An intelligent tire pressure early warning method for internet of vehicles

Abstract

The application provides a tire pressure intelligent early warning method for Internet of Vehicles, aiming at solving the problems of high false alarm rate and low early warning granularity of the existing tire pressure early warning method. The application first collects multi-dimensional data such as tire pressure, temperature and vehicle speed in real time, and constructs a multi-dimensional feature vector after temperature compensation; then initializes a sliding window self-adjusting incremental isolation forest model, filters abnormal samples through a credibility mechanism and locally updates the model, and dynamically adjusts the window size combined with the complexity of the driving scene; finally, based on the abnormal score, three-level early warning is triggered and pushed to the driver's end. The application effectively solves the defects of high false alarm rate, inability to identify early slow air leakage and low early warning granularity of the traditional method, realizes accurate, early and differentiated early warning of tire pressure abnormalities, and improves driving safety and early warning practicality.

Description

Technical Field

[0001] This invention relates to the field of vehicle network fault early warning, and in particular to a vehicle network tire pressure intelligent early warning method. Background Technology

[0002] With continuous technological advancements, the Internet of Vehicles (IoV) has become a hot topic in the field of intelligent transportation. IoV connects vehicles to the internet, creating an intelligent interconnected platform that integrates resources such as vehicles and roads. Among these features, fault warning is one of the most crucial aspects of IoV, and tire pressure monitoring is a key component of fault warning systems.

[0003] Traditional tire pressure monitoring systems include direct tire pressure monitoring and indirect tire pressure monitoring. Indirect tire pressure monitoring uses the difference in ABS wheel speed sensors to determine tire pressure abnormalities. This method has low accuracy and has been gradually phased out. In direct tire pressure monitoring, each tire has a built-in pressure / temperature sensor to monitor tire pressure abnormalities. It is mainly divided into static threshold method, that is, the tire pressure is lower than a certain threshold and an alarm is triggered; relative difference method, that is, the maximum difference in tire pressure of multiple tires of the vehicle is greater than a certain threshold and an alarm is triggered; and simple trend method, that is, the tire pressure drops to a certain threshold within a certain time and an alarm is triggered. Traditional tire pressure monitoring systems have the following problems: (1) High false alarm rate, unable to distinguish between real leaks and normal fluctuations. For example, low tire pressure in winter can easily trigger false alarms; (2) Unable to detect early slow leaks. After a tire is punctured by a nail, as long as the pressure is higher than the threshold for several days, an alarm still cannot be triggered; (3) Low warning granularity, only outputting a binary output of alarm / no alarm. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a vehicle-to-everything (V2X) intelligent tire pressure warning method, the specific steps of which are as follows:

[0005] S1: Collect various vehicle data in real time through the vehicle terminal and report them to the vehicle network platform, process the various vehicle data, and construct multi-dimensional feature vectors in real time.

[0006] The vehicle data includes initial tire pressure, tire temperature, ambient temperature, vehicle speed, and vehicle acceleration.

[0007] The multidimensional feature vector is a structured data set that integrates different time dimensions and multidimensional parameters, used to comprehensively characterize the dynamic changes in tire pressure.

[0008] Further steps include the following sub-steps:

[0009] S11: The vehicle terminal collects various vehicle data according to the preset collection tasks and reports them to the vehicle network platform; The vehicle terminal collects raw tire pressure, tire temperature, ambient temperature, vehicle speed, and vehicle acceleration in real time according to the preset collection tasks and at the preset collection cycle frequency, and reports them to the vehicle network platform.

[0010] The vehicle-mounted terminal is a vehicle-mounted hardware device deployed on a vehicle, capable of data collection, preliminary processing, and remote communication; the vehicle network platform is a vehicle network service center capable of data reception, storage, processing, model calculation, and early warning push.

[0011] Preferably, the preset acquisition period is set to 30 seconds.

[0012] By collecting multi-dimensional data, it breaks through the limitations of traditional single tire pressure data collection, enriches the data dimensions, and avoids judgment bias caused by one-sided data.

[0013] S12: Process the collected vehicle data and construct a multi-dimensional feature vector; First, temperature compensation is performed on the original tire pressure based on the ideal gas law, and the compensated tire pressure is calculated using the following formula: ; in and These are the initial tire pressure and tire temperature at time t, respectively. It is the standard reference temperature, taken in the algorithm. =20. , All units are degrees Celsius; The time t refers to the time point at which various vehicle data are collected according to the preset collection cycle of S11. In other words, time t is the time point at which various vehicle data are collected once every preset collection cycle.

[0014] The compensated tire pressure at time t was calculated. By accurately offsetting the effect of temperature on tire pressure through temperature compensation, it greatly improves the problem of false alarms caused by naturally low tire pressure in winter.

[0015] Then, a multidimensional feature vector is constructed, including the following five parameters (1)-(5): (1) Rate of change of tire pressure in the past 10 minutes It captures dynamic changes in tire pressure and identifies slow leaks such as nail punctures at an early stage; among which... This is the tire pressure after nearly 10 minutes of compensation. The change value; (2) Pearson correlation coefficient between tire pressure and temperature over the past 30 minutes : , among which It is the average tire pressure after compensation over the past 30 minutes. It is the average tire temperature over the past 30 minutes; (3) The deviation of the tire pressure in the past hour from the historical baseline under the same working conditions : Query historical data for the same vehicle speed (v) and ambient temperature range, and calculate the compensated tire pressure for the past hour. mean Tire pressure deviation value for: ; Avoid misjudging normal fluctuations under different operating conditions by using tire pressure deviation values; (4) Standard deviation of tire pressure fluctuation over the past 5 minutes : ,in = ,in This represents the average tire pressure after compensation over the past 5 minutes. It represents the number of multidimensional feature vectors constructed within the last 5 minutes. This represents the compensated tire pressure value at the i-th acquisition time within the last 5 minutes, where i is the index of the constructed multidimensional feature vector (i=1,2,...). By identifying drastic fluctuations through the standard deviation of tire pressure fluctuations, it helps determine whether a sensor malfunction or a sudden leak has occurred, while also ruling out minor, normal fluctuations. (5) Average vehicle speed in the last 5 minutes; calculate the average vehicle speed in the last 5 minutes based on the speeds v of all vehicles collected in the last 5 minutes.

[0016] The five parameters in the aforementioned multidimensional feature vector are designed based on complementary multi-timescale and multi-correlation dimensions, which improves the model's adaptability to complex scenarios, expands the ability to monitor equipment anomalies beyond tire pressure warning, and further enhances the reliability and practicality of the warning method.

[0017] S2: The multidimensional feature vector constructed from safety data is used as a multidimensional feature sample. The AWS-IIF model is initialized based on the multidimensional feature sample. Untrustworthy multidimensional feature samples are filtered by calculating the confidence score. The sliding window data is updated online to obtain the updated AWS-IIF model adapted to real-time vehicle conditions and driving scenarios, and the sliding window size is dynamically adjusted.

[0018] The safety data refers to data selected from various vehicle data that shows no repair records for one consecutive week. The AWS-IIF model is a sliding window self-adjusting incremental isolated forest model, which is an isolated forest model that combines sliding window dynamic adjustment and incremental update. It is used to learn normal tire pressure patterns in real time, accurately quantify the degree of abnormality, and the sliding window stores the most recent valid data. Incremental updates enable dynamic model adaptation.

[0019] Further steps include the following sub-steps:

[0020] S21: Initialize the AWS-IIF model using secure data; After S1 completes the collection of various vehicle data and the construction of multi-dimensional feature vectors, for the first modeling scenario of the vehicle, safety data of vehicles that have no maintenance records for one consecutive week are selected to provide a clean data source for the model to learn the normal tire pressure feature pattern.

[0021] n multidimensional feature vectors were constructed from the filtered secure data. As n multidimensional feature samples .

[0022] Then use the n multidimensional feature samples Building the initial AWS-IIF model ,in It refers to the number of trees; Each tree in (k=1,2,..., Constructed through recursive random partitioning, with the partitioning rule as follows: , ; Where d is the feature dimension, each separator randomly selects one dimension, and the separator point is the mean distribution between the maximum and minimum values ​​of the selected dimension; Indicates uniform distribution. Let represent the set of values ​​for all samples on the q-th feature.

[0023] Preferably, n takes the value 256 in the algorithm. The value of d is 100 in the algorithm, and the value of d is 5 in the algorithm.

[0024] For the input multidimensional feature samples Its abnormal scores The path length is calculated from the isolated forest path length, specifically as follows:

[0025] in, It is a multidimensional feature sample The path length (t=1,2,...,n) in a single tree, i.e. the number of edges from the root node to the leaf node; It is a multidimensional feature sample The average path length across all trees in the forest; ) is the normalization factor, and n is the number of training samples, defined as: , Abnormal scores The closer it is to 1, the more abnormal it is.

[0026] This step initializes the AWS-IIF model based on security data to ensure the initial identification accuracy of the model; and calculates anomaly scores to quantify the degree of anomaly, providing a core basis for graded early warning.

[0027] S22: Calculate the real-time data credibility score to filter out untrustworthy multidimensional feature samples, update the sliding window data, and trigger a local update of the AWS-IIF model to obtain the updated AWS-IIF model; To prevent anomalous samples from contaminating the AWS-IIF model, a credibility mechanism for physical rule constraints is introduced, setting the following conditions (A) and (B) to verify the rule function. : (A) Tire pressure drops but tire temperature rises, specifically: Take all multidimensional feature samples over 10 minutes and calculate the Pearson correlation coefficient between the compensated tire pressure and tire temperature. ; in, This represents the compensated tire pressure corresponding to the i-th multidimensional feature sample within the past 10 minutes. The value is the tire temperature corresponding to the i-th multidimensional feature sample in the last 10 minutes; i is the index of all multidimensional feature samples in the last 10 minutes (i=1,2,…,N, where N is the total number of multidimensional feature samples in the last 10 minutes).

[0028] When Pearson correlation coefficient If the value is greater than 0.3, then the i-th multidimensional feature sample within the last 10 minutes is considered abnormal, satisfying condition (A). = false indicates that the i-th multidimensional feature sample in the last 10 minutes is unreliable.

[0029] (B) The vehicle speed is 0, but the tire pressure fluctuates drastically, specifically: Take all multidimensional feature samples from the most recent 5 minutes when the vehicle speed v is 0, and calculate the slope of the pressure linear regression: ;

[0030] in This refers to the collection time point corresponding to the i-th multidimensional feature sample within the most recent 5 minutes when the vehicle speed v is 0. This refers to the arithmetic mean of all multidimensional feature samples collected over the last 5 minutes when the vehicle speed v is 0. This refers to the compensated tire pressure corresponding to the i-th multi-dimensional feature sample within the last 5 minutes when the vehicle speed v is 0. It refers to the arithmetic mean of the compensated tire pressure corresponding to all multidimensional feature samples within the most recent 5 minutes when the vehicle speed v is 0.

[0031] When the slope of the linear regression of pressure If the value is greater than 0.8, it is determined to be a sensor malfunction or tire leak, and is considered abnormal data, meeting condition (B). = false indicates that the i-th multidimensional feature sample within the last 5 minutes where the vehicle speed v is 0 is unreliable.

[0032] The rule function This is a sample credibility verification function designed based on the physical laws of vehicle operation. Its core function is to determine whether multi-dimensional feature samples conform to the logical laws of normal vehicle conditions. The output is a Boolean value (only true or false), which is used to calculate the credibility score. The core basis.

[0033] Furthermore, if the multidimensional feature samples conform to physical laws, that is, they do not satisfy either condition (A) or condition (B), then = true indicates that the multidimensional feature sample is reliable; Credibility Score The calculation formula is: ; in Multidimensional feature samples Input initial model The calculated abnormal scores; If multidimensional feature samples Credibility score If the value is greater than 0.6, then update the sliding window data according to steps a)-c) as follows: a) Input the current multidimensional feature samples Add to window; b) If the window size is full, remove the oldest multidimensional feature sample according to the first-in-first-out principle; c) Randomly select 10% of the trees in the forest and perform local reconstruction, while leaving the other trees unchanged.

[0034] This step ultimately yields an updated AWS-IIF model adapted to real-time vehicle conditions and driving scenarios. By locally reconstructing the trees in the forest, it avoids abnormal samples from contaminating the model, improves the real-time performance of the AWS-IIF model, adapts to dynamic changes in vehicle conditions, and further reduces false alarms.

[0035] S23: Calculate the complexity index of the driving scenario and dynamically adjust the size of the sliding window; The driving scenario complexity index The calculation formula is: ;

[0036] in yes The variance of vehicle speed over time, where yes The variance of acceleration over time, It is the absolute value of the Pearson correlation coefficient between vehicle speed v and vehicle acceleration a, calculated in the same way as in S22. same; , , These are all weight values; Better, weight , , The algorithm takes values ​​of 0.5, 0.4, and 0.1 respectively. Values ​​are taken over 5 minutes.

[0037] A smaller value indicates a smoother vehicle condition, and the sliding window can be enlarged. A larger value indicates a more complex vehicle condition. The sliding window should be narrowed to avoid polluting the model with noisy data.

[0038] Then, the sliding window size is dynamically adjusted using the following linear piecewise mapping function: ;

[0039] in, This is the current maximum capacity of the sliding window. This is the maximum capacity of the sliding window. This is the lower limit of the sliding window capacity. It is the complexity index of the current driving scenario. and These are the minimum and maximum complexity thresholds, respectively.

[0040] Ideally, the algorithm should take... =1000, =500, =50, =100.

[0041] This step reduces noise interference in complex scenarios and ensures data volume in stable scenarios, so as to balance the accuracy of early warning in different scenarios.

[0042] S3: Input the multidimensional feature vector generated in real time by S1 into the updated AWS-IIF model obtained by S2, calculate the anomaly score of the current multidimensional feature vector according to the preset period, and perform early warning classification judgment.

[0043] The multidimensional feature vector generated in real time by S1 is a multidimensional feature vector that is constructed and updated in real time after the various data of the vehicle collected in real time in S1 are calculated and processed. It is the input data that the updated AWS-IIF model can directly recognize.

[0044] The latest constructed multidimensional feature vector of the vehicle is obtained every preset period to obtain the current multidimensional feature vector. And following the calculation logic of the abnormal score in S21, based on the current multidimensional feature vector Calculate its abnormal score ; If satisfied If the value is greater than 0.5 and continues for 3 preset periods, a Level 1 warning will be triggered. If satisfied If the value is greater than 0.7 and continues for two preset periods, a Level 2 warning will be triggered. If satisfied If the value is greater than 0.9 and continues for one preset period, a Level 3 warning will be triggered. Preferably, the preset period is set to 3 minutes.

[0045] Based on the triggered warning level, the vehicle-to-everything (V2X) platform pushes a warning message to the driver's device, prompting them to check the vehicle's tire pressure.

[0046] Furthermore, the Level 1 warning is a mild anomaly warning, with slow changes, indicating an early and slight abnormal trend in tire pressure with low risk. The vehicle-to-everything (V2X) platform prompts the driver to pay attention to the tire pressure status. The Level 2 warning is a moderate anomaly warning, with faster changes, indicating an escalation of the tire pressure abnormality, with increased leakage rate or deviation, posing a certain risk. The V2X platform prompts the driver to promptly check for tire leaks or sensor malfunctions. The Level 3 warning is a severe anomaly warning, indicating a drastic abnormality in tire pressure with extremely high risk, potentially endangering driving safety. The V2X platform prompts the driver to immediately stop at a safe location and take emergency measures.

[0047] By using a three-tiered early warning system, the granularity of warnings is improved; different levels correspond to different degrees of urgency, enabling drivers to respond accurately and avoid overreacting or ignoring risks.

[0048] The beneficial effects of the intelligent tire pressure warning method for vehicle networking of the present invention are as follows: (1) This invention eliminates temperature interference through temperature compensation, judges the difference between working conditions by multi-dimensional feature fusion, filters abnormal samples by credibility mechanism, verifies the continuous cycle of three-level early warning, and reduces the false alarm rate through multi-dimensional collaboration, so that the driver only receives real and effective early warnings and reduces unnecessary parking inspections and driving interference; (2) This invention introduces the pressure change rate and the deviation feature of the historical baseline under the same working conditions to accurately capture the slow downward trend of tire pressure. When the tire pressure is still higher than the traditional threshold, early slow leaks can be identified, giving the driver enough time to deal with the problem and avoiding tire damage and driving safety risks; (3) This invention outputs quantitative abnormal scores through the AWS-IIF model, sets up a three-level early warning mechanism, and provides graded prompts according to the degree of abnormality and duration, so that the driver can clearly know the degree of urgency and take appropriate measures, which avoids excessive panic and prevents the neglect of serious faults; In addition, the incremental update of the model and the dynamic adjustment of the window design also bring additional advantages of strong scene adaptability and high real-time performance, ensuring the accuracy and efficiency of early warning in long-term use. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the overall workflow of a vehicle-to-everything (V2X) tire pressure intelligent warning method according to the present invention. Detailed Implementation

[0050] To provide a further understanding of the purpose, structure, features, and functions of the present invention, detailed descriptions are provided below with reference to specific embodiments.

[0051] like Figure 1 As shown, the present invention provides a vehicle-to-everything (V2X) intelligent tire pressure warning method, comprising the following steps:

[0052] S1: Collect various vehicle data in real time through the in-vehicle terminal and report it to the vehicle network platform, and process the vehicle data to construct a multi-dimensional feature vector in real time; including the following sub-steps:

[0053] S11: The vehicle terminal collects various vehicle data according to the preset collection tasks and reports them to the vehicle network platform; A commercial truck was selected as the implementation target. The vehicle is equipped with an onboard terminal that meets automotive-grade standards. The onboard terminal collects various vehicle data in real time according to a preset data collection task, with a preset collection cycle of 30 seconds, including the original tire pressure of the vehicle's four tires. Tire temperature The data includes ambient temperature, vehicle speed v, and vehicle acceleration a. During the data collection process, a unique timestamp is added to each data item to ensure time sequence consistency. Then, the collected vehicle data is reported to the vehicle network platform in real time through a 4G / 5G automotive-grade communication module according to a preset reporting protocol. After receiving the collected vehicle data, the vehicle-to-everything (V2X) platform stores it in a structured database and archives it by timestamp and tire number to form a traceable raw dataset.

[0054] This step overcomes the limitations of traditional tire pressure monitoring, which only collects single tire pressure data, by collecting multi-dimensional correlated data, thus avoiding judgment bias caused by incomplete data.

[0055] S12: Process the collected vehicle data and construct a multi-dimensional feature vector; The vehicle-to-everything (V2X) platform first calls a preset temperature compensation algorithm, based on the ideal gas law, to adjust the original tire pressure. Perform temperature calibration according to the formula. Perform the calculation, where That's the original tire pressure. It's the tire temperature. =20℃ is the standard reference temperature, and the final compensated tire pressure is obtained. .

[0056] Subsequently, based on the compensated tire pressure Using other multi-source data, a structured multidimensional feature vector containing five parameters is constructed: The first parameter is the rate of pressure change over the past 10 minutes. By extracting all compensated tire pressures from the 10 minutes preceding the current moment. Calculate the total change during this period. With time span The ratio provides characteristic support for the early identification of slow leaks; The second parameter is the Pearson correlation coefficient between tire pressure and temperature over the past 30 minutes. The average tire pressure after compensation was calculated over the past 30 minutes. and tire temperature mean Substitute into the correlation coefficient formula The results are used to identify invalid data to reduce the risk of false alarms; The third parameter is the deviation of the tire pressure over the past hour from the historical baseline under the same operating conditions. By querying historical data stored on the platform, the system filters historical data that are within the same range as the current vehicle speed (v) and ambient temperature to calculate the compensated tire pressure for the past hour. mean Then calculate the tire pressure deviation value. To avoid misjudging normal fluctuations under different operating conditions; The fourth parameter is the standard deviation of tire pressure fluctuation over the past 5 minutes. First, calculate the average tire pressure after compensation over the past 5 minutes. = Then substitute it into the standard deviation formula Calculated It is used to identify drastic fluctuations in tire pressure; The fifth parameter is the average vehicle speed over the past 5 minutes. By calculating the arithmetic mean of the speeds v of all vehicles over the past 5 minutes, a basis is provided for subsequent working condition matching and driving scenario complexity calculation.

[0057] The final output is a multi-dimensional feature vector that associates the timestamp with the tire number. The temperature compensation process at this stage accurately offsets the effect of temperature on tire pressure, improving the false alarm problem caused by naturally low tire pressure in winter. The pressure change rate and the deviation value of the baseline under the same working conditions provide key feature support for solving the problem of failing to detect early slow leaks.

[0058] S2: The multidimensional feature vector constructed from safety data is used as multidimensional feature samples. The AWS-IIF model is initialized based on the multidimensional feature samples, and untrustworthy multidimensional feature samples are filtered out by calculating a confidence score. The sliding window data is updated online to obtain an updated AWS-IIF model adapted to real-time vehicle conditions and driving scenarios, and the sliding window size is dynamically adjusted. This includes the following sub-steps:

[0059] S21: Initialize the AWS-IIF model using secure data; The safety data refers to data selected from various vehicle data that has no maintenance records for one consecutive week, ensuring that the data corresponds to the normal tire pressure status. n multidimensional feature vectors were constructed from the filtered secure data. As n multidimensional feature samples Each multidimensional feature sample contains 5 feature parameters, and the feature dimension d=5.

[0060] With n multidimensional feature samples Build an initial AWS-IIF model with 100 trees for the training data. Each tree is constructed by recursively random partitioning, that is, one feature dimension q is randomly selected from 5 feature dimensions, and a partition point p is randomly selected between the minimum and maximum values ​​of the feature value range. Subsample sets are divided according to the relationship between the feature value and the partition point. This process is repeated recursively until the subsample sets meet the preset termination condition. At the same time, the rules for calculating anomaly scores are clearly defined for input multidimensional feature samples. Calculate its path length in each tree. And take the average value As multidimensional feature samples The average path length across all trees in the forest is calculated using a normalization factor based on the number of training samples. Substitute into the formula Obtain abnormal scores ; Abnormal scores The value range is [0,1], and the closer it is to 1, the more abnormal the sample is. The final output is the initial AWS-IIF model. The initial model is built using safety data to ensure it has the basic ability to accurately identify normal tire pressure patterns, while the quantitative logic of abnormal scores provides the core basis for subsequent graded warnings.

[0061] S22: Calculate the real-time data credibility score to filter out untrustworthy multidimensional feature samples, update the sliding window data, and trigger a local update of the AWS-IIF model to obtain the updated AWS-IIF model; Multidimensional feature samples based on input Initial AWS-IIF model And in the current sliding window, initiate the credibility verification and model update process. Credibility verification is performed through a rule function. Specifically, if any of the following conditions (A) and (B) are met, then the currently real-time input sample is determined. Unreliable: Condition (A) is to extract the compensated tire pressure corresponding to the multidimensional feature samples within the previous 10 minutes. With tire temperature Calculate the Pearson correlation coefficient between the two. ,like If the value is greater than 0.3, then the input multidimensional feature sample is determined. Unreliable; Condition (B) is to extract multidimensional feature samples where the vehicle speed v is 0 within the 5 minutes prior to arrival, and calculate the slope of the pressure linear regression. ,like If the value is greater than 0.8, the sample is deemed unreliable, and it is only considered reliable if the multidimensional feature sample does not satisfy both conditions (A) and (B) at the same time.

[0062] Then the input multidimensional feature samples Input initial model Calculate abnormal scores According to the formula Calculate the credibility score , like If the value is greater than 0.6, then the current input multidimensional feature sample will be... Add to the window; if the sliding window reaches its current maximum capacity, remove the oldest multidimensional feature sample according to the first-in-first-out principle, and randomly select 10% of the trees in the initial model for local reconstruction, while the remaining trees remain unchanged, to obtain the updated AWS-IIF model.

[0063] This step filters untrusted samples through a credibility mechanism constrained by physical rules, preventing abnormal samples from contaminating the model and further reducing the false alarm rate. Meanwhile, the partial update of the AWS-IIF model improves real-time performance and adapts to the dynamic changes in vehicle operating status.

[0064] S23: Calculate the complexity index of the driving scenario and dynamically adjust the size of the sliding window; Based on the current moment and moving forward The system uses data on vehicle speed (v) and acceleration (a) over a 5-minute period to calculate the driving scenario complexity index and dynamically adjusts the sliding window size. First, it calculates... Variance of vehicle speed over time acceleration variance and the absolute value of the correlation coefficient between vehicle speed and acceleration According to the formula Calculate the scene complexity index ; where weight =0.5、 =0.4、 =0.1, The smaller the value, the smoother the vehicle's condition; the larger the value, the more complex the vehicle's condition.

[0065] The maximum window capacity is then adjusted using the following linear piecewise mapping function: ,in =1000、 =500、 =50、 =100, in stable scenarios, the window is expanded to ensure data volume, while in complex scenarios, the window is reduced to avoid noise pollution.

[0066] This step dynamically adjusts the sliding window size based on the complexity of the driving scenario, enabling the updated AWS-IIF model to adapt to different driving scenarios, avoiding misjudgments caused by noisy data in complex scenarios, and further improving the accuracy of early warnings.

[0067] S3: Input the multidimensional feature vector generated in real time by S1 into the updated AWS-IIF model obtained by S2, calculate the anomaly score of the current multidimensional feature vector according to the preset period, and perform early warning classification judgment.

[0068] The task is triggered at a preset interval of 3 minutes, and the latest constructed multidimensional feature vector is extracted as the current sample every 3 minutes. ,Will Inputting the data into the updated AWS-IIF model, the anomaly score is calculated according to the logic defined in S21. abnormal scores .

[0069] The following three-level early warning classification judgment is then performed: like If the tire pressure is >0.5 and remains above 0.5 for 9 minutes, a Level 1 warning (mild abnormality warning) is triggered. The vehicle networking platform pushes a warning message to the driver's in-vehicle terminal display and mobile APP to remind them to pay attention to the tire pressure status. like If the tire pressure is >0.7 and persists for 6 minutes, a Level 2 warning (moderate abnormality warning) is triggered, indicating that the degree of tire pressure abnormality has escalated, the leakage rate or deviation has increased, and there is a certain risk. The vehicle networking platform pushes a message to remind you to go to a safe area as soon as possible to check whether the tires or sensors are abnormal. like If the tire pressure is >0.9 and remains above 0.9 for 3 minutes, a Level 3 warning (serious anomaly warning) is triggered, indicating a severe abnormality in tire pressure with an extremely high risk that may endanger driving safety. The vehicle networking platform will push a message prompting you to immediately stop at a safe location and contact emergency repair. This step outputs the corresponding level of warning messages and push results. Level 1 warnings capture the slow downward trend of tire pressure, providing early warning even when tire pressure is still above the traditional threshold, allowing for timely detection of early slow leaks. The three-level warning classification is based on the degree of abnormality and duration, replacing the traditional binary output, allowing drivers to clearly understand the urgency of the risk and improving the granularity of the warnings. Simultaneously, continuous periodic verification filters out instantaneous anomalies, and combined with the updated model's accurate identification capabilities, ensures the authenticity and effectiveness of the warning messages.

[0070] This invention provides a vehicle-to-everything (V2X) tire pressure intelligent warning method that introduces a sliding window self-adjusting incremental isolated forest algorithm to achieve unsupervised, online, and pollution-resistant abnormal trend learning; constructs multi-dimensional feature vectors to improve discrimination accuracy; establishes a three-level warning classification logic, and comprehensively determines the risk level based on anomaly score, pressure drop rate, duration, and vehicle speed; and relies on the V2X architecture to support real-time warnings on the vehicle and data collaboration with the cloud.

[0071] The present invention has been described in the above-described embodiments; however, these embodiments are merely examples for implementing the present invention. It must be noted that the disclosed embodiments do not limit the scope of the present invention. Conversely, any modifications and refinements made without departing from the spirit and scope of the present invention are within the scope of patent protection of the present invention.

Claims

1. A vehicle-to-everything (V2X) intelligent tire pressure warning method, characterized in that, Includes the following steps: S1: Collect various vehicle data in real time through the vehicle terminal and report them to the vehicle network platform, process the various vehicle data, and construct multi-dimensional feature vectors in real time; The vehicle data includes original tire pressure, tire temperature, ambient temperature, vehicle speed, and vehicle acceleration. The vehicle-mounted terminal collects various vehicle data in real time at a preset collection cycle; The multidimensional feature vector is a structured data set that integrates different time dimensions and multidimensional parameters, used to comprehensively characterize the dynamic changes in tire pressure. S2: Use the multidimensional feature vector constructed from safety data as multidimensional feature samples, initialize the AWS-IIF model based on the multidimensional feature samples, filter untrusted multidimensional feature samples by calculating the confidence score, update the sliding window data online, obtain the updated AWS-IIF model adapted to real-time vehicle conditions and driving scenarios, and dynamically adjust the sliding window size. The safety data refers to data selected from various vehicle data that shows no repair records for one consecutive week. The AWS-IIF model is a sliding window self-adjusting incremental isolated forest model, which is an isolated forest model that integrates sliding window dynamic adjustment and incremental update. Step S2 includes the following sub-steps: S21: Initialize the AWS-IIF model using secure data; For the initial vehicle modeling scenario, safety data from vehicles with no maintenance records for one consecutive week are selected. Then, n multidimensional feature vectors are constructed from the selected safety data. As n multidimensional feature samples ; Then use the n multidimensional feature samples Building the initial AWS-IIF model ,in It refers to the number of trees; And calculate the input multidimensional feature samples abnormal scores ; S22: Calculate the real-time data credibility score to filter out untrustworthy multidimensional feature samples, update the sliding window data, and trigger a local update of the AWS-IIF model; Introduce a credibility mechanism for physical rule constraints to verify the rule function. According to the rule function Determine the multidimensional feature sample of the current input Does it conform to the logical patterns of normal vehicle conditions? The output is a boolean value. = true indicates multidimensional feature samples Credible; = false indicates multidimensional feature samples Unreliable; Subsequently according to and Calculate multidimensional feature samples Credibility score ,and ； when Update the sliding window data when the value is greater than 0.6; S23: Calculate the complexity index of the driving scenario and dynamically adjust the size of the sliding window; The driving scenario complexity index The calculation formula is: ； in yes The variance of vehicle speed over time, where yes The variance of acceleration over time, It is the absolute value of the correlation coefficient between vehicle speed and acceleration. , , It is weight; Then, the sliding window size is dynamically adjusted using the following linear piecewise mapping function: ； in, This is the current maximum capacity of the sliding window. This is the maximum capacity of the sliding window. This is the lower limit of the sliding window capacity. It is the complexity index of the current driving scenario. and These are the minimum and maximum complexity thresholds, respectively. S3: Input the multidimensional feature vector generated in real time by S1 into the updated AWS-IIF model obtained by S2, calculate the anomaly score of the current multidimensional feature vector according to the preset period, and perform early warning classification judgment. The multidimensional feature vector generated in real time by S1 is a multidimensional feature vector that is constructed and updated in real time after the vehicle data collected by the S1 process is calculated and processed. It is the input data that the updated AWS-IIF model can directly recognize. The latest constructed multidimensional feature vector of the vehicle is obtained every preset period to obtain the current multidimensional feature vector. and according to Calculate its abnormal score ,pass The size comparison is used to determine the warning level, and different warning levels correspond to different degrees of urgency. The vehicle network platform will then provide the corresponding warning level prompts.

2. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 1, characterized in that, Step S1 includes the following sub-steps: S11: The vehicle terminal collects various vehicle data according to the preset collection tasks and reports them to the vehicle network platform; The vehicle-mounted terminal collects raw tire pressure in real time at a preset collection cycle. Tire temperature The system collects ambient temperature, vehicle speed v, and vehicle acceleration a, and reports these data to the vehicle network platform. S12: Process the collected vehicle data and construct a multi-dimensional feature vector; Temperature compensation is performed on the original tire pressure based on the ideal gas law, according to the formula... Calculate the compensated tire pressure ,in Standard reference temperature; The multidimensional feature vector includes the following parameters: tire pressure change rate over the past 10 minutes, and the Pearson correlation coefficient between tire pressure and tire temperature over the past 30 minutes. The deviation of tire pressure over the past hour from the historical baseline under the same operating conditions. Standard deviation of tire pressure fluctuation over the past 5 minutes Average vehicle speed over the past 5 minutes.

3. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 2, characterized in that, In step S12, the Pearson correlation coefficient of tire pressure and temperature over the past 30 minutes is calculated. ,in The tire pressure after compensation within the last 30 minutes The mean, Tire temperature over the past 30 minutes The mean; calculate At that time, query historical data for the same vehicle speed and ambient temperature range, and calculate the compensated tire pressure. mean Tire pressure deviation value ; Standard deviation of tire pressure fluctuation over the past 5 minutes ,in = ,in This represents the average tire pressure after compensation over the past 5 minutes. It represents the number of multidimensional feature vectors constructed within the last 5 minutes. This represents the compensated tire pressure value at the i-th data collection time within the last 5 minutes, where i = 1, 2, ... .

4. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 1, characterized in that, In step S21, Each tree in Constructed through recursive random partitioning, with the following partitioning rule: , ; Where k=1,2,..., d is the feature dimension. Each separator randomly selects one dimension, and the separator points are the mean distribution between the maximum and minimum values ​​of the selected dimension. Indicates uniform distribution. Represents the set of values ​​for all samples on the q-th feature; For the input multidimensional feature samples Its abnormal scores The path length is calculated from the isolated forest path length, specifically as follows: ; in, It is a multidimensional feature sample The path length in a single tree, which is the number of edges from the root node to the leaf node; It is a sample The average path length across all trees in the forest; where t = 1, 2, ..., n; It is the normalization factor, where n is the number of training samples, defined as: , Abnormal scores The closer it is to 1, the more abnormal it is.

5. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 1, characterized in that, In step S22, the rule function is validated based on the following conditions (A) and (B). : Condition (A) Take all multidimensional feature samples over 10 minutes and calculate the Pearson correlation coefficient between compensated tire pressure and tire temperature. ； When Pearson correlation coefficient If the value is greater than 0.3, it is considered abnormal and meets condition (A). Condition (B) Take all samples from the most recent 5 minutes when the vehicle speed is 0, and calculate the slope of the linear regression for pressure. ; When the slope of the linear regression of pressure If the value is greater than 0.8, it is determined to be a sensor malfunction or tire leakage, which is considered abnormal data and meets condition (B). If neither condition (A) nor condition (B) is satisfied, then = true; When condition (A) or condition (B) is met, = false.

6. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 1, characterized in that, In step S22, when When the value is greater than 0.6, update the sliding window data as follows: a) Input the current multidimensional feature samples Add to window; b) If the window size is full, remove the oldest multidimensional feature sample according to the first-in-first-out principle; c) Randomly select 10% of the trees in the forest and perform local reconstruction, while leaving the other trees unchanged.

7. The vehicle-to-everything (V2X) tire pressure intelligent warning method according to claim 1, characterized in that, In step 3, based on the calculated abnormal scores... Perform the following early warning classification judgment: If satisfied If the value is greater than 0.5 and continues for 3 preset periods, a Level 1 warning will be triggered. If satisfied If the value is greater than 0.7 and continues for two preset periods, a Level 2 warning will be triggered. If satisfied If the value is greater than 0.9 and continues for one preset period, a Level 3 warning will be triggered. The Level 1 warning is a mild anomaly warning, and the vehicle networking platform prompts the driver to pay attention to the tire pressure status; The Level 2 warning is a moderate anomaly warning, and the vehicle networking platform prompts the driver to check whether the tires are leaking air or whether the sensors are malfunctioning. The Level 3 warning is a severe anomaly warning, and the vehicle-to-everything (V2X) platform prompts the driver to immediately stop at a safe location and take emergency measures.

Images