Joint prediction method and system for drag coefficient and surface pressure field of automobile

Abstract

The invention provides a joint prediction method and system for a wind resistance coefficient and a surface pressure field of an automobile, and relates to the technical field of automobile aerodynamics. According to the joint prediction method, the pressure field and the wind resistance coefficient are predicted in the same network in parallel, and the wind resistance coefficient obtained by pressure field integration and derived from the pressure field and the wind resistance coefficient obtained by network direct regression are calibrated by physical consistency constraint. The model has physical confidence and interpretability while ensuring numerical precision, so that a key stress area can be positioned and serves an automobile modeling iteration process.

Description

Technical Field

[0001] This invention relates to the field of automotive aerodynamics, specifically to a method and system for jointly predicting automotive drag coefficient and surface pressure field. Background Technology

[0002] The aerodynamic performance of a vehicle's overall shape directly affects its energy consumption and stability. In engineering, this includes considerations such as the vehicle's drag coefficient (…). To evaluate the merits of a design, a surface pressure field is needed to determine the contribution of the "windward zone" to drag, which can then be used for design optimization and local modifications.

[0003] Existing methods for predicting automotive drag coefficients and surface pressure fields mainly fall into two categories. The first category is the traditional CFD / RANS-based method for predicting drag and pressure fields. This method typically uses CAD and STL geometry as input, performs mesh generation, sets up turbulence models, configures boundary conditions and numerical parameters, solves the RANS using the finite volume method, obtains the surface pressure and shear stress fields, and then integrates to obtain the drag coefficient. The process chain is as follows: geometry → mesh → governing equations and turbulence closed loop → numerical solution → field variables → aerodynamic post-processing. The advantages of this approach are strong physical interpretation and the ability to output a complete pressure field; the disadvantages are long computation cycles and high costs in complex shapes and multi-scheme scenarios, sensitivity to mesh and model selection, and difficulty in supporting high-frequency iteration and selection in the early stages of automotive styling. The second category involves existing, common methods that use deep learning-based surrogate models to predict drag and pressure fields. These methods directly predict the drag coefficient and pressure field path of a car from its point cloud using deep learning models. Examples include direct point cloud regression, representing the shape with 3D point clouds, using networks like PointNet / PointNet++ for feature encoding, and regressing against other aerodynamic parameters using a multilayer perceptron. While these methods can support high-frequency iteration and selection in the early stages of car styling, they are not universally applicable.

[0004] However, existing methods for prediction using surrogate models mostly focus on single-indicator or single-field prediction, lacking collaborative modeling between objectives within the same framework, making it difficult to simultaneously provide predictions for the entire vehicle. The mechanism for predicting and maintaining physical consistency between surface pressure fields and the prediction is difficult to implement, even when predicting multiple targets simultaneously. This can easily lead to situations where the numerical values ​​are accurate individually but inconsistent with each other, thus reducing the usability of the project. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a joint prediction method and system for vehicle drag coefficient and surface pressure field, solving the technical problem that existing prediction methods using surrogate models cannot guarantee the physical consistency between the vehicle drag coefficient and surface pressure field.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for jointly predicting the drag coefficient and surface pressure field of an automobile, comprising: A training dataset including vehicle point cloud data and label data is obtained, and the vehicle point cloud data in the training dataset is preprocessed to obtain a sampling point matrix and shape parameter vector; then the sampling point matrix is ​​uniformly sampled to obtain a uniform point matrix. The initial joint prediction model is trained by sampling point matrix, shape parameter vector and uniform point matrix, and the loss value between the output of the joint prediction model and the label data is optimized to obtain the joint prediction model. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

[0007] Preferably, the vehicle drag prediction subnetwork includes a PointNet++ module, an MLP module, and a feature fusion module. The PointNet++ module is used to process the uniform point matrix to obtain point cloud structural features, the MLP module is used to encode the shape parameters to obtain shape features, and the feature fusion module is used to align and fuse the point cloud structural features and the shape features.

[0008] Preferably, the process of aligning and fusing the point cloud structural features and shape features by the feature fusion module includes: The linear mapping of the feature fusion module is Then cross attention is The formula for the cross-attention feature fusion module is: in, p =0.1 indicates the dropout rate of Dropout; , representing the structural features of the point cloud; , represents the shape feature; C represents the number of feature channels, B Indicates the batch size.

[0009] Preferably, the vehicle surface pressure field prediction subnetwork includes a first normalization layer, a physical attention layer, a second normalization layer, and a feedforward layer. After the first normalization layer processes the sampling point matrix, the output data passes through the physical attention layer. The output of the physical attention layer and the sampling point matrix are concatenated with residuals to obtain residual features. The residual features are processed by the second normalization layer and then input to the feedforward layer. The output of the feedforward layer and the residual features are concatenated with residuals to obtain pressure field features.

[0010] Preferably, the physical attention layer includes a first linear layer, a multi-head attention module, and a second linear layer; The multi-head attention module includes a slice layer, a label-specific attention layer, and a reverse slice layer.

[0011] Preferably, the step of integrating the predicted surface pressure field to obtain the drag coefficient derived from the pressure field includes: Given the unit vector of the incoming flow direction air density ,speed Reference windward area A Dynamic pressure ; Pressure subnet outputs surface pressure field Then, discrete integration yields the drag coefficient derived from the pressure field. : in, Indicates the correction factor; This represents the total number of points on the surface of the car's point cloud; Indicates the first i The external normal vector of each point; Indicates the first i The area of ​​each point.

[0012] Preferably, the formula for calculating the loss value between the output of the joint prediction model and the label data is as follows: The first item is wind resistance monitoring, the second item is pressure monitoring, and the third item is physical consistency constraint. , , These are the weighting coefficients; The drag coefficient derived from the pressure field is obtained by integrating the pressure field. The drag coefficient predicted by the joint forecasting model; Represents the true value of the drag coefficient; The joint prediction model predicts the first... i Pressure field on the surface of a car at a point; Indicates the first i The true value of the pressure field on the car surface at a point.

[0013] Secondly, the present invention provides a joint prediction system for vehicle drag coefficient and surface pressure field, comprising: The data acquisition and preprocessing module is used to acquire a training dataset including vehicle point cloud data and label data, preprocess the vehicle point cloud data in the training dataset to obtain a sampling point matrix and shape parameter vector; and perform uniform sampling on the sampling point matrix to obtain a uniform point matrix. The model training module is used to train the initial joint prediction model using the sampling point matrix, shape parameter vector, and uniform point matrix, optimize the loss value between the model output and the label data, and obtain the joint prediction model. The joint prediction model is used to process the sampling point matrix, shape parameter vector, and uniform point matrix of the vehicle to be evaluated to obtain the vehicle's drag coefficient and surface pressure field. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

[0014] Thirdly, the present invention provides a computer-readable storage medium storing a computer program for the joint prediction of vehicle drag coefficient and surface pressure field, wherein the computer program causes a computer to execute the joint prediction method of vehicle drag coefficient and surface pressure field as described above.

[0015] Fourthly, the present invention provides an electronic device, comprising: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing a joint prediction method for vehicle drag coefficient and surface pressure field as described above.

[0016] (III) Beneficial Effects This invention provides a method and system for jointly predicting the drag coefficient and surface pressure field of an automobile. Compared with the prior art, it has the following advantages: The joint prediction method of this invention predicts pressure fields in parallel within the same network. With drag coefficient The drag coefficient derived from the pressure field and obtained by integrating the pressure field is constrained by physical consistency. drag coefficient directly regressed from the network Calibration is performed to ensure that the model has physical credibility and interpretability while maintaining numerical accuracy, thereby enabling the location of key stress areas and serving the automotive styling iteration process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A block diagram of a method for jointly predicting the drag coefficient and surface pressure field of a vehicle; Figure 2 This is a block diagram of the subnetwork for predicting the pressure field on the vehicle surface. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This application provides a method and system for jointly predicting the drag coefficient and surface pressure field of a vehicle. It solves the technical problem that existing methods using surrogate models cannot guarantee the physical consistency between the drag coefficient and the surface pressure field. The method enables parallel prediction of the pressure field and drag coefficient in the same network. It also uses physical consistency constraints to calibrate the drag coefficient derived from the pressure field by integrating the pressure field with the drag coefficient directly regressed from the network. This ensures that the model has physical confidence and interpretability while maintaining numerical accuracy, thereby enabling the location of key stress areas and serving the vehicle styling iteration process.

[0021] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows: Existing methods for predicting drag and pressure fields using deep learning-based surrogate models often focus on predicting single indicators or single field quantities. They lack collaborative modeling between objectives within the same framework, making it difficult to simultaneously provide predictions of both the overall vehicle drag coefficient and surface pressure field while maintaining their physical consistency. When predicting simultaneously, drag coefficient and pressure field are regressed in parallel to minimize errors. While this provides two types of numerical outputs, it doesn't explicitly model the closed-form physical relationship between the pressure field integral and the drag coefficient. In other words, even when predicting both parameters simultaneously, the physical relationship that "the drag coefficient derived from the pressure field integral should be consistent with the predicted drag coefficient" is not explicitly included in the constraint. This results in situations where each parameter is numerically accurate but inconsistent with the others, reducing engineering usability. Most existing methods adopt a purely data-driven strategy, lacking the design of "physical characteristics" and "physical losses" consistent with external flow dynamics. Meanwhile, point cloud and mesh representations are sensitive to sampling density and distribution, lacking adaptive sampling strategies for key areas such as high curvature, leading to the loss of some details; the evaluation system focuses on errors such as MSE / MAE, lacking indicators and benchmarks to measure the consistency between the predicted drag coefficient and the drag coefficient obtained by integrating the pressure field, making it difficult to connect with the existing automotive styling iteration process and guide local optimization.

[0022] To overcome the shortcomings of existing technologies, this invention proposes a joint prediction method for vehicle drag coefficient and surface pressure field. This joint prediction method predicts the pressure field in parallel within the same network. With drag coefficient The drag coefficient derived from the pressure field and obtained by integrating the pressure field is constrained by physical consistency. drag coefficient directly regressed from the network Calibration is performed to ensure that the model has physical credibility and interpretability while maintaining numerical accuracy, thereby enabling the location of key stress areas and serving the automotive styling iteration process.

[0023] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0024] This invention provides a method for jointly predicting the drag coefficient and surface pressure field of a vehicle, such as... Figure 1 As shown, it includes: S1. Obtain a training dataset including vehicle point cloud data and label data, and preprocess the vehicle point cloud data in the training dataset to obtain a sampling point matrix and shape parameter vector; and perform uniform sampling on the sampling point matrix to obtain a uniform point matrix. S2. The initial joint prediction model is trained using the sampling point matrix, shape parameter vector, and uniform point matrix. The loss value between the model output and the label data is optimized to obtain the joint prediction model. The joint prediction model is used to process the sampling point matrix, shape parameter vector, and uniform point matrix of the vehicle to be evaluated to obtain the vehicle drag coefficient and the vehicle surface pressure field. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

[0025] The joint prediction method of this invention predicts pressure fields in parallel within the same network. With drag coefficient The drag coefficient derived from the pressure field and obtained by integrating the pressure field is constrained by physical consistency. drag coefficient directly regressed from the network Calibration is performed to ensure that the model has physical credibility and interpretability while maintaining numerical accuracy, thereby enabling the location of key stress areas and serving the automotive styling iteration process.

[0026] It should be noted that after the joint prediction model is trained, the physical consistency constraint calculation module no longer functions during actual use.

[0027] The structure and training process of the joint prediction model are described in detail below: In this embodiment of the invention, in order to make the training of the joint prediction model faster and more stable, the vehicle drag prediction sub-network is initialized in advance. The initialization process of the vehicle drag prediction sub-network is as follows: The system uses a 3D point cloud (STL) of the vehicle as input and employs curvature-based adaptive sampling. Given a fixed number of sampling points, it retains aerodynamically sensitive areas with high curvature based on curvature indicators, reduces redundant points in flat areas, and ensures high-fidelity representation of key surface details. Simultaneously, it automatically extracts global shape parameters (vehicle length L, vehicle width W, and vehicle height H) from the point cloud. The sampled point cloud is used to extract local-point cloud structure features using PointNet++ in the vehicle drag prediction subnetwork. The shape parameters are encoded using MLP in the same subnetwork. These two feature sets are then fused by the feature fusion module in the vehicle drag prediction subnetwork before being output as the drag coefficient via a regression head. The training phase optimizes using MSE. with truth value ,get Rapid prediction results provide a stable initialization for subsequent joint training.

[0028] The curvature sampling process is as follows: Given a car point cloud, , The goal is to prioritize high curvature regions when extracting M samples, for each point... Constructing the covariance matrix ,Pick eigenvalues ,by As a point The curvature index, and the covariance matrix are: in, For point of Nearest neighbor set It is the neighborhood mean. This represents the total number of points on the surface of the car's point cloud; for Nearest neighbor set The point in the middle.

[0029] The feature fusion process in the feature fusion module can be described as follows: For a batch of samples, the point cloud structural features are obtained through the PointNet+ module. The global shape dimensions (L, W, H) are increased in dimension by MLP and replicated to each point shape feature. Let the linear mapping of the feature fusion module be... Then cross attention is The formula for the cross-attention feature fusion module is: in, p =0.1 indicates the dropout rate of Dropout; C represents the number of feature channels.B Indicates the batch size.

[0030] It should be noted that the whole vehicle drag prediction sub-network includes PointNet++, MLP, and feature fusion module. During the initialization process, the network, along with the regression head used during initialization, can be used as a standalone drag prediction model to quickly predict the drag coefficient.

[0031] After initializing the vehicle drag prediction subnetwork, the joint prediction model can be trained using a VTK file containing the vehicle's 3D point cloud data, as follows: The 3D point cloud data of the vehicle in the VTK file is preprocessed as follows: The vehicle's 3D point cloud data is simplified by curvature-based sampling to obtain a point set, which is then uniformly sampled. Simultaneously, global shape information is automatically extracted from the point cloud. The specific process is consistent with the sampling and shape parameter information extraction in the vehicle drag prediction sub-network, and will not be elaborated here.

[0032] The vehicle drag prediction subnetwork uses a fusion regression of "uniformly sampled point cloud + shape parameters", while the vehicle surface pressure field prediction subnetwork uses a Translator point-level regression structure to obtain preliminary predictions. To ensure physical consistency between the outputs of the two subnetworks, the physical relationship of "pressure field integral - drag coefficient" is introduced as a constraint in the joint training.

[0033] Integrating the pressure field yields the drag coefficient derived from it: Given the unit vector of the incoming flow direction air density ,speed Reference windward area A Dynamic pressure Pressure subnet output point-level pressure field Then, discrete integration yields: in, Indicates the correction factor; This represents the total number of points on the surface of the car's point cloud; Indicates the first i The external normal vector of each point; Indicates the first i The area of ​​each point.

[0034] The loss function of the joint prediction model during training is as follows: The first item is wind resistance monitoring, the second item is pressure monitoring, and the third item is physical consistency constraint. , , These are the weighting coefficients; The drag coefficient derived from the pressure field is obtained by integrating the pressure field. The drag coefficient predicted by the joint forecasting model; Represents the true value of the drag coefficient; The joint prediction model predicts the first... i Pressure field on the surface of a car at a point; Indicates the first i The true value of the pressure field on the car surface at a point.

[0035] Automotive surface pressure field prediction subnetwork, such as Figure 2 As shown, it includes a first normalization layer, a physical attention layer, a second normalization layer, and a feedforward layer. After the first normalization layer processes the sampling point matrix, the output data passes through the physical attention layer. The output of the physical attention layer and the sampling point matrix are concatenated with residuals to obtain residual features. The residual features are processed by the second normalization layer and then input to the feedforward layer. The output of the feedforward layer and the residual features are concatenated with residuals to obtain pressure field features.

[0036] The physical attention layer includes a first linear layer, a multi-head attention module, and a second linear layer; the multi-head attention module includes a slice layer, a label-specific attention layer, and a reverse slice layer.

[0037] The following uses a specific dataset to verify the validity of the embodiments of the present invention: Verification Experiment 1: To verify the effectiveness of the method disclosed in the embodiments of the present invention, this verification experiment used the DriveAerNet++ dataset, which contains 8,000 car designs. Each design was subjected to high-fidelity CFD simulation, covering typical car body categories such as fastback, notchback, and estateback.

[0038] (1) In the wind resistance prediction experiment, each sample was downsampled to 20,000 points based on curvature; the model was trained for 300 epochs with a batch size of 1. The data was divided into: a training set of 5,600 vehicles, a validation set of 800 vehicles, and a test set of 1,600 vehicles. The results show that the wind resistance prediction technology proposed in this invention has significant advantages on the DriveAerNet++ dataset. Specifically, under the condition of sampling 20,000 points, the verification process compared the method of this embodiment with traditional methods such as PointNet and GCNN. The results are shown in Table 1. Table 1 compares the method of this invention with traditional methods such as PointNet and GCNN. (2) In the joint prediction experiment of drag and pressure field, joint training and prediction were performed on the publicly available DriverNet++ dataset. The curvature of the 3D car mesh was sampled to 10w points according to curvature, and a simplified car mesh was generated based on this. The overall shape parameters of the vehicle (length, width, and height) were extracted. The simplified mesh was uniformly sampled and input into the whole vehicle drag prediction sub-network along with the shape parameters. At the same time, the simplified mesh was fully sampled and input into the surface pressure field prediction sub-network. The experimental results show that under the conditions of fluid density of 1.184 kg / m^3, inflow velocity of 30 m / s, and inflow direction along the x-axis in the DriverNet++ dataset, the drag coefficient and the MRE of the pressure field are: MSE is MAE is The consistency coefficient was 0.9607, and the good consistency rate was 96.2%. This metric was used as the physical consistency benchmark for the public dataset DriveAerNet++. Experiments showed that the joint prediction network output... C d It exhibits good physical consistency with the surface pressure field.

[0039] Verification Experiment 2: This experiment uses a dataset of a specific brand of automobiles, which contains a much richer variety of vehicle types than publicly available datasets, including but not limited to sedans, SUVs, off-road vehicles, MPVs, and pickup trucks. These vehicle types exhibit significant differences in geometry, drag sources, vortex distribution, chassis design, and spoiler layout. The results show that the drag prediction technique proposed in this invention demonstrates excellent generalization ability on this dataset for diverse vehicle types, and ablation experiments with different numbers of sampling points show that a sampling point count of 5w yields the best prediction results.

[0040] Table 2 shows the performance of the method of the present invention with different numbers of sampling points. This invention also provides a joint prediction system for vehicle drag coefficient and surface pressure field, comprising: The data acquisition and preprocessing module is used to acquire a training dataset including vehicle point cloud data and label data, preprocess the vehicle point cloud data in the training dataset to obtain a sampling point matrix and shape parameter vector; and perform uniform sampling on the sampling point matrix to obtain a uniform point matrix. The model training module is used to train the initial joint prediction model using the sampling point matrix, shape parameter vector, and uniform point matrix, optimize the loss value between the model output and the label data, and obtain the joint prediction model. The joint prediction model is used to process the sampling point matrix, shape parameter vector, and uniform point matrix of the vehicle to be evaluated to obtain the vehicle's drag coefficient and surface pressure field. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

[0041] It is understood that the joint prediction system for vehicle drag coefficient and surface pressure field provided in this embodiment of the invention corresponds to the joint prediction method for vehicle drag coefficient and surface pressure field described above. The explanation, examples, and beneficial effects of the relevant content can be referred to the corresponding content in the joint prediction method for vehicle drag coefficient and surface pressure field, and will not be repeated here.

[0042] This invention also provides a computer-readable storage medium storing a computer program for the joint prediction of vehicle drag coefficient and surface pressure field, wherein the computer program causes a computer to execute the joint prediction method of vehicle drag coefficient and surface pressure field as described above.

[0043] This invention also provides an electronic device, including: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing a joint prediction method for vehicle drag coefficient and surface pressure field as described above.

[0044] In summary, compared with existing technologies, it has the following beneficial effects: 1. The joint prediction method of this invention predicts pressure fields in parallel within the same network. and And with physical consistency constraints, the integral obtained from the pressure field Direct regression of the network Calibration is performed to ensure that the model has physical credibility and interpretability while maintaining numerical accuracy, thereby enabling the location of key stress areas and serving the automotive styling iteration process.

[0045] 2. The drag coefficient prediction in this embodiment of the invention does not require CFD mesh generation and numerical solution, eliminating the need for cumbersome geometric repair, turbulence modeling, and boundary condition configuration. It can be directly used for repeated iterations and scheme selection in the early stages of styling, significantly reducing manpower, computing power, and time costs. Simultaneously, it employs a dual-channel representation of "local geometry + global shape," automatically extracting the vehicle's shape parameters (length, width, and height) from the point cloud. These parameters are then encoded separately with the local-point cloud structural features of the curvature sampling point cloud, and cross-attention fusion is performed in the network. This approach considers both detail and contour priors, improving the cross-vehicle generalization of drag prediction.

[0046] 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 apparatus 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 apparatus. 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 apparatus that includes said element.

[0047] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for jointly predicting the drag coefficient and surface pressure field of a vehicle, characterized in that, include: Obtain a training dataset that includes vehicle point cloud data and label data, and preprocess the vehicle point cloud data in the training dataset to obtain a sampling point matrix and shape parameter vector; The sampling point matrix is ​​then uniformly sampled to obtain a uniform point matrix; The initial joint prediction model is trained by sampling point matrix, shape parameter vector and uniform point matrix, and the loss value between the output of the joint prediction model and the label data is optimized to obtain the joint prediction model. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

2. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in claim 1, characterized in that, The vehicle drag prediction subnetwork includes a PointNet++ module, an MLP module, and a feature fusion module. The PointNet++ module is used to process the uniform point matrix to obtain point cloud structural features. The MLP module is used to encode the shape parameters to obtain shape features. The feature fusion module is used to align and fuse the point cloud structural features and the shape features.

3. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in claim 2, characterized in that, The process of aligning and fusing the point cloud structural features and shape features by the feature fusion module includes: The linear mapping of the feature fusion module is Then cross attention is The formula for the cross-attention feature fusion module is: in, p =0.1 indicates the dropout rate of Dropout; , representing the structural features of the point cloud; , represents the shape feature; C represents the number of feature channels, B Indicates the batch size.

4. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in claim 1, characterized in that, The vehicle surface pressure field prediction subnetwork includes a first normalization layer, a physical attention layer, a second normalization layer, and a feedforward layer. After the first normalization layer processes the sampling point matrix, the output data passes through the physical attention layer. The output of the physical attention layer and the sampling point matrix are concatenated with residuals to obtain residual features. The residual features are processed by the second normalization layer and then input to the feedforward layer. The output of the feedforward layer and the residual features are concatenated with residuals to obtain pressure field features.

5. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in claim 4, characterized in that, The physical attention layer includes a first linear layer, a multi-head attention module, and a second linear layer; The multi-head attention module includes a slice layer, a label-specific attention layer, and a reverse slice layer.

6. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in any one of claims 1 to 5, characterized in that, The step of integrating the predicted surface pressure field to obtain the drag coefficient derived from the pressure field includes: Given the unit vector of the incoming flow direction air density ,speed Reference windward area A Dynamic pressure ; Pressure subnet outputs surface pressure field Then, discrete integration yields the drag coefficient derived from the pressure field. : in, Indicates the correction factor; This represents the total number of points on the surface of the car's point cloud; Indicates the first i The external normal vector of each point; Indicates the first i The area of ​​each point.

7. The method for jointly predicting the drag coefficient and surface pressure field of a vehicle as described in any one of claims 1 to 5, characterized in that, The formula for calculating the loss value between the output of the joint prediction model and the label data is as follows: The first item is wind resistance monitoring, the second item is pressure monitoring, and the third item is physical consistency constraint. , , These are the weighting coefficients; The drag coefficient derived from the pressure field is obtained by integrating the pressure field. The drag coefficient predicted by the joint forecasting model; Represents the true value of the drag coefficient; The joint prediction model predicts the first... i Pressure field on the surface of a car at a point; Indicates the first i The true value of the pressure field on the car surface at a point.

8. A joint prediction system for automobile drag coefficient and surface pressure field, characterized in that, include: The data acquisition and preprocessing module is used to acquire a training dataset including vehicle point cloud data and label data, preprocess the vehicle point cloud data in the training dataset to obtain a sampling point matrix and shape parameter vector; and perform uniform sampling on the sampling point matrix to obtain a uniform point matrix. The model training module is used to train the initial joint prediction model using the sampling point matrix, shape parameter vector, and uniform point matrix, optimize the loss value between the model output and the label data, and obtain the joint prediction model. The joint prediction model is used to process the sampling point matrix, shape parameter vector, and uniform point matrix of the vehicle to be evaluated to obtain the vehicle's drag coefficient and surface pressure field. The initial joint prediction model includes a vehicle drag prediction sub-network, a vehicle surface pressure field prediction sub-network, a multilayer perceptron, an image segmentation network, and a physical consistency constraint calculation module. Among them, the physical consistency constraint calculation module is used to integrate the predicted surface pressure field to obtain the drag coefficient derived from the pressure field. The whole vehicle drag prediction subnetwork is used to process the uniform point matrix and the shape parameter vector to obtain the drag characteristics; the vehicle surface pressure field prediction subnetwork is used to process the sampling point matrix to obtain the pressure field characteristics; the multilayer perceptron is used to process the drag characteristics and the drag coefficient derived from the pressure field to obtain the vehicle drag coefficient; the image segmentation network is used to process the pressure field characteristics and the drag coefficient derived from the pressure field to obtain the vehicle surface pressure field.

9. A computer-readable storage medium, characterized in that, It stores a computer program for the joint prediction of vehicle drag coefficient and surface pressure field, wherein the computer program causes a computer to execute the joint prediction method of vehicle drag coefficient and surface pressure field as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing a joint prediction method for vehicle drag coefficient and surface pressure field as described in any one of claims 1 to 7.

Images