Indirect tire pressure monitoring method and system based on multi-modal fusion
By extracting wheel speed signal features using a multimodal fusion method, the accuracy and reliability issues of indirect tire pressure monitoring systems under complex operating conditions are resolved, enabling more accurate and timely tire pressure monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- DIYIN AUTOMOTIVE TECH (SHANGHAI) CO LTD
- Filing Date
- 2025-10-30
- Publication Date
- 2026-07-24
AI Technical Summary
Existing indirect tire pressure monitoring systems have low monitoring accuracy and poor alarm reliability when the vehicle is in poor operating conditions, and are prone to false alarms or missed alarms, which affects the robustness of the system.
A multimodal fusion method is adopted to extract wheel speed signal features based on three dimensions: tooth count pulse, radius, and frequency domain. By acquiring wheel speed data during vehicle operation, the angular velocity decrease value, radius decrease value, and frequency features are calculated, and threshold comparison is used to mark abnormal tire pressure.
It improves the accuracy and timeliness of tire pressure monitoring, reduces false alarms and missed alarms, and enhances the robustness of the system.
Smart Images
Figure CN121375369B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the automotive field, and more particularly to an indirect tire pressure monitoring method and system based on multimodal fusion. Background Technology
[0002] Tire pressure monitoring is a system that monitors the tire pressure of a vehicle in real time. Its core function is to prevent problems such as tire blowouts, increased fuel consumption, and excessive tire wear caused by abnormal tire pressure.
[0003] Existing tire pressure monitoring systems are mainly divided into two types: direct tire pressure monitoring systems (dTPMS) and indirect tire pressure monitoring systems (iTPMS). Direct tire pressure monitoring systems monitor tire pressure by installing tire pressure sensors in the vehicle's tires. They are structurally complex, difficult to install and maintain, and costly. Most importantly, they suffer from unreliable signal transmission. Indirect tire pressure monitoring systems determine the wheel speed difference between different wheels by using wheel speed sensor signals from the vehicle's anti-lock braking system (ABS) or electronic stability control system (ESC), and then indirectly determine the tire pressure based on this wheel speed difference.
[0004] Indirect tire pressure monitoring systems (TPMS) offer advantages such as reduced costs and easier maintenance by eliminating the need for separate sensor installations. However, existing indirect TPMS suffer from low monitoring accuracy and alarm reliability that heavily depends on the vehicle's dynamic conditions. In poor vehicle operating conditions, these systems exhibit issues like low accuracy and unreliable alarms. For instance, when a vehicle is shifting gears, braking, making sharp turns, or experiencing uneven loads, sudden tire slippage and vehicle body oscillations can occur. The varying wheel speeds of different wheels generate non-tire pressure disturbances, leading to a high rate of false alarms or missed alarms in existing indirect TPMS systems. This severely weakens the system's robustness and negatively impacts the user experience. Summary of the Invention
[0005] To address the aforementioned technical problems, the present invention aims to provide an indirect tire pressure monitoring system and method based on multimodal fusion, which can extract wheel speed signal features based on three different dimensions: tooth count pulse, radius, and frequency domain, thereby improving the accuracy and timeliness of system monitoring.
[0006] To achieve the above objectives, the present invention aims to provide an indirect tire pressure monitoring method based on multimodal fusion, comprising:
[0007] Obtain wheel speed data of the vehicle's wheels during operation;
[0008] Obtaining angular velocity decrease, radius decrease, and frequency characteristics based on the wheel speed data includes: processing the wheel speed data to obtain wheel angular velocity data, wheel radius data, and wheel frequency data; comparing the wheel speed data corresponding to at least two wheels to obtain wheel speed difference data, the wheel speed difference data including angular velocity difference data, radius difference data, and frequency difference data; and comparing the wheel speed difference data with standard deviation data to obtain relative difference data, the relative difference data including angular velocity decrease, radius decrease, and frequency characteristics.
[0009] The relative difference data is compared with the corresponding threshold. In response to at least one difference data in the relative difference data satisfying the threshold, the corresponding wheel is marked as having abnormal tire pressure.
[0010] In some implementations, the method further includes the following steps before obtaining the angular velocity decrease value, radius decrease value, and frequency characteristic value based on the wheel speed data:
[0011] Preprocessing the wheel speed data includes:
[0012] The gradient descent method is used to estimate the first correction error, and the pulse signal generated by wheel rotation in the wheel speed data is corrected based on the first correction error to obtain the first correction pulse signal.
[0013] The cumulative angular velocity of the corresponding wheel is calculated based on the first corrected pulse signal;
[0014] Based on the yaw rate, wheelbase, and cumulative angular velocity in the wheel speed data, a wheel speed correction model is established for the vehicle when turning, and the first correction pulse signal is corrected based on the correction model to generate a second correction pulse signal.
[0015] In some implementations, the step of obtaining the angular velocity decrease value based on the wheel speed data includes:
[0016] Processing the wheel speed data to obtain the angular velocity decrease value includes:
[0017] A first corrected angular velocity is generated based on the second corrected pulse signal, and the first corrected angular velocity is normalized to obtain the relative angular velocity of each wheel.
[0018] The difference in relative angular velocity between two adjacent wheels is calculated sequentially, and the difference data is filtered to generate the first filtered data.
[0019] During the self-learning phase, the average value of the difference in relative angular velocity between two adjacent wheels is obtained and used as the first reference data.
[0020] During the monitoring phase, the first difference between the first filtered data and the first reference data is calculated; the first difference is then restored based on the feature matrix to obtain the angular velocity decrease value of each wheel.
[0021] In some embodiments, processing the wheel speed data further includes:
[0022] Processing the wheel speed data to obtain the radius reduction value includes:
[0023] The overall transmission ratio is calculated based on engine speed and wheel speed, and the wheel end torque is calculated in combination with the engine output torque.
[0024] Based on the wheel-end torque and standard wheel radius, estimate the traction force at the wheel end;
[0025] A first corrected angular velocity is generated based on the second corrected pulse signal, and the slip ratio of the corresponding wheel is calculated based on the first corrected angular velocity.
[0026] The relative radius difference of the wheels is determined by the least squares method based on the slip ratio and the traction force;
[0027] During the self-learning phase, the average value of the relative radius difference between the wheels is calculated as the second benchmark data;
[0028] During the monitoring phase, a second difference between the relative radius difference and the second benchmark data is calculated; based on the correspondence between the radius difference and the relative radius difference, the second difference is converted into a radius reduction value.
[0029] In some embodiments, processing the wheel speed data further includes:
[0030] Processing the wheel speed data to obtain frequency characteristics includes:
[0031] The second corrected pulse signal is subjected to bandpass filtering to retain the wheel speed signal in a specific frequency band;
[0032] A second-order autoregressive model is constructed based on the wheel speed signal in the reserved specific frequency band, and the resonance frequency is estimated using the least squares method.
[0033] During the self-learning phase, multiple sets of resonant frequencies are collected, and their average value is calculated as the third benchmark data.
[0034] During the monitoring phase, the decrease ratio of the actual measured resonant frequency relative to the third reference data is calculated; the frequency characteristic is output after performing Kalman filtering on the decrease ratio of the third reference data.
[0035] In some implementations, before comparing the relative difference data with the corresponding threshold, the method further includes:
[0036] The angular velocity decrease value, the radius decrease value, and the frequency feature are fused according to a preset ratio to generate fused difference data;
[0037] The fused difference data is compared with the corresponding threshold. In response to the fused data meeting the threshold, the corresponding wheel is marked as having abnormal tire pressure.
[0038] According to another aspect of this application, a multimodal fusion-based indirect tire pressure monitoring system is further provided, comprising:
[0039] Wheel speed data acquisition module, used to acquire wheel speed data of the vehicle's wheels during operation;
[0040] A wheel speed data processing module is used to obtain angular velocity decrease value, radius decrease value, and frequency characteristics based on the wheel speed data. The wheel speed data processing module includes a difference acquisition unit and a relative difference acquisition unit. The difference acquisition unit is used to compare the wheel speed data corresponding to at least two wheels to obtain wheel speed difference data, which includes angular velocity difference data, radius difference data, and frequency difference data. The relative difference acquisition unit is used to compare the wheel speed difference data with standard deviation data to obtain relative difference data, which includes angular velocity decrease value, radius decrease value, and frequency characteristics.
[0041] The abnormality alert module is used to compare the relative difference data with the corresponding threshold, and in response to at least one difference data in the relative difference data satisfying the threshold, the corresponding wheel is marked as having abnormal tire pressure.
[0042] In some embodiments, the wheel speed data processing module includes a preprocessing module, the preprocessing module being used for:
[0043] The gradient descent method is used to estimate the first correction error, and the pulse signal generated by wheel rotation in the wheel speed data is corrected based on the first correction error to obtain the first correction pulse signal.
[0044] The cumulative angular velocity of the corresponding wheel is calculated based on the first corrected pulse signal;
[0045] Based on the yaw rate, wheelbase, and cumulative angular velocity in the wheel speed data, a wheel speed correction model is established for the vehicle when turning, and the first correction pulse signal is corrected based on the correction model to generate a second correction pulse signal.
[0046] In some embodiments, the wheel speed data processing module further includes an angular velocity module, the angular velocity module being used for:
[0047] A first corrected angular velocity is generated based on the second corrected pulse signal, and the first corrected angular velocity is normalized to obtain the relative angular velocity of each wheel.
[0048] The difference in relative angular velocity between two adjacent wheels is calculated sequentially, and the difference data is filtered to generate the first filtered data.
[0049] During the self-learning phase, the average value of the difference in relative angular velocity between two adjacent wheels is obtained and used as the first reference data.
[0050] During the monitoring phase, the first difference between the first filtered data and the first reference data is calculated; the first difference is then restored based on the feature matrix to obtain the angular velocity decrease value of each wheel.
[0051] In some embodiments, the wheel speed data processing module further includes a radius module, the radius module being used for:
[0052] The overall transmission ratio is calculated based on engine speed and wheel speed, and the wheel end torque is calculated in combination with the engine output torque.
[0053] Based on the wheel-end torque and standard wheel radius, estimate the traction force at the wheel end;
[0054] A first corrected angular velocity is generated based on the second corrected pulse signal, and the slip ratio of the corresponding wheel is calculated based on the first corrected angular velocity.
[0055] The relative radius difference of the wheels is determined by the least squares method based on the slip ratio and the traction force;
[0056] During the self-learning phase, the average value of the relative radius difference between the wheels is calculated as the second benchmark data;
[0057] During the monitoring phase, a second difference between the relative radius difference and the second benchmark data is calculated; based on the correspondence between the radius difference and the relative radius difference, the second difference is converted into a radius reduction value.
[0058] In some embodiments, the wheel speed data processing module further includes a frequency module, the frequency module being used for:
[0059] The second corrected pulse signal is subjected to bandpass filtering to retain the wheel speed signal in a specific frequency band;
[0060] A second-order autoregressive model is constructed based on the wheel speed signal in the reserved specific frequency band, and the resonance frequency is estimated using the least squares method.
[0061] During the self-learning phase, multiple sets of resonant frequencies are collected, and their average value is calculated as the third benchmark data.
[0062] During the monitoring phase, the decrease ratio of the actual measured resonant frequency relative to the third reference data is calculated; the frequency characteristic is output after performing Kalman filtering on the decrease ratio of the third reference data.
[0063] In some embodiments, the wheel speed data processing module further includes a fusion module, the fusion module being used for:
[0064] The angular velocity decrease value, the radius decrease value, and the frequency feature are fused according to a preset ratio to generate fused difference data;
[0065] The fused difference data is compared with the corresponding threshold. In response to the fused data meeting the threshold, the corresponding wheel is marked as having abnormal tire pressure. Attached Figure Description
[0066] The preferred embodiments will now be described in a clear and easy-to-understand manner, in conjunction with the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of the present invention.
[0067] Figure 1 This is a schematic diagram of the flow structure of the indirect tire pressure monitoring method based on multimodal fusion according to a preferred embodiment of the present invention;
[0068] Figure 2 This is a schematic diagram of the data preprocessing process in the indirect tire pressure monitoring method based on multimodal fusion according to a preferred embodiment of the present invention;
[0069] Figure 3 This is a schematic diagram of the process structure for obtaining the angular velocity decrease value in the indirect tire pressure monitoring method based on multimodal fusion according to a preferred embodiment of the present invention;
[0070] Figure 4 This is a schematic diagram of the process structure for obtaining the radius reduction value in the indirect tire pressure monitoring method based on multimodal fusion according to a preferred embodiment of the present invention;
[0071] Figure 5 This is a schematic diagram of the process structure for obtaining frequency features in the indirect tire pressure monitoring method based on multimodal fusion according to a preferred embodiment of the present invention;
[0072] Figure 6 This is a block diagram of a preferred embodiment of the indirect tire pressure monitoring system based on multimodal fusion. Detailed Implementation
[0073] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.
[0074] To keep the drawings concise, each figure only schematically shows the parts relevant to the invention, and these do not represent the actual structure of the product. Furthermore, to facilitate understanding, in some figures, only one of components with the same structure or function is schematically depicted, or only one is labeled. In this document, "one" not only means "only one," but can also mean "more than one."
[0075] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0076] In this document, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0077] Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0078] refer to Figures 1 to 5 This application provides an indirect tire pressure monitoring method 100 based on multimodal fusion, which can extract wheel speed signal features based on three different dimensions: tooth count pulse, radius, and frequency domain, thereby improving the accuracy and timeliness of system monitoring. In some embodiments, the three dimensions of tooth count pulse, radius, and frequency domain can be fused to improve the overall detection performance. Specifically, the indirect tire pressure monitoring method 100 based on multimodal fusion provided in this application includes:
[0079] S101: Obtain wheel speed data of the vehicle's wheels during operation;
[0080] S102: Obtaining angular velocity decrease, radius decrease, and frequency characteristics based on the wheel speed data includes: processing the wheel speed data to obtain wheel angular velocity data, wheel radius data, and wheel frequency data; comparing the wheel speed data corresponding to at least two wheels to obtain wheel speed difference data, the wheel speed difference data including angular velocity difference data, radius difference data, and frequency difference data; comparing the wheel speed difference data with standard deviation data to obtain relative difference data, the relative difference data including angular velocity decrease, radius decrease, and frequency characteristics;
[0081] S103: Compare the relative difference data with the corresponding threshold. In response to at least one difference data in the relative difference data satisfying the threshold, mark the corresponding wheel as having abnormal tire pressure.
[0082] The indirect tire pressure monitoring method 100 based on multimodal fusion provided in this application is an indirect tire pressure monitoring method. During the monitoring process, the wheel speed data of each wheel is collected in real time by means of the wheel speed sensors of the vehicle's ABS system, and the rotation speed signals of each wheel are continuously compared by the ECU to achieve the purpose of tire pressure monitoring.
[0083] In step S101 above, wheel speed data of the vehicle's wheels during operation is acquired by wheel speed sensors in the ABS system. Wheel speed sensors in the ABS system are typically used in conjunction with gears on the wheels. These gears consist of multiple conductive teeth spaced apart along their outer circumference. The magnetic reluctance of the gear teeth changes along the outer circumference. As the gears rotate with the wheel, the wheel speed sensors detect the passage of the gears and output a tooth detection signal indicating the passage of each gear tooth. This signal is typically a square wave pulse signal.
[0084] In step S102 above, this application processes the square wave pulse signal from three dimensions: tooth count pulse, radius, and frequency domain, to obtain the wheel angular velocity data, the wheel radius data, and the wheel frequency data. The wheel speed data corresponding to at least two wheels are compared to obtain wheel speed difference data. Specifically, taking a four-wheeled vehicle as an example, the comparison process compares the wheel speed data of the two wheels mounted on the front axle, the two wheels mounted on the rear axle, the left front wheel and left rear wheel, and the right front wheel and right rear wheel. The wheel speed difference data obtained in step S102 is compared with the standard deviation data to obtain relative difference data.
[0085] In step 103 above, the relative difference data is compared with the corresponding threshold. When the relative difference data meets the threshold, the corresponding wheel is marked as having abnormal tire pressure.
[0086] refer to Figure 2Furthermore, in step S102 above, processing the wheel speed data includes: preprocessing the wheel speed data, the preprocessing steps including:
[0087] S1011: The gradient descent method is used to estimate the first correction error, and the pulse signal generated by wheel rotation in the wheel speed data is corrected based on the first correction error to obtain the first correction pulse signal;
[0088] S1012: Calculate the cumulative angular velocity of the corresponding wheel based on the first corrected pulse signal;
[0089] S1013: Based on the yaw rate, vehicle wheelbase, wheelbase and accumulated angular velocity in the wheel speed data, establish a wheel speed correction model when the vehicle is turning, and correct the first correction pulse signal based on the correction model to generate a second correction pulse signal.
[0090] In step S1011 above, the gears that rotate with the wheel may cause periodic errors in the pulse signal collected by the wheel speed sensor due to manufacturing tolerances, installation deviations, or wear, such as wheel speed fluctuations caused by uneven tooth pitch. This error is unrelated to tire pressure changes. Eliminating the influence of this error on the pulse signal can improve the accuracy of tire pressure monitoring.
[0091] Specifically, the wheel speed signal is collected when the vehicle is traveling at a constant speed, at which time the wheel rotation speed is... Stable; signal fluctuations are caused solely by gear ring errors. Set estimated correction values. The revised theoretical period is Loss function Defined as the sum of squares of the deviations between the "corrected period" and the "actual measurement period". ,in For the first The actual measurement cycle of each tooth. Regarding the loss function. Seeking information about The partial derivatives (gradient) give the direction of correction for the error estimate: The gradient is positive, indicating that The value is too large and needs to be reduced; a negative gradient indicates... The value is too small and needs to be increased. Update the error estimate along the reverse gradient direction, gradually decreasing the loss function. ,in The learning rate controls the iteration speed, allowing adjustments based on the specific scenario to avoid slow convergence or oscillations. When the loss function... The iteration stops when the number of iterations falls below a preset threshold or the maximum number of iterations is reached. This is the optimal estimate of the gear ring error, which is also the first correction error. ,in This is the corrected pulse data, also known as the first corrected pulse signal. This refers to pulse data obtained through wheel speed sensor measurements.
[0092] In step S1013 above, the acquired wheel speed signal also includes the vehicle's yaw rate, wheel track width, and wheelbase. Based on the yaw rate, wheel track width, wheelbase, and accumulated angular velocity in the wheel speed data, a wheel speed correction model for vehicle turning is established, and the first correction pulse signal is corrected based on the correction model to compensate for wheel speed during turning and eliminate the influence of vehicle turning on the wheel speed of adjacent wheels.
[0093] refer to Figure 3 The processing of the wheel speed data further includes: processing the wheel speed data to obtain the angular velocity decrease value, specifically including:
[0094] S1021: Generate a first corrected angular velocity based on the second corrected pulse signal, and normalize the first corrected angular velocity to obtain the relative angular velocity of each wheel;
[0095] S1022: Calculate the difference in relative angular velocity between two adjacent wheels in sequence, and filter the difference data to generate the first filtered data;
[0096] S1023: During the self-learning phase, the mean value of the difference between the relative angular velocities of two adjacent wheels is obtained, and this mean value is used as the first reference data.
[0097] S1024: During the monitoring phase, calculate the first difference between the first filtered data and the first reference data; restore the first difference based on the feature matrix to obtain the angular velocity decrease value of each wheel.
[0098] In step S1021 above, the first corrected angular velocity is normalized to obtain the relative angular velocity of each wheel. Specifically, the ratio of the angular velocity of the wheel to the sum of the angular velocities of all wheels is used to reduce the influence of vehicle speed on wheel speed signal.
[0099] In step S1022 above, the difference data of the relative angular velocity of two adjacent wheels is calculated in sequence, and the difference data is filtered to generate the first filtered data, so as to eliminate instantaneous interference, retain trend changes, and the filtered data is more stable.
[0100] In step S1023 above, there is a self-learning phase during vehicle operation to acquire the average value of the difference in relative angular velocity between two adjacent wheels in real time, and use this average value as the first reference data so that the measured difference in relative angular velocity can be compared during the monitoring phase. Preferably, during the self-learning phase, the road conditions are generally relatively smooth, and the tire pressure of each wheel is close to the standard tire pressure.
[0101] In step S1024 above, during the monitoring phase, which can also be understood as the actual operation phase, the first difference between the first filtered data and the first reference data is calculated; the first difference is restored based on the feature matrix to obtain the angular velocity decrease value of each wheel.
[0102] When triggering an alarm based on angular velocity drop, two alarm threshold levels and an alarm accumulation threshold are set. These two threshold levels can be a severe threshold (e.g., -17%) and a minor threshold (e.g., -9%). After acquiring the angular velocity drop values of each wheel, they are first sorted. When the three smallest angular velocity drop values are all less than the minor threshold, a three-wheeled tire anomaly flag is activated. When the smallest angular velocity drop value is less than the severe threshold and the second smallest angular velocity drop value is higher than the minor threshold, a single-wheeled tire anomaly flag is activated. If both the three-wheeled tire anomaly flag and the single-wheeled tire anomaly flag are 0, a diagonal tire anomaly flag is activated. The diagonal tire pressure difference is calculated and accumulated when the diagonal tire anomaly flag is activated. When the accumulated count reaches a threshold, an alarm is triggered, and the location of the abnormal tire is precisely pinpointed.
[0103] refer to Figure 4 Processing the wheel speed data further includes: processing the wheel speed data to obtain a radius reduction value, including:
[0104] S1031: Calculate the total transmission ratio based on engine speed and wheel speed, and calculate the wheel end torque in combination with engine output torque;
[0105] S1032: Based on the wheel-end torque and standard wheel radius, estimate the traction force at the wheel end;
[0106] S1033: Generate a first corrected angular velocity based on the second corrected pulse signal, and calculate the slip ratio of the corresponding wheel based on the first corrected angular velocity;
[0107] S1034: Determine the relative radius difference of the wheels using the least squares method based on the slip ratio and the traction force;
[0108] S1035: During the self-learning phase, the average value of the relative radius difference of the wheels is calculated as the second reference data;
[0109] S1036: During the monitoring phase, calculate the second difference between the relative radius difference and the second reference data; based on the correspondence between the radius difference and the relative radius difference, convert the second difference into a radius reduction value.
[0110] In step S1031 above, the wheel speed signal further includes engine speed and engine output torque. Based on the engine speed and wheel speed, the overall transmission ratio can be calculated, and based on the overall transmission ratio and the output torque, the wheel-end torque can be calculated. Transmission Ratio ,in Engine speed, Wheel speed. Wheel-end torque. ,in For engine output torque, For transmission efficiency.
[0111] In step S1032 above, the traction force at the wheel end can be estimated based on the torque at the wheel end and the standard wheel radius. ,in This refers to the standard wheel radius.
[0112] In step S1033 above, a first corrected angular velocity is generated based on the second corrected pulse signal, and the slip ratio of the corresponding wheel is calculated based on the first corrected angular velocity. Slip ratio ,in This is the vehicle's actual speed. This refers to the rolling speed of the wheel.
[0113] In step S1034 above, the relative radius difference of the wheels is determined by the least squares method based on the slip ratio and the traction force. That is, the relative slip ratio and the traction force are linearly fitted by the least squares method to estimate the fitting intercept, which is the relative radius difference.
[0114] In step S1035 above, during the self-learning phase, the average value of the relative radius difference of the wheels is calculated as the second reference data.
[0115] In the above-mentioned correction S1036, during the monitoring phase, the second difference between the relative radius difference and the second reference data is calculated; based on the correspondence between the radius difference and the relative radius difference, the second difference is converted into a radius reduction value.
[0116] When triggering an alarm based on the radius decrease value, the initial settings are first performed, defining the status identifier and mask value of each wheel; the abnormal counters of each wheel are dynamically updated according to the characteristic indicators; the four wheel counters are sequentially checked, and if the counter of a certain wheel exceeds the threshold, the alarm flag is activated to mark the tire pressure as abnormal.
[0117] refer to Figure 5Processing the wheel speed data further includes: processing the wheel speed data to obtain frequency characteristics, including:
[0118] S1041: Perform bandpass filtering on the second corrected pulse signal to retain the wheel speed signal in a specific frequency band;
[0119] S1042: Construct a second-order autoregressive model based on the wheel speed signal of the retained specific frequency band, and estimate the resonance frequency using the least squares method;
[0120] S1043: During the self-learning phase, multiple sets of resonant frequencies are collected, and their average value is calculated as the third reference data.
[0121] S1044: During the monitoring phase, calculate the decrease ratio of the actual measured resonant frequency relative to the third reference data; and output the frequency characteristic after performing Kalman filtering on the decrease ratio of the third reference data.
[0122] In step S1041 above, the second corrected pulse signal is subjected to bandpass filtering to retain the wheel speed signal in a specific frequency band, thereby preserving the information of the specific frequency band and eliminating interference frequencies.
[0123] In step S1042 above, a second-order autoregressive model is constructed based on the wheel speed signal in the retained specific frequency band, and the resonance frequency is estimated using the least squares method. The second-order autoregressive model is preferably an AR autoregressive model, which can predict the current value using historical data. The second-order AR model can be expressed as: ,in , For model parameters, This represents noise. The characteristic equation of the model is: The reciprocal of its root corresponds to the resonant frequency of the signal, through The transformation is mapped to the frequency domain. This is applied to the filtered wheel speed signal. Construct the error function. The optimal solution is obtained by using the least squares method. and Then calculate the resonant frequency. .
[0124] In step S1043 above, during the self-learning phase, multiple sets of resonant frequencies are collected, and their average value is calculated as the third reference data.
[0125] In step S1044 above, during the monitoring phase, the decrease ratio of the actually measured resonant frequency relative to the third reference data is calculated; the frequency characteristic is output after performing Kalman filtering on the decrease ratio of the third reference data. In this step, the resonant frequency may change abruptly due to instantaneous road surface disturbances (such as going over speed bumps). Kalman filtering, through "prediction-update" iteration and combined with the statistical characteristics of the signal (such as noise variance), can smooth high-frequency fluctuations and preserve trend changes.
[0126] When triggering alarms based on frequency characteristics, the system first checks whether the frequency characteristics of each wheel are below a preset threshold. If the frequency characteristics of all wheels are below the preset threshold, a global alarm flag is activated. When the combined conditions of "global alarm flag activation" are met, the cumulative count increments; otherwise, it decrements. When the cumulative count reaches a preset threshold, a system-wide and all-wheel abnormality alarm is triggered. This "increment-decrement" logic filters out transient anomalies to avoid false alarms; the decrement step size can be adjusted according to the required duration of the anomaly.
[0127] Furthermore, before comparing the relative difference data with the corresponding threshold, the indirect tire pressure monitoring method based on multimodal fusion further includes:
[0128] S104: The angular velocity decrease value, the radius decrease value, and the frequency feature are fused according to a preset ratio to generate fused difference data;
[0129] S105: Compare the fused difference data with the corresponding threshold. In response to the fused data meeting the threshold, mark the corresponding wheel as having abnormal tire pressure.
[0130] In step S104 above, the relative differences in angular velocity, radius, and frequency are fused according to a preset ratio based on these three types of features. This allows for the complementarity of multi-dimensional features, reducing the risk of misjudgment when a single feature is interfered with, and ultimately outputting more reliable fused difference data. (Fused data) ,in This represents the relative frequency difference. The relative difference in radius. This represents the relative difference in angles. In some implementations, the formula above... , as well as The proportions are all fixed values. In some implementations, based on road condition characteristics, , as well as The value also changes accordingly.
[0131] refer to Figure 6According to another aspect of this application, a multimodal fusion-based indirect tire pressure monitoring system 200 is further provided, comprising: a wheel speed data acquisition module 10, a wheel speed data processing module 20, and an anomaly alert module 30. The wheel speed data acquisition module 10 is used to acquire wheel speed data of the wheels when the vehicle is running; the wheel speed data processing module 20 is used to obtain angular velocity decrease value, radius decrease value, and frequency characteristics based on the wheel speed data; the wheel speed data processing module 20 includes a difference acquisition unit 21 and a relative difference acquisition unit 22, the difference acquisition unit 21 is used to compare the wheel speed data corresponding to at least two wheels to obtain wheel speed difference data, the wheel speed difference data includes angular velocity difference data, radius difference data, and frequency difference data; the relative difference acquisition unit 22 is used to compare the wheel speed difference data with standard deviation data to obtain relative difference data, the relative difference data includes angular velocity decrease value, radius decrease value, and frequency characteristics; the anomaly alert module 30 is used to compare the relative difference data with a corresponding threshold, and in response to at least one difference data in the relative difference data satisfying the threshold, the corresponding wheel is marked as having abnormal tire pressure.
[0132] The wheel speed data processing module 20 includes a preprocessing module 23, which is used to: estimate a first correction error using the gradient descent method; correct the pulse signal generated by wheel rotation in the wheel speed data based on the first correction error to obtain a first correction pulse signal; calculate the cumulative angular velocity of the corresponding wheel based on the first correction pulse signal; establish a wheel speed correction model when the vehicle is turning based on the yaw rate, vehicle wheelbase, wheelbase and the cumulative angular velocity in the wheel speed data; and correct the first correction pulse signal based on the correction model to generate a second correction pulse signal.
[0133] The wheel speed data processing module 20 further includes an angular velocity module 24, which is used to: generate a first corrected angular velocity based on the second corrected pulse signal; normalize the first corrected angular velocity to obtain the relative angular velocity of each wheel; sequentially calculate the difference data of the relative angular velocities of two adjacent wheels, and filter the difference data to generate first filtered data; in the self-learning stage, obtain the mean data of the difference data of the relative angular velocities of two adjacent wheels, and use the mean data as the first reference data; in the monitoring stage, calculate the first difference between the first filtered data and the first reference data; and restore the first difference based on the feature matrix to obtain the angular velocity decrease value of each wheel.
[0134] The wheel speed data processing module 20 further includes a radius module 25, which is used for: calculating the total transmission ratio based on engine speed and wheel speed, and calculating the wheel-end torque in combination with engine output torque; estimating the wheel-end traction force based on the wheel-end torque and standard wheel radius; generating a first corrected angular velocity based on the second corrected pulse signal, and calculating the slip ratio of the corresponding wheel based on the first corrected angular velocity; determining the relative radius difference of the wheels using the least squares method based on the slip ratio and the traction force; calculating the average value of the relative radius difference of the wheels as the second reference data during the self-learning phase; calculating the second difference between the relative radius difference and the second reference data during the monitoring phase; and converting the second difference into a radius reduction value based on the correspondence between the radius difference and the relative radius difference.
[0135] The wheel speed data processing module 20 further includes a frequency module 26, which is used to: perform bandpass filtering on the second corrected pulse signal to retain the wheel speed signal in a specific frequency band; construct a second-order autoregressive model based on the retained wheel speed signal in the specific frequency band, and estimate the resonance frequency using the least squares method; in the self-learning stage, collect multiple sets of resonance frequencies and calculate their mean as the third reference data; in the monitoring stage, calculate the decrease ratio of the actually measured resonance frequency relative to the third reference data; and output the frequency feature after performing Kalman filtering on the decrease ratio of the third reference data.
[0136] The wheel speed data processing module 20 further includes a fusion module 27, which is used to: fuse the angular velocity decrease value, the radius decrease value, and the frequency feature according to a preset ratio to generate fused difference data; compare the fused difference data with a corresponding threshold, and in response to the fused data meeting the threshold, mark the corresponding wheel as having abnormal tire pressure.
[0137] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the invention. The advantages of the present invention have been fully and effectively realized. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments; any variations or modifications can be made to the implementation of the present invention without departing from these principles.
Claims
1. An indirect tire pressure monitoring method based on multimodal fusion, characterized in that, include: Obtain wheel speed data of the vehicle's wheels during operation; The wheel speed data is preprocessed, including: estimating a first correction error using the gradient descent method; correcting the pulse signal generated by wheel rotation in the wheel speed data based on the first correction error to obtain a first correction pulse signal; calculating the cumulative angular velocity of the corresponding wheel based on the first correction pulse signal; establishing a wheel speed correction model for vehicle turning based on the yaw rate, vehicle track width, wheelbase, and the cumulative angular velocity in the wheel speed data; and correcting the first correction pulse signal based on the correction model to generate a second correction pulse signal. Based on the wheel speed data, the angular velocity decrease value, radius decrease value, and frequency characteristics are obtained, including: The step of obtaining the angular velocity decrease value based on the wheel speed data includes: generating a first corrected angular velocity based on the second corrected pulse signal, normalizing the first corrected angular velocity to obtain the relative angular velocity of each wheel; sequentially calculating the difference data of the relative angular velocities of two adjacent wheels, and filtering the difference data to generate first filtered data; in the self-learning stage, obtaining the mean data of the difference data of the relative angular velocities of two adjacent wheels, and using the mean data as the first reference data; in the monitoring stage, calculating the first difference between the first filtered data and the first reference data; and restoring the first difference based on the feature matrix to obtain the angular velocity decrease value of each wheel. Obtaining the radius reduction value based on the wheel speed data includes: calculating the total transmission ratio based on engine speed and wheel speed, and calculating the wheel-end torque based on engine output torque; estimating the wheel-end traction force based on the wheel-end torque and standard wheel radius; generating a first corrected angular velocity based on the second corrected pulse signal, and calculating the slip ratio of the corresponding wheel based on the first corrected angular velocity; determining the relative radius difference of the wheels using the least squares method based on the slip ratio and the traction force; calculating the average value of the relative radius difference of the wheels as the second reference data during the self-learning phase; calculating the second difference between the relative radius difference and the second reference data during the monitoring phase; and converting the second difference into a radius reduction value based on the correspondence between the radius difference and the relative radius difference. Obtaining frequency features based on the wheel speed data includes: performing bandpass filtering on the second corrected pulse signal to retain wheel speed signals in a specific frequency band; constructing a second-order autoregressive model based on the retained wheel speed signals in the specific frequency band, and estimating the resonant frequency using the least squares method; during the self-learning phase, collecting multiple sets of resonant frequencies and calculating their mean as the third reference data; during the monitoring phase, calculating the decrease ratio of the actually measured resonant frequency relative to the third reference data; and outputting the frequency features after performing Kalman filtering on the decrease ratio of the third reference data. The decrease in angular velocity, the decrease in radius, and the frequency feature are compared with corresponding thresholds. In response to at least one of the decrease in angular velocity, the decrease in radius, and the frequency feature satisfying the threshold, the corresponding wheel is marked as having abnormal tire pressure.
2. The indirect tire pressure monitoring method based on multimodal fusion according to claim 1, characterized in that, Before comparing the angular velocity decrease value, radius decrease value, and frequency feature with the corresponding threshold, the process further includes: The angular velocity decrease value, the radius decrease value, and the frequency feature are fused according to a preset ratio to generate fused difference data; The fusion difference data is compared with the corresponding threshold. In response to the fusion difference data meeting the threshold, the corresponding wheel is marked as having abnormal tire pressure.
3. An indirect tire pressure monitoring system based on multimodal fusion, characterized in that, include: Wheel speed data acquisition module, used to acquire wheel speed data of the vehicle's wheels during operation; The wheel speed data processing module is used to obtain the angular velocity decrease value, radius decrease value, and frequency characteristics based on the wheel speed data; The wheel speed data processing module includes a preprocessing module, which is used to: estimate a first correction error using the gradient descent method, and correct the pulse signal generated by wheel rotation in the wheel speed data based on the first correction error to obtain a first correction pulse signal; The cumulative angular velocity of the corresponding wheel is calculated based on the first corrected pulse signal; Based on the yaw rate, vehicle wheelbase, wheelbase, and accumulated angular velocity in the wheel speed data, a wheel speed correction model for vehicle turning is established, and the first correction pulse signal is corrected based on the correction model to generate a second correction pulse signal. The wheel speed data processing module further includes an angular velocity module, which is used to: generate a first corrected angular velocity based on the second corrected pulse signal, and normalize the first corrected angular velocity to obtain the relative angular velocity of each wheel; The difference in relative angular velocity between two adjacent wheels is calculated sequentially, and the difference data is filtered to generate the first filtered data. During the self-learning phase, the mean value of the difference in relative angular velocity between two adjacent wheels is obtained and used as the first reference data. During the monitoring phase, the first difference between the first filtered data and the first reference data is calculated; The first difference is reconstructed based on the feature matrix to obtain the angular velocity decrease value of each wheel; The wheel speed data processing module further includes a radius module, which is used to: calculate the total transmission ratio based on the engine speed and wheel speed, and calculate the wheel end torque in combination with the engine output torque; Based on the wheel-end torque and standard wheel radius, estimate the traction force at the wheel end; A first corrected angular velocity is generated based on the second corrected pulse signal, and the slip ratio of the corresponding wheel is calculated based on the first corrected angular velocity; the relative radius difference of the wheels is determined by the least squares method based on the slip ratio and the traction force. During the self-learning phase, the average value of the relative radius difference between the wheels is calculated as the second benchmark data; During the monitoring phase, a second difference between the relative radius difference and the second benchmark data is calculated; based on the correspondence between the radius difference and the relative radius difference, the second difference is converted into a radius reduction value; The wheel speed data processing module further includes a frequency module, which is used to: perform bandpass filtering on the second corrected pulse signal to retain wheel speed signals in a specific frequency band; construct a second-order autoregressive model based on the retained wheel speed signals in the specific frequency band, and estimate the resonance frequency using the least squares method; in the self-learning stage, collect multiple sets of resonance frequencies and calculate their mean as the third reference data; in the monitoring stage, calculate the decrease ratio of the actually measured resonance frequency relative to the third reference data. The frequency characteristics are output after Kalman filtering the descent ratio of the third reference data. The abnormality alert module is used to compare the angular velocity decrease value, the radius decrease value, and the frequency feature with the corresponding thresholds, and in response to at least one of the angular velocity decrease value, the radius decrease value, and the frequency feature satisfying the threshold, to mark the corresponding wheel as having abnormal tire pressure.
4. The indirect tire pressure monitoring system based on multimodal fusion according to claim 3, characterized in that, The wheel speed data processing module further includes a fusion module, the fusion module being used for: The angular velocity decrease value, the radius decrease value, and the frequency feature are fused according to a preset ratio to generate fused difference data; The fusion difference data is compared with the corresponding threshold. In response to the fusion difference data meeting the threshold, the corresponding wheel is marked as having abnormal tire pressure.