Methods, devices and vehicles for estimating tire wear depth

By employing multi-strategy parallel computing and adaptive weighting of the fusion model, the problem of insufficient accuracy in tire wear estimation is solved, achieving high-precision and reliable estimation under all operating conditions.

CN123104283APending Publication Date: 2026-11-17ZHEJIANG GEELY HLDG GRP CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611482599.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing single-strategy tire wear estimation schemes lack sufficient accuracy across the entire operating range, failing to meet the actual needs for precise tire wear monitoring.

Method used

Multiple independent estimation strategies based on different physical mechanisms are employed to compute wear depth estimates and uncertainty parameters in parallel. The results are then adaptively weighted and fused using a pre-trained fusion model to output the fused wear depth estimates and uncertainty parameters.

Benefits of technology

It improves the accuracy and reliability of tire wear depth estimation, overcomes the shortcomings of insufficient accuracy of a single strategy across the entire operating range, and achieves stable estimation under different operating conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104283A_ABST
    Figure CN123104283A_ABST
Patent Text Reader

Abstract

This application relates to a method, apparatus, and vehicle for estimating tire wear depth. The method includes: acquiring multi-source signals during vehicle operation; based on the multi-source signals, employing at least two independent estimation strategies to calculate wear depth estimates and uncertainty parameters corresponding to each estimation strategy in parallel; and inputting the wear depth estimates and uncertainty parameters corresponding to each estimation strategy into a pre-trained fusion model to obtain a fused wear depth estimate and fused uncertainty parameters. This application overcomes the shortcomings of insufficient accuracy of a single strategy across all operating conditions by employing at least two independent estimation strategies based on different physical mechanisms to estimate wear depth and uncertainty parameters in parallel, and improves estimation accuracy after fusion.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of vehicle intelligent technology, and in particular to a method, device and vehicle for estimating tire wear depth. Background Technology

[0002] Tire tread wear depth is a core indicator for assessing the remaining lifespan and safety performance of a tire. Excessive wear significantly reduces tire grip, increases braking distance, and raises the risk of tire blowout, directly impacting vehicle driving safety. Therefore, accurately estimating tire tread wear depth during vehicle operation is crucial for ensuring driving safety and optimizing tire maintenance strategies.

[0003] Existing technologies have developed various indirect tire wear estimation methods based on onboard sensor signals. These methods are based on different physical mechanisms, utilizing different physical phenomena generated during tire wear to estimate the degree of wear. Theoretically, any single strategy can provide a certain level of wear estimation accuracy under its effective operating conditions. However, each estimation strategy has its specific effective operating conditions and error blind spots. For example, it may be highly accurate under certain driving conditions, but may produce significant deviations under other conditions due to physical model failure or signal interference. Because a single strategy relies on a single physical mechanism, its applicability is inherently limited, and it cannot maintain stable estimation accuracy across the entire operating range. The estimation error is usually insufficient to meet the actual needs of precise tire wear monitoring.

[0004] There is currently no effective solution to the problem of insufficient accuracy of single-strategy estimation schemes in related technologies. Summary of the Invention

[0005] In view of this, this application provides a tire wear depth estimation method, apparatus and vehicle that can integrate multiple estimation strategies to estimate tire tread wear depth and improve estimation accuracy.

[0006] Firstly, this embodiment provides a method for estimating tire wear depth, including:

[0007] Acquire multi-source signals during vehicle operation;

[0008] Based on the multi-source signals, at least two independent estimation strategies are used respectively, and the wear depth estimate and uncertainty parameter corresponding to each estimation strategy are calculated in parallel.

[0009] The wear depth estimate and uncertainty parameter corresponding to each estimation strategy are input into the pre-trained fusion model to obtain the fused wear depth estimate and fusion uncertainty parameter.

[0010] Secondly, this embodiment provides a tire wear depth estimation device, including:

[0011] The signal acquisition module is used to acquire multi-source signals during vehicle operation.

[0012] The parallel estimation module is used to calculate the wear depth estimate and uncertainty parameter corresponding to each of the multi-source signals in parallel, using at least two independent estimation strategies respectively.

[0013] The fusion estimation module is used to input the wear depth estimate and uncertainty parameters corresponding to each of the estimation strategies into the pre-trained fusion model to obtain the fused wear depth estimate and the corresponding fusion uncertainty parameters.

[0014] Thirdly, this embodiment provides a vehicle, including:

[0015] A control device, the control device including a memory, a processor, and a computer program;

[0016] The computer program is stored in the memory and configured to be executed by the processor to implement the tire wear depth estimation method as described in the first aspect.

[0017] Compared with related technologies, the tire wear depth estimation method, device, and vehicle provided in this embodiment acquire multi-source signals during vehicle operation; based on the multi-source signals, at least two independent estimation strategies are employed to calculate the wear depth estimate and uncertainty parameter corresponding to each estimation strategy in parallel; the wear depth estimate and uncertainty parameter corresponding to each estimation strategy are input into a pre-trained fusion model to obtain a fused wear depth estimate and fused uncertainty parameter. Through this embodiment, by employing at least two independent estimation strategies based on different physical mechanisms to estimate the wear depth estimate and uncertainty parameter in parallel, the fusion overcomes the deficiency of insufficient accuracy of a single strategy across the entire operating condition range, thus improving the estimation accuracy.

[0018] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 A flowchart of Embodiment 1 of the tire wear depth estimation method provided in this application;

[0021] Figure 2A flowchart of Embodiment 2 of the tire wear depth estimation method provided in this application;

[0022] Figure 3 A flowchart of Embodiment 3 of the tire wear depth estimation method provided in this application;

[0023] Figure 4 A flowchart of Embodiment 4 of the tire wear depth estimation method provided in this application;

[0024] Figure 5 A flowchart of Embodiment 5 of the tire wear depth estimation method provided in this application;

[0025] Figure 6 A hardware structure diagram of the control equipment containing the tire wear depth estimation device provided in this application;

[0026] Figure 7 This is a schematic diagram of the tire wear depth estimation device provided in this application. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0028] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0029] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0030] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0031] Figure 1 A flowchart illustrating an embodiment of the tire wear depth estimation method provided in this application. (Refer to...) Figure 1The method provided in this embodiment may include:

[0032] Step S101: Acquire multi-source signals during vehicle operation.

[0033] Specifically, multi-source signals are collected through the vehicle's existing sensors. These multi-source signals include: wheel speed signals of each wheel collected by the wheel speed sensors of the Anti-lock Braking System (ABS) (sampling frequency 100Hz), six-axis inertial data of the Inertial Measurement Unit (IMU) (sampling frequency 200Hz), steering angle sensor signals (sampling frequency 50Hz), load estimation signals provided by the suspension height sensor (sampling frequency 50Hz), and vehicle dynamics information such as driving force / braking torque signals obtained from the CAN bus.

[0034] Step S102: Based on the multi-source signals, at least two independent estimation strategies are adopted respectively, and the wear depth estimate and uncertainty parameters corresponding to each estimation strategy are calculated in parallel.

[0035] Specifically, the independent estimation strategies employed all operate independently based on different physical mechanisms. Optional independent estimation strategies include: a first estimation strategy based on wheel speed ratio, a second estimation strategy based on resonance spectrum information, and a third estimation strategy based on friction energy. This embodiment can select any two or three independent estimation strategies for fusion. Each estimation strategy is deployed in parallel as a software module in the chassis domain controller, independently reading the required multi-source signals and performing calculations. The output of each estimation strategy includes the current tread wear depth value estimated by that strategy, and the corresponding uncertainty parameters (such as standard deviation and variance). The uncertainty parameters corresponding to the estimated value reflect the reliability of the estimation result under the current operating conditions; the larger the uncertainty parameter, the less reliable the current estimation result of that strategy.

[0036] Optionally, as a possible implementation, each independent estimation strategy performs four-wheel independent calculations, calculating the wheel-level wear depth estimate for each wheel separately. By performing a four-wheel consistency check, the difference between the maximum and minimum values ​​of the four wheel-level wear depth estimates is calculated. If the difference exceeds a preset consistency threshold (which is calibrated by the test bench, typically 2.0 mm), it is determined that there is uneven wear, and four-wheel aggregation is not performed. A marker bit is output to represent the uneven wear state. If the difference does not exceed the preset consistency threshold, the four wheel-level results are aggregated to obtain the overall wear depth estimate of the estimation strategy.

[0037] Step S103: Input the wear depth estimate and uncertainty parameters corresponding to each estimation strategy into the pre-trained fusion model to obtain the fused wear depth estimate and fusion uncertainty parameters.

[0038] Specifically, the wear depth estimates output by the three estimation strategies and their corresponding uncertainty parameters are combined to form a multi-dimensional feature vector, which is then input into the fusion model for fusion inference. In the fusion model, fusion weights are adaptively assigned based on the current uncertainty parameters of each estimation strategy. Estimation strategies with smaller uncertainty parameters contribute more to the fusion result, while the contributions of estimation strategies with larger uncertainty parameters are automatically suppressed.

[0039] Optionally, the fusion model can be a pre-trained Bayesian neural network. A Bayesian neural network is a type of neural network that models network weights as probability distributions rather than fixed values, and its output naturally contains uncertainty information. In addition to the wear depth estimates and uncertainty parameters output by each estimation strategy, ambient temperature, vehicle load, and cumulative mileage can also be fed into the fusion model as auxiliary features to obtain the fused wear depth estimate and fused uncertainty parameters. It should be noted that the fusion model is not limited to Bayesian neural networks. Weighted averaging (e.g., weighting the fusion based on the reciprocals of the uncertainty parameters of each path), linear or logarithmic opinion pooling (linearly or geometrically weighting the probability distributions of the outputs of each path), and evidence theory (fusion of multiple pieces of evidence through confidence allocation and evidence synthesis formulas) can also be used to achieve multi-path probability fusion.

[0040] Optionally, the outputs of each estimation strategy can be transmitted to the cloud in real time via 4G / 5G, where they are fused by a Gaussian process regression model with higher strength. This method utilizes the vehicle's historical wear curves and group data from the same model's fleet to achieve higher accuracy. This approach is suitable for vehicles with comprehensive vehicle networking capabilities.

[0041] Finally, by using the fused wear depth estimate and the fusion uncertainty parameter, the three decision levels of tire wear status can be reliably distinguished, and the conclusions of "normal" (e.g., wear depth fusion estimate greater than 3mm), "recommend replacing tire" (e.g., wear depth fusion estimate between 1.6mm and 3mm), and "dangerous" (e.g., wear depth fusion estimate less than 1.6mm) can be output.

[0042] This embodiment employs at least two independent estimation strategies based on different physical mechanisms in parallel estimation. This leverages the complementary error characteristics between different estimation strategies to overcome the insufficient accuracy of a single strategy across the entire operating range, thereby improving estimation accuracy. Furthermore, by inputting the uncertainty parameters of each strategy into the fusion model, adaptive weighted fusion based on current reliability is achieved, with the equivalent fusion weights of each strategy dynamically changing with the operating conditions. Finally, the fusion model outputs fusion uncertainty parameters, enabling downstream decision-making systems to determine the reliability of the estimation results.

[0043] Figure 2 A flowchart illustrating Embodiment Two of the tire wear depth estimation method provided in this application. (Refer to...) Figure 2 Based on the above embodiments, at least two independent estimation strategies, including a first estimation strategy based on wheel speed ratio, and the step of calculating the wear depth estimate and uncertainty parameter corresponding to each estimation strategy, may include:

[0044] Step S201: For each wheel, determine the wheel speed ratio at multiple times based on the wheel speed at multiple times, the vehicle's reference speed, and the wheel's nominal rolling radius.

[0045] Specifically, the ABS wheel speed sensors provide wheel speed signals for each wheel. The IMU calculates the vehicle's longitudinal reference speed by integrating the data. For each wheel i, based on the wheel speed, the reference vehicle speed, and the wheel's nominal rolling radius, the wheel speed ratio at time t can be determined. Optionally, one method of calculating the wheel speed ratio is as follows:

[0046] ;

[0047] in, This refers to the nominal rolling radius of the wheel (unit: mm). For example, the nominal rolling radius of a 235 / 45R18 tire is approximately 334 mm. Physically, the wheel speed ratio represents the ratio of the actual rolling radius of a wheel to its nominal rolling radius. When the tire tread is unworn, this ratio is close to 1.0. Because the wheel speed ratio fluctuates instantaneously due to factors such as tire slippage during transient conditions such as acceleration, deceleration, or turning, this step uses the wheel speed signals of the vehicle at multiple moments (within a continuous time window) to calculate the wheel speed ratio at each moment, forming a time series of the wheel speed ratio for each wheel.

[0048] Step S202: Filter the wheel speed ratios at multiple times to obtain the corresponding filter sequences for the wheels.

[0049] Specifically, a first-order low-pass filter is applied to the time series of wheel speed ratios for each wheel. The purpose of the low-pass filter is to remove instantaneous noise and short-term fluctuations (such as instantaneous changes in wheel speed caused by uneven road surfaces, acceleration, and deceleration) from the wheel speed ratio, and to extract the slow-varying component that reflects the long-term trend of the effective rolling radius of the tire. The filtered sequence for each wheel is then obtained after this filtering process. Each value in the filtered sequence represents the long-term trend value of the wheel speed ratio at that moment.

[0050] Step S203: Based on the nominal rolling radius and the current filter value corresponding to the current moment in the filter sequence, calculate the change in the effective rolling radius of the wheel as an estimate of the wheel's wear depth.

[0051] Specifically, when tread wear reduces the effective rolling radius of the tire, the long-term trend value of the wheel speed ratio will deviate from 1.0. The change in effective rolling radius at the current filtered value of the filtered sequence is shown in the figure. The calculation formula is:

[0052] ;

[0053] in, It can be used as an estimate of the wear depth of the wheel at the current time t. This allows us to obtain an estimate of the wear depth for each wheel.

[0054] Step S204: Determine the wear depth estimate corresponding to the first estimation strategy based on the wear depth estimate of each wheel.

[0055] Specifically, a four-wheel consistency check is performed on the wear depth estimates of each wheel, calculating the maximum and minimum values ​​of the wheel-level wear depth estimates for all four wheels, as well as the difference between the maximum and minimum values. If this difference is less than or equal to a preset consistency threshold, it is determined to be uniform wear. The arithmetic mean or weighted average of the four wheel-level wear depth estimates is then used as the overall wear depth estimate for the first estimation strategy. If the difference is greater than the consistency threshold, it is determined that tire wear is uneven, four-wheel aggregation is not performed, an invalid mark is output, and the estimation strategy result reduces the subsequent fusion weight.

[0056] Step S205: Based on the filter sequence corresponding to each wheel, determine the standard deviation of all filter sequences within the preset time window at the current moment, and use the standard deviation as the uncertainty parameter corresponding to the first estimation strategy.

[0057] Specifically, for the filtered sequence of each wheel, data within a preset time window (e.g., a window length corresponding to 500km of driving distance) prior to the current time t is taken, and the standard deviation of each filtered value within this window is calculated. A larger standard deviation indicates greater fluctuation in the long-term trend of the wheel speed ratio within that time window, making the estimation result of the first estimation strategy less reliable at the current time; a smaller standard deviation indicates a more stable long-term trend of the wheel speed ratio, making the estimation result more reliable. By combining the standard deviations of each filtered value, the uncertainty parameter output corresponding to the first estimation strategy is obtained. .

[0058] This embodiment utilizes the physical law that wear leads to a decrease in rolling radius, and employs a first estimation strategy based on wheel speed ratio to calculate the corresponding wear depth estimate and uncertainty parameters. This estimation strategy is highly sensitive to uniform wear, and suppresses instantaneous slippage and road surface interference through long-period filtering. The sliding window standard deviation can objectively reflect the road noise level under the current working conditions, and outputs reasonable uncertainty parameters for reference by the fusion model.

[0059] Figure 3 A flowchart illustrating Embodiment 3 of the tire wear depth estimation method provided in this application. (Refer to...) Figure 3 Based on the above embodiments, at least two independent estimation strategies, including a second estimation strategy based on resonance spectrum information, are included. The steps of calculating the wear depth estimate and uncertainty parameters corresponding to each estimation strategy may include:

[0060] Step S301: For each wheel, determine the vibration spectrum corresponding to the wheel based on the wheel speed at multiple times.

[0061] Specifically, the process is performed separately for each of the four wheels. A Short-Time Fourier Transform (STFT) is applied to the wheel speed time-domain signal to obtain frequency-domain data. In practice, a 1024-point STFT is performed after applying a Hanning window to the wheel speed signal, with a window overlap rate of 50%. The STFT converts the time-domain wheel speed signal into a time-frequency domain spectral representation, allowing the system to observe the energy distribution of the wheel speed signal at different frequencies at each moment. The horizontal axis of the vibration spectrum represents frequency, and the vertical axis represents the amplitude or energy intensity of the corresponding frequency component. Each wheel undergoes independent STFT processing; optionally, the frequency-domain data can be directly used as the vibration spectrum for each wheel.

[0062] Optionally, as a possible embodiment, the step of determining the vibration spectrum corresponding to the wheel based on the wheel speed at multiple moments may include:

[0063] (1) Perform time-frequency transformation on the wheel speed at multiple times to obtain frequency domain data.

[0064] Specifically, a short-time Fourier transform is performed on the wheel speed signal of each wheel to obtain frequency domain data containing frequency information.

[0065] (2) Obtain the temperature information of the tire and determine the temperature compensation factor based on the temperature information and the reference temperature.

[0066] Specifically, because the stiffness of tire rubber is temperature-sensitive—decreasing as temperature rises and increasing as temperature falls—this causes a temperature drift in the torsional resonance spectrum that is independent of wear. To eliminate this interference, this step acquires temperature information, which may include ambient temperature from an ambient temperature sensor and / or tire temperature. The tire temperature may come from the tire temperature sensor of a virtual tire pressure monitoring system or from a surface acoustic wave (SAW) tag temperature sensor. Optionally, in vehicles equipped with passive SAW tags, the SAW tag temperature reading is cross-validated with the temperature from the vibration spectrum module to help correct for temperature interference components in the spectrum shift. This can compress the uncertainty parameter of the second estimation strategy from ±2.0 mm to ±1.2 mm, ultimately improving the fusion accuracy to ±0.8 mm.

[0067] Based on the temperature information and reference temperature, determine the temperature compensation factor. One calculation is as follows:

[0068] ;

[0069] in, The temperature sensitivity coefficient of rubber stiffness is obtained through bench calibration. For temperature information, The reference temperature can be determined through bench testing, such as placing the tire in a constant and known temperature environment (e.g., a thermostat) in a laboratory. Alternatively, in the absence of specialized calibration, an industry-recognized ambient temperature (e.g., 20°C or 25°C) is typically used as the reference temperature.

[0070] (3) Correct the frequency values ​​in the frequency domain data according to the temperature compensation factor to obtain the vibration spectrum.

[0071] Specifically, the extracted resonance spectrum is normalized using a temperature compensation factor to correct the frequency values ​​in the frequency domain data, thereby obtaining the equivalent frequency at the reference temperature, and this equivalent frequency is used as the vibration spectrum.

[0072] Optionally, when the temperature information includes the tire's internal temperature, a more accurate tire temperature compensation model can be used; when the temperature information only includes the ambient temperature, the ambient temperature can be used as an approximation for tire temperature compensation. The specific form of the temperature compensation factor is not limited to the linear form described above. In another parallel scheme, a lookup table method or a polynomial fitting method can be used to determine the temperature compensation factor.

[0073] This embodiment eliminates the resonance spectrum drift caused by the temperature sensitivity of rubber stiffness through temperature compensation, avoids misjudging temperature fluctuations as tread wear, and uses the temperature-compensated equivalent frequency as the vibration spectrum, which can reduce the temperature correlation error of the second estimation strategy and make the wear estimation results consistent under different temperature conditions.

[0074] Step S302: Extract the resonance spectrum information of the wheel at the current moment from the vibration spectrum. Based on the resonance spectrum information of the wheel at the current moment and the pre-established mapping relationship between the resonance spectrum information and the tread wear depth, determine the estimated value of the wheel wear depth.

[0075] Specifically, within the 40Hz-80Hz frequency band, the frequency value corresponding to the energy peak is extracted from the vibration spectrum; this frequency is the first-order torsional resonance spectrum of the tire. The first-order torsional resonance spectrum of a tire refers to the first-order natural vibration frequency of the tire in the torsional direction. Its theoretical value is related to the tire's torsional stiffness and tread mass as follows:

[0076] ;

[0077] in, For tire torsional stiffness, This represents the tread mass. This relationship indicates that when tread wear leads to a decrease in tread mass, the torsional resonance spectrum undergoes a systematic upward shift.

[0078] A resonance spectrum-wear depth calibration curve for the tire model is pre-established through tire bench testing. Using tires of the same model with different wear depths, their torsional resonance spectra are measured, yielding multiple sets of (frequency, wear depth) data points, which are then fitted into an interpolation curve or mapping table. This calibration curve is stored in the vehicle's chassis domain controller. During real-time estimation, the currently extracted resonance spectrum information is input into this calibration curve, mapping to obtain the wear depth estimate for that wheel at the current moment. This mapping is performed independently for each wheel, yielding its own wear depth estimate.

[0079] Step S303: Determine the wear depth estimate corresponding to the second estimation strategy based on the wear depth estimate of each wheel.

[0080] Specifically, a four-wheel consistency check is performed on the wear depth estimates of each wheel, calculating the maximum and minimum values ​​of the wheel-level wear depth estimates for all four wheels, as well as the difference between the maximum and minimum values. If this difference is less than or equal to a preset consistency threshold, it is determined to be uniform wear. The arithmetic mean or weighted average of the four wheel-level wear depth estimates is then used as the overall wear depth estimate for the second estimation strategy. If the difference is greater than the consistency threshold, it is determined that tire wear is uneven, four-wheel aggregation is not performed, an invalid mark is output, and the estimation strategy result reduces the subsequent fusion weight.

[0081] Step S304: Determine the uncertainty parameters corresponding to the second estimation strategy based on the full width at half maximum (FWHM) of the vibration spectrum corresponding to all wheels.

[0082] Specifically, for the vibration spectrum of each wheel, the full width at half maximum (FWHM) of the spectral peak is measured at the peak of the resonance spectrum. The FWHM refers to the frequency width corresponding to half the amplitude of the spectral peak. A smaller FWHM indicates a sharper resonance peak, resulting in more accurate and reliable resonance spectrum extraction; a larger FWHM indicates a flatter spectral peak, making resonance spectrum extraction more susceptible to noise interference. The FWHMs of each wheel are aggregated (e.g., by taking the average or maximum value), and the aggregated FWHM is used as the uncertainty parameter corresponding to the second estimation strategy. .

[0083] Optionally, when temperature compensation is introduced in step S301 above, the temperature measurement itself has a measurement error, i.e., temperature error. This error will be propagated to form a temperature-compensated residual. The residuals are included in the uncertainty parameters of the second estimation strategy. In practice, the frequency domain uncertainty is calculated from the full width at half maximum (FWHM) of the resonance peaks. Combined with temperature compensation residual Synthetic second estimation strategy uncertainty parameters Uncertainty parameters The calculation is as follows:

[0084] ;

[0085] Among them, frequency domain uncertainty The calculation is as follows:

[0086] ;

[0087] Where 2.355 is the Gaussian peak conversion factor. Temperature compensation residual. The calculation is as follows:

[0088] ;

[0089] in, This is a temperature compensation factor. Typical in temperature-stable environments. ; when the temperature changes drastically or the road surface is rough Increased to 2.0 mm.

[0090] This embodiment extracts the torsional resonance spectrum through vibration spectrum analysis. Utilizing the physical law that tread wear causes resonance spectrum drift, a second estimation strategy is adopted for wear estimation. This estimation path can operate across the entire speed domain. Compared to the first estimation strategy based on wheel speed ratio, it still has effective output under low-speed conditions and has a certain sensitivity to identify uneven tire wear. Furthermore, this embodiment calculates the uncertainty based on the half-width at half-maximum (WHM) of the spectrum peaks. The more diffuse the resonance peak, the greater the uncertainty, which can objectively reflect the decrease in estimation reliability caused by road surface roughness and noise interference.

[0091] Figure 4 A flowchart of Embodiment 4 of the tire wear depth estimation method provided in this application. (Refer to...) Figure 4 Based on the above embodiments, at least two independent estimation strategies, including a third estimation strategy based on frictional energy, are included. The steps for calculating the wear depth estimate and uncertainty parameters corresponding to each estimation strategy may include:

[0092] Step S401: Obtain the vehicle's dynamics information and determine the force and motion state of each wheel based on the dynamics information.

[0093] Specifically, the dynamic information includes IMU six-axis data (longitudinal acceleration, lateral acceleration, yaw rate, sampling frequency 200Hz), driving force / braking torque signals acquired by the CAN bus, steering angle sensor signals (sampling frequency 50Hz), and wheel load estimation signals provided by the suspension height sensor (sampling frequency 50Hz).

[0094] Based on the above dynamic information, the force and motion states of each wheel are calculated. Specifically, the force state includes the longitudinal force on each wheel. and lateral force (Unit: N). Longitudinal forces can be calculated by dividing the driving force / braking torque by the effective rolling radius of the tire; lateral forces can be estimated using the vehicle's lateral acceleration and yaw rate, combined with the load distribution to each wheel. Motion states include the slip ratio of each wheel. and sideslip angle Slip ratio Calculated from the degree to which the wheel speed ratio deviates from 1.0; sideslip angle It is calculated from the steering angle and the vehicle's yaw rate. Optionally, when a new tire replacement event is triggered, the accumulated friction energy inside this strategy is reset to zero.

[0095] Step S402: Determine the instantaneous friction power of each wheel based on the force and motion state of each wheel.

[0096] Specifically, the instantaneous friction power of a single wheel The calculation formula is as follows:

[0097] ;

[0098] in, and These represent the longitudinal and lateral velocities of the wheel, respectively. The physical meaning of this formula is: the instantaneous frictional power between the tire and the road surface equals the sum of the longitudinal force and the work done by slippage, plus the sum of the lateral force and the work done by lateral deviation. During acceleration, deceleration, and cornering, the dissipation of frictional energy between the tire and the road surface is most significant; therefore, this strategy has higher estimation accuracy under these conditions.

[0099] Step S403: Based on the instantaneous friction power of each wheel, the cumulative friction energy of each wheel is accumulated.

[0100] Specifically, the instantaneous frictional power of each wheel is integrated over time to obtain the cumulative frictional energy of each wheel. Accumulated frictional energy The calculation is as follows:

[0101] ;

[0102] The integration start point is the moment when the tire begins to be used (i.e., the moment a new tire is installed), and the integration end point is the current time t. The accumulated frictional energy increases continuously with mileage and driving conditions, reflecting the total frictional energy input that the tire has experienced since it was put into use.

[0103] Step S404: Based on the preset mapping coefficient, the cumulative friction energy of each wheel is mapped to the estimated wear depth of each wheel.

[0104] Specifically, based on a simplified form of Arcard's wear law, which states that the wear volume is proportional to the accumulated frictional energy, an energy-wear mapping relationship is established, thereby obtaining an estimate of the wear depth for each wheel. as follows:

[0105] ;

[0106] in, The preset mapping coefficient is the energy-wear mapping coefficient (unit: mm / J), which is obtained by tire bench durability test calibration. It is a specific parameter of this tire model. This coefficient represents the wear depth increment corresponding to a unit friction energy input.

[0107] Step S405: Determine the wear depth estimate corresponding to the third estimation strategy based on the wear depth estimate of each wheel.

[0108] Specifically, a four-wheel consistency check is performed on the wear depth estimates of each wheel, calculating the maximum and minimum values ​​of the wheel-level wear depth estimates for all four wheels, as well as the difference between the maximum and minimum values. If this difference is less than or equal to a preset consistency threshold, it is determined to be uniform wear. The arithmetic mean or weighted average of the four wheel-level wear depth estimates is then used as the overall wear depth estimate for the third estimation strategy. If the difference is greater than the consistency threshold, it is determined that tire wear is uneven, four-wheel aggregation is not performed, an invalid mark is output, and the estimation strategy result reduces the subsequent fusion weight.

[0109] Step S406: Determine the uncertainty parameters corresponding to the third estimation strategy based on the cumulative duration corresponding to the cumulative friction energy of each wheel.

[0110] Specifically, since the calculation of instantaneous frictional power involves the fusion of signals from multiple sensors and the estimation of multiple physical quantities, measurement and model errors exist at each stage. These errors accumulate continuously with time integration. Therefore, the uncertainty parameter of the third estimation strategy accumulates and diverges as the integration time increases. The specific calculation is as follows:

[0111] ;

[0112] in, The initial uncertainty is denoted as ±0.5 mm (typically corresponding to the initial estimate uncertainty when a new tire is installed), and t is the cumulative duration of accumulated frictional energy (i.e., the driving time or mileage from the start of use of the current tire). This represents the divergence time constant (typically around 10,000 km of driving distance). The physical meaning of this formula is: uncertainty increases with the square root of the accumulated time; the longer the accumulated time, the greater the cumulative error of the integral, and the less reliable the estimation result. For example, when driving 10,000 km... Approximately ±1.0 mm, after 50,000 km of driving. It is approximately ±2.5mm.

[0113] This embodiment is based on Arcard's wear law and uses a third estimation strategy to estimate wear from the perspective of energy dissipation. It has high accuracy in acceleration and deceleration sections. By accumulating frictional energy through time integration, it achieves continuous tracking of tire wear input throughout its entire life cycle. Furthermore, the uncertainty parameter of this estimation strategy diverges with the accumulation time according to the square root law, accurately reflecting the growth characteristics of the integral accumulation error.

[0114] Optionally, as a possible implementation, the fusion model is a Bayesian neural network; the wear depth estimates and uncertainty parameters corresponding to each estimation strategy are input into the pre-trained fusion model to obtain the fused wear depth fusion estimate and the corresponding fusion uncertainty parameters, which may include:

[0115] Obtain the posterior distribution of the weights of the Bayesian neural network; sample multiple sets of network weights from the posterior distribution of weights, perform forward propagation with each set of network weights, map the wear depth estimates and uncertainty parameters corresponding to each input estimation strategy to the corresponding wear depth estimates, and obtain multiple wear depth estimates; use the mean of the multiple wear depth estimates as the fused wear depth estimate, and use the variance of the multiple wear depth estimates as the fused uncertainty parameter.

[0116] Specifically, in this embodiment, the Bayesian neural network has a two-layer fully connected network structure, with a hidden layer dimension of 64. The network weights are based on a standard normal distribution N(0,1) as the prior. The input feature vector is 9-dimensional. , These are ambient temperature, load, and mileage, respectively.

[0117] During the training phase, the posterior distribution of the network weights is learned through variational inference. Specifically, a parameterized simple distribution (usually Gaussian) is used to approximate the true, but difficult-to-calculate, posterior distribution of the weights. The variational parameters are optimized simultaneously during training. ,in The average of the weights. The standard deviation of the weights is obtained after softplus transformation. The optimization objective is a variational lower bound, which includes a relative entropy (Kullback-Leibler, KL) divergence term (penalizing the degree of deviation of the approximate posterior from the prior) and a likelihood term (encouraging the model to fit the training data). The training data includes the output values ​​of each strategy and their corresponding measured ground truth groove depths under different driving conditions (covering a temperature range of -30℃ to +50℃, various road conditions, and various tire brands). During training, the network's loss function includes both a prediction error term and a calibration error term; the former ensures that the fused estimate is statistically unbiased, while the latter ensures that the uncertainty parameters of the network output match the actual prediction error. Variational inference is used to train the network weights, obtaining the posterior distribution of each network weight.

[0118] After training, each network weight corresponds to a posterior distribution, rather than a fixed value, and is stored in the chassis domain controller for use during the inference phase. Obtaining the posterior distribution of the weights means obtaining the distribution parameters (mean and standard deviation) of these trained network weights.

[0119] During inference, K sets of weights are independently sampled from the posterior distribution of each network weight (e.g., K=20). For each network weight w, a specific value is sampled from its distribution, and a set of sampled values ​​for all network weights constitutes a complete set of network weights.

[0120] The wear depth estimates output by each estimation strategy at the current moment. and uncertainty parameters The input feature vector x is constructed by combining ambient temperature, load, and mileage (along with auxiliary information). Forward propagation is performed with K sets of network weights: For the k-th set of weights, the input feature vector x is input into the network, passing through a first fully connected layer and a second fully connected layer. The first fully connected layer uses Rectified Linear Unit (ReLU) activation, and the second fully connected layer provides a linear output, resulting in a predicted wear depth value. This process is repeated K times to obtain K predicted wear depth values.

[0121] Calculate the arithmetic mean and sample variance of the K wear depth predictions. The output of the Bayesian neural network is the fused wear depth estimate (i.e., the mean of the posterior distribution) and the fusion uncertainty parameter (i.e., the variance or standard deviation of the posterior distribution).

[0122] This embodiment uses a Bayesian neural network to perform probabilistic fusion of multiple estimation strategies. The output includes not only point estimates but also uncertainty quantification. The method of taking the mean and variance through multiple sampling ensures that the fusion result fully reflects the uncertainty of the model weights, realizing uncertainty-aware adaptive weighting, automatically trusting high-credibility paths and suppressing high-noise paths when operating conditions change. Finally, the model outputs a probability distribution, along with the estimation results and overall uncertainty, providing a basis for downstream hierarchical alarm decisions.

[0123] Optionally, as a possible embodiment, after obtaining the fused wear depth estimate and fusion uncertainty parameters, the above method further includes:

[0124] When the fusion uncertainty parameter is less than the first preset threshold, the fusion estimate of the wear depth is output.

[0125] When the fusion uncertainty parameter is greater than or equal to the first preset threshold and less than the second preset threshold, the estimation range is determined based on the wear depth fusion estimate and the estimation range is output.

[0126] When the fusion uncertainty parameter is greater than or equal to the second preset threshold, an alarm message is output; wherein, the second preset threshold is greater than the first preset threshold.

[0127] Specifically, the uncertainty parameter is integrated with the standard deviation. (Unit: mm). Optionally, the first preset threshold is set to 0.5 mm. When When the confidence level is 1, it indicates that the estimation results of the fusion model have a high degree of confidence, meaning that the estimation results of the three estimation strategies are highly consistent, and the uncertainty parameters of each strategy are small, with weak external interference (such as temperature fluctuations and signal noise). In this case, the system directly outputs the fusion estimate of wear depth in numerical form. For example, the dashboard displays "Current tread depth: 3.6mm".

[0128] When the fusion uncertainty parameter is greater than or equal to the first preset threshold and less than the second preset threshold, it indicates that the estimation result of the fusion model has a certain degree of uncertainty, meaning there is some divergence among the strategies, or some strategies are affected by external interference. In this case, the system outputs the result in the form of an estimation range. The estimation range is a 95% confidence interval, specifically based on the fusion estimate of wear depth. Centered on, with Using the standard, construct a 95% confidence interval. As an estimate range. For example, if , The estimated range is [3.6-1.96×0.7, 3.6+1.96×0.7]=[2.23mm, 4.97mm]. The dashboard displays "Current tread groove depth is approximately between 2.2mm and 5.0mm". 1.96 is the 0.05 quantile of a standard normal distribution, corresponding to a 95% confidence level. Optionally, the second preset threshold is set to 1.0mm.

[0129] When the fusion uncertainty parameter is greater than or equal to the second preset threshold, it indicates that the confidence level of the fusion model's estimation result is too low, meaning that the estimation results of the three strategies differ significantly, or that external interference is severe (such as drastic temperature changes or poor signal quality). In this case, the system will no longer display the specific wear depth value or range, but will instead output an alarm message, prompting the user that the current estimation result is unreliable and recommending that the user go to a shop to obtain an accurate tread wear depth through direct measurement methods (such as using a groove depth caliper). The alarm message can be displayed on the dashboard as a prompt such as "Low confidence level in tread wear estimation, recommended to go to a shop for inspection."

[0130] Optionally, the specific values ​​of the first and second preset thresholds can be adjusted according to the tire model, vehicle type, or usage scenario. For example, for applications requiring higher precision, the first preset threshold can be set to 0.3mm and the second preset threshold to 0.8mm.

[0131] This embodiment adopts a hierarchical output strategy based on the magnitude of the fusion uncertainty parameter: for low uncertainty, the value is displayed directly (high confidence); for medium uncertainty, the estimated range is displayed (indicating possible errors); and for high uncertainty, an alarm is output (to avoid misleading).

[0132] Optionally, as one possible embodiment, the method further includes:

[0133] Obtain the measured value of tire tread wear depth; update the fusion model based on the deviation between the measured value and the fused wear depth estimate.

[0134] Specifically, during regular vehicle maintenance, technicians use physical measuring tools such as groove calipers to directly measure the tread groove depth of each tire, obtaining the measured values ​​of the tread wear depth of all four wheels. These measured values ​​are uploaded to the vehicle's chassis domain controller via a diagnostic interface, and after data anonymization processing, they are uploaded to a cloud server as ground truth (GT) values ​​for subsequent calibration.

[0135] Calculate the deviation between the measured values ​​of each wheel and the fused estimate of wear depth output by the fusion model at the maintenance time. :

[0136] ;

[0137] Where i represents the wheel number. This is a fusion estimate of the wear depth output by the fusion model for this wheel. This is the measured value of the tread wear depth of the wheel.

[0138] when Within the normal range ( When the deviation between the estimated results of the fusion model and the measured values ​​is within an acceptable range, the model's architecture and parameters do not require significant adjustment. At this point, the wear rate estimation parameters for each wheel are updated using the measured values ​​from this maintenance. The wear rate is defined as the rate of change of wear depth with mileage; specifically, the difference in wear depth between the current maintenance and the previous maintenance is calculated. and mileage difference To obtain the true physical speed Simultaneously, the estimated wear rate is obtained by temporal differencing the wear depth estimates output by the fusion model within the same time period. The actual rate With estimated rate By comparing the two, the energy-wear mapping coefficients in the third estimation strategy are adjusted. This mechanism ensures that the estimated wear rate converges to the actual wear rate during long-term driving cycles when maintenance data is lacking, thus preventing the divergence of integral accumulation errors.

[0139] when Outside the normal range ( When the result is zero, it indicates that there is a systematic deviation between the estimation results of the fusion model and the measured values, requiring individualized fine-tuning of the model. In this case, the deviation is... It is added as a compensation parameter to the output layer of the fusion model. In actual execution, it will... This bias term is stored locally on the vehicle. In each subsequent inference process, the wear depth fusion estimate output by the fusion model is added to this bias term. This yields the final estimated value after individualized calibration. This bias term contains only a small number of parameters (approximately 2KB of data) and is wirelessly transmitted to the vehicle for updates via Over-The-Air (OTA) without requiring full retraining of the fusion model. During subsequent maintenance, if the deviation between the new measured value and the calibrated estimated value falls back to the normal range, the current bias term remains unchanged; if the deviation still exceeds the threshold, the bias term is recalculated and updated.

[0140] Optionally, when the vehicle is not equipped with cloud communication capabilities, the deviation compensation parameters can be directly written to the vehicle controller via diagnostic equipment during maintenance, without relying on OTA. In addition, after obtaining new ground truth values ​​after each maintenance, full retraining can be avoided. Instead, the individual deviation compensation vector (2KB) can be updated using Kalman filtering with single-point deviation and superimposed on the output of the base model, thus avoiding the storage and retraining overhead of a large amount of fleet data.

[0141] This embodiment introduces measured true values ​​through a maintenance calibration closed loop, enabling continuous model calibration and individualized adaptation; when the deviation is normal, the wear rate is updated to ensure the convergence of the prediction trend during maintenance-free periods; when the deviation exceeds the threshold, a bias term is superimposed, requiring only 2KB of OTA data to complete individualized fine-tuning without full retraining, greatly reducing OTA traffic and computational overhead; the estimation accuracy continues to improve with vehicle usage.

[0142] Figure 5 A flowchart of Embodiment 5 of the tire wear depth estimation method provided in this application. (Refer to...) Figure 5 The method may include:

[0143] Step S501: Acquire multi-source signals during vehicle operation.

[0144] Step S502: Based on the multi-source signals, a first estimation strategy based on the wheel speed ratio is adopted to obtain the wear depth estimate and uncertainty parameters corresponding to the first estimation strategy.

[0145] Step S503: Based on the multi-source signals, a second estimation strategy based on resonance spectrum information is adopted to obtain the wear depth estimate and uncertainty parameters corresponding to the second estimation strategy.

[0146] Step S504: Based on the multi-source signals, a third estimation strategy based on frictional energy is adopted to obtain the wear depth estimate and uncertainty parameters corresponding to the third estimation strategy.

[0147] Step S505: Input the wear depth estimate and uncertainty parameters corresponding to each estimation strategy, as well as the load, mileage and ambient temperature, into the pre-trained Bayesian network to obtain the fused wear depth estimate and the corresponding fused uncertainty parameters.

[0148] Step S506: When the fusion uncertainty parameter is less than the first preset threshold, output the wear depth fusion estimate; when the fusion uncertainty parameter is greater than or equal to the first preset threshold and less than the second preset threshold, determine the estimation range based on the wear depth fusion estimate and output the estimation range; when the fusion uncertainty parameter is greater than or equal to the second preset threshold, output alarm information.

[0149] Step S507: Obtain the measured value of tire tread wear depth; update the fusion model based on the deviation between the measured value and the fused wear depth estimate.

[0150] Specifically, the implementation principles and processes of steps S501 to S507 can be found in the descriptions in the previous embodiments, and will not be repeated here.

[0151] This embodiment achieves complementarity among the estimation strategies in the time domain (e.g., wheel speed ratio filtering period of approximately 1000km, spectrum of approximately 100km, and real-time updates of friction energy), operating condition domain (wheel speed ratio is effective at medium and high speeds, spectrum is effective across the entire speed domain, and friction energy has the highest accuracy in acceleration and deceleration sections), and error domain (wheel speed ratio exhibits slow drift, spectrum is affected by temperature fluctuations, and friction exhibits cumulative divergence). The fusion achieves full time domain coverage, full operating condition coverage, and error cancellation, thereby improving estimation accuracy. Furthermore, a tiered output strategy is adopted based on the magnitude of the fusion uncertainty parameter: low uncertainty results in direct numerical display (high confidence); medium uncertainty results in displaying the estimation range (indicating potential errors); and high uncertainty results in outputting an alarm (avoiding misleading information).

[0152] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0153] Corresponding to the aforementioned embodiment of a tire wear depth estimation method, this application also provides an embodiment of a tire wear depth estimation device.

[0154] An embodiment of the tire wear depth estimation device disclosed in this application can be applied to control devices (such as vehicle domain controllers). The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of the control device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 6 The diagram shown is a hardware structure diagram of the control equipment containing the tire wear depth estimation device of this application. Except for... Figure 6 In addition to the processor, memory, network interface, and non-volatile memory shown, the control device in the embodiment, which typically estimates the actual function of the device based on the tire wear depth, may also include other hardware, which will not be described in detail here.

[0155] Figure 7 This is a schematic diagram of the tire wear depth estimation device provided in this application. (Refer to...) Figure 7 The apparatus provided in this embodiment includes:

[0156] Signal acquisition module 10 is used to acquire multi-source signals during vehicle operation;

[0157] The parallel estimation module 20 is used to calculate the wear depth estimate and uncertainty parameters corresponding to each estimation strategy in parallel based on the multi-source signals, using at least two independent estimation strategies respectively.

[0158] The fusion estimation module 30 is used to input the wear depth estimates and uncertainty parameters corresponding to each estimation strategy into the pre-trained fusion model to obtain the fused wear depth fusion estimate and the corresponding fusion uncertainty parameters.

[0159] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0160] Optionally, as one possible embodiment, the parallel estimation module 20 is also used for:

[0161] For each wheel, the wheel speed ratio at multiple times is determined based on the wheel speed at multiple times, the vehicle's reference speed, and the wheel's nominal rolling radius.

[0162] The wheel speed ratios at multiple times are filtered to obtain the corresponding filter sequences for the wheels.

[0163] Based on the nominal rolling radius and the current filter value corresponding to the current moment in the filter sequence, the change in the effective rolling radius of the wheel is calculated as an estimate of the wear depth of the wheel.

[0164] Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the first estimation strategy;

[0165] Based on the filtering sequence corresponding to each wheel, the standard deviation of all filtering sequences within the preset time window at the current moment is determined, and the standard deviation is used as the uncertainty parameter corresponding to the first estimation strategy.

[0166] Optionally, as one possible embodiment, the parallel estimation module 20 is also used for:

[0167] For each wheel, the vibration spectrum corresponding to the wheel is determined based on the wheel speed at multiple times.

[0168] The resonance spectrum information of the wheel at the current moment is extracted from the vibration spectrum. Based on the resonance spectrum information of the wheel at the current moment and the pre-established mapping relationship between the resonance spectrum information and the tread wear depth, the estimated value of the wheel wear depth is determined.

[0169] Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the second estimation strategy;

[0170] The uncertainty parameters corresponding to the second estimation strategy are determined based on the half-width at half-maximum (WHM) of the vibration spectrum corresponding to all wheels.

[0171] Optionally, as one possible embodiment, the parallel estimation module 20 is also used for:

[0172] The wheel speed at multiple moments is transformed using time-frequency conversion to obtain frequency domain data.

[0173] Obtain tire temperature information and determine the temperature compensation factor based on the temperature information and reference temperature;

[0174] The vibration spectrum is obtained by correcting the frequency values ​​in the frequency domain data using a temperature compensation factor.

[0175] Optionally, as one possible embodiment, the parallel estimation module 20 is also used for:

[0176] Acquire the vehicle's dynamics information and determine the force and motion state of each wheel based on the dynamics information;

[0177] The instantaneous frictional power of each wheel is determined based on the force and motion state of each wheel.

[0178] The cumulative frictional energy of each wheel is obtained by accumulating the instantaneous frictional power of each wheel;

[0179] Based on the preset mapping coefficients, the cumulative friction energy of each wheel is mapped to the estimated wear depth of each wheel.

[0180] Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the third estimation strategy;

[0181] The uncertainty parameters corresponding to the third estimation strategy are determined based on the cumulative duration corresponding to the cumulative friction energy of each wheel.

[0182] Optionally, as one possible embodiment, the fusion estimation module 30 is also used for:

[0183] Obtain the posterior distribution of the weights of the Bayesian neural network;

[0184] Multiple sets of network weights are sampled from the weight posterior distribution, and forward propagation is performed with each set of network weights to map the wear depth estimate and uncertainty parameter corresponding to each input estimation strategy to the corresponding wear depth estimate, thereby obtaining multiple wear depth estimates.

[0185] The mean of multiple wear depth estimates is used as the fused wear depth estimate, and the variance of the multiple wear depth estimates is used as the fusion uncertainty parameter.

[0186] Optionally, as one possible embodiment, the fusion estimation module 30 is also used for:

[0187] When the fusion uncertainty parameter is less than the first preset threshold, the fusion estimate of the wear depth is output.

[0188] When the fusion uncertainty parameter is greater than or equal to the first preset threshold and less than the second preset threshold, the estimation range is determined based on the wear depth fusion estimate and the estimation range is output.

[0189] When the fusion uncertainty parameter is greater than or equal to the second preset threshold, an alarm message is output; wherein, the second preset threshold is greater than the first preset threshold.

[0190] Optionally, as one possible embodiment, the device further includes a maintenance calibration module for:

[0191] Obtain the measured value of tire tread wear depth;

[0192] The fusion model is updated based on the deviation between the measured values ​​and the fused estimate of wear depth.

[0193] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the methods in any of the embodiments provided in this application.

[0194] This application also provides a vehicle including a control device, the control device including a memory, a processor, and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the tire wear depth estimation method as described in the above embodiments.

[0195] It is understood that the method provided in this application can be applied to a domain controller in a vehicle. The domain controller may include a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the methods provided in this application.

[0196] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0197] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0198] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0199] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0200] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or alternative to other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0201] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A method for estimating tire wear depth, characterized in that, include: Acquire multi-source signals during vehicle operation; Based on the multi-source signals, at least two independent estimation strategies are used respectively, and the wear depth estimate and uncertainty parameter corresponding to each estimation strategy are calculated in parallel. The wear depth estimate and uncertainty parameter corresponding to each estimation strategy are input into the pre-trained fusion model to obtain the fused wear depth estimate and fusion uncertainty parameter.

2. The tire wear depth estimation method according to claim 1, characterized in that, The at least two independent estimation strategies include a first estimation strategy based on wheel speed ratio; The calculation of the wear depth estimate and uncertainty parameters corresponding to each of the estimation strategies includes: For each wheel, the wheel speed ratio at multiple times is determined based on the wheel speed of the wheel at multiple times, the reference speed of the vehicle, and the nominal rolling radius of the wheel. The wheel speed ratio of the wheel at multiple times is filtered to obtain the filter sequence corresponding to the wheel. Based on the nominal rolling radius and the current filter value in the filter sequence corresponding to the current moment, the effective rolling radius change of the wheel is calculated as the wear depth estimate of the wheel; Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the first estimation strategy; Based on the filtering sequence corresponding to each wheel, the standard deviation of all filtering sequences within the preset time window at the current moment is determined, and the standard deviation is used as the uncertainty parameter corresponding to the first estimation strategy.

3. The tire wear depth estimation method according to claim 1, characterized in that, The at least two independent estimation strategies include a second estimation strategy based on resonance spectrum information; The calculation of the wear depth estimate and uncertainty parameters corresponding to each of the estimation strategies includes: For each wheel, the vibration spectrum corresponding to the wheel is determined based on the wheel speed at multiple times. The resonance spectrum information of the wheel at the current moment is extracted from the vibration spectrum. Based on the resonance spectrum information of the wheel at the current moment and the pre-established mapping relationship between the resonance spectrum information and the tread wear depth, the wear depth estimate of the wheel is determined. Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the second estimation strategy; The uncertainty parameters corresponding to the second estimation strategy are determined based on the peak half-width of the vibration spectrum corresponding to all wheels.

4. The tire wear depth estimation method according to claim 3, characterized in that, The step of determining the vibration spectrum corresponding to the wheel based on the wheel speed at multiple moments includes: The wheel speed at multiple moments is transformed using time-frequency conversion to obtain frequency domain data. Obtain tire temperature information and determine a temperature compensation factor based on the temperature information and a reference temperature; The frequency values ​​in the frequency domain data are corrected according to the temperature compensation factor to obtain the vibration spectrum.

5. The tire wear depth estimation method according to claim 1, characterized in that, The at least two independent estimation strategies include a third estimation strategy based on frictional energy; the calculation of the wear depth estimate and uncertainty parameters corresponding to each estimation strategy includes: The dynamic information of the vehicle is acquired, and the force and motion state of each wheel is determined based on the dynamic information. Based on the force and motion states of each wheel, determine the instantaneous friction power of each wheel; The cumulative frictional energy of each wheel is obtained by accumulating the instantaneous frictional power of each wheel. Based on a preset mapping coefficient, the cumulative friction energy of each wheel is mapped to an estimated wear depth for each wheel. Based on the wear depth estimates of each wheel, determine the wear depth estimate corresponding to the third estimation strategy; The uncertainty parameters corresponding to the third estimation strategy are determined based on the cumulative duration corresponding to the cumulative friction energy of each wheel.

6. The tire wear depth estimation method according to claim 1, characterized in that, The fusion model is a Bayesian neural network; the step of inputting the wear depth estimate and uncertainty parameter corresponding to each estimation strategy into the pre-trained fusion model to obtain the fused wear depth estimate and the corresponding fusion uncertainty parameter includes: Obtain the posterior distribution of the weights of the Bayesian neural network; Multiple sets of network weights are sampled from the weight posterior distribution, and forward propagation is performed with each set of network weights to map the wear depth estimate and uncertainty parameter corresponding to each input estimation strategy to the corresponding wear depth estimate, thereby obtaining multiple wear depth estimates. The mean of multiple wear depth estimates is used as the fused wear depth estimate, and the variance of the multiple wear depth estimates is used as the fusion uncertainty parameter.

7. The tire wear depth estimation method according to claim 1, characterized in that, After obtaining the fused wear depth estimate and fusion uncertainty parameters, the method further includes: When the fusion uncertainty parameter is less than the first preset threshold, the fusion estimate of the wear depth is output; When the fusion uncertainty parameter is greater than or equal to the first preset threshold and less than the second preset threshold, the estimation range is determined based on the wear depth fusion estimate and the estimation range is output. When the fusion uncertainty parameter is greater than or equal to the second preset threshold, an alarm message is output; wherein, the second preset threshold is greater than the first preset threshold.

8. The tire wear depth estimation method according to claim 1, characterized in that, The method further includes: Obtain the measured value of tire tread wear depth; The fusion model is updated based on the deviation between the measured value and the fused estimate of the wear depth.

9. A tire wear depth estimation device, characterized in that, include: The signal acquisition module is used to acquire multi-source signals during vehicle operation. The parallel estimation module is used to calculate the wear depth estimate and uncertainty parameter corresponding to each of the multi-source signals in parallel, using at least two independent estimation strategies respectively. The fusion estimation module is used to input the wear depth estimate and uncertainty parameters corresponding to each of the estimation strategies into the pre-trained fusion model to obtain the fused wear depth estimate and the corresponding fusion uncertainty parameters.

10. A vehicle, characterized in that, include: A control device, the control device including a memory, a processor, and a computer program; The computer program is stored in the memory and configured to be executed by the processor to implement the tire wear depth estimation method as described in any one of claims 1 to 8.