Tire pattern depth prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of vehicles, and provides a tread pattern depth prediction method and device, electronic equipment and a storage medium, and the tread pattern depth prediction method comprises the steps: obtaining multi-dimensional features of a to-be-predicted tire; inputting the multi-dimensional features into a preset depth prediction model, and predicting the pattern depth of the to-be-predicted tire; wherein the depth prediction model comprises a hybrid network structure, and the hybrid network structure is obtained based on the self-attention mechanism of a Transform network and the fusion of a KAN network. The depth prediction model provided by the invention combines the advantages of the self-attention mechanism and the KAN network, the multi-dimensional features of the to-be-predicted tire are processed through the depth prediction model, and the pattern depth of the to-be-predicted tire is predicted.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to a tire tread depth prediction method, device, electronic equipment and storage medium. Background Art

[0002] With the continuous development of electric vehicle technology, the market share and popularity of electric vehicles have increased significantly. However, due to the unique transmission system and body layout structure of electric vehicles, their tires tend to wear faster than those of traditional fuel vehicles of the same specifications. This excessive wear not only affects vehicle performance, but also poses serious safety hazards. Especially for passenger cars and large trucks, excessive tire wear may lead to serious consequences such as loss of control and rollover.

[0003] With the rapid development of the automobile industry, the significance of active monitoring of tire wear has become increasingly prominent. As the only contact part between the vehicle and the road, the condition of the tire directly affects driving safety. As an important indicator for evaluating tire wear, tire tread depth is crucial for timely replacement of tires to avoid safety hazards. Therefore, a solution for detecting tire tread depth is urgently needed. Summary of the invention

[0004] The present invention provides a tire tread depth prediction method, device, electronic equipment and storage medium, which are used to solve the defects in the prior art and achieve the effect of detecting the tire tread depth.

[0005] The present invention provides a tire tread depth prediction method, comprising: Obtaining multi-dimensional features of the tire to be predicted; The multi-dimensional features are input into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0006] According to a tire tread depth prediction method provided by the present invention, the hybrid network structure includes a first network structure layer and a second network structure layer connected in sequence, the first network structure layer includes a first self-attention layer, a first residual connection layer, a first KAN network, a fully connected layer and a second residual connection layer connected in sequence, and the second network structure layer includes a second self-attention layer, a third residual connection layer, a second KAN network and a third KAN network connected in sequence.

[0007] According to a tire tread depth prediction method provided by the present invention, the step of obtaining multi-dimensional features of a tire to be predicted includes: Acquiring an acceleration signal corresponding to the tire to be predicted; calculating a dynamic load, a peak acceleration, a contact length ratio, a rolling period, a lag peak, and a lead peak based on the acceleration signal; The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, and the leading peak.

[0008] According to a tire tread depth prediction method provided by the present invention, the multi-dimensional features are obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling cycle, the lag peak and the leading peak, including: Acquire tire information corresponding to the tire to be predicted; wherein the tire information includes absolute air pressure; The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, the lead peak and the tire information.

[0009] According to a tire tread depth prediction method provided by the present invention, the process of constructing the depth prediction model includes: Obtaining a training set and a validation set, and preprocessing a plurality of tire samples in the training set and the validation set; wherein the tire samples include sample features and sample labels; Inputting the tire samples in the training set into a basic depth prediction model for iterative training until the training is completed, thereby obtaining the depth prediction model; wherein the basic depth prediction model includes a basic hybrid network structure obtained by fusion of a Transformer network and a KAN network; The depth prediction completed through training is verified by using the tire samples in the verification set.

[0010] According to a tire tread depth prediction method provided by the present invention, the preprocessing of the multiple tire samples in the training set and the validation set includes: The sample features and the sample labels of each of the tire samples are normalized.

[0011] According to a tire tread depth prediction method provided by the present invention, after inputting the multi-dimensional features into a preset depth prediction model and predicting the tread depth of the tire to be predicted, the method comprises: Based on the predicted pattern depth, a prediction error of the depth prediction model is calculated by a root mean square error formula, and the depth prediction model is evaluated based on the prediction error.

[0012] The present invention also provides a tire tread depth prediction device, comprising: A first acquisition module, configured to acquire multi-dimensional features of a tire to be predicted; A prediction module is configured to input the multidimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned tire tread depth prediction methods is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the tire tread depth prediction method described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above tire tread depth prediction methods.

[0016] The tire tread depth prediction method, device, electronic device and storage medium provided by the present invention can capture the relationship between multi-dimensional features in a global scope and dynamically adjust the weights of different features through the self-attention mechanism of the Transformer network. This enables the deep prediction model to effectively integrate multi-dimensional features and identify key factors affecting the tread depth in the prediction of tire tread depth. The KAN network can more effectively process nonlinear relationships and complex patterns in data. The deep prediction model provided by the present invention combines the advantages of the self-attention mechanism and the KAN network, processes the multi-dimensional features of the tire to be predicted through the deep prediction model, predicts the tread depth of the tire to be predicted, and then can accurately and real-time monitor the tire wear status under different working conditions, provide early warnings to users, and thus improve the driving safety of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a schematic flow chart of the tire tread depth prediction method provided by the present invention.

[0019] Figure 2 It is a schematic diagram of the structure of the depth prediction model in the embodiment provided by the present invention.

[0020] Figure 3 It is a curve diagram of the change of hysteresis peak value with sample number at different pattern depths in the embodiment provided by the present invention.

[0021] Figure 4 It is a curve diagram of the change of the leading peak value with the number of samples at different pattern depths in the embodiment provided by the present invention.

[0022] Figure 5 is a heat map of the correlation between different features in the embodiments provided by the present invention.

[0023] Figure 6 Schematic diagram of the relationship between MSE and training rounds in the embodiment provided by the present invention.

[0024] Figure 7 It is a schematic diagram of the prediction results of the depth prediction model on the test set in the embodiment provided by the present invention.

[0025] Figure 8 It is a structural schematic diagram of the tire tread depth prediction device provided by the present invention.

[0026] Fig. 9 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Combine the following Figure 1-Figure 9 The present invention describes a tire tread depth prediction method, device, electronic device and storage medium.

[0029] Figure 1 FIG. 1 is a flow chart of a tire tread depth prediction method according to an exemplary embodiment. Figure 1 As shown, in an exemplary embodiment, the tire tread depth prediction method includes steps 110 to 120, which are described in detail as follows.

[0030] Step 110, obtaining multi-dimensional features of the tire to be predicted.

[0031] In the embodiment of the present invention, the multi-dimensional features are features of the tire to be predicted in different dimensions.

[0032] Step 120, inputting the multi-dimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0033] In an embodiment of the present invention, the Transformer network can capture the relationship between data globally and dynamically adjust the weights of different features through the self-attention mechanism. This enables the Transformer to effectively fuse multi-dimensional features and identify the key factors affecting the tread depth in tire tread depth prediction. The KAN network (Kolmogorov–Arnold Network) is an emerging deep learning architecture that demonstrates superiority in multi-feature data processing with its unique structure and modeling capabilities. Inspired by the Kolmogorov-Arnold representation theorem, the KAN network has a learnable activation function on the edge ("weight"), which can more effectively process nonlinear relationships and complex patterns in the data.

[0034] In the embodiment of the present invention, a deep prediction model is constructed by combining the advantages of the Transformer and KAN networks. The multi-dimensional features of the tire to be predicted are processed through the deep prediction model to obtain the tread depth of the tire to be predicted, and then the tire wear status can be accurately and real-time monitored under different working conditions, providing early warnings to users, thereby improving the driving safety of the vehicle.

[0035] In an exemplary embodiment of the present invention, the hybrid network structure includes a first network structure layer and a second network structure layer connected in sequence, the first network structure layer includes a first self-attention layer, a first residual connection layer, a first KAN network, a fully connected layer and a second residual connection layer connected in sequence, and the second network structure layer includes a second self-attention layer, a third residual connection layer, a second KAN network and a third KAN network connected in sequence.

[0036] In the embodiment of the present invention, a hybrid network structure integrating the Transformer network and the KAN network is proposed, specifically integrating the self-attention mechanism in the Transformer network. The specific network architecture is as follows: Figure 2 As shown in the figure, the hybrid network structure is mainly composed of a self-attention layer, a KAN network, a fully connected layer, and a residual connection layer.

[0037] The hybrid network structure includes a first network structure layer and a second network structure layer. The first network structure layer is located at Figure 2 On the left side of the schematic diagram shown, the second network structure layer is located at Figure 2 The right side of the schematic diagram shown. In the first network structure layer, the first self-attention layer is used to capture the global dependencies between the multi-dimensional features of the input. Through the self-attention mechanism, the deep prediction model can effectively identify the complex associations between the multi-dimensional features of the input. Especially in terms of time series data, the attention mechanism can capture the dynamic relationship of features changing over time. The output of the first self-attention layer is added to the multi-dimensional features of the input through the first residual connection layer, so as to enhance the feature expression ability and avoid information loss while maintaining the original multi-dimensional features. This jump connection also helps to alleviate the gradient disappearance problem.

[0038] Subsequently, the features after residual connection enter the first KAN network. The KAN network replaces the Multi-Layer Perceptron (MLP) layer in the traditional Transformer network. The KAN network has a powerful nonlinear feature mapping capability, which can more accurately fit and transform the input features in a low-dimensional space. Since tire tread wear has certain nonlinear characteristics, the introduction of the KAN network enables the deep prediction model to better capture and describe this complex nonlinear relationship, thereby improving the accuracy of the prediction. The KAN network can effectively reduce overfitting during the training process, improve the generalization ability of the model, and enable it to maintain high-precision prediction performance under different working conditions.

[0039] In the first network structure layer, the output of the first KAN network is further passed through a fully connected layer, which linearly combines the input features and further compresses the dimensions to enhance the deep prediction model's ability to express features. The output of the fully connected layer is input to the second residual connection layer, which is residually connected with the input multi-dimensional features for fusion, further improving the nonlinear fitting ability of the deep prediction model.

[0040] In the second network structure layer, the output features of the first network structure layer are used as the input features of the second self-attention layer in the second network structure layer. The second self-attention layer is used to capture the global dependency between its input features. The output of the second self-attention layer is then added to the input features of the second self-attention layer through the third residual connection layer, so as to enhance the feature expression capability and avoid information loss while maintaining the original input features.

[0041] In the second network structure layer, a stacked structure of two layers of KAN networks is adopted. This design enables the model to perform deeper feature extraction at different feature levels. The output features of the third residual connection layer are passed through two KAN nonlinear mappings, so that the second network structure layer can capture more complex feature patterns, further improving the performance of the deep prediction model.

[0042] In the Transformer attention layer, the key, query, and value vectors are calculated as: ; in, , , are the weight matrices for query, key, and value, respectively, and x is the corresponding input feature.

[0043] The attention weight is calculated as: ; in, is the dimension of the key vector.

[0044] In the structure of the depth prediction model of the present invention, residual connection is used, which can effectively alleviate the gradient vanishing problem in the deep network and improve the training efficiency and performance of the model. Its mathematical expression is: ; in, represents the nonlinear transformation of the network layer, is the corresponding input feature, is the output after residual connection. The residual connection directly adds the input to the output, which enhances the information flow by retaining the original information of the input, making the network more stable and accurate in complex feature mapping tasks.

[0045] The KAN network is a new neural network structure based on the Kolmogorov-Arnold representation theorem. Unlike the traditional multi-layer perceptron model, the KAN network places learnable activation functions at the edge of the network rather than in the deep layers of the network. These activation functions are automatically adjusted according to the training data, allowing the network to remain flexible and reduce the need to implement complex network structures. Its mathematical expression is as follows: ; Among them, x is the corresponding n-dimensional input vector, and the internal function The domain is usually [0,1], and its range is the set of real numbers. , external function The domain and range are , the activation function related expression is: ; ; ; in, and are all learnable parameters, and b(x) is the bias activation function, which is initialized to the silu activation function. is a set of one-dimensional function combinations, is a predefined Base-spline basis function. During the training process, the spline parameters Continuously optimize to adjust the spline shape to fit the training data.

[0046] In an exemplary embodiment of the present invention, the step of obtaining the multi-dimensional features of the tire to be predicted includes the following steps, which are described in detail as follows.

[0047] Acquire an acceleration signal corresponding to the tire to be predicted.

[0048] Dynamic load, peak acceleration, contact length ratio, rolling period, lag peak, and lead peak are calculated based on the acceleration signal.

[0049] The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, and the leading peak.

[0050] In the embodiment of the present invention, Figure 3 The variation curves of the hysteresis peak value with the number of samples at different pattern depths (2mm, 3.5mm, 5mm, 6mm) are shown. The variation curves at different depths show obvious trend changes, indicating that the pattern depth has a significant effect on the hysteresis peak characteristics. Figure 4 The graph shows the changing curve of the leading peak under different pattern depths. The amplitudes of the changing curves under different depths are different, especially the amplitudes of 3.5mm and 6mm pattern depths are significantly different, which shows that the leading peak is very sensitive to changes in pattern depth and can reflect the wear characteristics of the tire.

[0051] A sensor is installed in the tire, which integrates and packages a micro-electro-mechanical system (MEMS), an accelerometer, a power supply module, and a data storage module. The sampling frequency of the sensor is 5000 Hz.

[0052] The acceleration signal corresponding to the tire to be predicted is obtained by the accelerometer. Assume that the acceleration signals of the x-axis and z-axis of the tire to be predicted intercepted from each cycle are and According to the acceleration signal, the dynamic load C1, the peak acceleration C2, the contact length ratio C3, the rolling period C4, the lag peak C5 and the leading peak C6 are calculated.

[0053] Specifically, the dynamic load C1 is calculated by the following formula: ; in, For each rotation cycle of the tire to be predicted The mean of represents the average vertical acceleration over a period of time, and R is the radius of the tire to be predicted.

[0054] The peak acceleration C2 is calculated by the following formula: ; in, For each rotation cycle of the tire to be predicted Maximum value of the signal.

[0055] The ground length ratio C3 is calculated by the following formula: ; Where T is The duration of one complete rotation cycle of the signal grounding process.

[0056] The rolling period C4 can be determined by analyzing the peak and valley time of the acceleration signal, and is therefore calculated using the following formula: ; in, and Respectively The occurrence time of the peak and valley points of the signal in the i-th rotation cycle.

[0057] The lagging peak value C5 and the leading peak value C6 are calculated by the following formulas respectively: ; .

[0058] In an exemplary embodiment of the present invention, the multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak and the leading peak, including the following steps, which are described in detail as follows.

[0059] Obtain tire information corresponding to the tire to be predicted; wherein the tire information includes absolute air pressure.

[0060] The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, the lead peak and the tire information.

[0061] In the embodiment of the present invention, Figure 5 It shows the average correlation heat map of different features of the tire pattern, among which C2 has the strongest correlation with C5 and C6, with correlation coefficients reaching -0.95 and 0.96 respectively, indicating that C2 of the tire has a significant impact on the peak characteristics of the acceleration signal. The correlation between C5 and C6 reaches -0.98, which are interrelated characteristic parameters. In addition, C2 also has a strong correlation with features such as C4, C1 and C3, which are -0.98 and 0.93 respectively, indicating that an increase in C2 may cause C4 to shorten, and an increase in C1 will cause C3 to increase. Therefore, multidimensional features are obtained based on C1, C2, C3, C4, C5, C6 and absolute air pressure. In the embodiment of the present invention, the above-mentioned 7 features are used as the input features of the depth prediction model to predict the tread depth of the tire to be predicted, and then the wear depth of the tire pattern is predicted, thereby improving the accuracy and stability of tire state prediction.

[0062] In an exemplary embodiment of the present invention, the process of constructing the depth prediction model includes the following steps, which are described in detail as follows.

[0063] A training set and a validation set are obtained, and a plurality of tire samples in the training set and the validation set are preprocessed; wherein the tire samples include sample features and sample labels.

[0064] The tire samples in the training set are input into a basic depth prediction model for iterative training until the training is completed, thereby obtaining the depth prediction model; wherein the basic depth prediction model includes a basic hybrid network structure obtained by fusion of a Transformer network and a KAN network.

[0065] The depth prediction completed through training is verified by using the tire samples in the verification set.

[0066] In the embodiment of the present invention, during the training process, a training set and a validation set are constructed, each of which includes a plurality of tire samples, each of which has a corresponding sample feature and a sample label, the sample feature is a multidimensional feature consistent with the above, and the sample label is the corresponding tire tread depth. Each tire sample is preprocessed, and the preprocessed sample features in the training set are input into the basic depth prediction model for iterative training. In each iteration, the obtained prediction result is compared with the actual sample label to determine the error between the two, and then the model parameters of the basic depth prediction model are optimized based on the error.

[0067] In this embodiment of the present invention, the mean square error (MSE) is used as a loss function to calculate the error between the prediction result and the sample label. The mean square error can be expressed by the following formula: ; in, is the sample label, To predict the results, is the number of samples.

[0068] Figure 6 The trend of the loss function (mean square error MSE) with the number of training rounds during model training is shown. It can be seen that with the increase of training rounds, the training loss of the model gradually decreases and tends to be stable, indicating that the model gradually improves the prediction accuracy during the parameter optimization process, verifying the convergence of the depth prediction model provided by the present invention in the pattern depth prediction task.

[0069] Figure 7 The comparison between the actual tread depth value and the predicted value based on the trained depth prediction model is shown. The black dots in the figure represent the actual measured tire tread depth values, and the gray dots are the prediction results of the model. It can be observed from the figure that the distribution of the predicted value and the actual value are relatively close, indicating that the depth prediction model can effectively capture the tread depth characteristics of different samples and has a high prediction accuracy. At the same time, the distribution stability of the predicted value between samples of different depths further proves the adaptability of the depth prediction model under various working conditions. In addition, the prediction error of the depth prediction model is further quantified by calculating the root mean square error (RMSE). The RMSE value of the depth prediction model is 0.5949, indicating that the deviation between the predicted result and the actual value is small, reflecting a high accuracy. The low RMSE value verifies that the model can accurately predict the tire tread depth in actual working conditions, providing strong technical support for real-time monitoring of tire status.

[0070] In an exemplary embodiment of the present invention, the preprocessing of the plurality of tire samples in the training set and the validation set includes the following steps, which are described in detail as follows.

[0071] The sample features and the sample labels of each of the tire samples are normalized.

[0072] In the embodiment of the present invention, it is necessary to perform normalization processing on the features, and then use RMSE as an evaluation index. After regression prediction, the predicted pattern depth is denormalized to obtain the pattern depth.

[0073] The normalization formula of feature x is expressed as: ; in, is the mean of feature x, is the standard deviation of feature x.

[0074] When training a deep prediction model, both sample features and sample labels need to be normalized. Then, after obtaining the predicted value, the predicted value needs to be denormalized. The expression is: ; in, is the predicted value, is the mean tire tread depth, is the standard deviation of tire tread depth.

[0075] In an exemplary embodiment of the present invention, after inputting the multi-dimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted, the method includes the following steps, which are described in detail as follows.

[0076] Based on the predicted pattern depth, a prediction error of the depth prediction model is calculated by a root mean square error formula, and the depth prediction model is evaluated based on the prediction error.

[0077] In the embodiment of the present invention, the root mean square error (RMSE) formula is: The error between the predicted tread depth and the actual tread depth is calculated by using the root mean square error formula to further quantify the prediction error of the depth prediction model. The lower prediction error indicates that the depth prediction model can accurately predict the tire tread depth in actual working conditions, providing strong technical support for real-time monitoring of tire status.

[0078] The tire tread depth prediction device provided by the present invention is described below. The tire tread depth prediction device described below and the tire tread depth prediction method described above can be referred to each other. It should be noted that the device provided in the following embodiment and the method provided in the above embodiment belong to the same concept, and the specific manner in which each module and unit performs the operation has been described in detail in the method embodiment, which will not be repeated here.

[0079] In an exemplary embodiment of the present invention, see Figure 8 , Figure 8 A tire tread depth prediction device is shown according to an exemplary embodiment, including the following modules.

[0080] A first acquisition module 810 is configured to acquire multi-dimensional features of a tire to be predicted; The prediction module 820 is configured to input the multi-dimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0081] In an exemplary embodiment of the present invention, the hybrid network structure includes a first network structure layer and a second network structure layer connected in sequence, the first network structure layer includes a first self-attention layer, a first residual connection layer, a first KAN network, a fully connected layer and a second residual connection layer connected in sequence, and the second network structure layer includes a second self-attention layer, a third residual connection layer, a second KAN network and a third KAN network connected in sequence.

[0082] In an exemplary embodiment of the present invention, the first acquisition module 810 includes: An acquisition submodule, configured to acquire an acceleration signal corresponding to the tire to be predicted; a calculation submodule configured to calculate a dynamic load, a peak acceleration, a contact length ratio, a rolling period, a lag peak, and a lead peak based on the acceleration signal; The multi-dimensional feature submodule is configured to obtain the multi-dimensional feature based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak value and the leading peak value.

[0083] In an exemplary embodiment of the present invention, the multidimensional feature submodule includes: An acquisition unit, configured to acquire tire information corresponding to the tire to be predicted; wherein the tire information includes absolute air pressure; The multi-dimensional feature unit is configured to obtain the multi-dimensional feature based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, the leading peak and the tire information.

[0084] In an exemplary embodiment of the present invention, a tire tread depth prediction device comprises: A second acquisition module is configured to acquire a training set and a validation set, and preprocess a plurality of tire samples in the training set and the validation set; wherein the tire samples include sample features and sample labels; A training module, configured to input the tire samples in the training set into a basic depth prediction model for iterative training until the training is completed, thereby obtaining the depth prediction model; wherein the basic depth prediction model includes a basic hybrid network structure obtained by fusion of a Transformer network and a KAN network; A verification module is configured to verify the depth prediction completed through training using the tire samples in the verification set.

[0085] In an exemplary embodiment of the present invention, the second acquisition module includes: The normalization processing submodule is configured to perform normalization processing on the sample features and the sample labels of each of the tire samples.

[0086] In an exemplary embodiment of the present invention, the tire tread depth prediction device further includes: The evaluation module is configured to calculate the prediction error of the depth prediction model through a root mean square error formula based on the predicted pattern depth, and evaluate the depth prediction model based on the prediction error.

[0087] Fig. 9 An example of a physical structure diagram of an electronic device is shown in FIG. Fig. 9 As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930 and a communication bus 940, wherein the processor 910, the communication interface 920 and the memory 930 communicate with each other via the communication bus 940. The processor 910 may call the logic instructions in the memory 930 to execute the tire tread depth prediction method, the method comprising: obtaining multi-dimensional features of the tire to be predicted; The multi-dimensional features are input into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0088] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0089] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the tire tread depth prediction method provided by the above methods, the method includes: obtaining multi-dimensional features of the tire to be predicted; The multi-dimensional features are input into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0090] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being implemented when executed by a processor to perform the tire tread depth prediction method provided by the above methods, the method comprising: obtaining multi-dimensional features of a tire to be predicted; The multi-dimensional features are input into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

[0091] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0092] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A tire tread depth prediction method, characterized in that: include: Obtaining multi-dimensional features of the tire to be predicted; The multi-dimensional features are input into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

2. The tire tread depth prediction method according to claim 1, characterized in that: The hybrid network structure includes a first network structure layer and a second network structure layer connected in sequence, the first network structure layer includes a first self-attention layer, a first residual connection layer, a first KAN network, a fully connected layer and a second residual connection layer connected in sequence, and the second network structure layer includes a second self-attention layer, a third residual connection layer, a second KAN network and a third KAN network connected in sequence.

3. The tire tread depth prediction method according to claim 1, characterized in that: The step of obtaining the multi-dimensional features of the tire to be predicted includes: Acquiring an acceleration signal corresponding to the tire to be predicted; calculating a dynamic load, a peak acceleration, a contact length ratio, a rolling period, a lag peak, and a lead peak based on the acceleration signal; The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, and the leading peak.

4. The tire tread depth prediction method according to claim 3, characterized in that: The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the hysteresis peak and the leading peak, including: Acquire tire information corresponding to the tire to be predicted; wherein the tire information includes absolute air pressure; The multi-dimensional feature is obtained based on the dynamic load, the peak acceleration, the contact length ratio, the rolling period, the lag peak, the lead peak and the tire information.

5. The tire tread depth prediction method according to claim 1, characterized in that: The construction process of the depth prediction model includes: Obtaining a training set and a validation set, and preprocessing a plurality of tire samples in the training set and the validation set; wherein the tire samples include sample features and sample labels; Inputting the tire samples in the training set into a basic depth prediction model for iterative training until the training is completed, thereby obtaining the depth prediction model; wherein the basic depth prediction model includes a basic hybrid network structure obtained by fusion of a Transformer network and a KAN network; The depth prediction completed through training is verified by using the tire samples in the verification set.

6. The tire tread depth prediction method according to claim 5, characterized in that: The preprocessing of the plurality of tire samples in the training set and the validation set comprises: The sample features and the sample labels of each of the tire samples are normalized.

7. The tire tread depth prediction method according to any one of claims 1 to 6, characterized in that: After inputting the multi-dimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted, the method comprises: Based on the predicted pattern depth, a prediction error of the depth prediction model is calculated by a root mean square error formula, and the depth prediction model is evaluated based on the prediction error.

8. A tire tread depth prediction device, characterized in that: include: A first acquisition module, configured to acquire multi-dimensional features of a tire to be predicted; A prediction module is configured to input the multidimensional features into a preset depth prediction model to predict the tread depth of the tire to be predicted; wherein the depth prediction model includes a hybrid network structure, and the hybrid network structure is obtained by fusion of the self-attention mechanism of the Transformer network and the KAN network.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the tire tread depth prediction method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the tire tread depth prediction method according to any one of claims 1 to 7 is implemented.

Images