Automobile tire modeling method and system and storage medium

Abstract

The invention relates to the technical field of modeling, in particular to an automobile tire modeling method and system and a storage medium, and the method comprises the steps: 1, obtaining tire contour data; step 2, extracting tire cross-section image features: extracting feature vectors of the two-dimensional tire cross-section image based on the point cloud image, and calculating the distance between the feature vectors; the similarity between the images is judged based on the distance, and the tire point cloud is segmented into a tire surface point cloud and a pattern groove point cloud according to the similarity; step 3, constructing a tread point cloud mapping matrix; 4, fitting a pattern boundary curve; step 5, determining the distribution of a tire material layer: segmenting the tire material layer by using a DeepLabV3 +-based network structure in combination with geometric features obtained from the tread point cloud mapping matrix to obtain the distribution condition of the tire material layer; and step 6, training and verifying the model, and obtaining an automobile tire finite element model after verification is passed. The precision of the finite element model of the automobile tire is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of modeling, and in particular to a method, system and storage medium for modeling a car tire. Background Art

[0002] In the field of vehicle technology, tires are the only medium for vehicles to contact the ground, and their performance directly affects the vehicle's handling, safety, and comfort. Therefore, accurate modeling of tires to accurately predict their mechanical properties under different working conditions has always been a research hotspot in the field of vehicle engineering.

[0003] Existing tire modeling methods mainly rely on experimental testing and data analysis to fit the mechanical properties of tires through a large amount of experimental data. However, this method is not only time-consuming and labor-intensive, but also difficult to fully consider the performance changes of tires under different wear conditions. In recent years, with the rapid development of computer technology, finite element simulation technology has gradually become an important means of tire modeling. Through finite element simulation, the stress conditions of tires under different working conditions can be simulated, thereby predicting their mechanical properties.

[0004] However, there are still some problems with the existing tire finite element simulation technology, especially in the tire vulcanization molding process and tire wear considerations. During the tire vulcanization molding process, due to the high temperature and high pressure, its profile and material layer will undergo certain deformation. This deformation not only affects the geometry of the tire, but also changes its internal stress distribution and material properties. If this factor is ignored during finite element simulation, there will be a large deviation between the simulation results and the actual situation, thereby reducing the accuracy of the simulation.

[0005] In addition, tires will wear out during use, and the degree of wear will directly affect the mechanical properties of the tires. Although existing tire modeling methods have begun to consider the impact of tire wear on mechanical properties, they often identify wear characteristics after the tire model has been established. Although this method can reflect the performance changes of tires after wear to a certain extent, it cannot consider the impact of wear on tire modeling from the source, and therefore cannot obtain a completely accurate tire dynamics model.

[0006] Therefore, it is necessary to develop a new automobile tire modeling method, system and storage medium. Summary of the invention

[0007] The object of the present invention is to provide a vehicle tire modeling method, system and storage medium, which can improve the accuracy of the vehicle tire finite element model.

[0008] In a first aspect, a method for modeling a vehicle tire according to the present invention comprises the following steps:

[0009] Step 1. Obtain tire contour data: Scan the tire to obtain a point cloud image of the inner and outer contours of the tire;

[0010] Step 2. Extracting tire cross-sectional image features: extracting feature vectors of the two-dimensional tire cross-sectional image based on the point cloud image, and calculating the distance between the feature vectors; judging the similarity between images based on the distance, and segmenting the tire point cloud into a tread point cloud and a groove point cloud according to the similarity;

[0011] Step 3. Constructing a tread point cloud mapping matrix: using a NURBS spline curve to fit the tread basis function curve, constructing a tread point cloud mapping matrix that maps the tread point cloud to the tire model surface;

[0012] Step 4: Fitting the pattern boundary curve: After segmentation, the pattern groove point cloud adopts a multi-segment dynamically optimized straight line arc approximation method to achieve the tangent curve fitting of each pattern boundary;

[0013] Step 5. Determine the distribution of tire material layers: Use the DeepLabV3+ network structure to segment the tire material layers and obtain the distribution of the tire material layers, that is, to construct a preliminary model of the car tire;

[0014] Step 6. Training and verifying the model: using tire image samples to train and verify the preliminary model of the automobile tire, and obtaining the automobile tire finite element model after passing the verification.

[0015] Optionally, in step 1, a 3D scanner is used to perform a 3D scan on the tire to obtain a point cloud map of the inner and outer contours of the tire, and the resolution of the 3D scanner reaches 0.1 mm or higher. Using a high-resolution 3D scanner to scan the tire can ensure that the obtained point cloud map has high precision and details, providing a reliable data basis for subsequent tire modeling. This high-precision data acquisition method helps to improve the accuracy and authenticity of the tire model.

[0016] Optionally, in step 2, based on the point cloud image, an AlexNet network is used to extract image features of the two-dimensional tire cross-section image and establish a feature vector. Using the AlexNet network to extract the feature vector of the two-dimensional tire cross-section image can fully utilize the deep learning capability of the network to extract rich feature information from the image.

[0017] Optionally, the hierarchical structure of the AlexNet network includes features from low-level to high-level, specifically including edges, textures, shapes, and complex objects, which helps to more accurately segment the tire point cloud and improve the detail representation of the tire model.

[0018] Optionally, in step 5, the network structure of DeepLabV3+ includes an encoder and a decoder, the encoder uses a lightweight MobileNet V3_Large as a backbone network, and mines basic features through a void space pyramid pooling module; the decoder restores the high-level semantic information extracted by the encoder to the scale of the original image to reconstruct the edges and details of the tire material layer. Using the DeepLabV3+ network structure to segment the tire material layer can make full use of the deep learning and semantic segmentation capabilities of the DeepLabV3+ network. Through the collaboration of the encoder and the decoder, the feature information of the tire material layer can be efficiently extracted, and a tire material layer model with clear edges and details can be reconstructed. In addition, using a lightweight MobileNet V3_Large as a backbone network can also improve the operating efficiency and real-time performance of the model.

[0019] Optionally, in step 5, before segmenting the material layer of the tire, the following operations are further performed:

[0020] The low-dimensional feature map extracted by the backbone network is cascaded with the high-dimensional feature map obtained by the four-fold upsampling operation to fuse the spatial information of the tire surface texture, edge lines, and seams, as well as the corresponding channel information of brightness, color changes, and texture features. Before the tire material layer is segmented, the low-dimensional feature map is cascaded and fused with the high-dimensional feature map, which can make full use of feature information of different scales and improve the accuracy and robustness of the tire material layer segmentation. This fusion method helps to capture the detailed information of the tire surface texture, edge lines, seams, and the corresponding brightness, color changes, and texture features, thereby generating a more realistic tire model.

[0021] Optionally, in step 6, the tire image samples are actual tire image samples. Using actual tire image samples for model training and verification can ensure the similarity and accuracy between the tire model and the actual tire. This modeling method based on actual data helps to improve the practicality and reliability of the tire model and provide strong support for tire design and optimization.

[0022] Optionally, in step 6, the actual tire image samples are divided into a training set and a validation set, the preliminary model is trained using the training set, the preliminary model is validated using the validation set, the model performance is comprehensively evaluated by the loss function value, average intersection-over-union ratio and accuracy, and after the validation is passed, the automobile tire finite element model is obtained. The actual tire image samples are divided into a training set and a validation set, and the model performance is comprehensively evaluated using evaluation indicators such as the loss function value, average intersection-over-union ratio and accuracy, which can ensure the accuracy and reliability of the tire model. At the same time, the verified tire finite element model can provide a reliable foundation for subsequent tire simulation analysis and optimization design. This modeling and verification method driven by actual tire data helps to improve the efficiency and accuracy of tire design and optimization.

[0023] In a second aspect, a car tire modeling system according to the present invention comprises a controller and a memory, wherein the memory stores a computer-readable program, and when the computer-readable program is called by the controller, the steps of the car tire modeling method according to the present invention can be executed.

[0024] In a third aspect, a storage medium according to the present invention stores a computer-readable program, and when the computer-readable program is called, the steps of the automobile tire modeling method according to the present invention can be executed.

[0025] Beneficial effects of the present invention:

[0026] (1) The present invention uses real tires for modeling, which can fully reflect the deformation of the tire profile and material layer during the vulcanization molding process, which cannot be fully captured by traditional design parameters. Therefore, the use of real tires for modeling makes the automobile tire finite element model closer to the physical properties of the actual tire, thereby significantly improving the accuracy and authenticity of the simulation.

[0027] (2) The present invention uses a convolutional neural network to extract the features of the actual two-dimensional tire cross-sectional image, and determines the similarity between images by calculating the distance between the feature vectors of the images, thereby improving the accuracy of the actual cross-sectional profile of the tire. This technology not only improves the efficiency of establishing the finite element model of the automobile tire, but also makes the simulation results more accurate and reliable.

[0028] (3) The present invention uses the DeepLabV3+ network structure to segment the material layer of the tire, and can accurately obtain the material layer distribution of the tire. This technology provides key data support for the accurate establishment of the finite element model of the automobile tire, so that the finite element model of the automobile tire can more accurately reflect the material structure and mechanical properties inside the tire.

[0029] In summary, the present invention significantly improves the accuracy and authenticity of finite element simulation of automobile tires by using real tires for modeling and combining advanced deep learning technology, providing strong technical support for tire design, optimization and performance evaluation. This innovative achievement not only reduces R&D costs and time, but also promotes the continuous progress and development of tire modeling and simulation technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flowchart of the automobile tire modeling method described in an embodiment of the present application;

[0031] Figure 2 A schematic diagram of image synthesis in an embodiment of the present application;

[0032] Figure 3 This is a flowchart of image retrieval in the embodiment of this application;

[0033] Figure 4 This is a schematic diagram of the DeepLabV3+ network structure in an embodiment of the present application;

[0034] Figure 5 This is a schematic diagram of key semantic category labels of a tread rubber dataset in an embodiment of the present application;

[0035] Figure 6 A curve diagram showing the change of the loss function values ​​of the training set and the validation set with the number of training rounds in the embodiment of the present application;

[0036] Figure 7 A curve chart showing the average intersection-over-union ratio of the validation set as a function of the number of training rounds in the embodiment of the present application;

[0037] Figure 8 A curve chart showing the change in accuracy of the verification set with the number of training rounds in the embodiment of the present application;

[0038] Fig. 9 This is a principle block diagram of the automobile tire modeling system described in the embodiment of the present application;

[0039] In the figure: 1- memory, 2- controller. DETAILED DESCRIPTION

[0040] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0041] like Figure 1 As shown, in an embodiment of the present application, a method for modeling a car tire includes the following steps:

[0042] Step 1. Obtain tire contour data: Scan the tire to obtain a point cloud image of the inner and outer contours of the tire.

[0043] Step 2. Extract tire cross-section image features: Extract feature vectors of the two-dimensional tire cross-section image based on the point cloud image, and calculate the distance between feature vectors; judge the similarity between images based on the distance, and segment the tire point cloud into tread point cloud and groove point cloud based on the similarity.

[0044] Step 3. Constructing a tread point cloud mapping matrix: Using a NURBS spline curve to fit the tread basis function curve, construct a tread point cloud mapping matrix that maps the tread point cloud to the tire model surface.

[0045] Step 4: Fitting the pattern boundary curve: After segmentation, the pattern groove point cloud adopts a multi-segment dynamically optimized straight line and arc approximation method to achieve the tangent curve fitting of each pattern boundary.

[0046] Step 5. Determine the distribution of tire material layers: Use the DeepLabV3+ network structure to segment the tire material layers and obtain the distribution of the tire material layers, that is, construct a preliminary model of the car tire.

[0047] Step 6. Training and validating the model: Use tire image samples to train and validate the preliminary model of the automobile tire, and obtain the automobile tire finite element model after the validation.

[0048] In a possible embodiment, the tire profile data is obtained by:

[0049] Use a 3D scanner to scan the tire in 3D and get a point cloud map of the inner and outer contours of the tire. In order to ensure the high accuracy of the point cloud data, it is recommended to use a high-resolution 3D scanner with a resolution of 0.1 mm or higher to ensure that the detailed features of the tire can be captured.

[0050] In a possible embodiment, the tire cross-section image features are extracted as follows:

[0051] In order to compare the 2D tire cross-section image with the actual cross-section outer contour, the AlexNet network is used to extract the image features of the 2D tire cross-section image and establish a feature vector. When extracting image features, the hierarchical structure of the AlexNet network contains features from low to high levels, including details such as edges, textures, shapes, and complex objects. In the process of extracting image features, the first few layers of the AlexNet network extract low-level edge and texture features, the middle layers capture contours and shapes, and the last few layers extract more advanced semantic complex object features.

[0052] When establishing feature vectors, they specifically include feature vectors of multiple dimensions, such as activation values ​​after processing through convolutional layers and pooling layers. Feature vectors include edge features, texture features, local structural features, and geometric shape features of the image. The extracted feature vectors can fully represent the key information of the two-dimensional tire cross-section image, and use these feature vectors to represent the corresponding image. The similarity between images is determined by calculating the distance between feature vectors, and then the tire point cloud is segmented into tread point cloud and groove point cloud based on the similarity.

[0053] like Figure 2 As shown in the figure, A represents the synthesized image area, B represents the image to be synthesized, and A' and B' are the parts of the overlapping area belonging to image A and image B respectively. In the overlapping area, the smaller the difference in feature information of the two images, the higher the similarity, and vice versa.

[0054] In a possible embodiment, the distance between feature vectors is calculated as:

[0055]

[0056] Where d(x,y) is the distance between feature vectors; i is the feature vector dimension of the image, i = 1, 2, ..., m, and m is the total dimension of the feature vector. i is the feature vector of the image to be retrieved, y i is the image feature vector in the image database (i.e., the multiple images scanned in step 1); the distance between the images in each dimension is obtained by comparing the feature vector distance between the image to be retrieved and the image database. The smaller the distance, the higher the similarity.

[0057] D j =qT j (2)

[0058] Assume that the total number of images in the image database is N, q is the feature vector of the feature area of ​​the image to be retrieved, T j is the feature vector of the image fragment corresponding to the jth image in the image database near the feature area of ​​the image to be retrieved, j = 1, 2, ..., N, D j Represents the distance difference between the feature vector of the image to be retrieved and the feature vector of the jth image in the image database. D is sorted by a sorting algorithm. jThe values ​​are compared to find and output n similar images. The purpose of finding and outputting n similar images is to find the image fragment that is most similar to the image to be retrieved, so as to compare and enhance the accuracy of local details of the image to be retrieved, thereby improving the accuracy of image processing results. The trained finite element model of automobile tires is used to extract image features from the image database and establish a vector index library. The feature vector of the image to be retrieved is extracted through the AlexNet network. Formula (2) is used to calculate and output n images similar to the image to be retrieved. Then, the distance calculated by formula (1) is used to compare the n images to calculate the degree of similarity matching.

[0059] like Figure 3 As shown, in a possible embodiment, the neighborhood retrieval process is as follows: input the image to be retrieved; use the trained AlexNet network to extract the feature vectors of each image in the image database, perform PCA (principal component analysis) dimensionality reduction processing, and establish a two-dimensional tire cross-section image database; calculate the Euclidean distance matching feature vector, and after retrieval, place the image to be retrieved in the specified area according to the specified image size, and then use the Dijkstra algorithm to stitch the new and old images to ensure a smooth transition between the new and old images and reduce the boundary inconsistency problem that may exist in the image.

[0060] In a possible embodiment, the distribution of the tire material layer is determined as follows:

[0061] In order to accurately determine the material distribution of each layer of the tire, the network structure of DeepLabV3+ is proposed to optimize the training data set.

[0062] like Figure 4As shown in the figure, the network structure of DeepLabV3+ is divided into two parts: encoder and decoder. The encoder first extracts basic visual features through the backbone network, also known as low-dimensional feature maps. These features include but are not limited to edge features, texture features, color features, and shape features. Through the multi-level processing of the convolutional neural network, these feature maps contain the preliminary structure and texture information of the image. The low-level convolution kernel (such as 3x3 convolution) can capture local edge and texture information, and the middle and high-level convolution kernels can recognize more complex shape and structure information. In this process, in order to meet the requirements of computational efficiency and segmentation performance, the original backbone network Xception is improved to a lighter MobileNet V3_Large. MobileNet V3_Large uses deep separable convolution to further enhance the efficiency of feature extraction and reduce the amount of calculation while ensuring accuracy requirements. In addition, these basic features are deeply mined through the atrous spatial pyramid pooling module (ASPP). In this process, more detailed feature information is extracted through 1×1 convolution; feature information of different scales is detected using a series of 3×3 dilated convolutions with different dilation rates, thereby effectively capturing and fusing multi-scale information and enhancing the ability to extract objects of different sizes in the image; and the global context information of the image is captured through global average pooling.

[0063] In addition, these basic features are further mined through the dilated spatial pyramid pooling module. The dilated spatial pyramid pooling module effectively captures multi-scale information in the image through convolution operations of different scales (i.e., different dilation rates). Specifically: more detailed feature information is extracted through 1×1 convolution. These detailed feature information include the particle size of tire rubber, subtle surface texture changes, etc., which can improve feature resolution in local areas. A series of 3×3 dilated convolutions with different dilation rates are used to detect feature information of different scales. By selecting multiple different dilation rates (such as 1, 2, 4, 8), the information of the image can be fully captured from details to the whole, but it is generally most appropriate to select 3-4 different dilation rates to achieve a balanced feature fusion effect.

[0064] The global context information of the image is captured through global average pooling. The global context information is reflected in the tire image as the overall material distribution and large-scale structural information. Specifically, through global average pooling, the average feature value of all pixels in the image can be comprehensively considered, and the global information at different positions is included, thereby improving the model's understanding of the overall structure and segmentation accuracy.

[0065] The high-level semantic information extracted by the encoder is restored to the scale of the original image to accurately reconstruct the edges and details of the object. This process first effectively cascades the low-dimensional feature map extracted by the backbone network with the high-dimensional feature map obtained by four-fold upsampling, fusing the spatial information including the texture, edge lines, seams, etc. of the tire surface with the channel information of the corresponding brightness, color changes, and texture features to enhance the recognition of the edges and details of objects such as tread patterns, carcass layers, and tire defects in the image, thereby improving the accuracy of semantic segmentation. Secondly, the number of channels of the output feature map is adjusted through 3×3 convolution to match the number of target categories, and four-fold upsampling is used to output the segmentation result of the same size as the original image. In this way, the decoder can remap the deep features extracted by the convolutional neural network back to the resolution of the original image, while ensuring that the edges and details in the image are accurately restored, ultimately improving the overall effect and accuracy of semantic segmentation.

[0066] In a possible embodiment, the training and verification model is specifically:

[0067] Semantic segmentation training requires a large number of image samples with pixel-level labels (8-bit grayscale images). By cutting the tire, we obtain a tire section sample set, which is manually labeled. We annotate 15 different semantic categories of tread rubber in detail, including rubber elements such as inner liner, carcass, belt, cap, tread, and apex. Figure 5 As shown, each category is equipped with a unique identification ID and corresponding RGB color coding for easy identification and application. The sample set consists of 2975 images for training, 500 images for verification, and 1525 images for testing.

[0068] like Figure 6 As shown in the figure, it can be seen that in the initial stage, the loss function value drops rapidly to about 0.174 (after about 15 rounds), and then continues to decrease at a slower rate and converges, and the final training loss is about 0.075. This shows that the automobile tire finite element model is effectively learned and gradually approaches the optimal performance. The validation set loss also shows a downward trend and finally stabilizes at about 0.101, showing the good generalization ability of the automobile tire finite element model on unseen data.

[0069] like Figure 7 and Figure 8 As shown, the average intersection-over-union ratio and accuracy on the validation set reached 74.65% and 94.74% respectively, highlighting the superior performance of the DeepLabV3+ model in processing tread rubber material layer images.

[0070] like Fig. 9As shown, in an embodiment of the present application, a car tire modeling system includes a controller 2 and a memory 1, wherein the memory 1 stores a computer-readable program, and when the computer-readable program is called by the controller, it can execute the steps of the car tire modeling method in the embodiment of the present application.

[0071] In an embodiment of the present application, a storage medium stores a computer-readable program, and when the computer-readable program is called, the steps of the automobile tire modeling method described in the embodiment of the present application can be executed.

[0072] In an embodiment of the present application, the storage medium may be a tangible storage medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. The storage medium may be a machine-readable signal storage medium or a machine-readable storage medium. The storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0073] The above-mentioned embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above-mentioned embodiments. Any other changes, modifications, substitutions, combinations and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement modes and shall be included in the protection scope of the present invention.

Claims

1. A method for modeling a car tire, characterized in that: The following steps are involved: Step 1. Obtain tire contour data: Scan the tire to obtain a point cloud image of the inner and outer contours of the tire; Step 2. Extracting tire cross-sectional image features: extracting feature vectors of the two-dimensional tire cross-sectional image based on the point cloud image, calculating the distance between the feature vectors, judging the similarity between the images based on the distance, and segmenting the tire point cloud into a tread point cloud and a groove point cloud based on the similarity; Step 3. Constructing a tread point cloud mapping matrix: using the NURBS algorithm to fit the basis function curve of the tread into a NURBS curve, establishing a mapping relationship between the tread point cloud and the tire model surface according to the NURBS curve, organizing the mapped point cloud data into a preset data structure, and forming a tread point cloud mapping matrix; Step 4. Fitting the pattern boundary curve: using a multi-segment dynamically optimized straight line arc approximation method for the segmented pattern groove point cloud to achieve tangent curve fitting of each pattern boundary, and using the structural information in the tread point cloud mapping array to assist in locating the pattern boundary; Step 5. Determine the distribution of tire material layers: Combine the geometric features obtained from the tread point cloud mapping array, use the algorithm based on the DeepLabV3+ network structure to segment the material layers of the tire, obtain the distribution of the tire material layers, and build a preliminary model of the automobile tire; Step 6. Training and verifying the model: using tire image samples to train and verify the preliminary model of the automobile tire, and obtaining the automobile tire finite element model after passing the verification.

2. The automobile tire modeling method according to claim 1, characterized in that: In the step 1, a 3D scanner is used to perform a 3D scan on the tire to obtain a point cloud image of the inner and outer contours of the tire.

3. The automobile tire modeling method according to claim 1, characterized in that: In the step 2, based on the point cloud image, the image features of the two-dimensional tire cross-section image are extracted using the AlexNet network, and a feature vector is established.

4. The automobile tire modeling method according to claim 3, characterized in that: The hierarchical structure of the AlexNet network contains features from low-level to high-level, specifically including edges, textures, shapes, and complex objects.

5. The automobile tire modeling method according to claim 1, characterized in that: In step 5, the network structure of DeepLabV3+ includes an encoder and a decoder. The encoder uses a lightweight MobileNet V3_Large as a backbone network and mines basic features through a dilated spatial pyramid pooling module. The decoder restores the high-level semantic information extracted by the encoder to the scale of the original image to reconstruct the edges and details of the tire material layer.

6. The automobile tire modeling method according to claim 1, characterized in that: In step 5, before the material layer of the tire is segmented, the following operations are further performed: The low-dimensional feature map extracted by the backbone network is cascaded with the high-dimensional feature map obtained through a four-fold upsampling operation to fuse spatial information including the texture, edge lines, and seams of the tire surface, as well as channel information of corresponding brightness, color changes, and texture features.

7. The automobile tire modeling method according to claim 1, characterized in that: In step 6, the tire image samples are actual tire image samples.

8. The automobile tire modeling method according to claim 7, characterized in that: In step 6, the actual tire image samples are divided into a training set and a validation set, the preliminary model is trained using the training set, and the preliminary model is validated using the validation set, the model performance is comprehensively evaluated by the loss function value, average intersection-over-union ratio and accuracy, and after the validation is passed, the automobile tire finite element model is obtained.

9. A car tire modeling system, characterized in that: The invention comprises a controller (2) and a memory (1), wherein the memory (1) stores a computer-readable program, and when the computer-readable program is called by the controller (2), the steps of the automobile tire modeling method according to any one of claims 1 to 8 can be executed.

10. A storage medium, characterized in that: A computer-readable program is stored therein, and when the computer-readable program is called, the steps of the automobile tire modeling method according to any one of claims 1 to 8 can be executed.

Images