Tire pattern wear estimation method and system based on time-frequency diagram feature fusion of intelligent tire array sensor

Abstract

The invention provides a tire pattern wear estimation method and system based on intelligent tire array sensor time-frequency diagram feature fusion. The method comprises the following steps: forming a sensor array by adopting an acceleration sensor and a piezoelectric sensor, and collecting one-dimensional time sequence signals of tires with different wear degrees under various working conditions; receiving the collected one-dimensional time sequence signal, and transmitting the collected one-dimensional time sequence signal to an upper computer; the upper computer converts a one-dimensional time sequence signal into a two-dimensional time-frequency diagram by adopting a wavelet so as to realize fusion of time domain and frequency domain characteristics of the acceleration sensor and the piezoelectric sensor; and designing a convolutional neural network model, inputting the time-frequency graph generated by the multi-source sensor into the convolutional neural network model in a multi-channel form for training, performing tire pattern wear classification by using the trained convolutional neural network model, automatically extracting features in the two-dimensional time-frequency graph, and obtaining a tire pattern wear value. And more accurate estimation of the wear of the tire pattern is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of tire tread wear, and in particular to a tire tread wear estimation method based on time-frequency graph feature fusion of an intelligent tire array sensor. Background Art

[0002] Tire tread wear is a key factor affecting vehicle safety and performance. As tires age, their tread depth decreases, significantly reducing their grip, drainage, and handling stability, increasing the risk of traffic accidents. Therefore, accurate and real-time monitoring of tire tread wear is crucial for ensuring driving safety and optimizing tire life. In recent years, intelligent tire technology has rapidly developed. By embedding sensors within the tire, dynamic information about the tire's contact with the road can be acquired in real time, providing a new technical approach for accurately estimating tire tread wear.

[0003] Currently, methods for estimating tire tread wear mainly include finite element simulation analysis, lidar scanning of the tread, and machine vision image processing detection. In addition, smart tire technology based on embedded sensors is also gradually emerging. Among them, acceleration sensors and PVDF piezoelectric sensors are widely used in tire tread wear estimation due to their performance advantages. Acceleration sensors can effectively capture changes in tire acceleration during driving and are insensitive to tire deformation, while PVDF piezoelectric sensors are sensitive to changes in tire load but not to changes in speed. These sensors provide data support for estimating the degree of tire tread wear by collecting tire response signals under different working conditions. However, the existing technology usually only performs time domain or frequency domain analysis on the signal collected by a single sensor to achieve the estimation of the degree of tire tread wear.

[0004] Although existing sensor technologies and analysis methods have made certain progress in tire tread wear estimation, there are still some limitations. First, when using accelerometers or PVDF piezoelectric sensors alone, due to the limitations of the sensor's own characteristics, it is impossible to fully reflect the wear of the tire under different working conditions. For example, accelerometers are sensitive to vehicle speed but not to load, while piezoelectric sensors are the opposite, which leads to differences in the estimation accuracy of a single sensor under different working conditions. Secondly, the analysis of sensor signals in the existing technology is mostly limited to separate analysis in the time domain or frequency domain, and fails to fully utilize the complementarity of time-frequency domain features. This single analysis method cannot effectively integrate the advantages of different types of sensors, thereby limiting the accuracy and robustness of tire tread wear estimation. Therefore, how to effectively fuse accelerometers and PVDF piezoelectric sensors and perform joint analysis in the time-frequency domain to improve the accuracy and reliability of tire tread wear estimation is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] To address the aforementioned technical issues, a tire tread wear estimation method and system based on time-frequency feature fusion of intelligent tire array sensors is provided. This invention extracts features from signals from acceleration sensors and PVDF piezoelectric sensors, then fuses and trains these features to produce more accurate tire tread wear values, enabling a more precise estimation of tire tread wear.

[0006] The technical means adopted in the present invention are as follows:

[0007] A tire tread wear estimation method based on time-frequency graph feature fusion of intelligent tire array sensors, comprising:

[0008] S1. Using acceleration sensors and piezoelectric sensors to form a sensor array, collecting one-dimensional time series signals of tires with different degrees of wear under various working conditions;

[0009] S2, receiving the collected one-dimensional time series signal, and transmitting the collected one-dimensional time series signal to the host computer;

[0010] S3, the host computer uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency diagram, realizing the fusion of the time domain and frequency domain characteristics of the acceleration sensor and piezoelectric (PVDF) sensor;

[0011] S4. Design a convolutional neural network (CNN) model, and fuse the two-dimensional time-frequency graphs generated by multi-source sensors into the CNN model for training. Use the trained CNN model to classify wear, automatically extract features from the two-dimensional time-frequency graph, and obtain the tire tread wear value.

[0012] Furthermore, step S1 specifically includes:

[0013] Arrange the acceleration sensor and the piezoelectric sensor linearly along the tread grooves and close to the central area of ​​the tire contact surface;

[0014] Lead wires through the valve stem of the wheel hub, connect the acceleration sensor and the piezoelectric sensor to the signal acquisition device respectively, and use a mobile power supply to power the signal acquisition device;

[0015] Collect one-dimensional time series signals of tires with different degrees of wear under various working conditions.

[0016] Furthermore, step S2 specifically includes:

[0017] The collected one-dimensional timing signal is sent to the router through the FPC soft antenna and the local area network;

[0018] The router then transmits the one-dimensional timing signal to the host computer on the PC.

[0019] Furthermore, step S3 specifically includes:

[0020] Based on the collected one-dimensional time series signals, including acceleration signals and piezoelectric signals, a third-order Butterworth filter is used for filtering. An empirical expression of the tire rolling speed corresponding to the filter's -3dB frequency is fitted to eliminate incomplete cycles in the data.

[0021] The complex Morlet wavelet is used to perform continuous wavelet transform (CWT) on the filtered signal to extract the time-frequency feature map containing the instantaneous frequency and amplitude changes. The expression of the complex Morlet wavelet is as follows:

[0022]

[0023] in, represents the wavelet function, f b represents the bandwidth parameter, f c represents the center frequency, and t represents the time variable.

[0024] Furthermore, in step S4, the designed convolutional neural network model includes 3 convolutional layers, 3 pooling layers, 1 flattening layer, and 2 fully connected layers, where:

[0025] The convolution kernel size in the convolution layer is 3×3, the pooling layer performs the maximum pooling operation, the pooling window size is 2×2, and the number of convolution kernels in the three convolution layers is 32, 64, and 128, respectively. The maximum pooling operation is performed after each convolution layer, and then the flattening operation is performed to process the multi-dimensional tensor. Finally, it is input to the fully connected layer to convert the extracted features into the final classification results.

[0026] Furthermore, in step S4, the training process of the convolutional neural network model is as follows:

[0027] The collected data were preprocessed and divided into training set and validation set using stratified sampling method with a ratio of 8:2;

[0028] The fused two-dimensional time-frequency graphs of the six channels are input into the convolutional neural network model, and the convolutional neural network model is trained using the back-propagation algorithm.

[0029] By gradually increasing the number of training rounds, the trend of the loss function of the training set is monitored, and the hyperparameters of the convolutional neural network model are adjusted according to the training results. Combined with Batch Normalization processing, ReLU activation function, Adam optimizer and CrossEntropyLoss loss function, network training is completed on the PyTorch platform.

[0030] Furthermore, in step S4, the process of using the trained convolutional neural network model to perform wear classification is as follows:

[0031] The trained convolutional neural network model is used to automatically extract features and classify the samples in the verification set to identify the degree of tire tread wear.

[0032] The present invention also provides a tire tread wear estimation system based on the tire tread wear estimation method and time-frequency map feature fusion of intelligent tire array sensors, comprising: a data acquisition module, a data receiving module, a feature fusion module, and a wear identification module, wherein:

[0033] The data acquisition module uses an acceleration sensor and a piezoelectric sensor to form a sensor array to collect one-dimensional time series signals of tires with different degrees of wear under various working conditions;

[0034] The data receiving module receives the collected one-dimensional time series signal and transmits the collected one-dimensional time series signal to the host computer;

[0035] The feature fusion module uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency diagram, realizing the fusion of the time domain and frequency domain features of the acceleration sensor and the piezoelectric (PVDF) sensor;

[0036] The wear recognition module designs a convolutional neural network model, inputs the two-dimensional time-frequency graph into the convolutional neural network model for training, uses the trained convolutional neural network model to perform wear classification, automatically extracts features from the two-dimensional time-frequency graph, and obtains the tire tread wear value.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] 1. The tire tread wear estimation method provided by this invention utilizes signals from an accelerometer and a PVDF piezoelectric sensor, leveraging the strengths of both sensors under different operating conditions. This overcomes the limitations of single-sensor tire wear estimation accuracy. The accelerometer is sensitive to changes in vehicle speed, while the PVDF piezoelectric sensor is sensitive to changes in load. This combination of the two more comprehensively reflects the forces acting on the tire during actual driving, significantly improving the accuracy of tire tread wear estimation.

[0039] 2. The tire tread wear estimation method provided by the present invention adopts a method of joint analysis of time and frequency domain features, which can extract characteristic information of tire wear signals from multiple dimensions. Compared with single time domain or frequency domain analysis, it can more effectively capture subtle changes in tire wear and further improve estimation accuracy.

[0040] 3. The tire tread wear estimation method provided by the present invention uses a CNN neural network to automatically learn and adapt to tire wear characteristics under different working conditions. It has stronger adaptability to complex actual driving environments (such as different road types, vehicle loads, etc.) and enhances the robustness of the system.

[0041] 4. The tire tread wear estimation method provided by this invention uses a sensor array to collect real-time dynamic data about tires during driving. Combined with the rapid processing capabilities of a neural network, this method enables real-time monitoring and dynamic response to tire tread wear. This means that tire wear information can be obtained promptly while the vehicle is driving, providing early warning of potential safety hazards and providing more reliable safety protection for the driver.

[0042] Based on the above reasons, the present invention can be widely promoted in the fields of tire tread wear and the like. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 any creative labor.

[0044] Figure 1 Flow chart of the method of the present invention.

[0045] Figure 2 This is the CNN neural network structure diagram of the present invention.

[0046] Figure 3 The z-axis acceleration waveform and its corresponding time-frequency diagram provided by the embodiment of the present invention.

[0047] Figure 4 The PVDF original signal waveform and its corresponding time-frequency diagram provided by the embodiment of the present invention

[0048] Figure 5 This is the system architecture diagram of the present invention DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only 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 making creative efforts should fall within the scope of protection of the present invention.

[0050] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or apparatuses.

[0051] like Figure 1 As shown, the present invention provides a tire tread wear estimation method based on time-frequency graph feature fusion of intelligent tire array sensors, comprising:

[0052] S1. Using acceleration sensors and piezoelectric sensors to form a sensor array, collecting one-dimensional time series signals of tires with different degrees of wear under various working conditions;

[0053] S2, receiving the collected one-dimensional time series signal, and transmitting the collected one-dimensional time series signal to the host computer;

[0054] S3, the host computer uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency diagram, realizing the fusion of the time domain and frequency domain characteristics of the acceleration sensor and piezoelectric (PVDF) sensor;

[0055] S4. Design a convolutional neural network (CNN) model and fuse the two-dimensional time-frequency graphs of multi-source sensors into the CNN model for training. Use the trained CNN model to classify wear and automatically extract features from the two-dimensional time-frequency graph to obtain the tire tread wear value.

[0056] In specific implementation, as a preferred embodiment of the present invention, step S1 specifically includes:

[0057] Accelerometers and piezoelectric sensors are arranged linearly along the tread grooves, near the center of the tire contact surface. In this embodiment, the PVDF sensor is located on the left side, and the accelerometer on the right side, secured with tape. Sensors are located every 90 degrees within the tire. Embedded within the tire, these sensors do not affect normal operation and are resistant to damage and drop, resulting in high durability. This sensor placement improves the real-time accuracy of tire wear estimation.

[0058] Lead wires through the valve stem of the wheel hub, connect the acceleration sensor and the piezoelectric sensor to the signal acquisition device respectively, and use a mobile power supply to power the signal acquisition device;

[0059] One-dimensional time series signals are collected from tires with varying degrees of wear under various operating conditions. In this embodiment, sensor signal data is collected at vehicle speeds of 20 km / h, 40 km / h, and 60 km / h and loads of 4 kN, 5 kN, and 6 kN. Each frame of 1024 data points is normalized.

[0060] In specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes:

[0061] The collected one-dimensional timing signal is sent to the router through the FPC soft antenna and the local area network. In this embodiment, wireless transmission is used to avoid the abnormal situation caused by loose interfaces that may be caused by wired transmission, and can also reduce energy consumption.

[0062] The router then transmits the one-dimensional timing signal to the host computer on the PC.

[0063] In specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes:

[0064] Based on the collected one-dimensional time series signals, including acceleration signals and piezoelectric signals, a third-order Butterworth filter is used for filtering. The empirical expression of the tire rolling speed corresponding to the filter -3dB frequency is fitted. Incomplete cycles in the data are eliminated, and the cutoff frequency is 2 times the vehicle speed divided by the sampling frequency. Thus, a relatively complete acceleration waveform is obtained after processing, as shown in the following figure: Figure 3 The PVDF waveform is good and does not require filtering. The complete acceleration waveform is shown in Figure 4 .

[0065] The complex Morlet wavelet is used to perform continuous wavelet transform (CWT) on the filtered signal to extract the time-frequency feature map containing the instantaneous frequency and amplitude changes. The expression of the complex Morlet wavelet is as follows:

[0066]

[0067] in, represents the wavelet function, f b represents the bandwidth parameter, f c represents the center frequency, and t represents the time variable. In this embodiment, wavelet transform is used to generate 128×128 time-frequency plots for the time series signals of the PVDF sensor and the accelerometer sensor, respectively, to extract the time-frequency features of the non-stationary signals. Due to the differences in the responses of the two types of sensors to speed and load, the frequency domain distributions differ. However, a unified frequency scale is used to standardize the time-frequency plots, facilitating subsequent feature fusion and neural network input.

[0068] In specific implementation, as a preferred embodiment of the present invention, in step S4, as Figure 2 As shown in Figure 2, the designed convolutional neural network model includes 3 convolutional layers, 3 pooling layers, 1 flattening layer, and 2 fully connected layers, where:

[0069] The convolution kernel size in the convolution layer is 3×3, the pooling layer performs the maximum pooling operation, the pooling window size is 2×2, and the number of convolution kernels in the three convolution layers is 32, 64, and 128, respectively. The maximum pooling operation is performed after each convolution layer, and then the flattening operation is performed to process the multi-dimensional tensor. Finally, it is input to the fully connected layer to convert the extracted features into the final classification results.

[0070] In specific implementation, as a preferred embodiment of the present invention, in step S4, the training process of the convolutional neural network model is as follows:

[0071] The collected data were preprocessed and divided into training set and validation set using stratified sampling method with a ratio of 8:2;

[0072] The fused two-dimensional time-frequency graphs of the six channels are input into the convolutional neural network model, and the convolutional neural network model is trained using the back-propagation algorithm.

[0073] By gradually increasing the number of training rounds, the trend of the loss function of the training set is monitored, and the hyperparameters of the convolutional neural network model are adjusted according to the training results. Combined with Batch Normalization processing, ReLU activation function, Adam optimizer and CrossEntropyLoss loss function, network training is completed on the PyTorch platform.

[0074] In specific implementation, as a preferred embodiment of the present invention, in step S4, the process of wear classification using the trained convolutional neural network model is as follows:

[0075] The trained convolutional neural network model is used to automatically extract features and classify the samples in the verification set to identify the degree of tire tread wear.

[0076] Corresponding to the tire tread wear estimation method in this application, this application also provides a tire tread wear estimation system based on the fusion of time-frequency graph features of intelligent tire array sensors, such as Figure 5 As shown, it includes: a data acquisition module, a data receiving module, a feature fusion module, and a wear identification module, wherein:

[0077] The data acquisition module uses an acceleration sensor and a piezoelectric sensor to form a sensor array to collect one-dimensional time series signals of tires with different degrees of wear under various working conditions;

[0078] The data receiving module receives the collected one-dimensional time series signal and transmits the collected one-dimensional time series signal to the host computer;

[0079] The feature fusion module uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency diagram, realizing the fusion of the time domain and frequency domain features of the acceleration sensor and the piezoelectric (PVDF) sensor;

[0080] The wear recognition module designs a convolutional neural network model, inputs the two-dimensional time-frequency graph into the convolutional neural network model for training, uses the trained convolutional neural network model to perform wear classification, automatically extracts features from the two-dimensional time-frequency graph, and obtains the tire tread wear value.

[0081] In summary, the advantages of the present invention include: First, by deploying accelerometers and piezoelectric sensors to collect tire data, the multi-dimensional data provides a more comprehensive research perspective. Second, the fusion of the data features of the two sensors improves the accuracy of tire wear estimation. Third, the fusion of the sensor's time-frequency domain features can simultaneously reveal the local characteristics of the signal in the time and frequency dimensions, enabling a comprehensive understanding of the signal's dynamic characteristics. Fourth, this high-precision tire wear estimation is crucial for improving the performance, durability, and safety of tire wear assessment.

[0082] 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 above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors, characterized in that: include: S1. Using acceleration sensors and piezoelectric sensors to form a sensor array, collecting one-dimensional time series signals of tires with different degrees of wear under various working conditions; S2, receiving the collected one-dimensional time series signal, and transmitting the collected one-dimensional time series signal to the host computer; S3. The host computer uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency graph, realizing the fusion of the time domain and frequency domain characteristics of the acceleration sensor and piezoelectric sensor; S4. Design a convolutional neural network model and input the multi-channel fusion of the two-dimensional time-frequency graph obtained by multi-source sensors into the convolutional neural network model for training. Use the trained convolutional neural network model to perform wear classification, automatically extract features from the two-dimensional time-frequency graph, and obtain the tire tread wear value.

2. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 1, characterized in that: Step S1 specifically includes: Arrange the acceleration sensor and the piezoelectric sensor linearly along the tread grooves and close to the central area of ​​the tire contact surface; Lead the wires through the valve of the wheel hub, connect the acceleration sensor and the piezoelectric sensor to the signal acquisition device respectively, and use a mobile power supply to power the signal acquisition device; Collect one-dimensional time series signals of tires with different degrees of wear under various working conditions.

3. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 1, characterized in that: Step S2 specifically includes: The collected one-dimensional timing signal is sent to the router through the FPC soft antenna and the local area network; The router then transmits the one-dimensional timing signal to the host computer on the PC.

4. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 1, characterized in that: Step S3 specifically includes: Based on the collected one-dimensional time series signals, including acceleration signals and piezoelectric signals, a third-order Butterworth filter is used for filtering. An empirical expression of the tire rolling speed corresponding to the filter's -3dB frequency is fitted to eliminate incomplete cycles in the data. The complex Morlet wavelet is used to perform continuous wavelet transform on the filtered signal to extract the time-frequency feature map containing the instantaneous frequency and amplitude changes. The expression of the complex Morlet wavelet is as follows: in, represents the wavelet function, f b represents the bandwidth parameter, f c represents the center frequency, and t represents the time variable.

5. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 1, characterized in that: In step S4, the designed convolutional neural network model includes 3 convolutional layers, 3 pooling layers, 1 flattening layer, and 2 fully connected layers, where: The convolution kernel size in the convolution layer is 3×3, the pooling layer performs the maximum pooling operation, the pooling window size is 2×2, and the number of convolution kernels in the three convolution layers is 32, 64, and 128, respectively. The maximum pooling operation is performed after each convolution layer, and then the flattening operation is performed to process the multi-dimensional tensor. Finally, it is input to the fully connected layer to convert the extracted features into the final classification results.

6. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 1, characterized in that: In step S4, the training process of the convolutional neural network model is as follows: The collected data were preprocessed and divided into training set and validation set using stratified sampling method with a ratio of 8:2; The fused two-dimensional time-frequency graphs of the six channels are input into the convolutional neural network model, and the convolutional neural network model is trained using the back-propagation algorithm. By gradually increasing the number of training rounds, the trend of the loss function of the training set is monitored, and the hyperparameters of the convolutional neural network model are adjusted according to the training results. Combined with Batch Normalization processing, ReLU activation function, Adam optimizer and CrossEntropyLoss loss function, network training is completed on the PyTorch platform.

7. The tire tread wear estimation method based on time-frequency feature fusion of intelligent tire array sensors according to claim 6, characterized in that: In step S4, the process of using the trained convolutional neural network model to perform wear classification is as follows: The trained convolutional neural network model is used to automatically extract features and classify the samples in the verification set to identify the degree of tire tread wear.

8. A tire tread wear estimation system based on time-frequency map feature fusion of intelligent tire array sensors implemented based on the tire tread wear estimation method according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, data receiving module, feature fusion module, wear identification module, including: The data acquisition module uses an acceleration sensor and a piezoelectric sensor to form a sensor array to collect one-dimensional time series signals of tires with different degrees of wear under various working conditions; The data receiving module receives the collected one-dimensional time series signal and transmits the collected one-dimensional time series signal to the host computer; The feature fusion module uses wavelet to convert the one-dimensional time series signal into a two-dimensional time-frequency graph, realizing the fusion of the time domain and frequency domain features of the acceleration sensor and the piezoelectric sensor; The wear recognition module designs a convolutional neural network model and inputs the multi-channel fusion of the two-dimensional time-frequency graph obtained by multi-source sensors into the convolutional neural network model for training. The trained convolutional neural network model is used to classify wear, automatically extract features from the two-dimensional time-frequency graph, and obtain the tire tread wear value.

Images