CN118395886A - Flow field prediction method based on token selection transducer - Google Patents

Abstract

The invention discloses a flow field prediction method based on token selection transform, which comprises the following steps: acquiring an airfoil shape dataset and a real flow field dataset; extracting airfoil geometric parameters through a token selection transducer network; constructing a parameter fusion network to fuse physical information characteristics; training a flow field prediction network based on a multi-layer perceptron; and predicting the pressure field and the speed field on different airfoil data by using a trained flow field prediction model. The invention solves the problems of inaccurate focus and coarser feature extraction granularity when extracting the geometrical parameters of the wing profile by introducing a self-attention mechanism and a token selection mechanism; by introducing a physical information fusion network, the expression capacity and generalization capacity of the model are enhanced; the method has good prediction precision and prediction efficiency under different wing section and flow field conditions, and can provide rapid guidance for the tasks of aerodynamic shape design, aerodynamic performance analysis and the like of the aircraft.

Description

A flow field prediction method based on token selection Transformer

Technical Field

The present invention belongs to the fields of computational fluid dynamics and artificial intelligence technology, and specifically relates to a flow field prediction method based on token selection Transformer.

Background technique

Flow field prediction is the process of predicting and simulating the motion state of a fluid. It has a wide range of applications in chemical engineering, climate, and aerodynamics. Taking aerodynamics as an example, obtaining flow field data near an aircraft airfoil is a crucial step in the design and optimization of an aircraft. Initially, the main means of obtaining airfoil flow field data was through wind tunnel testing. Although this method has relatively accurate results, the experimental design requires a lot of prior knowledge, and the experimental cycle is long. The resource overhead required to build wind tunnel equipment is large, so it is mostly used in the later stages of aircraft design. With the rise of high-performance computing and numerical simulation methods, computational fluid dynamics (CFD) methods have gradually become the main means of simulating flow fields near airfoils. CFD methods mainly discretize continuous fluid equations, such as the Navier-Stokes equations, and numerically approximate these equations to obtain flow field data. However, CFD solutions under complex conditions require a large number of iterative processes, and the requirements for CPU and memory are also high.

In recent years, with the widespread application of artificial intelligence and neural networks, methods based on deep learning and data-driven have become a new means of obtaining flow field data. This method only requires a certain amount of time to train the network model in the early stage, and then the trained model can be used to complete the flow field prediction of different airfoils within a few seconds. However, the existing deep learning flow field prediction methods also have some limitations: many CNN-based flow field prediction methods project the flow field data into a uniformly distributed Cartesian grid. This pixelation method will cause a large amount of flow field details to be missing, especially near the airfoil surface; and the existing airfoil geometric parameter acquisition method cannot accurately capture the most effective geometric features of the airfoil. Therefore, how to more accurately extract airfoil features and predict flow field details is the direction that the flow field prediction method based on deep learning needs to be optimized.

Summary of the invention

In view of the shortcomings of the prior art, the present invention proposes a flow field prediction method based on token selection Transformer, which can extract more accurate and effective airfoil features and improve the flow field prediction accuracy, especially near the airfoil surface.

In order to solve the above technical problems, the present invention is implemented in the following ways:

A flow field prediction method based on token selection Transformer includes the following steps:

S1, obtaining an airfoil shape data set and a real flow field data set;

S2, extracting airfoil geometric parameters through token selection Transformer network;

S3, construct parameter fusion network to fuse physical information features;

S4, training a flow field prediction network based on a multi-layer perceptron;

S5. Use the trained flow field prediction model to predict the pressure field and velocity field on different airfoil data.

Furthermore, the step S1 specifically includes the following steps:

S11, selecting a reference airfoil from the UIUC airfoil database, performing airfoil fitting by using a non-uniform rational B-spline interpolation method, and selecting a control point on the airfoil fitting curve as a new x-coordinate and y-coordinate of the current airfoil;

S12, performing maximum and minimum normalization processing on all new airfoil coordinates in step S11, and the expression is as follows:

Among them, x _i , y _i represent the coordinates of the current airfoil, x _min , y _min are the global minimum values, x _max , y _max are the global maximum values, and x _j , y _i are the normalized coordinates;

S13, generating an airfoil shape grayscale image according to the normalized coordinates, wherein the pixel value on the airfoil geometry curve is 1, the pixel value not on the airfoil geometry curve is 0, and the pixel values at other positions are in the range [0,1], thereby obtaining an airfoil shape data set;

S14, dividing the computational grid, performing CFD solution on the airfoil samples in the airfoil data set, obtaining real data of different airfoil samples under different flow conditions, and forming a real flow field data set.

Furthermore, the step S2 specifically comprises the following steps:

S21, serialize the airfoil images in the airfoil shape data set obtained in step S1, divide and flatten each airfoil image pic _i into blocks, and convert it into a series of two-dimensional tokens pic _i , where the expressions of pic _i and _pi are as follows:

in, represents a real number set, H, W and C represent the width, height and number of channels of the airfoil image respectively, P represents the width and height of the selected partition block, and N = H*W/P ² represents the number of two-dimensional tokens divided;

S22. Add a learnable embedding CLStoken to the acquired token sequence, with the dimension consistent with other token dimensions, and then obtain the complete token sequence representation I of each airfoil image, the specific expression of which is as follows:

Wherein, p _cls represents CLStoken, p ₀ , p ₁ , ... p _n-1 represent the N tokens obtained in the above, and D = P ² *C represents the dimension of the token;

S23, reduce the dimension of I through a multi-layer perceptron network, so that the dimension of each token is reduced from D to I' is obtained, and its specific expression is as follows:

S24, p′ _cls in step S23 is used as the global feature, p′ ₀ , p′ ₁ , .... p′ _n-1 are used as local features, and the global feature p′ _cls is concatenated with each local feature p′ _i to obtain the concatenated feature Its expression is as follows:

S25. Record the N+1 splicing features obtained in step S24 as Then input it into another multi-layer perceptron network followed by a normalization layer to calculate the importance score vector S of the feature, which is expressed as follows:

Among them, MLP() represents the function of the multi-layer perceptron network, and Softmax() represents the normalized exponential function, which converts the output of MLP into a probability value between [0,1];

S26. Use the importance score vector S to filter out the token index M with the top K importance among all N+1 tokens. The expression is as follows:

M∈{0,1} ^(N+1)*K

Each element in M is a one-hot indicator, and its value is selected from the set {0, 1};

S27, use the transpose of the matrix M in step S26 and I to perform matrix multiplication to obtain the most important K airfoil parameter features of the given airfoil

Furthermore, the step S3 specifically comprises the following steps:

S31. For each grid point coordinate, calculate the shortest distance d to the airfoil geometry curve, which is expressed as follows:

Where (x, y) represents the X-axis coordinate and Y-axis coordinate of the grid point. Indicates the X-axis and Y-axis coordinates of a point on the airfoil geometry curve;

S32. Construct a parameter fusion network, input the horizontal and vertical coordinates of each grid point, the shortest distance from the coordinate to the airfoil geometric boundary, and the normalized Reynolds coefficient and angle of attack into a multi-layer perceptron network for feature fusion and dimension upgrading, and finally obtain the physical information feature p _phy , which is expressed as follows:

in, and They represent the normalized Reynolds coefficient and angle of attack respectively, and D represents the dimension of the fused physical information feature, which is consistent with the dimension of each airfoil feature extracted in step S2.

Furthermore, the step S4 specifically comprises the following steps:

S41. Rank the top K airfoil parameter features by importance The physical information feature p _phy is used as the input of the flow field prediction network based on the multi-layer perceptron. The output of the flow field prediction network is the flow field data corresponding to the grid point coordinates, and its expression is as follows:

Among them, f _predict () represents the function of the flow field prediction network, and They represent the predicted values of the velocity component in the x direction, the velocity component in the y direction, and the pressure, respectively;

S42. For the flow field prediction network, the loss function Loss _MLP is designed as follows:

Where N represents the total number of samples, _ui , _vi , _pi represent the true values of the x-direction velocity component, the y-direction velocity component and the pressure;

S43. Train the model on the GPU server, set different model parameters and iteration steps for multiple rounds of training, and select the model with the smallest prediction error on the test data as the final flow field prediction model.

Furthermore, the specific method of step S5 is as follows:

Using the trained flow field prediction model, different airfoil characteristics, Reynolds number, and angle of attack information are input, and the pressure field and velocity field near the airfoil under the corresponding airfoil and flow conditions are output; the flow field data is visualized using processing software to facilitate subsequent work such as airfoil optimization, aerodynamic performance analysis, and aerodynamic noise prediction.

Compared with the prior art, the present invention has the following beneficial effects:

With the help of Transformer's self-attention mechanism and through the token selection module, the airfoil features can be screened more accurately, while parameter redundancy can be avoided to a certain extent; normalized coordinate values are extracted for different airfoil shapes and standardized images are generated to facilitate subsequent neural network processing; the flow field grid position information and the flow field parameter physical information are feature fused using a neural network and aligned with the airfoil parameter features, which can enhance the expression and generalization capabilities of the model to a certain extent; compared with the existing deep learning flow field prediction methods, the present invention improves the accuracy of flow field prediction, especially near the surface of the airfoil structure.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic flow chart of the flow field prediction method of the present invention.

Detailed ways

The specific implementation of the present invention is further described in detail below with reference to the accompanying drawings and specific examples.

As shown in FIG1 , a flow field prediction method based on token selection Transformer includes the following steps:

S1, obtaining an airfoil shape data set and a real flow field data set;

S2, extracting airfoil geometric parameters through token selection Transformer network;

S3, construct parameter fusion network to fuse physical information features;

S4, training a flow field prediction network based on a multi-layer perceptron;

S5. Use the trained flow field prediction model to predict the pressure field and velocity field on different airfoil data.

Furthermore, the step S1 specifically includes the following steps:

S11, selecting a reference airfoil from the UIUC airfoil database, performing airfoil fitting by using a non-uniform rational B-spline interpolation method, and selecting a control point on the airfoil fitting curve as a new x-coordinate and y-coordinate of the current airfoil;

S12, performing maximum and minimum normalization processing on all new airfoil coordinates in step S11, and the expression is as follows:

Among them, x _i , y _i represent the coordinates of the current airfoil, x _min , y _min are the global minimum values, x _max , y _max are the global maximum values, and x _j , y _j are the normalized coordinates;

S13, generating an airfoil shape grayscale image according to the normalized coordinates, wherein the pixel value on the smooth curve (airfoil geometry curve) formed by the airfoil coordinate points is 1, the pixel value not on the airfoil geometry curve is 0, and the pixel values at other positions are in the range [0,1], thereby obtaining an airfoil shape data set;

S14, dividing the calculation grid, performing CFD solution on the airfoil samples in the airfoil data set, obtaining the real flow field data of different airfoil samples under different flow conditions, and forming a real flow field data set.

Furthermore, the step S2 specifically comprises the following steps:

S21, serialize the airfoil images in the airfoil shape data set obtained in step S1, divide and flatten each airfoil image pic _i into blocks, and convert it into a series of two-dimensional tokens pic _i , where the expressions of pic _i and _pi are as follows:

in, represents a real number set, H, W and C represent the width, height and number of channels of the airfoil image respectively, P represents the width and height of the selected partition block, and N = H*W/P ² represents the number of two-dimensional tokens divided;

S22, adding a learnable embedding CLStoken to the token sequence obtained in step S21, with the dimension consistent with other token dimensions, and then obtaining a complete token sequence representation I of each airfoil image, the specific expression of which is as follows:

Wherein, p _cls represents CLStoken, p ₀ , p ₁ , ... p _n-1 represent the N tokens obtained in the above, and D = P ² *C represents the dimension of the token;

S23, reduce the dimension of I through a multi-layer perceptron network, so that the dimension of each token is reduced from D to I' is obtained, and its specific expression is as follows:

Among them, p′ _cls , p′ ₀ , p′ ₁ , ..., p′ _n-1 are the representations of N+1 tokens including CLStoken after dimensionality reduction;

S24, p′ _cls in step S23 is used as the global feature, p′ ₀ , p′ ₁ , .... p′ _n-1 are used as local features, and the global feature p′ _cls is concatenated with each local feature p′ _i to obtain the concatenated feature Its expression is as follows:

Among them, i is the subscript of each local feature, and its value range is [0, N-1]. The global feature p′ _cls is concatenated with itself to obtain N+1 concatenated features in total;

S25. Record the N+1 splicing features obtained in step S24 as Then input it into another multi-layer perceptron network followed by a normalization layer to calculate the importance score vector S of the feature, which is expressed as follows:

Among them, MLP() represents the function of the multi-layer perceptron network, and Softmax() represents the normalized exponential function, which converts the output of MLP into a probability value between [0,1];

S26. Use the importance score vector S to filter out the token index M with the top K importance among all N+1 tokens. The expression is as follows:

M∈{0,1} ^(N+1)*K

Each element in M is a one-hot indicator, and its value is selected from the set {0, 1};

S27, use the transpose of the matrix M in step S26 and the token representation I obtained in step S22 to perform matrix multiplication to obtain the most important K airfoil parameter features of the given airfoil

Furthermore, the step S3 specifically comprises the following steps:

S31. For each grid point coordinate, calculate the shortest distance d to the airfoil geometry curve, which is expressed as follows:

Where (x, y) represents the X-axis coordinate and Y-axis coordinate of the grid point. Indicates the X-axis and Y-axis coordinates of a point on the airfoil geometry curve;

S32. Construct a parameter fusion network, input the horizontal and vertical coordinates of each grid point, the shortest distance from the coordinate to the airfoil geometric boundary, and the normalized Reynolds coefficient and angle of attack into a multi-layer perceptron network for feature fusion and dimension upgrading, and finally obtain the physical information feature p _phy , which is expressed as follows:

in, and They represent the normalized Reynolds coefficient and angle of attack respectively, and D represents the dimension of the fused physical information feature, which is consistent with the dimension of each airfoil feature extracted in step S2.

Furthermore, the step S4 specifically comprises the following steps:

S41. Rank the top K airfoil parameter features by importance The physical information feature p _phy is used as the input of the flow field prediction network based on the multi-layer perceptron. The output of the flow field prediction network is the flow field data corresponding to the grid point coordinates, and its expression is as follows:

Among them, f _predict () represents the function of the flow field prediction network, and They represent the predicted values of the velocity component in the x direction, the velocity component in the y direction, and the pressure, respectively;

S42. For the flow field prediction network, the loss function Loss _MLP is designed as follows:

Where N represents the total number of samples, _ui , _vi , _pi represent the true values of the x-direction velocity component, the y-direction velocity component and the pressure;

S43. Train the model on the GPU server, set different model parameters and iteration steps for multiple rounds of training, and select the model with the smallest prediction error on the test data as the final flow field prediction model.

Furthermore, the specific method of step S5 is as follows:

Using the trained flow field prediction model, different airfoil characteristics, Reynolds number, and angle of attack information are input, and the pressure field and velocity field near the airfoil under the corresponding airfoil and flow conditions are output; the flow field data is visualized using processing software to facilitate subsequent work such as airfoil optimization, aerodynamic performance analysis, and aerodynamic noise prediction.

The above description is only an implementation mode of the present invention. It is stated again that for ordinary technicians in this technical field, several improvements can be made to the present invention without departing from the principle of the present invention. These improvements are also included in the protection scope of the claims of the present invention.

Claims (6)

1. A flow field prediction method for selecting a transducer based on a token is characterized by comprising the following steps of: the method comprises the following steps:

S1, acquiring an airfoil shape dataset and a real flow field dataset;

S2, extracting geometrical parameters of the airfoil through a token selection transducer network;

s3, constructing a parameter fusion network fusion physical information characteristic;

s4, training a flow field prediction network based on a multi-layer perceptron;

S5, predicting a pressure field and a speed field on different wing profile data by using the trained flow field prediction model.

2. The token-based transform flow field prediction method of claim 1, wherein:

The step S1 specifically comprises the following steps:

s11, selecting a reference airfoil from a UIUC airfoil database, performing airfoil fitting by a non-uniform rational B-spline interpolation method, and selecting control points on an airfoil fitting curve as new x coordinates and y coordinates of the current airfoil;

S12, carrying out maximum and minimum normalization processing on all new airfoil coordinates in the step S11, wherein the expression is as follows:

Wherein x _i、y_i represents the coordinates of the current airfoil, x _min、y_min is the global minimum, x _max、y_max is the global maximum, and x _j、y_j is the normalized coordinates;

s13, generating an airfoil shape gray scale map according to the normalized coordinates, wherein the pixel point value on the airfoil geometric curve is 1, the pixel value not on the airfoil geometric curve is 0, and the pixel values at other positions are in the range of [0,1], so as to obtain an airfoil shape data set;

S14, dividing a calculation grid, and carrying out CFD solving on airfoil samples in the airfoil data set to obtain real data of different airfoil samples under different flow conditions, so as to form a real flow field data set.

3. The token-based transform flow field prediction method of claim 1, wherein:

the step S2 specifically includes the following steps:

s21, serializing the wing-shaped images in the obtained wing-shaped shape data set, dividing and flattening each wing-shaped image pic _i in a block mode, and converting the wing-shaped images into a series of two-dimensional tokens p _i,pic_i and p _i as follows:

Wherein, Representing a real number set, H, W and C respectively representing the width, the height and the channel number of an original image, P representing the width and the height of a selected dividing block, and N=H×W/P ² representing the number of divided two-dimensional tokens;

S22, adding a leachable embedding CLStoken in the acquired token sequence, keeping the dimension consistent with other token dimensions, and further obtaining a complete token sequence representation I of each airfoil image, wherein the specific expression is as follows:

wherein P _cls represents CLS token, P ₀,p₁,...p_n-1 represents N tokens obtained in the above, and d=p ² ×c represents the token dimension;

S23, reducing the dimension of each token from D to I through a multi-layer perceptron network Obtaining I', wherein the specific expression is as follows:

S24, p '_cls in the step S23 is used as a global feature, p' ₀,p′₁,....p′_n-1 is used as a local feature, and the global feature p '_cls and each local feature p' _i are spliced to obtain a spliced feature The expression is as follows:

S25, marking the N+1 splicing characteristics obtained in the step S24 as And then inputting the importance score vector S to a multi-layer perceptron network of another subsequent normalization layer, and calculating the importance score vector S of the feature, wherein the expression is as follows:

Wherein, MLP () represents the function of the multi-layer perceptron network, softmax () represents the normalized exponential function, converting the output of MLP into a probability value between [0,1 ];

S26, screening out a token index M with the importance arranged in the front K bits in all N+1 tokens by using an importance score vector S, wherein the expression is as follows:

M∈{0，1}^(N+1)*K

wherein each element in M is a one-hot indicator, and the value is selected from the set {0,1 };

S27, performing matrix multiplication on the transpose of the matrix M in the step S26 and the I to obtain the most important K airfoil parameter characteristics of the given airfoil

4. The token-based transform flow field prediction method of claim 1, wherein:

The step S3 specifically includes the following steps:

s31, calculating the shortest distance d between each grid point coordinate and the wing profile geometric curve, wherein the shortest distance d is expressed as follows:

Where (X, Y) denotes the X-axis coordinates and Y-axis coordinates of the grid points, X-axis coordinates and Y-axis coordinates representing points on the airfoil geometry;

S32, constructing a parameter fusion network, inputting the shortest distance from the abscissa and ordinate of each grid point to the geometrical boundary of the wing profile, the normalized Reynolds coefficient and attack angle into a multi-layer perceptron network for feature fusion and dimension ascending, and finally obtaining a physical information feature p _phy, wherein the expression is as follows:

Wherein, AndAnd D represents the feature dimension of the physical information obtained by fusion and is consistent with each airfoil feature dimension extracted in the step S2.

5. The token-based transform flow field prediction method of claim 1, wherein:

the step S4 specifically includes the following steps:

S41, ranking the importance K airfoil parameter features The physical information characteristic p _phy is used as the input of a flow field prediction network based on a multi-layer perceptron, the output of the flow field prediction network is flow field data corresponding to grid point coordinates, and the expression is as follows:

Where f _predict () represents a function of the flow field prediction network, AndPredictive values respectively representing an x-direction velocity component, a y-direction velocity component, and a pressure;

S42, for a flow field prediction network, designing a Loss function Loss _MLP expression as follows:

Where N represents the total number of samples, u _i,v_i,p_i represents the x-direction velocity component, the y-direction velocity component and the true value of pressure;

S43, training a model on the GPU server, setting different model parameters and iteration steps to perform multi-round training, and selecting a model with the minimum prediction error on test data as a final flow field prediction model.

6. The token-based transform flow field prediction method of claim 1, wherein:

the specific method of the step S5 is as follows:

Using the trained flow field prediction model, inputting different airfoil characteristics, reynolds number and attack angle information, and outputting a pressure field and a velocity field near the airfoil under the corresponding airfoil and flow conditions; and using processing software to perform flow field data visualization processing.