Automobile wind resistance coefficient prediction method and system based on modeling cas data

Abstract

The present disclosure provides a vehicle wind resistance coefficient prediction method and system based on modeling CAS data, relates to the technical field of artificial intelligence, acquires vehicle three-dimensional model modeling data, pre-processes the three-dimensional model modeling data; constructs graph network data by using the pre-processed three-dimensional model modeling data; inputs the graph network data into a GNN network model, iteratively updates the features of nodes and edges to learn the complex relationship between nodes, in each graph convolution layer, the node features are updated according to the features of the neighbor nodes, acquires the process of transmission and aggregation of the three-dimensional model modeling data in the graph, extracts the complex correlation between nodes through the iteration of multiple layers of graph convolution layers, selects the average relative error as the loss function of the model, and outputs the risk coefficient label of the three-dimensional model modeling data corresponding to the vehicle type; the present disclosure greatly shortens the prediction time and reduces the calculation cost.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of artificial intelligence, and particularly relates to a vehicle drag coefficient prediction method and system based on styling CAS data. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The drag coefficient of a vehicle is an important indicator of its aerodynamic performance, and has a significant impact on the economy, stability and safety of the vehicle. Traditional drag coefficient prediction methods rely on complex numerical simulation and testing, which is time-consuming, labor-intensive and costly. These methods require a large amount of calculation and experiment, resulting in high cost and long development cycle. To solve this problem, a new type of vehicle drag coefficient prediction method based on artificial intelligence technology has emerged. This method uses deep learning algorithms to analyze and model the appearance shape data of the vehicle, and combines large-scale simulation data for training, thereby achieving accurate prediction of the vehicle drag coefficient.

[0004] In the current deep learning algorithm, the drag coefficient of the vehicle is predicted by using model parameterized data through a convolutional neural network, but this method requires extracting the parameters of the model, and the process of extracting the model parameters is relatively complex, which is not conducive to the actual operation of engineers and cannot quickly obtain results. SUMMARY

[0005] To solve the above problems, the present disclosure proposes a vehicle drag coefficient prediction method and system based on styling CAS data, which uses styling CAS data to predict the drag performance of the vehicle, uses a graph neural network model to extract and represent the features of the grid structure, and inputs the feature vector of the graph into the GNN model for prediction, greatly shortening the prediction time and reducing the calculation cost.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions:

[0007] The vehicle drag coefficient prediction method based on styling CAS data comprises:

[0008] Obtaining vehicle three-dimensional model styling data, and preprocessing the three-dimensional model styling data;

[0009] Constructing graph network data using the preprocessed three-dimensional model styling data;

[0010] The graph network data is input into a GNN network model, complex relationships between nodes are learned by iteratively updating features of nodes and edges, in each graph convolution layer, node features are updated according to features of neighbor nodes, a process of passing and aggregating three-dimensional model modeling data in a graph is obtained, complex correlation relationships between nodes are extracted through iteration of multiple graph convolution layers, an average relative error is selected as a loss function of the model, and a risk coefficient label of three-dimensional model modeling data corresponding to a vehicle model is output.

[0011] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0012] The automobile wind resistance coefficient prediction system based on modeling CAS data comprises:

[0013] A data acquisition module is configured to acquire three-dimensional model modeling data of a vehicle, pre-process the three-dimensional model modeling data, and construct graph network data by using the pre-processed three-dimensional model modeling data.

[0014] A prediction module is configured to input the graph network data into a GNN network model, learn complex relationships between nodes by iteratively updating features of nodes and edges, update node features according to features of neighbor nodes in each graph convolution layer, obtain a process of passing and aggregating three-dimensional model modeling data in a graph, extract complex correlation relationships between nodes through iteration of multiple graph convolution layers, select an average relative error as a loss function of the model, and output a risk coefficient label of three-dimensional model modeling data corresponding to a vehicle model.

[0015] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0016] A non-transitory computer-readable storage medium is configured to store computer instructions, and the computer instructions are executed by a processor to implement the automobile wind resistance coefficient prediction method based on modeling CAS data.

[0017] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0018] An electronic device comprises a processor, a memory and a computer program, wherein the processor is connected with the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the automobile wind resistance coefficient prediction method based on modeling CAS data.

[0019] Compared with the prior art, the present disclosure has the beneficial effects that:

[0020] The automobile wind resistance coefficient prediction method based on styling CAS data of the present disclosure can convert the three-dimensional geometric model of the automobile into the grid format required by the computational fluid dynamics (CFD) solver by using the computer-aided design (CAD) software. This can provide accurate automobile appearance data for subsequent numerical simulation. Secondly, the method uses deep learning algorithms to analyze and model the styling data of the automobile. By inputting these data into the deep learning model for training, the complex relationship between the automobile shape and wind resistance can be learned, and the prediction of the wind resistance coefficient can be realized. The method is trained in combination with large-scale simulation data. By collecting and labeling a large amount of wind resistance coefficient data of different automobile models, a rich training set is constructed to realize accurate prediction of the automobile wind resistance coefficient. Compared with traditional methods, the method has higher prediction accuracy and efficiency, and can provide more scientific and effective guidance for automobile design and optimization.

[0021] The automobile wind resistance coefficient prediction method based on styling CAS data of the present disclosure can directly use three-dimensional styling CAS data to predict the wind resistance coefficient of the automobile; the data set uses CFD simulation data, and the types of vehicles are numerous, increasing the diversity of the data and improving the authenticity and reliability of the model; the graph neural network of SAGEconv is used to train and predict the three-dimensional styling data, which can realize accurate prediction of the automobile wind resistance coefficient and improve the accuracy of the prediction results; compared with traditional numerical simulation and test methods, the present disclosure uses existing three-dimensional model data and CFD simulation calculation results for training, without the need for frequent actual tests and large amounts of data collection, reducing the consumption of resources and energy, and meeting the requirements of environmental protection and sustainable development.

[0022] The present disclosure can quickly perform multiple iterations and optimization through the establishment of an AI model for wind resistance performance prediction, which can assist automobile designers and engineers in obtaining the best design scheme in a short time, improving design efficiency and work efficiency; it can be applied to different types and sizes of automobiles, including cars, trucks, SUVs and various vehicle models. At the same time, according to actual needs, the model can be modified and adjusted according to specific problems to adapt to different application scenarios and needs. BRIEF DESCRIPTION OF DRAWINGS

[0023] The drawings accompanying the specification of the present disclosure serve to provide a further understanding of the present disclosure, and the illustrative embodiments of the present disclosure and their descriptions serve to explain the present disclosure, and do not constitute an improper limitation on the present disclosure.

[0024] Figure 1 The flowchart of the wind resistance coefficient prediction method of the present disclosure embodiment;

[0025] Figure 2 The flowchart of the model data preprocessing of the present disclosure. DETAILED DESCRIPTION

[0026] The present disclosure is further described below with reference to the accompanying drawings and examples.

[0027] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the present disclosure. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0029] Example 1

[0030] In an embodiment of the present disclosure, a method for predicting the wind resistance coefficient of a vehicle based on styling CAS data is provided, comprising:

[0031] Step 1: Obtain the three-dimensional model styling data of the vehicle, and pre-process the three-dimensional model styling data;

[0032] Step 2: Construct graph network data using the pre-processed three-dimensional model styling data;

[0033] Step 3: Input the graph network data into the GNN network model, learn the complex relationship between nodes by iteratively updating the features of nodes and edges, and update the node features according to the features of the neighbor nodes in each graph convolution layer, obtain the process of transmission and aggregation of the three-dimensional model styling data in the graph, extract the complex correlation between nodes through the iteration of multiple layers of graph convolution layers, select the average relative error as the loss function of the model, and output the risk coefficient label of the corresponding three-dimensional model styling data of the vehicle.

[0034] As an embodiment, the method for predicting the wind resistance coefficient of a vehicle based on styling CAS data of the present disclosure uses styling CAS data to train an AI model to carry out a wind resistance performance prediction method, uses a graph neural network model to extract and represent the features of the grid structure, inputs the feature vector of the graph into the GNN model for prediction, and the specific training steps include:

[0035] Data acquisition: First, acquire the three-dimensional modeling data of the car, including car, SUV and MPV models, a total of 2000 data, which are obtained through modeling software design or existing car model database, and the corresponding wind resistance coefficient label is calculated through CFD simulation. The simulation process first prepares the geometric model and ensures the accuracy and integrity of the model ear; then the model is meshed through STARCCM+ software, and the boundary conditions of the fluid domain are set to build the simulation environment; then set the turbulence model, set the finite volume method of STARCCM+ software for solving; after completing the settings, the simulation can be started. STAR-CCM+ will iteratively solve the fluid dynamics equation in the calculation process, and calculate the wind resistance coefficient of the car at different speeds.

[0036] Data pre-extraction: A series of preprocessing operations are performed on the three-dimensional model data. First, convert the CATIA data to STL mesh data, and process and standardize the mesh data to ensure data quality and consistency.

[0037] Among them, the data of the obtained car model and the corresponding wind resistance coefficient are cleaned to remove abnormal data to ensure the accuracy of the data;

[0038] A series of preprocessing operations are performed on the cleaned and accurate data, including unit standardization, half-car symmetry, mesh simplification and data normalization. If the unit of the car model is not unified, it needs to be converted to meters for length unit unification; the width of the model needs to be used to determine whether it is a half-car, if it is a half-car, it needs to be symmetrical to a whole car; the mesh data processed by STARCCM+ is too large, and the number of meshes of each data is inconsistent, so it needs to be simplified to the same specification; then the simplified mesh is converted into the input format of the graph neural network, which is represented by edges and nodes, so that the dataset with wind resistance coefficient label can be obtained.

[0039] According to the ratio of 7:1.5:1.5, the dataset is divided into training set, validation set and test set.

[0040] Establishing a prediction model: Using deep learning method graph neural network (GNN) to establish a prediction model, the extracted features are used as input and the wind resistance coefficient is used as output.

[0041] The GNN network model structure is composed of 3 graph convolution layers and 2 fully connected layers. In each graph convolution layer, the node features are updated according to the features of its neighbor nodes. This process can be understood as the process of information transmission and aggregation in the graph. Through the iteration of multiple layers of graph convolution layers, the model can gradually learn and extract the complex associations between nodes.

[0042] Model training and validation: Use the pre-processed dataset to train the model, and validate and optimize the model to ensure its good generalization ability.

[0043] According to the deep learning of graph neural network (GNN) model, the automobile grid data is modeled and learned. Each node in the automobile grid data usually represents a specific position or unit of the grid, and each node has a feature vector, which can include geometric features, connection features, physical features and material features; the edge represents the connection or relationship between nodes, which can be a physical contact surface, a fluid flow path, etc. GNN learns the complex relationship between nodes by iteratively updating the features of nodes and edges. The model structure of GNN is composed of 3 graph convolution layers and 2 fully connected layers. In each graph convolution layer, the node features are updated according to the features of its neighbor nodes. The graph convolution network uses node features, neighbor node features and edge features to learn the complex relationship between nodes in the graph structure. Through feature extraction and update in the graph convolution layer, this process can be understood as the process of information transmission and aggregation in the graph. Through the iteration of multiple graph convolution layers, the model can gradually learn and extract the complex association between nodes;

[0044] The training set and the validation set are trained in the way of cross-validation, which can detect and avoid overfitting problem. In each cross-validation, different training set and validation set are used, which can more comprehensively evaluate the performance and generalization ability of the model.

[0045] In the training process, the mean absolute error (MAE) is selected as the loss function of the model, and the formula of MAE is defined as follows: Where N is the number of samples, y i is the true value of the i th sample, and is the predicted value of the i

[0046] th sample.

[0046] Prediction stage: for each sample i, the model calculates the predicted value i according to the input feature x Calculate the error: for each sample i, calculate the absolute error i between its true value y and predicted value Sum: sum the absolute errors of all samples; average: divide the sum by the number of samples N to get the average absolute error.

[0047] The trained model is evaluated using the test set. Evaluation indicators such as mean relative error, trend accuracy, and fitting determination coefficient are used to evaluate the pros and cons of the model; the prediction ability of the model is judged by calculating the mean relative error, trend accuracy, and fitting determination coefficient of the prediction results and simulation values of the test set, where the smaller the mean relative error, the closer the prediction results are to the simulation values; the greater the trend accuracy, the more consistent the change trend of the prediction results and the simulation values; the closer the fitting determination coefficient is to 1, the higher the fitting degree of the prediction results and the simulation values. According to the evaluation results, the model is optimized and the parameters are adjusted. The structure of the model is adjusted, and the number of network layers is increased or decreased, the regularization parameter and the inactivation rate are changed, and other methods are used to improve the performance of the model; the model hyperparameters can also be optimized, including batch size, epoch, and learning rate (lr) parameters, and selecting appropriate parameters can improve the training effect of the model;

[0048] Prediction: Apply the trained model to actual car design and optimization, predict the wind resistance coefficient according to the input car geometry and physical characteristics, and provide the corresponding results and suggestions.

[0049] Among them, the trained AI model is used for wind resistance performance prediction. Input new modeling CAS data, output the corresponding wind resistance coefficient through the model, to quickly evaluate the wind resistance performance of different car models.

[0050] Embodiment 2

[0051] An embodiment of the present disclosure provides a car wind resistance coefficient prediction system based on modeling CAS data, comprising:

[0052] A data acquisition module is configured to acquire three-dimensional model data of a car and pre-process the three-dimensional model data; and construct graph network data using the pre-processed three-dimensional model data.

[0053] A prediction module is configured to input the graph network data into a GNN network model, iteratively update the features of nodes and edges to learn the complex relationships between nodes, update the features of nodes according to the features of their neighbor nodes in each graph convolution layer, obtain the process of transmission and aggregation of three-dimensional model data in the graph, extract the complex correlation between nodes through iteration of multiple graph convolution layers, select the mean relative error as the loss function of the model, and output the risk coefficient label of the three-dimensional model data of the corresponding car model.

[0054] Embodiment 3

[0055] In an embodiment of the present disclosure, a non-transitory computer readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the method for predicting the wind resistance coefficient of a vehicle based on styling CAS data.

[0056] Embodiment 4

[0057] In an embodiment of the present disclosure, an electronic device is provided, comprising a processor, a memory, and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device implements the method for predicting the wind resistance coefficient of a vehicle based on styling CAS data.

[0058] The present disclosure is described with reference to the flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.

[0059] These computer program instructions can also be loaded onto a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable data processing device to produce a computer-implemented process, so that the instructions executed by the computer or other programmable data processing device provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The steps for implementing the functions specified in one or more flows and / or blocks.

[0060] Although the specific embodiments of the present disclosure are described above with reference to the accompanying drawings, the present disclosure is not limited thereto, and various modifications or variations can be made by those skilled in the art without creative labor, which are still within the protection scope of the present disclosure.

Claims

1. A method for predicting the drag coefficient of a vehicle based on styling CAS data, characterized in that, include: Acquire the three-dimensional model styling data of the car, and preprocess the three-dimensional model styling data; Construct graph network data using preprocessed 3D model data; The graph network data is input into the GNN network model. By iteratively updating the features of nodes and edges, the complex relationships between nodes are learned. In each graph convolutional layer, the node features are updated according to the features of its neighboring nodes. The process of 3D model styling data being transferred and aggregated in the graph convolutional layer is obtained. Through the iteration of multiple graph convolutional layers, the complex relationships between nodes are extracted. The average relative error is selected as the loss function of the model, and the risk coefficient label of the 3D model styling data of the corresponding vehicle model is output. The risk coefficient label is obtained by using the three-dimensional model data of the car to perform fluid dynamics simulation or wind tunnel test to obtain the drag coefficient label of the corresponding car model. The GNN network model consists of three graph convolutional layers and two fully connected layers. In each graph convolutional layer, the node features are updated based on the features of its neighboring nodes.

2. The method for predicting the drag coefficient of a vehicle based on styling CAS data as described in claim 1, characterized in that, The preprocessing of the three-dimensional model modeling data includes: cleaning the three-dimensional model modeling data to remove abnormal data, and then performing preprocessing operations on the cleaned data.

3. The method for predicting the drag coefficient of a vehicle based on styling CAS data as described in claim 2, characterized in that, The cleaned data is then subjected to unit standardization, half-vehicle symmetry, grid simplification, and data normalization to obtain a dataset labeled with drag coefficient.

4. The method for predicting the drag coefficient of a vehicle based on styling CAS data as described in claim 3, characterized in that, The dataset labeled with drag coefficients was divided into training, validation, and test sets in a ratio of 7:1.5:1.

5. The training and validation sets were trained using cross-validation. During training, the mean relative error was selected as the loss function of the model, and the backpropagation algorithm and optimizer were used to update the parameters.

5. The method for predicting the drag coefficient of a vehicle based on styling CAS data as described in claim 1, characterized in that, The GNN network model is evaluated using a test set. The mean relative error, maximum error, and trend accuracy metrics are used to evaluate the performance of the GNN network model. Based on the evaluation results, the GNN network model is optimized and its parameters are tuned, including the batch size, number of iterations, and learning rate.

6. A vehicle drag coefficient prediction system based on styling CAS data, characterized in that, include: The data acquisition module is used to acquire the three-dimensional model styling data of the car and to preprocess the three-dimensional model styling data. Construct graph network data using preprocessed 3D model data; The prediction module is used to input the graph network data into the GNN network model, learn the complex relationships between nodes by iteratively updating the features of nodes and edges, and in each graph convolutional layer, the node features are updated according to the features of its neighboring nodes. The process of 3D model styling data being transferred and aggregated in the graph convolutional layer is obtained. Through the iteration of multiple graph convolutional layers, the complex relationships between nodes are extracted, the average relative error is selected as the loss function of the model, and the risk coefficient label of the 3D model styling data of the corresponding vehicle model is output. The risk coefficient label is obtained by using the three-dimensional model data of the car to perform fluid dynamics simulation or wind tunnel test to obtain the drag coefficient label of the corresponding car model. The GNN network model consists of three graph convolutional layers and two fully connected layers. In each graph convolutional layer, the node features are updated based on the features of its neighboring nodes.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the vehicle drag coefficient prediction method based on styling CAS data as described in any one of claims 1-5.

8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the vehicle drag coefficient prediction method based on styling CAS data as described in any one of claims 1-5.

Images