result corresponding to the to-be-predicted object through a wing prediction model, the wing prediction model being trained based on a plurality of cross-sectional images and aerodynamic parameter labels of a plurality of three-dimensional wing samples, the aerodynamic parameter labels of the plurality of three-dimensional wing samples being calculated based on a RANS method coupled with a transition model; and when the to-be-predicted object is a three-dimensional blade, outputting an aerodynamic parameter prediction result corresponding to the to-be-predicted object through a blade prediction model, the blade prediction model being trained based on a plurality of cross-sectional images and aerodynamic parameter labels of a plurality of three-dimensional blade samples, the aerodynamic parameter labels of the plurality of three-dimensional blade samples being calculated based on a LES method. Embodiments of the present application aim to improve the speed of accurately predicting the aerodynamic parameters of a wing or a blade.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method for predicting aerodynamic parameters, a storage medium, and a computer program product. Background Technology

[0002] In the early aerodynamic design of aircraft wings or other low-speed blades, wind tunnel experiments and theoretical calculations were mainly relied upon. However, wind tunnel experiments can only simulate relatively simple operating conditions, and their high cost prevents their large-scale application.

[0003] Subsequently, with the development of computational fluid dynamics, numerical simulation algorithms can be used to solve the aerodynamic performance of most wings or blades relatively accurately. CFD simulation software can be used to obtain the flow field information and important aerodynamic parameters of aircraft or wind turbine blades. However, although numerical simulation methods have reduced experimental costs, they have the problem of long simulation time. Aerodynamic optimization requires repeated adjustment and verification of the geometric parameters of aircraft wings or wind turbine blades, and the long simulation time will seriously slow down this process.

[0004] Therefore, improving the speed of accurately predicting the aerodynamic parameters of wings or blades is an urgent problem to be solved. Summary of the Invention

[0005] This application provides an aerodynamic parameter prediction method, storage medium, and computer program product, aiming to improve the speed of accurately predicting the aerodynamic parameters of wings or blades.

[0006] In a first aspect, embodiments of this application provide a method for predicting aerodynamic parameters, the method comprising: Obtain the object to be predicted, which includes a three-dimensional wing and a three-dimensional blade; When the object to be predicted is a three-dimensional air wing, multiple cross-sectional images corresponding to the object to be predicted are input into the air wing prediction model, and the air wing prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The air wing prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional air wing samples, and the aerodynamic parameter labels of the multiple three-dimensional air wing samples are calculated based on the RANS method of the coupled transition model. When the object to be predicted is a three-dimensional blade, multiple cross-sectional images corresponding to the object to be predicted are input into the blade prediction model, and the blade prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The blade prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional blade samples. The aerodynamic parameter labels of the multiple three-dimensional blade samples are calculated based on the LES method.

[0007] Optionally, the method further includes: The geometric contours of multiple three-dimensional wing samples are obtained, and the aerodynamic parameter labels corresponding to each three-dimensional wing sample are calculated by the RANS method based on the coupled transition model. Construct a three-dimensional wing sample dataset, which includes multiple three-dimensional wing samples. Each three-dimensional wing sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to that three-dimensional wing. A first convolutional neural network is constructed, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional wing. The first convolutional neural network is iteratively trained based on the three-dimensional wing sample dataset to obtain the wing prediction model.

[0008] Optionally, the geometric contours of multiple three-dimensional wing samples are obtained, and the aerodynamic parameter labels corresponding to each three-dimensional wing sample are calculated using the RANS method based on the coupled transition model, including: For any three-dimensional wing sample, obtain the geometric contour corresponding to the three-dimensional wing sample, and input the SA model of the coupled transition model, wherein the SA model is a model based on the RANS method; The aerodynamic parameter labels corresponding to the three-dimensional wing sample are calculated using the SA model of the coupled transition model.

[0009] Optionally, the calculation formula for the SA model of the coupled transition model is:

[0010] in, The equations are obtained by solving the convection transport equations. The coefficient of dynamic viscosity, For density, Indicates kinematic viscosity; Geometric coordinates; and These are the coupled generation and dissipation terms, respectively; For additional generated items; These are empirical parameters; This represents the average speed.

[0011] Optionally, the method further includes: Obtain the geometric contours of multiple three-dimensional blade samples, and calculate the aerodynamic parameter labels corresponding to each three-dimensional blade sample based on the LES method. Construct a three-dimensional blade sample dataset, which includes multiple three-dimensional blade samples. Each three-dimensional blade sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to the three-dimensional blade. A second convolutional neural network is constructed, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional leaf. The second convolutional neural network is iteratively trained based on the three-dimensional leaf sample dataset to obtain the leaf prediction model.

[0012] Optionally, after outputting the predicted aerodynamic parameters of the object to be predicted through the wing prediction model, the method further includes: In response to the target operation, the target aerodynamic parameters corresponding to the object to be predicted are obtained, and the target aerodynamic parameters include at least one of the following parameter types: target drag coefficient, target pressure coefficient and maximum lift-to-drag ratio; Based on the target aerodynamic parameters corresponding to the object to be predicted and the aerodynamic parameter prediction results, the geometric parameters of the object to be predicted are iteratively updated until the object to be predicted after iterative updates satisfies the target aerodynamic parameters.

[0013] Optionally, after outputting the predicted aerodynamic parameters of the object to be predicted through the wing prediction model, the method further includes: In response to the target operation, the target aerodynamic parameters corresponding to the object to be predicted are obtained, and the target aerodynamic parameters include at least one of the following parameter types: target drag coefficient, target pressure coefficient and maximum lift-to-drag ratio; Based on the target aerodynamic parameters corresponding to the object to be predicted and the aerodynamic parameter prediction results, the geometric parameters of the object to be predicted are iteratively updated until the iteratively updated object to be predicted satisfies the target aerodynamic parameters, including: Calculate the target difference based on the target aerodynamic parameters corresponding to the object to be predicted and the prediction results of the aerodynamic parameters; When the target difference is greater than the preset target threshold, the geometric parameters of the object to be predicted are adjusted, and the object to be predicted after the geometric parameters are adjusted is taken as the object to be predicted. Based on the wing prediction model or the blade prediction model, the aerodynamic parameter prediction result corresponding to the current object to be predicted is calculated. When the target difference is less than the target threshold, the iteration stops, and the current prediction results of the geometric and aerodynamic parameters of the object to be predicted are output.

[0014] Optionally, the formula for calculating the target difference is:

[0015] Where MSE is the target difference; n is the number of parameter types in the target aerodynamic parameters; The i-th aerodynamic parameter in the predicted aerodynamic parameter results; Let i be the i-th aerodynamic parameter among the target aerodynamic parameters.

[0016] In a second aspect, embodiments of this application provide a computer device, including: at least one processor and a memory, the memory storing a computer program executable on the processor, wherein the processor executes the computer program to perform the aerodynamic parameter prediction method described in the first aspect of the embodiment.

[0017] Thirdly, embodiments of this application provide a non-volatile readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the aerodynamic parameter prediction method described in the first aspect of the embodiment.

[0018] Fourthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the aerodynamic parameter prediction method described in the first aspect of the embodiments.

[0019] Beneficial effects: In this method, when predicting the aerodynamic parameters of a high-speed 3D wing, a wing prediction model is used. The wing prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple 3D wing samples. The aerodynamic parameter labels of each 3D wing sample are calculated based on the RANS method of the coupled transition model. The RANS method of the coupled transition model can obtain more accurate aerodynamic parameters of the 3D wing. The wing prediction model learns the relationship between multiple cross-sectional images and aerodynamic parameter labels of the 3D wing samples. In subsequent prediction processes, the wing prediction model can quickly and accurately generate aerodynamic parameter prediction results for high-speed 3D wings.

[0020] When predicting the aerodynamic parameters of a low-speed three-dimensional blade, a blade prediction model is used. The blade prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional blade samples. The aerodynamic parameter labels of each three-dimensional blade sample are calculated based on the LES method. The blade prediction model learns the relationship between multiple cross-sectional images and aerodynamic parameter labels of the three-dimensional blade samples. In subsequent predictions, the blade prediction model can quickly generate the aerodynamic parameter prediction results of the three-dimensional blade.

[0021] Compared to existing aerodynamic parameter calculation methods that all use RANS, this method first uses different methods to calculate the aerodynamic parameter labels for high-speed 3D wings and low-speed 3D blades, which can improve the accuracy of aerodynamic parameter calculation for 3D wings and 3D blades. Furthermore, based on the different methods used to calculate the aerodynamic parameter labels, the wing prediction model and blade prediction model are trained and classified, which can quickly predict the aerodynamic parameters of 3D wings and 3D blades, thereby improving the speed of accurately predicting the aerodynamic parameters of wings or blades. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0023] Figure 1 A flowchart illustrating the steps of the aerodynamic parameter prediction method provided in an embodiment of this application is shown. Figure 2 A schematic diagram of the unfolding direction provided in an embodiment of this application is shown; Figure 3 A flowchart illustrating the steps of training the wing prediction model provided in an embodiment of this application is shown. Figure 4 This paper shows a schematic diagram of the structure of the first convolutional neural network provided in an embodiment of this application; Figure 5 A flowchart illustrating the training process of the blade prediction model provided in this application embodiment is shown. Figure 6 This illustration shows a schematic diagram of the genetic algorithm optimization process provided in an embodiment of this application; Figure 7 A schematic diagram of a non-volatile readable storage medium provided in an embodiment of this application is shown; Figure 8 A schematic diagram of a computer program product provided in an embodiment of this application is shown. Detailed Implementation

[0024] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been presented in the various embodiments of this application to enable readers to better understand this application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can be implemented. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0026] PCA: Principal Components Analysis, aims to transform multi-indicator data into a few comprehensive indicators using the concept of dimensionality reduction.

[0027] RANS: Reynolds equation, a computational method in computational fluid dynamics that averages physical quantities in a flow field over time and then substitutes these averaged quantities into the traditional Navier-Stokes equations for solving the problem.

[0028] LES: Large Eddy Simulation, is a computational method in computational fluid dynamics. It divides eddies in the flow field into large eddies and small eddies at a certain scale. Large eddies are solved directly using the Navier-Stokes equations, while small eddies are solved by constructing a new subgrid stress model.

[0029] Transition: The process by which the flow in the model changes from laminar to turbulent.

[0030] In the complex process of designing and developing large aircraft such as airplanes, aerodynamic optimization of wing blades is a core element, significantly impacting the overall aerodynamic performance of the aircraft. As the requirements for comprehensive aircraft performance continue to increase, improving aerodynamic parameters such as lift-to-drag ratio and lift-to-drag coefficient through aerodynamic optimization of aircraft wings has become an important and urgently needed area for improvement. Furthermore, for smaller aircraft such as low-altitude drones and low-speed blades like wind turbine blades, aerodynamic optimization is still essential to improve their aerodynamic performance.

[0031] In the early aerodynamic design of aircraft wings or other low-speed blades, wind tunnel experiments and theoretical calculations were mainly relied upon. However, wind tunnel experiments can only simulate relatively simple operating conditions, and their high cost prevents their large-scale application.

[0032] With the development of computational fluid dynamics, numerical simulation algorithms can be used to solve the aerodynamic performance of most wings or blades relatively accurately. CFD simulation software can be used to obtain the flow field information and important aerodynamic parameters of aircraft or wind turbine blades. Aerodynamic design is divided into two directions: inverse design and optimization design. Inverse design obtains the wing or blade geometry that satisfies the given pressure or velocity distribution on the wing or blade surface. Optimization design takes specified aerodynamic performance parameters (such as maximizing lift-to-drag ratio) as the optimization objective and finally uses optimization algorithms to adjust the shape of the wing or blade. However, although numerical simulation methods reduce experimental costs, they have the problem of long simulation time. Aerodynamic optimization requires repeated adjustment and verification of the geometric parameters of aircraft wings or wind turbine blades, and the long simulation time will seriously slow down this process.

[0033] Therefore, how to improve the speed of accurately predicting the aerodynamic parameters of wings or blades is an urgent problem to be solved; in order to improve the speed of accurately predicting the aerodynamic parameters of wings or blades, this application provides an aerodynamic parameter prediction method.

[0034] Reference Figure 1 The diagram illustrates a flowchart of a method for predicting aerodynamic parameters according to an embodiment of this application. The method may include the following steps: S101: Obtain the object to be predicted, which includes a three-dimensional wing and a three-dimensional blade.

[0035] In the actual implementation process, the object to be predicted is first obtained. The object to be predicted includes a three-dimensional wing and a three-dimensional blade. For example, the user can provide the three-dimensional shape of the three-dimensional wing or the three-dimensional blade. Existing image recognition technology can identify whether the object to be predicted belongs to a three-dimensional wing or a three-dimensional blade. Alternatively, prediction options for the three-dimensional wing and prediction options for the three-dimensional blade can be set separately. The user can upload multiple cross-sectional images of the three-dimensional wing to be predicted through the prediction option for the three-dimensional wing, and multiple cross-sectional images of the three-dimensional blade to be predicted through the prediction option for the three-dimensional blade.

[0036] Reference Figure 2 The diagram shows a spanwise schematic provided in the embodiments of this application. The cross-sectional image refers to the contour image containing the geometric features of a three-dimensional wing or a three-dimensional blade. For any three-dimensional wing or three-dimensional blade, there are cross-sectional images with different spanwise directions. The spanwise direction refers to the direction in which it extends. By cutting the three-dimensional wing or three-dimensional blade at different spanwise positions, different cross-sectional images can be obtained.

[0037] S102: When the object to be predicted is a three-dimensional airfoil, multiple cross-sectional images corresponding to the object to be predicted are input into the airfoil prediction model, and the airfoil prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The airfoil prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional airfoil samples. The aerodynamic parameter labels of the multiple three-dimensional airfoil samples are calculated based on the RANS method of the coupled transition model.

[0038] The aerodynamic parameters included in this embodiment include: pressure coefficient, drag coefficient, and lift coefficient, etc.

[0039] This method calculates aerodynamic parameter labels for multiple 3D wing samples using the RANS method based on a coupled transition model. Then, a wing prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of the multiple 3D wing samples. The wing prediction model predicts the aerodynamic parameters of the 3D wing. This eliminates the need to use numerical simulation algorithms to calculate the aerodynamic parameters of the 3D wing each time, thereby reducing the computational cost of prediction and accelerating the efficiency of aerodynamic parameter prediction.

[0040] Furthermore, for high-speed aircraft wings, the LES method requires a very large amount of mesh at high speeds. Therefore, for high-speed 3D wings, this embodiment uses the RANS method of coupled transition model to calculate the aerodynamic parameters of the 3D wing as training labels, which can obtain more accurate aerodynamic parameter calculation results for the 3D wing. Then, the wing prediction model learns the relationship between multiple cross-sectional images of the 3D wing sample and the aerodynamic parameter labels calculated by the RANS method of coupled transition model, which can quickly obtain more accurate aerodynamic parameter prediction results for the 3D wing.

[0041] Reference Figure 3 The diagram illustrates the steps of training a wing prediction model according to an embodiment of this application. In one feasible implementation, the training process of the wing prediction model includes: A1: Obtain the geometric contours of multiple 3D wing samples, and calculate the aerodynamic parameter labels corresponding to each 3D wing sample using the RANS method based on the coupled transition model.

[0042] First, multiple 3D wing samples are obtained. Commonly used 3D wing samples can be selected as 3D wing samples. The geometric contour network of the 3D wing samples is used as input. The velocity distribution around each 3D wing is simulated using open-source CFD simulation software. The aerodynamic parameter labels of each 3D wing sample can be calculated from the velocity distribution.

[0043] In this method, the RANS method based on the coupled transition model is used to calculate the aerodynamic parameter labels of each three-dimensional wing sample in open-source CFD simulation software.

[0044] Specifically, the core equation in the RANS method—the momentum equation—has the following form:

[0045] In the formula, For density, For time, For pressure, Geometric coordinates The average velocity is represented by the subscripts i, j, and k, which indicate the components of the traversal velocity in three directions in three-dimensional space. The short horizontal line above indicates the time average. get, and These are the total velocity and the pulsating velocity, respectively. For the Kronecker function, It characterizes centrifugal force and Coriolis force, and is applicable to problems such as wind turbine blades that require a rotating coordinate system.

[0046] In the formula, Let be the viscous stress tensor, and its expression is: ,in, Indicates kinematic viscosity; In the formula, Let Reynolds stress tensor be the expression for: ,in, It is turbulent eddy viscosity.

[0047] The RANS method mainly involves averaging some physical quantities over time and then substituting them into the traditional Navier-Stokes equations. This makes the traditional Navier-Stokes equations less closed, and the reason for this is that... Solve for it.

[0048] Therefore, further modeling is required. There are many forms of modeling. In this embodiment, for any three-dimensional wing sample, the geometric contour corresponding to the three-dimensional wing sample is obtained, and the SA model of the coupling transition model is input. The SA model is a model based on the RANS method. Through the SA model of the coupling transition model, the aerodynamic parameter labels corresponding to the three-dimensional wing sample are calculated.

[0049] Specifically, in the SA model, Represented as ,in For empirical parameter terms, This is solved by the convection transport equation, which takes the following form:

[0050] in, The coefficient of dynamic viscosity, and These are the generating and dissipating terms of the equation, respectively, which are also obtained from empirical formulas. These are empirical parameters.

[0051] Current RANS methods lack the ability to predict transition processes, therefore an additional transition equation is needed to solve for the intermittent factor. .

[0052] The transition equation takes the following form:

[0053] in, and These are the generating and dissipating terms of the transition equation, respectively.

[0054] After coupling the transition equation with the traditional SA model equation, the form of the SA model equation coupled with the transition model becomes:

[0055] in, and These are the coupled generating and dissipating terms, respectively. The transformed intermittent factor, This is an additional generation item used to ensure the accuracy of calculation results under certain special operating conditions.

[0056] In the equations of the SA model of the above coupled transition model, based on the intermittent factor Different values ​​can be used to switch between laminar and turbulent states, thus enabling more accurate aerodynamic parameter labels to be calculated for high-speed three-dimensional wing samples.

[0057] A2: Construct a three-dimensional wing sample dataset, which includes multiple three-dimensional wing samples. Each three-dimensional wing sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to that three-dimensional wing.

[0058] After generating corresponding aerodynamic parameter labels for each 3D wing sample, a 3D wing sample dataset is constructed. Each 3D wing sample includes multiple cross-sectional images and aerodynamic parameter labels calculated using the RANS method based on a coupled transition model.

[0059] A3: Construct a first convolutional neural network, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional wing.

[0060] Reference Figure 4 The diagram illustrates the structure of a first convolutional neural network provided in this embodiment. The input object of the input layer of the first convolutional neural network is multiple cross-sectional images of any three-dimensional wing. For example, the cross-sectional images include the contour map of the cross section of the three-dimensional wing in any spanwise direction. After the input layer, a dimensionality reduction layer is also included. The dimensionality reduction layer uses the PCA method to capture the common features between cross-sectional images of different spanwise positions of any three-dimensional wing. Since there is continuity between different cross-sectional images of the same three-dimensional wing, the dimensionality reduction layer can extract common feature data, thereby reducing the amount of data through dimensionality reduction, reducing the data processing cost of the first convolutional neural network, and speeding up the data processing speed.

[0061] Following the dimensionality reduction layer is the convolutional backbone network layer, which consists of three parts: convolutional layers, pooling layers, and fully connected layers. The convolutional layers use kernels of a certain size to sequentially apply to the input data, thereby extracting different features. Since convolution operations generate a large amount of data, pooling layers compress redundant data, reducing computational cost and preventing overfitting. Finally, the fully connected layers integrate the local features from the preceding convolution, pooling, and non-linear activation operations to obtain complete feature information.

[0062] The fully connected layer is followed by an output layer, which outputs the aerodynamic parameters of any three-dimensional wing.

[0063] A4: The first convolutional neural network is iteratively trained based on the three-dimensional wing sample dataset to obtain the wing prediction model.

[0064] Specifically, a first convolutional neural network is trained based on a three-dimensional wing sample dataset. Multiple cross-sectional images of multiple three-dimensional wing samples are input into the first convolutional neural network. The first convolutional neural network is iteratively trained by using the deviation between the aerodynamic parameter prediction results of multiple three-dimensional wing samples output by the first convolutional neural network and the aerodynamic parameter labels of multiple three-dimensional wing samples.

[0065] In practice, the first convolutional neural network can be iteratively trained based on an optimized optimizer such as Adam, ReLU, or Leaky ReLU, or iteratively trained based on a mean squared error loss function with an added L2 regularization term.

[0066] Custom loss functions can also be used to improve the accuracy of model predictions; this implementation does not impose any restrictions.

[0067] During network construction, different hyperparameters of the first convolutional neural network can be adjusted based on the deviation between the aerodynamic parameter prediction results and the aerodynamic parameter labels, as well as the convergence during training. If the prediction results show overfitting, the number of pooling layers can be increased. If the prediction results have poor convergence or poor accuracy, the number of hidden layers in the fully connected neural network or the number of neurons in each hidden layer can be increased.

[0068] When the first convolutional neural network reaches the target state through iterative training, such as when the number of iterations reaches the target number or the value of the error function is less than the calibration value, training can be stopped, and the trained wing prediction model can be obtained.

[0069] In the application of the prediction model, for any three-dimensional wing to be predicted, multiple cross-sectional images of the three-dimensional wing are input into the wing prediction model. The wing prediction model can quickly predict the aerodynamic parameters of the three-dimensional wing.

[0070] S103: When the object to be predicted is a three-dimensional blade, multiple cross-sectional images corresponding to the object to be predicted are input into the blade prediction model, and the blade prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The blade prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional blade samples. The aerodynamic parameter labels of the multiple three-dimensional blade samples are calculated based on the LES method.

[0071] This method is based on the aerodynamic parameter labels of multiple three-dimensional blade samples calculated by the LES method. Then, a blade prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional blade samples. The blade prediction model predicts the aerodynamic parameters of the three-dimensional blade. This eliminates the need to use numerical simulation algorithms to calculate the aerodynamic parameters of the three-dimensional blade each time, thereby reducing the computational cost of prediction and accelerating the efficiency of aerodynamic parameter prediction.

[0072] Reference Figure 5 The diagram illustrates the steps of training the blade prediction model provided in this application embodiment. In one feasible implementation, the training process of the blade prediction model includes: B1: Obtain the geometric contours of multiple three-dimensional blade samples, and calculate the aerodynamic parameter labels corresponding to each three-dimensional blade sample based on the LES method.

[0073] First, multiple 3D blade samples are obtained. Commonly used 3D blades can be selected as 3D blade samples. The geometric contour network of the 3D blade samples is used as input. The velocity distribution around each 3D blade is simulated using open-source CFD simulation software. The aerodynamic parameter labels of each 3D blade sample can be calculated from the velocity distribution.

[0074] In this embodiment, the aerodynamic parameter labels of each three-dimensional blade sample are calculated in open-source CFD simulation software based on the LES method.

[0075] Specifically, the momentum equation for the LES method takes the following form:

[0076] in, , This represents the velocity of a large-scale eddy; the tilde above the physical quantity indicates a filtering operation. and These are the total velocity and the velocity of the small-scale vortex, respectively. For the subgrid stress tensor, additional model equations are also required to solve it.

[0077] B2: Construct a three-dimensional blade sample dataset, which includes multiple three-dimensional blade samples. Each three-dimensional blade sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to that three-dimensional blade.

[0078] After generating corresponding aerodynamic parameter labels for each three-dimensional blade sample, a three-dimensional blade sample dataset is constructed. Each three-dimensional blade sample includes multiple cross-sectional images and aerodynamic parameter labels calculated based on the LES method.

[0079] B3: Construct a second convolutional neural network, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional leaf.

[0080] The network structure of the second convolutional neural network is the same as that of the first convolutional neural network, and will not be described again here. In the second convolutional neural network, a dimensionality reduction layer based on the PCA method is also included after the input layer. The dimensionality reduction layer captures the common features between cross-sectional images of different spanwise positions of any three-dimensional blade. Since there is continuity between different cross-sectional images of the same three-dimensional blade, the dimensionality reduction layer can extract common feature data, thereby reducing the amount of data through dimensionality reduction, which can reduce the data processing cost of the second convolutional neural network and speed up the data processing speed.

[0081] B4: The second convolutional neural network is iteratively trained based on the three-dimensional leaf sample dataset to obtain the leaf prediction model.

[0082] The process of training the second convolutional neural network based on the three-dimensional leaf sample dataset is the same as the process of training the first convolutional neural network in step A4, and will not be repeated here.

[0083] When the second convolutional neural network reaches the target state through iterative training, such as when the number of iterations reaches the target number or the value of the error function is less than the calibration value, training can be stopped, and the trained blade prediction model can be obtained.

[0084] In the application of the prediction model, for any three-dimensional blade to be predicted, multiple cross-sectional images of the three-dimensional blade are input into the blade prediction model. The blade prediction model can quickly predict the aerodynamic parameters of the three-dimensional blade.

[0085] This method first obtains aerodynamic parameter labels for a 3D wing and a low-speed 3D blade using different methods. Specifically, the aerodynamic parameter labels for the 3D wing are obtained using the RANS method with a coupled transition model, while the aerodynamic parameter labels for the 3D blade are obtained using the LES method. Compared to existing aerodynamic parameter calculation methods that all use the RANS method, this method allows for the selection of a more suitable aerodynamic parameter calculation method based on the speed differences between the 3D wing and the 3D blade, thus obtaining more accurate aerodynamic parameter labels for the 3D wing and the low-speed 3D blade.

[0086] Then, by training a wing prediction model with more accurate aerodynamic parameter labels for the 3D wing and multiple cross-sectional images of the 3D wing, the wing prediction model learns the relationship between multiple cross-sectional images of the 3D wing sample and the aerodynamic parameter labels. In subsequent prediction processes, the wing prediction model can quickly and accurately generate aerodynamic parameter prediction results for high-speed 3D wings, eliminating the need to calculate the aerodynamic parameters of each 3D wing sequentially based on numerical simulation methods, thereby accelerating the prediction speed of the aerodynamic parameters of 3D wings.

[0087] By training a blade prediction model using more accurate aerodynamic parameter labels for 3D blades and multiple cross-sectional images of 3D blades, the blade prediction model learns the relationship between multiple cross-sectional images of 3D blade samples and aerodynamic parameter labels. In subsequent prediction processes, the blade prediction model can quickly and accurately generate aerodynamic parameter prediction results for low-speed 3D blades, eliminating the need to calculate the aerodynamic parameters of each 3D blade sequentially based on numerical simulation methods, thereby accelerating the prediction speed of aerodynamic parameters for 3D blades.

[0088] In one feasible implementation, after outputting the predicted aerodynamic parameters of the object to be predicted through the wing prediction model, the method further includes: C1: In response to the target operation, acquire the target aerodynamic parameters corresponding to the object to be predicted, wherein the target aerodynamic parameters include at least one of the following parameter types: target drag coefficient, target pressure coefficient, and maximum lift-to-drag ratio.

[0089] Once the geometry of any three-dimensional wing or blade is determined, the corresponding aerodynamic parameters are predicted using the wing prediction model and the blade prediction model. Then, the target aerodynamic parameters are set as constraints, allowing the geometry of the three-dimensional wing or blade to be adjusted so that the aerodynamic parameters of the adjusted three-dimensional wing or blade approach or equal the target aerodynamic parameters.

[0090] C2: Based on the target aerodynamic parameters corresponding to the object to be predicted and the aerodynamic parameter prediction results, the geometric parameters of the object to be predicted are iteratively updated until the object to be predicted after iterative updates satisfies the target aerodynamic parameters.

[0091] Specifically, in each optimization iteration, the target difference is calculated based on the target aerodynamic parameters corresponding to the object to be predicted and the predicted aerodynamic parameters.

[0092] The formula for calculating the target difference is:

[0093] Where MSE is the target difference; n is the number of parameter types in the target aerodynamic parameters; The i-th aerodynamic parameter in the predicted aerodynamic parameter results; Let i be the i-th aerodynamic parameter among the target aerodynamic parameters.

[0094] When the target difference is greater than the preset target threshold, the geometric parameters of the object to be predicted are adjusted, and the object to be predicted after the geometric parameters are adjusted is taken as the object to be predicted. Based on the wing prediction model or the blade prediction model, the aerodynamic parameter prediction result corresponding to the current object to be predicted is calculated. When the target difference is less than the target threshold, the iteration stops, and the current prediction results of the geometric and aerodynamic parameters of the object to be predicted are output.

[0095] During the iterative update of the geometric parameters of the object to be predicted, existing methods such as the common gradient method, genetic algorithm, and particle swarm optimization algorithm can be used, and this embodiment does not impose any restrictions.

[0096] Taking the genetic algorithm as an example, the optimization process of the genetic algorithm is as follows: (1) Initialize and generate the first generation population; (2) Calculate the fitness of each individual in the population; (3) Determine whether the fitness of the current population meets the requirements. If it does, output the optimization result. If it does not meet the requirements, continue to the subsequent steps; (4) Select the best individual; (5) Generate the next generation population through crossover and mutation. Keep looping until the fitness meets the requirements.

[0097] Reference Figure 6 The illustration shows a schematic diagram of the genetic algorithm optimization process provided in the embodiment of this application. For example, in this embodiment, multiple three-dimensional wings can be used as the first generation population. The aerodynamic parameters of each three-dimensional wing are predicted by the wing prediction model. Then, the fitness is based on the target difference between the predicted aerodynamic parameters of each three-dimensional wing and the target aerodynamic parameters. If the fitness of any three-dimensional wing, i.e. the target difference, is less than or equal to the target threshold, then the three-dimensional wing with the smallest fitness is taken as the optimized three-dimensional wing.

[0098] If the fitness of any three-dimensional wing, i.e. the target difference, is greater than the target threshold, multiple three-dimensional wings can be cross-crossed and mutated to generate the next generation population. The aerodynamic parameter prediction results of each three-dimensional wing in the current population can be predicted by the wing prediction model. This process continues until there is a three-dimensional wing whose fitness, i.e. the target difference, is less than or equal to the target threshold. Then, the three-dimensional wing with the lowest fitness is taken as the optimized three-dimensional wing.

[0099] The method provided in this embodiment has at least the following beneficial effects: 1. For high-speed 3D airfoils, the RANS method using coupled transition models can obtain more accurate aerodynamic parameter labels for 3D airfoils, and the trained airfoil prediction model can generate more accurate aerodynamic parameter prediction results for 3D airfoils; for low-speed 3D blades, the LES method can obtain more accurate aerodynamic parameter labels for 3D airfoils, and the trained blade prediction model can generate more accurate aerodynamic parameter prediction results for 3D blades. 2. Based on the wing prediction model, the aerodynamic parameters of the three-dimensional wing are predicted, and based on the blade prediction model, the aerodynamic parameters of the three-dimensional blade are predicted. Compared with calculating the aerodynamic parameters of the three-dimensional wing or three-dimensional blade one by one using numerical simulation, the computational cost can be reduced and the speed of accurately predicting the aerodynamic parameters of the wing or blade can be improved. 3. Integrate the PCA method into the framework of convolutional neural networks, and use the PCA method to reduce the dimensionality of the input data so that the trained wing prediction model and blade prediction model can be applied to three-dimensional problems; 4. Combining wing prediction models and blade prediction models with optimization algorithms can improve the speed of optimizing the geometry of three-dimensional wings and three-dimensional blades.

[0100] This application also provides a computer device, including: at least one processor and a memory, the memory storing a computer program executable on the processor, wherein the processor executes the computer program to perform the aerodynamic parameter prediction method described in the embodiment.

[0101] Reference Figure 7 The diagram illustrates a non-volatile readable storage medium 700 provided in an embodiment of this application. The non-volatile readable storage medium 700 stores a computer program 701, wherein the computer program 701, when executed by a processor, performs the aerodynamic parameter prediction method described in the embodiment.

[0102] Reference Figure 8 The diagram illustrates a computer program product 800 provided in an embodiment of this application. The computer program product 800 includes a computer program / instruction 801, which, when executed by a processor, implements the aerodynamic parameter prediction method described in the embodiment.

[0103] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0104] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0105] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0106] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0108] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0109] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0110] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting aerodynamic parameters, characterized in that, The method includes: Obtain the object to be predicted, which includes a three-dimensional wing and a three-dimensional blade; When the object to be predicted is a three-dimensional air wing, multiple cross-sectional images corresponding to the object to be predicted are input into the air wing prediction model, and the air wing prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The air wing prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional air wing samples, and the aerodynamic parameter labels of the multiple three-dimensional air wing samples are calculated based on the RANS method of the coupled transition model. When the object to be predicted is a three-dimensional blade, multiple cross-sectional images corresponding to the object to be predicted are input into the blade prediction model, and the blade prediction model outputs the aerodynamic parameter prediction results corresponding to the object to be predicted. The blade prediction model is trained based on multiple cross-sectional images and aerodynamic parameter labels of multiple three-dimensional blade samples. The aerodynamic parameter labels of multiple three-dimensional blade samples are calculated based on the LES method. The method further includes: The geometric contours of multiple three-dimensional wing samples are obtained, and the aerodynamic parameter labels corresponding to each three-dimensional wing sample are calculated by the RANS method based on the coupled transition model. The process involves acquiring the geometric contours of multiple three-dimensional wing samples, and using the RANS method based on a coupled transition model to calculate the aerodynamic parameter labels for each three-dimensional wing sample, including: For any three-dimensional wing sample, obtain the geometric contour corresponding to the three-dimensional wing sample, and input the SA model of the coupled transition model, wherein the SA model is a model based on the RANS method; The aerodynamic parameter labels corresponding to the three-dimensional wing sample are calculated using the SA model of the coupling transition model. The calculation formula for the SA model of the coupled transition model is as follows: in, The equations are obtained by solving the convection transport equations. The coefficient of dynamic viscosity, For density, Indicates kinematic viscosity; Geometric coordinates; and These are the coupled generation and dissipation terms, respectively; For additional generated items; These are empirical parameters; Average speed; The method further includes: Obtain the geometric contours of multiple three-dimensional blade samples, and calculate the aerodynamic parameter labels corresponding to each three-dimensional blade sample based on the LES method. Construct a three-dimensional blade sample dataset, which includes multiple three-dimensional blade samples. Each three-dimensional blade sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to the three-dimensional blade. A second convolutional neural network is constructed, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional leaf. The second convolutional neural network is iteratively trained based on the three-dimensional leaf sample dataset to obtain a leaf prediction model. The plurality of cross-sectional images are cross-sectional images at different spanning positions.

2. The method according to claim 1, characterized in that, The method further includes: Construct a three-dimensional wing sample dataset, which includes multiple three-dimensional wing samples. Each three-dimensional wing sample includes multiple cross-sectional images and aerodynamic parameter labels corresponding to that three-dimensional wing. A first convolutional neural network is constructed, which includes an input layer, a dimensionality reduction layer, a convolutional backbone network layer, and an output layer. The dimensionality reduction layer is used to extract common features from multiple cross-sectional images corresponding to any three-dimensional wing. The first convolutional neural network is iteratively trained based on the three-dimensional wing sample dataset to obtain the wing prediction model.

3. The method according to claim 1, characterized in that, After outputting the predicted aerodynamic parameters of the object to be predicted through the wing prediction model, the method further includes: In response to the target operation, the target aerodynamic parameters corresponding to the object to be predicted are obtained, and the target aerodynamic parameters include at least one of the following parameter types: target drag coefficient, target pressure coefficient and maximum lift-to-drag ratio; Based on the target aerodynamic parameters corresponding to the object to be predicted and the aerodynamic parameter prediction results, the geometric parameters of the object to be predicted are iteratively updated until the iteratively updated object to be predicted satisfies the target aerodynamic parameters, including: Calculate the target difference based on the target aerodynamic parameters corresponding to the object to be predicted and the prediction results of the aerodynamic parameters; When the target difference is greater than the preset target threshold, the geometric parameters of the object to be predicted are adjusted, and the object to be predicted after the geometric parameters are adjusted is taken as the object to be predicted. Based on the wing prediction model or the blade prediction model, the aerodynamic parameter prediction result corresponding to the current object to be predicted is calculated. When the target difference is less than the target threshold, the iteration stops, and the current prediction results of the geometric and aerodynamic parameters of the object to be predicted are output.

4. The method according to claim 2, characterized in that, The formula for calculating the target difference is: Where MSE is the target difference; n is the number of parameter types in the target aerodynamic parameters; The i-th aerodynamic parameter in the predicted aerodynamic parameter results; Let i be the i-th aerodynamic parameter among the target aerodynamic parameters.

5. A computer device, characterized in that, include: At least one processor and a memory storing a computer program executable on the processor, wherein the processor executes the computer program to perform the aerodynamic parameter prediction method according to any one of claims 1-4.

6. A non-volatile readable storage medium, characterized in that, The non-volatile readable storage medium stores a computer program, wherein the computer program, when executed by a processor, performs the aerodynamic parameter prediction method according to any one of claims 1-4.

7. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the aerodynamic parameter prediction method according to any one of claims 1-4.

Images