An intelligent vehicle tracking management system and method based on artificial intelligence

Abstract

The application discloses an Internet of Things vehicle tracking management system and method based on artificial intelligence, relates to the technical field of vehicle tracking management, and obtains vehicle motion data through a vehicle-mounted Internet of Things terminal, obtains vehicle motion characteristics after preprocessing the vehicle motion data, extracts an equal number of fault samples and normal samples for any fault type, calculates the fault difference of each characteristic, screens relevant characteristics and determines the influence coefficient on the fault, calls the corresponding relevant characteristics for each fault, calculates a real-time running deviation index, judges the fault, determines the vehicle running state by considering the time attenuation coefficient and triggers an early warning, executes fault positioning when the early warning is triggered, automatically verifies the maintenance effect after vehicle maintenance, recommends the optimal maintenance point of the corresponding fault type within a preset range for the vehicle based on historical maintenance data, and the application adjusts the importance of historical data according to the vehicle state, and improves the response speed when the vehicle has abnormal signs.

Description

Technical Field

[0001] This invention relates to the field of vehicle tracking and management technology, specifically to an IoT vehicle tracking and management system and method based on artificial intelligence. Background Technology

[0002] With the deep integration of vehicle-to-everything (V2X) technology and artificial intelligence, IoT-based vehicle tracking and management has become a core component of smart transportation and fleet operations. Real-time collection of vehicle operation data via in-vehicle terminals enables remote monitoring of vehicle status and fault warnings, which is crucial for ensuring driving safety, reducing operating costs, and improving vehicle management efficiency. However, existing IoT-based vehicle fault warning technologies still have many shortcomings that urgently need to be addressed. The core issues lie in insufficient warning accuracy and incomplete management processes, severely restricting the practical application of this technology.

[0003] Existing methods generally use global features for fault analysis, without screening specific relevant features for different fault types. All vehicle operation features have the same impact on the determination of all faults. The interference of a large number of irrelevant features not only increases the computational burden of the system, but also leads to frequent false alarms and missed alarms. Summary of the Invention

[0004] The purpose of this invention is to provide an IoT vehicle tracking management system and method based on artificial intelligence to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: an AI-based IoT vehicle tracking and management method, comprising the following steps: S1. Obtain vehicle motion data through the vehicle-mounted Internet of Things terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. S2. For any fault type, extract an equal number of fault samples and normal samples, calculate the fault difference degree of each feature, screen relevant features and determine the influence coefficient on the fault. S3. When the vehicle is running, the corresponding relevant features are invoked for each type of fault to calculate the real-time running deviation index; S4. After each monitoring time interval, analyze the deviation index, make fault judgments, consider the time decay coefficient to determine the vehicle's operating status and trigger an early warning. S5. When a warning is triggered, fault location is performed, and the repair effect is automatically verified after the vehicle is repaired. S6. Based on historical maintenance data, recommend the optimal repair point for the corresponding fault type within a preset range for the vehicle; By measuring the correlation between fault difference quantification features and faults, analyzing operational deviations using fault-specific features, adopting normalized weighted fault judgment coefficients, introducing dynamic adaptive attenuation coefficients, and constructing an intelligent management full lifecycle based on dynamic data.

[0006] Furthermore, in step S1, after authorization, the vehicle-mounted IoT terminal collects dynamic three-dimensional coordinate data and timestamps of the terminal installation points in the vehicle in real time to obtain vehicle motion data. This vehicle motion data includes the vehicle's speed and acceleration in any direction, the opening and travel of any pedal, and the vehicle's electrical data, including but not limited to battery voltage, engine current, and sensor voltage. The vehicle motion data is preprocessed, synchronized in time, and normalized to obtain N vehicle motion characteristics. Authorized collection of multi-dimensional vehicle operation data through the vehicle-mounted IoT terminal comprehensively covers the vehicle's motion state, control component state, and electrical operation state, enabling complete capture of subtle changes during vehicle operation. Preprocessing, synchronizing, and normalizing the collected data effectively eliminates noise interference and time deviations during data collection, unifies the standards and dimensions of different data types, and forms standardized vehicle motion characteristics. This process provides comprehensive, accurate, and reliable basic data for subsequent fault feature screening and status judgment, ensuring the accuracy of fault warnings from the source. Simultaneously, the authorized collection method fully protects user data privacy and information security.

[0007] Further, in step S2, for any fault type, all fault samples within a preset time window prior to the fault occurrence are extracted from the fault sample library. Each fault sample includes the vehicle motion characteristics corresponding to the faulty vehicle. Normal samples operating within a preset time window of vehicles that have not experienced faults are extracted from the normal operation sample library, with the number of normal samples equal to the number of fault samples. Each normal sample includes the vehicle motion characteristics corresponding to a normal vehicle within any normal operation time period equal to the length of the fault sample time window. The average value of each vehicle motion characteristic in all fault samples and the average value in all normal samples are calculated. The fault difference degree of each vehicle motion characteristic is calculated, where the fault difference degree of any vehicle motion characteristic is the average value of the vehicle motion characteristic within the preset time window of all fault samples. The difference between the average value of the faulty samples and the average value of all normal samples is divided by the average value of all normal samples, and the absolute value is taken. If the difference is greater than a preset correlation threshold, the vehicle motion feature is determined to be a relevant feature of the fault type; otherwise, the vehicle motion feature is determined to be an irrelevant feature of the fault type. For any relevant feature of a fault type, the influence coefficient of any relevant feature on the fault type is the fault difference of the feature divided by the sum of the fault differences of all relevant features. By comparing the feature differences between faulty samples and normal samples, features strongly correlated with specific fault types are objectively screened, and interference from irrelevant features is eliminated, significantly reducing the computational load of subsequent analysis and improving the efficiency and relevance of fault analysis. Calculating the influence of features on the fault by the mean difference between samples can quantify the contribution of different features in the fault occurrence process, avoiding bias caused by subjective experience judgment. At the same time, the sample extraction method of quantity matching effectively avoids feature selection errors caused by sample imbalance, laying a scientific and objective feature foundation for subsequent operational deviation calculation and fault judgment.

[0008] Furthermore, in step S3, when any vehicle is running, real-time vehicle motion characteristics are obtained after user authorization. For any fault type, the corresponding relevant features are called, where the relevant features are {A1, A2, ..., Am, ..., AM}, where M is the number of relevant features and Am is the m-th relevant feature. This yields the vehicle's running deviation index B for the fault type. ; Where Cm is the influence coefficient of the m-th related feature on the fault type, and am is the average value of the m-th related feature in all normal samples. The operation deviation index is calculated by calling the corresponding related features for different fault types, avoiding the resource waste caused by full feature calculation, significantly improving the response speed of real-time calculation, and meeting the needs of real-time monitoring during vehicle operation. The operation deviation index integrates the degree of deviation of multiple related features from the normal operating state, transforming multi-dimensional scattered data into a unified quantitative indicator, intuitively reflecting the health level of the vehicle relative to the standard operating state. The calculation method based on normal sample benchmarks ensures the consistency of judgment standards under different vehicles and different operating scenarios, providing a reliable intermediate basis for subsequent fault classification and early warning.

[0009] Furthermore, in step S4, the fault is analyzed once after each monitoring time interval. Taking the current time point as the endpoint, the operating deviation index of K preset time windows is obtained. The operating deviation index closest to the current time point is denoted as BK. The operating deviation indices of the K preset time windows are {B1, B2, ..., Bk, ..., BK}, thereby obtaining the fault judgment coefficient D at the current time point. Where λ is the time decay coefficient, λ = λ0 + γ * DP / D1, and λ0 is the basic decay coefficient, ranging from 0.8 to 1. The larger λ0 is, the slower the weight decay of historical data, the higher the system stability, and the more effectively it can filter false alarms caused by instantaneous data fluctuations. The smaller λ0 is, the faster historical data is forgotten, and the more sensitive the system is to short-term changes. DP is the fault judgment coefficient obtained a monitoring time ago, and D1 is the preset anomaly judgment threshold. When the fault judgment coefficient D is greater than D1, the vehicle is judged to be in an abnormal state, the risk of the aforementioned fault type is predicted, and the system is triggered. The fault judgment mechanism provides early warnings for the specified fault types. Otherwise, if the fault judgment coefficient D is greater than r1*D1, the vehicle is judged to be in a sub-healthy state. r1 is a preset buffer coefficient, ranging from 0.5 to 0.8. The larger r1 is, the narrower the buffer zone, the later the sub-health warning, and the less daily interference. The smaller r1 is, the earlier the sub-health warning, and the earlier potential risks can be detected. Otherwise, the vehicle is judged to be in normal operating condition. The fault judgment coefficient is calculated using a multi-time window operating deviation index, which can comprehensively capture the changing trend of the vehicle's operating status and effectively avoid misjudgments and omissions caused by data fluctuations at a single moment. A time decay mechanism is introduced, which gives greater weight to the recent operating status in fault judgment, conforming to the objective law of the gradual development of vehicle faults. The graded judgment mechanism can accurately distinguish between the normal, sub-healthy, and abnormal operating states of the vehicle, avoiding unnecessary interference to users caused by excessive warnings, and timely detection of potential risks in the early stage of faults, issuing early warnings in advance, reserving sufficient maintenance time for users, and reducing safety hazards and economic losses caused by sudden faults.

[0010] Furthermore, in step S5, when a fault warning of any fault type is triggered, fault root cause localization is performed. The individual deviation contribution coefficient Em of the m-th related feature under the fault type is calculated, Em = Cm * |Am - am|, and the m0 related features with the highest individual deviation contribution coefficients are obtained as recommended troubleshooting features for the fault, where m0 is the preset number of recommended troubleshooting features. After the vehicle completes fault repair, repair effect verification is automatically performed, the repair completion time T0 is recorded, the average fault judgment coefficient Daft for K consecutive preset time windows after repair completion is calculated, and the average fault judgment coefficient Dbef for K consecutive preset time windows before repair completion is calculated, and the repair effect coefficient S is obtained, S = (Dbef - Daft) / Dbef. If R is greater than the preset repair judgment coefficient, the repair for the fault type is judged to be effective; otherwise, the repair is judged to be ineffective, and a second inspection is prompted. When a fault warning is triggered, root cause localization is automatically performed. By quantifying the deviation contribution of each related feature, the core factors most likely to cause the fault are quickly identified, providing repair personnel with a clear direction for troubleshooting, significantly reducing the time cost of fault troubleshooting, and improving repair efficiency. After maintenance is completed, the system automatically performs effect verification, objectively evaluating the maintenance effectiveness based on a comparison of operational data before and after maintenance, thus avoiding the subjectivity and uncertainty of manual judgment. When the maintenance effect does not meet expectations, it promptly prompts for secondary inspection, which can effectively prevent the secondary risks caused by fault recurrence. At the same time, the accumulated maintenance verification data also provides support for the optimization of subsequent fault diagnosis.

[0011] Furthermore, in step S6, based on the vehicle's repair locations recorded by the Internet of Things (IoT), when any vehicle malfunctions, the repair point with the highest historical average repair effectiveness coefficient for that malfunction type within a preset repair range is identified as the vehicle's optimal repair point. Based on the vehicle repair data recorded by the IoT, recommending the repair point with the best historical repair effectiveness to the user when a vehicle malfunctions helps the user quickly find reliable repair services, avoiding problems such as poor repair quality, excessively long repair cycles, or excessively high costs caused by blindly choosing repair points. The real-time identification recommendation method ensures the timeliness and accuracy of information, while the comprehensive evaluation method based on historical repair effectiveness objectively reflects the overall service level and professional capabilities of the repair point. This mechanism not only improves the user's car ownership experience and repair convenience but also promotes healthy competition in the repair industry and drives the continuous improvement of overall repair service quality.

[0012] An AI-based IoT vehicle tracking and management system includes: an IoT data acquisition module, an intelligent feature association module, a real-time deviation calculation module, a fault early warning and judgment module, a fault location verification module, and an intelligent maintenance recommendation module. The IoT data acquisition module is used to acquire vehicle motion data through the vehicle IoT terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. The intelligent feature association module is used to extract an equal number of fault samples and normal samples for any fault type, calculate the fault difference degree of each feature, screen relevant features, and determine the influence coefficient on the fault. The real-time deviation calculation module is used to calculate the real-time running deviation index for each type of fault when the vehicle is running. The fault warning and judgment module is used to analyze the deviation index after each monitoring time interval, make fault judgments, determine the vehicle operating status by considering the time decay coefficient, and trigger a warning. The fault location verification module is used to perform fault location when an early warning is triggered, and automatically verify the repair effect after vehicle repair. The intelligent maintenance recommendation module is used to recommend the optimal maintenance point for the corresponding fault type within a preset range for the vehicle based on historical maintenance data.

[0013] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: On the one hand, it eliminates the drawbacks of existing technologies that use global features for fault analysis. By quantifying the correlation between features and faults, it selects specific relevant features for each type of fault, and only uses features that are actually related to the fault in subsequent calculations. This invention eliminates the interference of a large number of irrelevant features on fault determination, fundamentally reducing the occurrence of false alarms and false negatives. At the same time, it simplifies the calculation dimensions, effectively reduces the computational pressure on the system, and improves the real-time response capability of the system. On the one hand, the fault judgment coefficient is calculated using a normalized weighted method, completely eliminating the influence of the time decay parameter on the numerical range of the judgment result. This allows the judgment threshold to be set independently without the need for parameter binding and calibration, significantly improving the system's debuggability and versatility. On the other hand, a dynamic adaptive time decay mechanism is introduced, which can automatically adjust the emphasis on historical data according to the vehicle's real-time status, ensuring the stability of the warning when the vehicle is running normally and improving the response speed when abnormal signs appear.

[0014] On the other hand, after triggering an alert, the system automatically locates the root cause of the fault, providing repair personnel with a clear direction for troubleshooting; after vehicle repair is completed, it automatically quantifies and verifies the repair effect, promptly identifying repair quality issues; and simultaneously, it recommends the optimal repair resources for users based on historical repair data. This end-to-end intelligent management significantly shortens the fault handling cycle, reduces vehicle operating costs, and substantially improves the user experience. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1This is a structural diagram of an IoT vehicle tracking and management system based on artificial intelligence according to the present invention; Figure 2 This is a flowchart of an IoT vehicle tracking and management method based on artificial intelligence according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 and Figure 2 This invention provides a technical solution: an AI-based IoT vehicle tracking and management method, comprising the following steps: S1. Obtain vehicle motion data through the vehicle-mounted Internet of Things terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. S2. For any fault type, extract an equal number of fault samples and normal samples, calculate the fault difference degree of each feature, screen relevant features and determine the influence coefficient on the fault. S3. When the vehicle is running, the corresponding relevant features are invoked for each type of fault to calculate the real-time running deviation index; S4. After each monitoring time interval, analyze the deviation index, make fault judgments, consider the time decay coefficient to determine the vehicle's operating status and trigger an early warning. S5. When a warning is triggered, fault location is performed, and the repair effect is automatically verified after the vehicle is repaired. S6. Based on historical maintenance data, recommend the optimal repair point for the corresponding fault type within a preset range for the vehicle; By measuring the correlation between fault difference quantification features and faults, analyzing operational deviations using fault-specific features, adopting normalized weighted fault judgment coefficients, introducing dynamic adaptive attenuation coefficients, and constructing an intelligent management full lifecycle based on dynamic data.

[0018] In step S1, after authorization, the vehicle-mounted IoT terminal collects dynamic three-dimensional coordinate data and timestamps of the terminal installation points in the vehicle in real time to obtain vehicle motion data. The vehicle motion data includes the vehicle's speed and acceleration in any direction, the opening and travel of any pedal, and the vehicle's electrical data, including but not limited to battery voltage, engine current, and sensor voltage. The vehicle motion data is preprocessed, time-synchronized, and normalized to obtain N vehicle motion features.

[0019] In step S2, for any fault type, all fault samples within a preset time window prior to the fault occurrence are extracted from the fault sample library. Each fault sample includes the vehicle motion characteristics corresponding to the faulty vehicle. Normal samples operating within a preset time window of vehicles that have not experienced faults, equal in number to the number of fault samples, are extracted from the normal operation sample library. Each normal sample includes the vehicle motion characteristics corresponding to a normal vehicle within any normal operation time period equal to the time window length of the fault samples. The average value of each vehicle motion characteristic in all fault samples and the average value in all normal samples are calculated. The fault difference degree of each vehicle motion characteristic is calculated. The fault difference degree of any vehicle motion characteristic is the difference between the average value of the vehicle motion characteristic in all fault samples and the average value in all normal samples, divided by the average value in all normal samples, and the absolute value is taken. When the difference degree is greater than a preset correlation threshold, the vehicle motion characteristic is determined to be a relevant feature of the fault type; otherwise, the vehicle motion characteristic is determined to be an unrelated feature of the fault type. For any relevant feature of a fault type, the influence coefficient of any relevant feature on the fault type is the fault difference degree of the feature divided by the sum of the fault difference degrees of all relevant features.

[0020] In step S3, when any vehicle is running, real-time vehicle motion characteristics are obtained after user authorization. For any fault type, the corresponding relevant features are called. The relevant features are {A1, A2, ..., Am, ..., AM}, where M is the number of relevant features and Am is the m-th relevant feature. Then, the vehicle's running deviation index B for the fault type is obtained. ; Where Cm is the influence coefficient of the m-th related feature on the fault type, and am is the average value of the m-th related feature in all normal samples.

[0021] In step S4, the fault is analyzed once after each monitoring time interval. Taking the current time point as the endpoint, the operating deviation index of K preset time windows is obtained. The operating deviation index closest to the current time point is denoted as BK. The operating deviation indices of the K preset time windows are {B1, B2, ..., Bk, ..., BK}, and then the fault judgment coefficient D at the current time point is obtained. Where λ is the time decay coefficient, λ=λ0+γ*DP / D1, where λ0 is the basic decay coefficient, with a value between 0.8 and 1, DP is the fault judgment coefficient obtained one monitoring time ago, and D1 is the preset anomaly judgment threshold. When the fault judgment coefficient D is greater than D1, the vehicle is judged to be in an abnormal state, the risk of the fault type is predicted, and the fault warning of the fault type is triggered; otherwise, if the fault judgment coefficient D is greater than r1*D1, the vehicle is judged to be in a sub-healthy state, where r1 is the preset buffer coefficient, with a value between 0.5 and 0.8; otherwise, the vehicle is judged to be in normal operating condition.

[0022] In step S5, when a fault warning for any fault type is triggered, fault root cause localization is performed. The individual deviation contribution coefficient Em of the m-th related feature under the fault type is calculated, Em = Cm * |Am - am|. Then, the m0 related features with the highest individual deviation contribution coefficients are obtained as recommended troubleshooting features for the fault, where m0 is the preset number of recommended troubleshooting features. After the vehicle completes fault repair, the repair effect verification is automatically performed. The repair completion time T0 is recorded. The average fault judgment coefficient Daft for K consecutive preset time windows after the repair is completed is calculated. The average fault judgment coefficient Dbef for K consecutive preset time windows before the repair is completed is calculated. Then, the repair effect coefficient S is obtained, S = (Dbef - Daft) / Dbef. If R is greater than the preset repair judgment coefficient, the repair for the fault type is judged to be effective; otherwise, the repair is judged to be ineffective, and a second inspection is prompted.

[0023] In step S6, the vehicle's repair location is recorded based on the Internet of Things. When any fault occurs in the vehicle, the repair point with the highest historical average repair effectiveness coefficient for the fault type within the preset repair range is marked as the vehicle's optimal repair point.

[0024] An AI-based IoT vehicle tracking and management system includes: an IoT data acquisition module, an intelligent feature association module, a real-time deviation calculation module, a fault early warning and judgment module, a fault location verification module, and an intelligent maintenance recommendation module. The IoT data acquisition module is used to acquire vehicle motion data through the vehicle IoT terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. The intelligent feature association module is used to extract an equal number of fault samples and normal samples for any fault type, calculate the fault difference degree of each feature, screen relevant features, and determine the influence coefficient on the fault. The real-time deviation calculation module is used during vehicle operation to call the corresponding relevant features for each type of fault and calculate the real-time operating deviation index. The fault warning and judgment module is used to analyze the deviation index after each monitoring time interval, make fault judgments, determine the vehicle operating status by taking into account the time decay coefficient, and trigger a warning. The fault location verification module is used to perform fault location when an early warning is triggered, and automatically verify the repair effect after vehicle repair. The intelligent maintenance recommendation module is used to recommend the optimal repair point for the corresponding fault type within a preset range for vehicles based on historical maintenance data.

[0025] Example 1: Vehicle operation data is collected in real time via an in-vehicle IoT terminal installed in the vehicle's central control unit, after authorization from the vehicle owner. This includes the dynamic three-dimensional coordinates and timestamps of the terminal's installation point, as well as seven vehicle motion data points: longitudinal speed, longitudinal acceleration, brake pedal travel, accelerator pedal opening, battery voltage, engine current, and lateral acceleration. The collected data undergoes outlier removal, timestamp alignment, and normalization to obtain seven standardized vehicle motion features. For a type of braking system fault, 100 fault samples of the same vehicle model within one hour prior to the occurrence of a braking system fault are extracted from a cloud-based fault sample database. Simultaneously, 100 normal operation samples of the same model vehicles that have not experienced braking faults within one hour are extracted from a normal operation sample database. The average value of the seven vehicle motion features across all fault samples and normal samples is calculated, and then the fault difference degree of each feature is calculated. The calculated fault difference degrees of brake pedal travel and longitudinal deceleration are greater than a preset correlation threshold, thus being identified as relevant features of the braking system fault. The difference degrees of the other five features are lower than the threshold, thus being identified as irrelevant features. The fault difference between the two related features was normalized, resulting in an influence coefficient of 0.6 for brake pedal travel and 0.4 for longitudinal deceleration. When the truck was driving on urban roads, the system acquired real-time vehicle motion features every 10 seconds and used the two related features mentioned above for braking system faults. At a certain moment, the real-time value of brake pedal travel was 0.35, and the real-time value of longitudinal deceleration was 0.8. The average values ​​of the two features in normal samples were 0.2 and 1.2, respectively. Substituting these values ​​into the formula, the real-time running deviation index was calculated, yielding a result of 0.25. The system performed a fault analysis every minute, acquiring the running deviation indices for three consecutive time windows, with the current time as the endpoint, resulting in indices of 0.12, 0.18, and 0.25, respectively. The base attenuation coefficient was set to 0.9, the adjustment coefficient to 0.05, the fault judgment coefficient of the previous monitoring time interval to 0.15, and the anomaly judgment threshold to 0.3. The calculated dynamic adaptive time attenuation coefficient was 0.925, and the normalized weighted fault judgment coefficient was approximately 0.187. The buffer coefficient is set to 0.6, and the corresponding sub-health threshold is 0.18. Therefore, the vehicle is determined to be in a sub-healthy state due to the braking system malfunction, triggering a level 2 warning.

[0026] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An AI-based IoT vehicle tracking and management method, characterized in that: The method includes the following steps: S1. Obtain vehicle motion data through the vehicle-mounted Internet of Things terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. S2. For any fault type, extract an equal number of fault samples and normal samples, calculate the fault difference degree of each feature, screen relevant features and determine the influence coefficient on the fault. S3. When the vehicle is running, the corresponding relevant features are invoked for each type of fault to calculate the real-time running deviation index; S4. After each monitoring time interval, analyze the deviation index, make fault judgments, consider the time decay coefficient to determine the vehicle's operating status and trigger an early warning. S5. When a warning is triggered, fault location is performed, and the repair effect is automatically verified after vehicle repair. S6. Based on historical maintenance data, recommend the optimal repair point for the corresponding fault type within a preset range for the vehicle.

2. The IoT vehicle tracking and management method based on artificial intelligence according to claim 1, characterized in that: By measuring the correlation between fault difference quantification features and faults, analyzing operational deviations using fault-specific features, adopting normalized weighted fault judgment coefficients, introducing dynamic adaptive attenuation coefficients, and constructing an intelligent management full lifecycle based on dynamic data.

3. The IoT vehicle tracking and management method based on artificial intelligence according to claim 2, characterized in that: In step S1, after authorization, the vehicle-mounted IoT terminal collects dynamic three-dimensional coordinate data and timestamps of the terminal installation points in the vehicle in real time to obtain vehicle motion data. The vehicle motion data includes the vehicle's speed and acceleration in any direction, the opening and travel of any pedal, and the vehicle's electrical data. The vehicle motion data is preprocessed, time-synchronized, and normalized to obtain N vehicle motion features.

4. The IoT vehicle tracking and management method based on artificial intelligence according to claim 3, characterized in that: In step S2, for any fault type, all fault samples within a preset time window prior to the fault occurrence are extracted from the fault sample library. Each fault sample includes the vehicle motion characteristics corresponding to the faulty vehicle. Normal samples operating within a preset time window of vehicles that have not experienced faults, equal in number to the number of fault samples, are extracted from the normal operation sample library. Each normal sample includes the vehicle motion characteristics corresponding to a normal vehicle within any normal operation time period equal to the length of the fault sample time window. The average value of each vehicle motion characteristic in all fault samples and the average value in all normal samples are calculated. The fault difference degree of each vehicle motion characteristic is calculated. The fault difference degree of any vehicle motion characteristic is the difference between the average value of the vehicle motion characteristic in all fault samples and the average value in all normal samples, divided by the average value in all normal samples, and the absolute value is taken. When the difference degree is greater than a preset correlation threshold, the vehicle motion characteristic is determined to be a relevant feature of the fault type; otherwise, the vehicle motion characteristic is determined to be an unrelated feature of the fault type. For any relevant feature of a fault type, the influence coefficient of any relevant feature on the fault type is the fault difference degree of the feature divided by the sum of the fault difference degrees of all relevant features.

5. The IoT vehicle tracking and management method based on artificial intelligence according to claim 4, characterized in that: In step S3, when any vehicle is running, real-time vehicle motion characteristics are obtained after user authorization. For any fault type, the corresponding relevant features are called. The relevant features are {A1, A2, ..., Am, ..., AM}, where M is the number of relevant features and Am is the m-th relevant feature. Then, the vehicle's running deviation index B for the fault type is obtained. ； Where Cm is the influence coefficient of the m-th related feature on the fault type, and am is the average value of the m-th related feature in all normal samples.

6. The IoT vehicle tracking and management method based on artificial intelligence according to claim 5, characterized in that: In step S4, the fault is analyzed once after each monitoring time interval. Taking the current time point as the endpoint, the operating deviation index of K preset time windows is obtained. The operating deviation index closest to the current time point is denoted as BK. The operating deviation indices of the K preset time windows are {B1, B2, ..., Bk, ..., BK}, and then the fault judgment coefficient D at the current time point is obtained. Where λ is the time decay coefficient, λ=λ0+γ*DP / D1, where λ0 is the basic decay coefficient, DP is the fault judgment coefficient obtained one monitoring time ago, and D1 is the preset anomaly judgment threshold. When the fault judgment coefficient D is greater than D1, the vehicle is judged to be in an abnormal state, the risk of the fault type is predicted, and the fault warning of the fault type is triggered; otherwise, if the fault judgment coefficient D is greater than r1*D1, the vehicle is judged to be in a sub-healthy state, where r1 is the preset buffer coefficient; otherwise, the vehicle is judged to be in normal operating condition.

7. The IoT vehicle tracking and management method based on artificial intelligence according to claim 6, characterized in that: In step S5, when a fault warning for any fault type is triggered, fault root cause localization is performed. The individual deviation contribution coefficient Em of the m-th related feature under the fault type is calculated, Em = Cm * |Am - am|. Then, the m0 related features with the highest individual deviation contribution coefficients are obtained as recommended troubleshooting features for the fault, where m0 is the preset number of recommended troubleshooting features. After the vehicle completes fault repair, the repair effect verification is automatically performed. The repair completion time T0 is recorded. The average fault judgment coefficient Daft for K consecutive preset time windows after the repair is completed is calculated. The average fault judgment coefficient Dbef for K consecutive preset time windows before the repair is completed is calculated. Then, the repair effect coefficient S is obtained, S = (Dbef - Daft) / Dbef. If R is greater than the preset repair judgment coefficient, the repair for the fault type is judged to be effective; otherwise, the repair is judged to be ineffective, and a second inspection is prompted.

8. The IoT vehicle tracking and management method based on artificial intelligence according to claim 7, characterized in that: In step S6, the vehicle's repair location is recorded based on the Internet of Things. When any fault occurs in the vehicle, the repair point with the highest historical average repair effectiveness coefficient for the fault type within the preset repair range is marked as the vehicle's optimal repair point.

9. An AI-based IoT vehicle tracking and management system, wherein the system is applied to the AI-based IoT vehicle tracking and management method described in any one of claims 1-8, characterized in that: The system includes: an IoT data acquisition module, an intelligent feature association module, a real-time deviation calculation module, a fault early warning and judgment module, a fault location verification module, and an intelligent maintenance recommendation module; The IoT data acquisition module is used to acquire vehicle motion data through the vehicle IoT terminal, and obtain vehicle motion characteristics after preprocessing the vehicle motion data. The intelligent feature association module is used to extract an equal number of fault samples and normal samples for any fault type, calculate the fault difference degree of each feature, screen relevant features, and determine the influence coefficient on the fault. The real-time deviation calculation module is used to calculate the real-time running deviation index for each type of fault when the vehicle is running. The fault warning and judgment module is used to analyze the deviation index after each monitoring time interval, make fault judgments, determine the vehicle operating status by considering the time decay coefficient, and trigger a warning. The fault location verification module is used to perform fault location when an early warning is triggered, and automatically verify the repair effect after vehicle repair. The intelligent maintenance recommendation module is used to recommend the optimal maintenance point for the corresponding fault type within a preset range for the vehicle based on historical maintenance data.

Images