Heavy vehicle safety detection method and device, equipment and medium thereof

Abstract

The invention relates to a heavy vehicle use safety detection method and device, equipment and a medium, and the method comprises the steps: obtaining vehicle use data uploaded by a heavy vehicle, and the vehicle use data comprises fact control data and operation activity data generated when the heavy vehicle is driven in a specific time period; mining a vehicle use data feature set based on the fact control data and the operation activity data, and determining risk degrees of the heavy vehicle corresponding to various risk events according to the vehicle use data feature set; the event weight distribution matched with the commercial type of the heavy vehicle is called to weight the risk degrees of various risk events so as to determine the vehicle use safety state of the heavy vehicle; and triggering a vehicle use value-added service process based on the vehicle use safety state. By analyzing the vehicle data in real time and intelligently responding to the safety state change, the risk assessment process is optimized, and the efficiency and accuracy of heavy vehicle safety management are improved.

Description

Technical Field

[0001] The present application relates to vehicle safety technology, and in particular to a heavy-duty vehicle safety detection method and its device, equipment, and medium. Background Art

[0002] As the number of heavy vehicles increases, traffic accidents occur more frequently, and society is paying more and more attention to vehicle safety. Logistics companies and construction companies hope to achieve differentiated pricing and reduce claims costs through refined risk assessment. At the same time, consumers and society are paying more and more attention to the safety and service quality of heavy vehicles, requiring a more transparent and fair scoring system to enhance trust in transportation and engineering services. Against this background, the emergence of a heavy vehicle risk scoring system has become inevitable. It is not only related to traffic safety issues, but also meets the needs of the logistics and engineering industries, while using technological advances to raise public awareness.

[0003] Among the existing technologies, the field of heavy vehicle risk scoring mainly adopts three methods: behavioral scoring model, accident history scoring and customer feedback and scoring. The behavioral scoring model evaluates the driver's driving safety by monitoring the driver's real-time behavior. It relies on the vehicle's OBD system or on-board sensors to collect data, such as sudden acceleration, sudden braking, etc., and then obtains the driver's score through cloud analysis algorithms. This method can capture the driver's driving habits in real time, provide timely feedback, and promote safe driving, but it is limited by the limited parameters that the equipment can report, which may lead to inaccurate data and difficult to unify the scoring standards. The accident history score assesses risks based on the driver's past accident records and traffic violations. It relies on the databases of insurance companies and traffic management departments. Although it can make risk assessments based on historical data, it may not reflect the driver's behavioral changes in real time and may ignore the driver's improvement and growth. The customer feedback and scoring method evaluates the driver's service quality and safety through customer feedback and scoring systems, which directly reflects the customer's real experience, but customer feedback is subjective and may be affected by factors such as emotions and individual experiences, affecting the fairness of the score.

[0004] The problems and shortcomings of existing technologies are that they often rely on a single data source or scoring method and cannot comprehensively and comprehensively assess the risks of commercial vehicles. In addition, existing technologies fail to fully explore and utilize vehicles and various valid data, lack in-depth mining and analysis of these data, resulting in inaccurate risk assessment. At the same time, existing technologies fail to distinguish different types of commercial vehicles for risk assessment and lack personalized capabilities. Finally, existing technologies are also not very good at utilizing risk assessment results, which limits the practicality and commercial value of technical solutions. Summary of the invention

[0005] The primary purpose of the present application is to solve at least one of the above problems and provide a heavy vehicle safety detection method and its device, equipment, and medium.

[0006] According to one aspect of the present application, a heavy vehicle safety detection method is provided, comprising the following steps: Obtaining vehicle usage data uploaded by a heavy vehicle, wherein the vehicle usage data includes actual control data and operational activity data generated when the heavy vehicle is driven within a specific period of time; Mining a vehicle data feature set based on the factual manipulation data and operational activity data; Determining the risk degree of the heavy-duty vehicle corresponding to various risk events according to the vehicle usage data feature set, including: calling multiple vehicle usage data feature sets formed by the heavy-duty vehicle in a specific period from a vehicle usage information table; encoding the multiple vehicle usage data feature sets in the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; inputting the multi-feature serialized data into a preset risk event identification model, and determining the deep semantic information of the multi-feature serialized data mapped to the probability corresponding to multiple preset risk events as the corresponding risk degree; Calling the event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle; The vehicle use value-added service process is triggered based on the vehicle use safety status.

[0007] According to another aspect of the present application, a heavy vehicle safety detection device is provided, comprising: A data acquisition module, configured to acquire vehicle usage data uploaded by a heavy vehicle, wherein the vehicle usage data includes actual control data and operational activity data generated by the heavy vehicle being driven within a specific period of time; A feature mining module configured to mine a vehicle data feature set based on the factual control data and the operational activity data; The risk assessment module is configured to determine the risk degree of the heavy-duty vehicle corresponding to various risk events according to the vehicle use data feature set, including: calling multiple vehicle use data feature sets formed by the heavy-duty vehicle in a specific period from the vehicle use information table; encoding the multiple vehicle use data feature sets in the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; inputting the multi-feature serialized data into a preset risk event identification model, and determining the probability corresponding to mapping the deep semantic information of the multi-feature serialized data to multiple preset risk events as the corresponding risk degree; A state determination module is configured to call an event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety state of the heavy vehicle; The safety application module is configured to trigger a vehicle value-added service process based on the vehicle safety status.

[0008] According to another aspect of the present application, a heavy-duty vehicle safety detection device is provided, comprising a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the heavy-duty vehicle safety detection method described in the present application.

[0009] According to another aspect of the present application, a non-volatile readable storage medium is provided, which stores a computer program implemented according to the heavy vehicle safety detection method in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the method are executed.

[0010] The technical solution of this application has many advantages, including but not limited to the following aspects: First, this application realizes a comprehensive analysis of vehicle usage data by comprehensively utilizing the actual control data and operational activity data generated by heavy vehicles being driven during a specific period of time. Compared with the existing practice of relying on a single data source, the use of this comprehensive data can more comprehensively reflect the driving status and risk status of the vehicle. By mining the feature set of vehicle usage data, this application can deeply analyze and identify the key factors that affect vehicle safety, thereby providing a more accurate risk assessment.

[0011] Secondly, this application determines the risk level of heavy vehicles corresponding to various risk events based on the vehicle usage data feature set. Compared with the behavior scoring model, accident history scoring and customer feedback and scoring methods in the prior art, this method can more dynamically evaluate the safety status of the vehicle. It not only takes into account the real-time behavior of the driver, but also includes information on vehicle performance, environmental factors, etc., making the risk assessment more comprehensive and detailed.

[0012] Furthermore, this application uses the event weight distribution that matches the commercial type of heavy vehicles to weight the risk of various risk events to determine the vehicle's safety status. Compared with the existing technology that lacks personalization capabilities, this personalized weighting method can provide more accurate safety assessments based on the characteristics and usage scenarios of different vehicles, thereby achieving more effective risk management.

[0013] Finally, this application triggers the value-added service process based on the vehicle safety status. By combining risk assessment results with value-added services, it not only improves the practicality of risk management, but also provides more business opportunities for logistics companies, insurance companies, etc. For example, based on the risk assessment results, personalized safety training can be provided for drivers, accurate insurance pricing basis can be provided for insurance companies, or optimization suggestions can be provided for fleet management, thereby improving the service quality and efficiency of the entire industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The network architecture for the exemplary deployment of the heavy vehicle operation platform of this application; Figure 2 A flowchart of an embodiment of a heavy vehicle safety detection method of the present application; Figure 3 This is a principle block diagram of the heavy vehicle safety detection device of the present application; Figure 4 This is a schematic diagram of the structure of a heavy-duty vehicle safety detection device used in this application. DETAILED DESCRIPTION

[0015] To facilitate understanding of the application scenarios of the present application, an exemplary heavy-duty vehicle operation platform is described. Figure 1 This is the network architecture of the heavy-duty vehicle operation platform of the present application. In this network architecture, the heavy-duty vehicle operation platform is composed of multiple key components, including application servers 81, business servers 82, and vehicle systems of each heavy-duty vehicle 80 connected to the platform. These components work together to achieve real-time monitoring, data collection, risk assessment, and value-added service provision of heavy-duty vehicles.

[0016] Heavy vehicles 80 cover a wide range of areas, including but not limited to trucks, engineering vehicles, and muck trucks, which play an important role in logistics, transportation, engineering construction, etc. This operating platform is characterized by its high degree of integration and intelligence, and can process a large amount of real-time data from vehicles and provide instant analysis and feedback.

[0017] The application server 81 is responsible for processing data from the vehicle system and providing a user interface for platform administrators and drivers. It runs various value-added vehicle services, including but not limited to any one or more of real-time monitoring services, fault warning services, driving behavior analysis services, route optimization services and insurance services. The real-time monitoring service can track the location and status of the vehicle in real time to ensure the safe operation of the vehicle. The fault warning service predicts potential technical problems by analyzing the vehicle's sensor data to reduce unexpected downtime. The driving behavior analysis service evaluates the driver's behavior patterns and provides suggestions for improving driving habits. The route optimization service provides the driver with the fastest route by analyzing traffic conditions and route information. The insurance service provides customized insurance products to insurance companies based on the risk assessment results of the vehicle.

[0018] The business server 82 is responsible for processing the business logic of the platform, including user management, order processing, payment processing, etc. It works closely with the application server to ensure that the business process of the platform is smooth. The business server is also responsible for storing and managing the user data and transaction records of the platform to ensure data security and privacy. These users can be drivers, car owners, or other downstream business management users.

[0019] The vehicle computer system is an intelligent terminal installed on each heavy-duty vehicle80, which is responsible for collecting the vehicle's control data and operational activity data and uploading this data to the application server. The vehicle computer system usually includes GPS positioning equipment, vehicle sensors, communication modules, etc., which work together to ensure the accuracy and real-time nature of the data. The vehicle computer system is also responsible for receiving instructions and information from the application server, such as navigation updates, fault warnings and driving suggestions, and feeding this information back to the driver.

[0020] The heavy-duty vehicle operation platform of this application realizes comprehensive monitoring and management of heavy-duty vehicles through its network architecture, improves the safety and operation efficiency of vehicles, and also provides a wealth of value-added services for vehicle users. The application of this platform can not only improve the safety management level of heavy-duty vehicles, but also bring innovative business value to the logistics and engineering industries.

[0021] See also Figure 2 According to a heavy-duty vehicle safety detection method provided by the present application, it can be implemented as a computer program product running in an application server to form a corresponding heavy-duty vehicle safety detection system. In one embodiment, it includes the following steps: Step S3100: obtaining vehicle usage data uploaded by a heavy vehicle, wherein the vehicle usage data includes actual control data and operational activity data generated when the heavy vehicle is driven within a specific period of time; In order to realize the safety inspection of heavy-duty vehicles, it is necessary to obtain the vehicle usage data uploaded by heavy-duty vehicles. The vehicle usage data consists of two parts: factual control data and operational activity data.

[0022] Actual control data refers to data directly related to driving operations generated by heavy vehicles during driving. This type of data includes the number of vehicle braking, forward braking, blind spot braking, etc., which reflect the driver's control behavior of the vehicle under specific circumstances. In addition, it also includes the number of severe fatigue driving, various alarms such as forward collision alarms, sudden acceleration alarms, sudden deceleration alarms, etc., as well as the number of vehicle rollover alarms, etc. These data can reveal the emergency situations and potential risks that the vehicle may encounter during driving. The actual control data is collected in real time through the vehicle's sensors and diagnostic systems, providing direct evidence for evaluating driving safety.

[0023] Operational activity data is defined as data related to the use and operation of heavy vehicles generated during the operation process. This includes time data such as total online time, online parking time, total parking time, driving time, etc., which reflect the frequency of vehicle use and driving time. Mileage data such as mileage, urban road mileage, highway mileage, and speed data such as maximum speed, average driving speed, and average speed provide relevant information on vehicle operating efficiency and driving speed. GPS positioning data involves the location information of the vehicle, including the length of stay in the city, whether it is driving on the highway, whether it is speeding, etc. This data is obtained through the reverse geolocation query service and is used to calculate the vehicle's operating conditions and driving behavior. Operational activity data not only involves the operating status of the vehicle, but also includes the vehicle's usage pattern and operating efficiency, providing important reference information for a comprehensive assessment of the vehicle's safety status.

[0024] In terms of technical implementation, the acquisition of vehicle usage data can be adapted to the communication protocols used by heavy vehicles, such as CAN bus, Ethernet, etc., and the corresponding data transmission technology used by the corresponding protocol is also combined to obtain vehicle usage data, such as 4G / 5G network, Wi-Fi, etc. The collection of vehicle usage data must not only ensure the integrity and accuracy of the data, but also take into account the real-time nature of the data, so as to be able to respond to possible risk events in a timely manner.

[0025] In a specific embodiment, the acquisition of vehicle usage data can be achieved through a vehicle computer system, which is configured with corresponding hardware and software, can communicate with various sensors and control units of heavy vehicles, and package and upload the collected data to an application server. For example, the vehicle computer system can read vehicle diagnostic data through the vehicle's OBD-II interface, and obtain location information through an integrated GPS module, and then transmit these data to the server through the network.

[0026] In practical applications, the acquisition of vehicle usage data may involve a variety of different data types and sources. For example, in addition to basic control data and operational data, it may also include vehicle maintenance records, driver driving habits data, etc. The collection and analysis of these data provide a rich information basis for subsequent risk assessment and safety testing. In this way, the technical solution of this application can achieve comprehensive monitoring and evaluation of the safety status of heavy vehicles.

[0027] Step S3200: mining a vehicle data feature set based on the factual control data and the operational activity data; By conducting in-depth analysis of the collected factual control data and operational activity data, key feature data that can represent the safety status of the vehicle can be extracted to form the vehicle use data feature set. The mining of vehicle use data feature sets is a process of data preprocessing and feature engineering. It uses optional technical means such as statistical analysis and machine learning algorithms to identify feature data that have an important impact on risk assessment from the original vehicle use data.

[0028] Fact manipulation data and operational activity data contain a large amount of raw information, but not all information contributes directly to risk assessment. Therefore, data mining technology is needed to filter and transform these data to form a meaningful feature set. For example, through statistical analysis of the number of braking times and the number of sudden acceleration alarms, the driver's driving behavior characteristics, such as "aggressive driving" or "smooth driving", can be extracted. Similarly, through the analysis of mileage and average speed, the vehicle's usage frequency and driving speed characteristics, such as "high frequency use" or "speeding", can be obtained.

[0029] In terms of technical implementation, data mining can select any one or more algorithms from a variety of algorithms, including but not limited to decision trees, cluster analysis, principal component analysis, etc. These algorithms can identify patterns and associations from a large amount of complex data, thereby helping to build a risk event identification model. For example, a decision tree algorithm can be used to build a model based on historical data, which can predict future risk events based on the driver's driving behavior and vehicle operation activity data.

[0030] In addition, feature selection and dimensionality reduction techniques can be used in the data mining process, which help remove redundant features and noise and retain the most valuable information. For example, correlation analysis can identify features that are highly correlated with risk events, while dimensionality reduction techniques such as principal component analysis can reduce the number of features while retaining most of the variability of the data.

[0031] Step S3300: determining the risk level of the heavy vehicle corresponding to various risk events according to the vehicle usage data feature set; After completing the mining of the vehicle usage data feature set, the next step is to determine the risk level of heavy vehicles corresponding to various risk events. This step uses a machine learning model to extract deep semantic information from the vehicle usage data feature set and map it to the preset risk events to determine the corresponding risk level of each risk event.

[0032] The extraction of deep semantic information is based on the mined vehicle data feature set, and the data is deeply analyzed using machine learning algorithms to identify and understand complex patterns and associations in the data. For example, by training a random forest model, patterns related to high-risk driving behaviors, such as frequent sharp turns or sudden deceleration, can be identified from driving behavior characteristics, and these patterns can be associated with specific risk events.

[0033] Next, the extracted deep semantic information is mapped to multiple preset risk events. Preset risk events may include, but are not limited to, speeding, fatigue driving, aggressive driving, etc. The mapping process may include comparing specific data in the vehicle data feature set with the known features of these risk events to determine the degree of association between the vehicle's current state and each risk event. For example, if a vehicle shows frequent rapid acceleration and deceleration behavior, this may be mapped to the risk event of aggressive driving, and the risk level of the risk event can be determined accordingly.

[0034] In terms of technical implementation, this step can be completed by building a risk event identification model, the input of which is the vehicle usage data feature set, and the output is the risk level of each risk event. For example, a neural network model can be used to process the high-dimensional data generated by encoding the vehicle usage data feature set, and output the risk score of each risk event.

[0035] In a specific embodiment, the risk event identification model can be deployed on an application server, receive data from the vehicle system, and calculate the risk level in real time. The output of the model can be a set of risk scores, each score corresponding to a risk event, which can be used directly or indirectly for safety decisions, such as triggering an early warning system or adjusting insurance premiums.

[0036] Assigning a quantitative risk level to each risk event provides a scientific basis for subsequent risk management and safety decision-making. This approach not only improves the accuracy of risk assessment, but also enhances the dynamic monitoring capability of the safety status of heavy vehicles.

[0037] In one embodiment, this step can be concretized to include: calling multiple vehicle usage data feature sets formed by the heavy vehicle in a specific period from the vehicle usage information table; encoding the multiple vehicle usage data feature sets in the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; inputting the multi-feature serialized data into a preset risk event identification model, and determining the deep semantic information of the multi-feature serialized data mapped to the probability corresponding to multiple preset risk events as the corresponding risk degree, see below for details.

[0038] Step S3400: Calling the event weight distribution that matches the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle; After determining the risk level of various risk events for heavy vehicles, the event weight distribution that matches the commercial type of the heavy vehicle can be called to weight the risk level of various risk events to comprehensively determine the safety status of the heavy vehicle. This step is a key link in the risk assessment process, which involves giving different levels of importance to different risk events based on the specific purpose and operating environment of the vehicle.

[0039] The event weight distribution is based on an in-depth understanding of the operating characteristics of heavy-duty vehicles and historical data analysis. Different types of commercial vehicles, such as engineering vehicles, logistics vehicles, and muck trucks, have different operating environments, load conditions, driving routes, and other factors, so the types and degrees of risks they face are also different. For example, engineering vehicles may work more frequently on construction sites, so the risk of blind spot braking and rollover may require a higher weight; while logistics vehicles may be more concerned about the risks of speeding and fatigue driving.

[0040] In terms of technical implementation, this step requires the construction of a weighting system that contains a series of preset weight values ​​corresponding to different commercial vehicle types and risk events. The weights can be determined based on a variety of factors such as historical accident data analysis, expert experience, and regulatory requirements. For example, by analyzing historical accident data, it can be found that a certain type of vehicle has a higher accident rate when driving at night, so a higher weight can be assigned to the risk of nighttime operation hours.

[0041] In a specific embodiment, the weight system call and the weighted calculation of the risk can be performed in the application server. After receiving the risk score data from the vehicle system, the application server selects a suitable event weight distribution according to the commercial type of the vehicle, and then performs a weighted calculation on the risk of each risk event. The result of the weighted calculation is a comprehensive risk score, which reflects the overall safety status of the vehicle in the current state, and can directly or indirectly determine the safety status of the vehicle.

[0042] The mapping relationship between the commercial type and the event weight distribution can be set in advance. In one embodiment, before calling the event weight distribution that matches the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle, it includes: classifying the heavy vehicle according to its operating characteristics, load range, and driving route, and dividing and setting the commercial type to which the heavy vehicle belongs; obtaining the preset weight corresponding to each risk event based on each commercial type to form the event weight distribution of the commercial type; storing the mapping relationship data between each commercial type and its event weight distribution in the database for future use. This embodiment allows the operation management personnel to divide various commercial types in advance according to experience, and then preset weights for each commercial type corresponding to each risk event to form the event weight distribution corresponding to each commercial type, and then store the mapping relationship data between the commercial type and the event weight distribution in the database for direct use in step S3400.

[0043] In addition, the design of the weighting system should also consider flexibility and scalability to adapt to changes in risks in different regions and time periods. For example, in winter, some regions may need to assign higher weights to risk events of road icing. Therefore, the weighting system should allow the weight value to be adjusted according to actual conditions to ensure the accuracy and timeliness of risk assessment.

[0044] Through this step, various risk events can be reasonably weighted according to the commercial type and operational characteristics of heavy-duty vehicles, thereby obtaining a comprehensive score that fully reflects the safety status of the vehicle. This comprehensive score provides an important reference for subsequent safety decisions, risk management, and service optimization, and helps to improve the safety management level and operational efficiency of heavy-duty vehicles.

[0045] Step S3500: triggering a vehicle value-added service process based on the vehicle safety status.

[0046] After determining the safety status of heavy vehicles, the corresponding value-added service process can be triggered based on the safety status of the vehicle. The purpose of this step is to use the results of risk assessment to provide customized services to improve the safety performance and operational efficiency of the vehicle, while also providing additional value to vehicle users and related stakeholders.

[0047] The vehicle value-added service process includes but is not limited to safety warnings, driving behavior improvement suggestions, scheduling of maintenance and repair services, adjustment of insurance rates, etc. These services can be activated based on the vehicle's safety status to ensure more intensive support and intervention during periods of higher risk.

[0048] For example, if the vehicle's safety status shows a high risk of fatigue driving, the system can automatically trigger the safety warning service and send a warning to the driver through the vehicle computer system to remind him to take a rest. At the same time, the system can also provide suggestions for improving driving behavior and provide specific measures to reduce fatigue driving by analyzing driving data, such as adjusting driving time or increasing the frequency of rest.

[0049] In terms of maintenance and repair services, if the vehicle's safety status indicates that the vehicle has technical problems or potential failures, the system can automatically dispatch repair services and remind the driver to send the vehicle to a repair station for inspection. This proactive maintenance service helps reduce unexpected failures and accidents and improves vehicle reliability.

[0050] For insurance companies, the vehicle's safety status can be used to dynamically adjust insurance premiums. If the vehicle's safety status is represented by a weighted score and the score remains high, the insurance company can offer a discount on insurance premiums to reward safe driving behavior. Conversely, if the score is low, the insurance company may increase the rate to reflect the higher risk level.

[0051] In addition, the vehicle value-added service process can also include route optimization suggestions, providing drivers with the safest driving route based on the vehicle's safety status and real-time traffic conditions. This can not only reduce the risk of accidents, but also improve transportation efficiency.

[0052] In a specific embodiment, the vehicle value-added service process is controlled by a computer program in an application server, which automatically triggers corresponding services based on the vehicle's safety score. For example, the application server can be connected to the vehicle's navigation system, maintenance service system, and insurance company's system to achieve automation and integration of services.

[0053] In this way, the vehicle value-added service process not only improves the safety management level of heavy-duty vehicles, but also provides practical economic and operational benefits to vehicle users and related stakeholders. The implementation of this service enables the effective application of risk assessment results and enhances the responsiveness and service quality of the entire heavy-duty vehicle management ecosystem.

[0054] It is not difficult to understand from the above embodiments that the technical solution of the present application has many advantages, including but not limited to the following aspects: First, this application realizes a comprehensive analysis of vehicle usage data by comprehensively utilizing the actual control data and operational activity data generated by heavy vehicles being driven during a specific period of time. Compared with the existing practice of relying on a single data source, the use of this comprehensive data can more comprehensively reflect the driving status and risk status of the vehicle. By mining the feature set of vehicle usage data, this application can deeply analyze and identify the key factors that affect vehicle safety, thereby providing a more accurate risk assessment.

[0055] Secondly, this application determines the risk level of heavy vehicles corresponding to various risk events based on the vehicle usage data feature set. Compared with the behavior scoring model, accident history scoring and customer feedback and scoring methods in the prior art, this method can more dynamically evaluate the safety status of the vehicle. It not only takes into account the real-time behavior of the driver, but also includes information on vehicle performance, environmental factors, etc., making the risk assessment more comprehensive and detailed.

[0056] Furthermore, this application uses the event weight distribution that matches the commercial type of heavy vehicles to weight the risk of various risk events to determine the vehicle's safety status. Compared with the existing technology that lacks personalization capabilities, this personalized weighting method can provide more accurate safety assessments based on the characteristics and usage scenarios of different vehicles, thereby achieving more effective risk management.

[0057] Finally, this application triggers the value-added service process based on the vehicle safety status. By combining risk assessment results with value-added services, it not only improves the practicality of risk management, but also provides more business opportunities for logistics companies, insurance companies, etc. For example, based on the risk assessment results, personalized safety training can be provided for drivers, accurate insurance pricing basis can be provided for insurance companies, or optimization suggestions can be provided for fleet management, thereby improving the service quality and efficiency of the entire industry.

[0058] Based on any embodiment of the method of the present application, the risk level of the heavy vehicle corresponding to various risk events is determined according to the vehicle usage data feature set, including: Step S3310: calling a plurality of vehicle usage data feature sets formed by the heavy vehicle in a specific period from the vehicle usage information table; The vehicle usage information table is a database that stores the vehicle usage data feature sets of multiple heavy vehicles collected by the application server. For each heavy vehicle, the vehicle usage data feature sets collected in different time periods can be associated with the feature identifier of the heavy vehicle and stored in the vehicle usage information table. These feature sets are previously extracted from the factual control data and operational activity data through the data mining step, and they represent the operation and driving characteristics of heavy vehicles in a specific period. These feature sets are called for further analysis to determine the risk status of the vehicle. For example, if the risk status of the vehicle in the past month needs to be evaluated, the system will retrieve all relevant feature sets in this month from the vehicle usage information table.

[0059] Step S3320: encoding multiple vehicle usage data feature sets within the specific period into multi-feature serialized data according to the corresponding specific period and specific type of feature data; This step ensures that the vehicle usage data feature set called from the vehicle usage information table can be input in a format suitable for machine learning model processing to conduct in-depth risk analysis.

[0060] In the previous step, multiple vehicle usage data feature sets generated by heavy vehicles in a specific period are extracted from the vehicle usage information table. These feature sets may include various data such as vehicle speed, acceleration, number of brakes, driving time, parking time, mileage, etc. in different time periods. These vehicle usage data feature sets reflect the operation and driving behavior of the vehicle and are the basis for evaluating the safety status of the vehicle.

[0061] In this step, these vehicle usage data feature sets need to be encoded in a specific format. The purpose of encoding is to convert multi-dimensional data features into a unified serialization format that can be accepted and processed by the subsequent risk event identification model. Serialized data refers to the conversion of data structures or object states into a format that can be stored or transmitted, for example, converting data into JSON, XML or other data exchange formats. Specifically, each specific type of feature data in the vehicle usage data feature set can be encoded accordingly.

[0062] During the encoding process, each vehicle usage data feature set is assigned a specific identifier to distinguish different data features and corresponding time periods. For example, one feature set may contain vehicle travel data within a day, while another feature set may contain data within a week. In this way, it can be ensured that each feature set can be accurately identified and associated with the corresponding time period.

[0063] In terms of technical implementation, this step can be completed by writing specific data processing scripts to complete the encoding process. These scripts can automatically extract specific types of feature data from the vehicle information table, encode them according to predetermined rules, and generate serialized data suitable for input into the risk event identification model. For example, Python, Java or other programming languages ​​can be used to implement this step, and the data processing and serialization libraries provided by these languages ​​can be used to simplify the encoding process.

[0064] This step ensures that the data is in a uniform format, easy to process, and provides accurate and complete input data for subsequent risk assessment. In this way, the efficiency and accuracy of risk assessment can be improved, providing strong data support for the safe operation of heavy vehicles.

[0065] Step S3330: input the multi-feature serialized data into a preset risk event identification model, and determine the probability corresponding to a plurality of preset risk events mapped to the deep semantic information of the multi-feature serialized data as the corresponding risk degree.

[0066] By inputting the multi-feature serialized data into the preset risk event identification model, the risk level of heavy vehicles corresponding to various risk events can be determined. The risk event identification model is responsible for parsing the input serialized data and outputting the probability of each risk event. These probability values ​​are the risk level.

[0067] In this embodiment, the risk event recognition model uses a recurrent neural network (RNN) as the backbone network, followed by a multi-classifier structure. Recurrent neural networks are particularly suitable for processing serialized data because they can capture the time dependency in the data, which is particularly important for analyzing vehicle driving data. For example, RNN can identify patterns consisting of a series of operations (such as continuous sharp turns or frequent sharp accelerations), which may indicate the driver's specific behavioral habits or current driving status.

[0068] After the RNN processes the serialized data, its output is fed into a multi-classifier. The task of the multi-classifier is to determine the probability that the input data corresponds to various preset risk events. These risk events may include speeding, fatigue driving, aggressive driving, etc. The multi-classifier calculates a probability value for each risk event by learning the association between data features and risk events. This value represents the possibility of each risk event occurring under the current vehicle data feature set.

[0069] For example, if a vehicle shows frequent rapid acceleration and deceleration behavior, the RNN can recognize this pattern and output a high probability value for the "aggressive driving" risk event in the multi-classifier. Similarly, if the vehicle is driven for a long time at night, the model may output a high probability value for the "drowsy driving" risk event.

[0070] In terms of technical implementation, RNN can choose different variants according to specific needs, such as long short-term memory network (LSTM) or gated recurrent unit (GRU), which can process long sequence data more effectively and reduce long-term dependency problems. The multi-classifier can be a softmax classifier or other machine learning models suitable for handling multi-classification problems.

[0071] In a specific embodiment, the training of the risk event recognition model requires a large amount of labeled data, which includes the historical vehicle usage data feature set and its corresponding risk event labels. Through training, the model learns how to identify the pattern of risk events from the feature set and predict the risk level of new data. After the model training is completed, it can be deployed on the application server to receive serialized data from the vehicle system in real time and output risk assessment results.

[0072] By combining this risk event identification model, the technical solution of this embodiment can accurately assess the risk status of heavy vehicles and provide a quantitative risk level for each risk event. This method not only improves the accuracy and real-time performance of risk assessment, but also provides strong data support for further risk management and decision-making. In this way, the operational risks of heavy vehicles can be more effectively monitored and managed, thereby improving road safety and vehicle operation efficiency.

[0073] Based on any embodiment of the method of the present application, mining a vehicle data feature set based on the factual control data and the operational activity data includes: Step S3210, obtaining a vehicle information table, wherein the vehicle information table is used to store mapping relationship data between feature identifiers of a plurality of heavy vehicles and vehicle data feature sets, wherein the vehicle data feature sets include control dynamic feature data and operating dynamic feature data corresponding to the heavy vehicles; The vehicle usage information table is used to store the mapping relationship data between the feature identification of heavy vehicles and their vehicle usage data feature sets. The vehicle usage information table is the basis for organizing and storing data. It contains the identification information of heavy vehicles, such as vehicle ID or license plate number, and the associated control dynamic feature data and operating dynamic feature data. The control dynamic feature data and operating dynamic feature data are specific feature data obtained by data mining the factual control data and operational control data in the application vehicle data, respectively. They can reflect the vehicle's control behavior and operating status. The vehicle usage information table allows for rapid retrieval and updating, providing convenient data access for subsequent feature data mining and risk assessment. In this embodiment, when data mining is required, the vehicle usage information table can be directly called to obtain the vehicle usage information table so that the vehicle usage data feature set obtained by data mining can be stored in the vehicle usage information table.

[0074] Step S3220, associating a plurality of dynamic control feature data included in the actual control data of the heavy vehicle with the feature identifier of the heavy vehicle and the specific time period and storing them in the vehicle use information table; The actual control data is directly related to the driving operation and real-time status of the vehicle, including but not limited to vehicle speed, acceleration, braking conditions, steering angle, etc. Given that the actual control data is intuitive in representing driving behavior, the specific data in the actual operation data can be directly used as the corresponding dynamic control feature data, or a simpler processing method can be used to obtain characteristic values ​​that can represent the vehicle control behavior after simple processing and analysis based on the actual control data, such as emergency braking frequency, speeding times, etc. as dynamic control feature data.

[0075] Each heavy-duty vehicle will have a unique characteristic identifier, such as a vehicle ID or license plate number, to ensure that the data can be accurately associated with a specific vehicle. Based on this, the dynamic characteristic data of the operation will be marked with the characteristic identifier of the heavy-duty vehicle associated with the corresponding specific time period and stored in the vehicle information table, so that the behavior patterns of vehicles in different time periods can be distinguished in subsequent analysis.

[0076] Step S3230: For any one or more of the driving duration data, driving speed data, driving mileage data, and activity area data contained in the operational activity data, apply the feature mining rules in the preset mining rule library to determine the corresponding operating dynamic feature data, and associate the operating dynamic feature data with the feature identifier of the heavy vehicle and the specific time period and store them in the vehicle usage information table.

[0077] This step aims to associate the dynamic characteristic data contained in the heavy vehicle operation data with the vehicle's characteristic identification and store it in the vehicle information table together with the specific time period. This step is an important part of the data mining process, which ensures that the operation data can be effectively organized and stored for subsequent analysis and risk assessment.

[0078] Operation dynamic characteristic data is extracted from the operation activity data and can reflect the operation status and efficiency of the vehicle. These data include but are not limited to total online time, online parking time, total parking time, driving time, urban road driving time, highway driving time, night operation time, early morning driving time, daytime operation time, overtime driving time, serious overtime driving time, mileage, urban road mileage, highway mileage, night mileage, maximum speed, average driving speed, average speed, urban road average speed, highway average speed, driving overtime number, fatigue driving number, speeding number, operation number, same-city operation number, city stay time, etc.

[0079] Operation dynamic feature data can be quickly mined using feature mining rules in the feature rule base. The feature rule base is a set of predefined rules that guide the extraction and transformation of data from raw operational activity data to form meaningful operation dynamic feature data through specific feature mining rules. These rules may be based on statistical methods, domain expert knowledge, or historical data analysis. For example, the feature rule base may contain a feature mining rule for identifying and calculating the average speed of a vehicle within a specific time period, or identifying the length of time a vehicle has traveled on a highway. Therefore, feature mining rules refer to methods for extracting features from raw data using specific algorithms or techniques. These rules ensure the consistency and repeatability of the data mining process, while also improving the efficiency and accuracy of feature extraction.

[0080] Feature rule bases and feature mining rules are used to process operational activity data to extract feature data that can represent vehicle operation dynamics. These rules define how to identify key information from raw data, how to calculate and transform data, and how to organize data into feature sets. For example, one rule in the rule base may guide the system to calculate the average daily mileage of a vehicle, while another rule may guide the system to identify the number of speeding events of a vehicle.

[0081] By applying the feature rule base and feature mining rules, key operational dynamic feature data can be extracted from a large amount of operational activity data, and these data can be associated with the vehicle's feature identification and specific time period. These data are then stored in the vehicle information table to provide support for subsequent risk assessment and decision-making. As a centralized database, the vehicle information table makes data retrieval, update and management more efficient and convenient.

[0082] After the above process, the feature mining of the heavy vehicle usage data can be completed, so that the heavy vehicle usage data feature set corresponding to a specific period of time is constructed and stored in the vehicle usage information table for subsequent call.

[0083] The unique and significant technical advantage of this embodiment lies in its systematic and automated feature mining process, which achieves efficient extraction and organization of heavy-duty vehicle handling dynamics and operating dynamics feature data through the structured storage of the vehicle information table and the application of the feature rule base. By standardizing the feature mining process, this embodiment not only improves the efficiency and accuracy of data mining, but also enhances the consistency and repeatability of feature data. In addition, the rapid retrieval and update capabilities of the vehicle information table make real-time monitoring and risk assessment of heavy vehicles more convenient, thereby providing strong data support for vehicle safety management and operational optimization. Compared with other embodiments that may rely on manual processing or non-standardized processes, this structured and automated feature mining method can respond to risk changes more quickly and predict safety incidents more accurately, thereby improving the overall intelligence level of heavy-duty vehicle management.

[0084] Based on any embodiment of the method of the present application, the method further includes: Step S4100: All vehicle usage data feature sets generated historically for a specific commercial type of heavy-duty vehicle, the applied feature mining rules, the mined operational dynamic feature data, the risk level of each risk event and its weight in the weight distribution, and the vehicle usage safety status are used as a sample data set, and are inserted into a preset rule generation prompt template to obtain a corresponding rule prompt text; In order to realize the automatic generation of feature mining rules, the preset rule generation prompt template can be used to generate rule prompt text in combination with historical data, and the rule prompt text can be input into the large language model to automatically discover and generate feature mining rules. To this end, a certain target commercial type of heavy-duty vehicle, that is, a set of all vehicle usage data feature sets generated historically for a specific commercial type of heavy-duty vehicle, the applied feature mining rules, the mined dynamic feature data of the operation, the risk level of each risk event and its weight in the weight distribution, and the vehicle safety status, etc., can be used as a sample data set and inserted into the preset rule generation prompt template to obtain the corresponding rule prompt text.

[0085] The rule generation prompt template is a predefined structure used to guide and help generate rule prompt text. In one embodiment, it contains a series of placeholders, which will be replaced by specific data and information in actual applications. For example, the template may contain placeholders such as "historical vehicle usage data feature set", "feature mining rules", "operating dynamic feature data", "risk degree of risk event" and "weight". When generating the rule prompt text, these placeholders will be replaced by specific information such as the actual historical vehicle usage data feature set, the feature mining rules for specific applications, the mining of operating dynamic feature data, the risk degree of risk events and their weights, etc., to form a complete rule prompt text.

[0086] The rule prompt text is the result of filling in the rule generation prompt template. It contains all the necessary information for subsequent data processing and rule prediction. For example, the rule prompt text may describe in detail the distribution of risk events of a certain type of heavy vehicles in the past year, as well as the corresponding risk level and weight. It will also include the dynamic characteristics of these vehicles, such as average speed, mileage, etc.

[0087] In this way, the rule prompt text not only provides rich contextual information for the large language model, but also enables the model to better understand and learn the correlations and patterns between data. This method of combining specific data and preset templates improves the accuracy and efficiency of rule generation, and also provides a solid foundation for the prediction of subsequent feature mining rules. In this way, historical data and rules can be systematically used to generate more accurate and effective rule prompt text to provide support for subsequent data processing and analysis.

[0088] Step S4200: input the rule prompt text into the large language model, control the large language model to extract the rule instructions inherited from the rule generation prompt template in the rule prompt text, and refer to the example data set to predict new feature mining rules; By inputting the generated rule prompt text into the large language model, the model's ability can be used to predict new feature mining rules based on the rule extraction instructions in the rule prompt text. This can fully utilize the natural language processing and understanding capabilities of the large language model to automatically optimize the generation of feature mining rules.

[0089] Big language models are a type of artificial intelligence model that has been trained with a large amount of text data. They can understand and generate natural language texts and perform a variety of complex language tasks such as language translation, text summarization, and question-answering systems. In this step, the big language model is used as a tool to understand and analyze the content in the rule prompt text and extract key information from it to generate new feature mining rules.

[0090] The rule extraction instructions inherited from the rule generation prompt template in the rule prompt text are specific guiding statements that tell the large language model which parts of the text to focus on and how to extract or generate new rules from them. These instructions can include instructions for identifying specific patterns, extracting key data features, or predicting new rules based on historical data and risk.

[0091] For example, the rule extraction instruction can guide the large language model to identify feature data associated with high-risk events and propose a new feature mining rule based on this data, such as "For events where the night driving speed exceeds the average speed, add a feature marker to identify the risk of fatigue driving." The large language model will refer to the example data sets in the rule prompt text, including historical vehicle data feature sets, applied feature mining rules, and dynamic feature data, to understand and generate such new rules.

[0092] Step S4300: Mark the new feature mining rule as an inactive candidate rule and store it in the mining rule base.

[0093] The new feature mining rules predicted by the large language model are stored in the mining rule base and marked as unenabled candidate rules, ensuring the systematic management and subsequent verification and application of the newly generated rules.

[0094] The mining rule base is a database or storage system that centrally stores and manages all feature mining rules. It contains enabled rules, unenabled candidate rules, and possible historical rule versions. This library not only provides a physical location for the storage of rules, but also supports the retrieval, update, and maintenance of rules.

[0095] After being predicted by the large language model, new feature mining rules need to go through a verification process before they can be officially enabled. Therefore, these rules are initially marked as "not enabled", meaning that they have not been tested in actual applications and require further evaluation and testing. This marking process is automated, ensuring that the rules can be correctly identified and distinguished before they are officially applied.

[0096] Storing unactivated candidate rules has multiple purposes. First, it enables version control and history tracking of rules, so that the evolution of rules can be recorded and traced back. Second, it allows rule developers and data scientists to conduct detailed review and testing of new rules to verify their effectiveness and accuracy. In addition, unactivated candidate rules can also serve as a resource library for rule optimization and improvement, providing a reference for future rule generation.

[0097] The unique and significant technical advantage of this embodiment is that it innovatively combines the capabilities of preset templates and large language models to generate and optimize feature mining rules in an automated and intelligent manner. Compared with other embodiments, this method significantly improves the efficiency and accuracy of feature mining, while reducing the need for manual participation and potential subjective bias.

[0098] First, by using a preset rule generation prompt template and a sample data set to generate a rule prompt text, this embodiment can ensure the structure and consistency of the rule generation process. This structured approach makes it more efficient to extract useful information from a large amount of complex historical data because it provides a clear framework to guide the generation of rules.

[0099] Secondly, by using a large language model to understand and analyze rule prompt text and predict new feature mining rules, this embodiment demonstrates a powerful natural language processing capability. This capability enables the model to recognize and learn complex associations and patterns between data, thereby generating more accurate and adaptable feature mining rules.

[0100] In addition, the newly predicted feature mining rules are stored as unactivated candidate rules in the mining rule base, and this embodiment provides a flexible and scalable rule management mechanism. This mechanism not only ensures that new rules can be fully verified and tested before formal application, but also facilitates version control and history tracking of rules, providing a solid foundation for continuous optimization and improvement of rules.

[0101] In general, this embodiment realizes the automation and intelligence of feature mining rule generation by combining templated data preparation and advanced analysis capabilities of large language models, improves the efficiency and quality of data processing, and also provides greater flexibility and scalability for rule management and application. These advantages make this embodiment have obvious technical advancement and application value in the field of vehicle safety detection and risk assessment.

[0102] Based on any embodiment of the method of the present application, the method further includes: Step S5100: taking the vehicle usage data feature set corresponding to each heavy vehicle in each specific period in the vehicle usage information table as sample data, and taking the risk degree of each risk event corresponding to the vehicle usage data feature set as the supervisory label corresponding to the training sample; The vehicle usage data feature set includes the vehicle's dynamic control feature data and operating dynamic feature data in a specific period of time, which can fully describe the vehicle's operating status. Using these feature sets as sample data can provide the model with rich input features for training and optimization. At the same time, using risk as a supervisory label means that the goal of the model is to predict the probability of a risk event corresponding to each sample data. This supervised learning method can improve the model's prediction accuracy, enabling it to better identify and evaluate risk events.

[0103] Step S5200: mining corresponding driving dynamics feature data from the vehicle usage data feature set of each sample data according to the candidate rules in the mining rule base, and adding the data to the vehicle usage data feature set to form a training sample; According to the candidate rules in the mining rule base, the corresponding business dynamics feature data is mined from the vehicle usage data feature set of each sample data, and added to the vehicle usage data feature set to form a complete training sample, thereby further enriching the feature set of the training sample to include more information related to risk events. Candidate rules are new rules that are not enabled in the mining rule base, which can guide the extraction of new business dynamics feature data from the original data. By applying these rules, new data patterns and associations can be discovered, providing more information for the model, thereby improving the model's predictive and generalization capabilities.

[0104] Step S5300: input each training sample into the risk event recognition model, call the supervisory label corresponding to the training sample to perform iterative training on the risk event recognition model until convergence; Each training sample is input into the risk event recognition model, and the corresponding supervisory label is called to iteratively train the model until convergence. Through continuous iterative training, the model can learn how to identify and predict risk events from input features. The supervisory label provides the correct results predicted by the model, allowing the model to adjust parameters and optimize performance by comparing the predicted results with the actual results. Iterative training continues until the performance of the model reaches a stable state, that is, convergence. This process can improve the accuracy and robustness of the model, enabling it to provide reliable risk assessment in practical applications.

[0105] Step S5400: Put the converged risk event identification model back online, and delete the mark of the candidate rule in the mining rule base to make it an enabled practical rule.

[0106] The converged risk event identification model is put back online, and the candidate rule marks in the mining rule base are deleted to make it an enabled practical rule. This step marks the official activation of the new model and new rules. The converged model has been verified and tested to prove that it has good prediction performance and stability, so it can replace the old model and be used for actual risk assessment tasks. At the same time, the deletion of the candidate rule marks means that these rules have passed verification and can be used as formal rules in the feature mining process. This step ensures the continuous updating and optimization of the model and rules, and improves the intelligence level and responsiveness of the entire system.

[0107] This embodiment realizes the continuous optimization and automatic verification of the heavy-duty vehicle risk assessment model and feature mining rules, forming a closed-loop automatic upgrade system, thereby significantly improving the model prediction capability and ensuring the intelligence level and responsiveness of the system.

[0108] First, by using the vehicle usage data feature set and its corresponding risk level as training samples and supervision labels, this embodiment provides a rich set of input-output pairs with clear objectives for the risk event identification model. This supervised learning method enables the model to directly learn how to predict the probability of risk events from feature data, thereby improving the model's prediction accuracy. As new data is continuously added, the model can continue to learn and adapt, keeping its prediction capabilities up to date.

[0109] Secondly, this embodiment enriches the feature set of training samples by selecting candidate rules from the mining rule base, and realizes the automatic verification of these rules. The application of candidate rules not only increases the dimension of the feature set, but also provides more information for the model to identify risk events. In this way, the effectiveness of new rules can be evaluated by the performance of the model during the training process without manual intervention, which greatly improves the efficiency and accuracy of rule verification.

[0110] Finally, the risk event identification model that has been iteratively trained to convergence is put back online, and the verified valid candidate rules are marked as enabled practical rules. This embodiment realizes the closed-loop automatic upgrade of models and rules. This process not only ensures that the model always uses the latest and verified rules for risk assessment, but also ensures that the intelligence level and responsiveness of the system can be continuously improved with environmental changes and technological advances.

[0111] It can be seen that the unique technical effect of this embodiment is that it can continuously optimize the predictive ability of the model, automatically verify the validity of candidate rules, and form a closed-loop automatic upgrade system. These advantages make this embodiment have significant application value and broad promotion prospects in the field of vehicle safety management.

[0112] Based on any embodiment of the method of the present application, calling the event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle includes: Step S3410: According to the commercial type of the heavy vehicle, a matching event weight distribution is retrieved from a preset mapping table, wherein the event weight distribution indicates weights corresponding to different risk events; The mapping table is a structured database or document that predefines the risk event weights for different commercial vehicle types. These weights reflect the importance of different risk events to the safety status of each type of vehicle. For example, for engineering vehicles that often work on construction sites, blind spot braking and rollover risk events may be assigned higher weights. The retrieval process ensures the personalization and accuracy of weight assignment, allowing risk assessment to be optimized for specific vehicle types.

[0113] The event weight distribution of each commercial type of heavy-duty vehicle can be preset individually, thereby enabling personalized management and maintenance of the event weight distribution of different commercial types of heavy-duty vehicles, so that the event weight distribution matches the commercial type.

[0114] Step S3420: Apply a preset weighted formula to perform weighted summation on the risk levels of each risk event of the heavy vehicle to obtain a comprehensive risk score; The weighted formula takes into account the risk of each risk event and its corresponding weight in the corresponding event weight distribution, and mathematically calculates a comprehensive risk score. This score combines all relevant risk factors of the vehicle and provides a comprehensive quantitative indicator of the vehicle's safety status. For example, the weighted formula may be a linear combination, where the risk of each risk event is multiplied by its weight, and then all products are summed to obtain a comprehensive risk score.

[0115] Step S3430: Determine the risk level corresponding to the heavy vehicle according to the comprehensive risk score to characterize the vehicle safety status of the heavy vehicle.

[0116] The comprehensive risk score calculated in the previous step can be used directly to indicate the vehicle safety status in other embodiments. In this embodiment, the comprehensive risk score can also be used to first determine the risk level corresponding to the heavy vehicle, and then use the risk level to characterize the vehicle safety status. The risk level is determined by dividing the threshold of the comprehensive risk score, and different score ranges correspond to different risk levels. For example, the score may be divided into several levels such as "low risk", "medium risk" and "high risk". Each level corresponds to a series of safety management measures or suggestions. For example, for vehicles with high risk levels, more frequent safety inspections or driving behavior interventions may be required.

[0117] The unique and significant technical effect of this embodiment is that it can provide personalized risk assessment and safety management for heavy-duty vehicles of different commercial types. By retrieving the event weight distribution that matches the commercial type of the vehicle from a preset mapping table, this embodiment ensures the accuracy and personalization of the weight distribution, so that the risk assessment can be optimized for the usage scenarios and risk characteristics of specific vehicles. By applying a weighted formula to calculate the comprehensive risk score and determining the risk level based on the score, this embodiment not only provides a comprehensive quantitative indicator of the vehicle's safety status, but also realizes a detailed division of the vehicle's safety status, thereby allowing more targeted safety management measures to be taken. This method improves the accuracy and practicality of risk assessment, makes the safety management of heavy vehicles more scientific and effective, and can be dynamically adjusted according to real-time data, enhancing the adaptability and foresight of the entire fleet management.

[0118] Based on any embodiment of the method of the present application, triggering a vehicle value-added service process based on the vehicle safety status includes: Step S3510: Detect whether the vehicle safety status is an abnormal state. If it is an abnormal state, start the timing of the future time period and send an alarm message to the user account of the corresponding heavy vehicle; Abnormal vehicle safety status may mean that the comprehensive risk score of a heavy vehicle or its safety level is lower than a preset threshold, or that the heavy vehicle has a specific risk pattern through the vehicle safety status, which indicates that the vehicle is in a high-risk state. For example, abnormal status may be caused by continuous risk events, high-risk driving behavior, or vehicle technical problems. The detection process can be completed by comparing the real-time updated vehicle safety status with the preset safety threshold, or by analyzing the frequency and severity of risk events to identify abnormal patterns.

[0119] Once the vehicle safety status is detected to be abnormal, the system will start timing for the future time period. This time period is preset and may be a few hours, days or longer, depending on the vehicle's usage scenario and safety requirements, to observe and verify the changing trend of the vehicle's safety status. At the same time, the system will send an alarm message to the user account of the corresponding heavy vehicle. The alarm message is a real-time notification that can include a detailed description of the vehicle's abnormal status, possible causes, recommended countermeasures, etc. This information can be sent to the vehicle driver, fleet manager or safety monitoring center through SMS, mobile application push, email and other methods.

[0120] The purpose of the warning message is to remind relevant personnel to pay attention to the safety status of the vehicle and take necessary preventive or countermeasures. For example, if the vehicle enters an abnormal state due to fatigue driving, the warning message may remind the driver to rest or suggest that the fleet manager arrange a replacement driver.

[0121] Step S3520: continuously monitor the corresponding vehicle safety status of the heavy vehicle in a preset future time period, and when the vehicle safety status of the heavy vehicle in the future time period is abnormal again, call the preset vehicle value-added service interface to start the vehicle value-added service process; In the preset future time period, the vehicle usage data feature set of heavy vehicles is collected and analyzed in real time to continuously evaluate the safety status of the vehicle. If the safety test results show that the vehicle usage safety status of the heavy vehicle is abnormal again or continuously in the future time period, the preset vehicle usage value-added service interface can be called to start the vehicle usage value-added service process. The vehicle usage value-added service interface is a programmed interface that connects the vehicle safety monitoring system and a series of value-added services, such as emergency rescue, maintenance service scheduling, safe driving training, etc.

[0122] Calling the vehicle value-added service interface means that a series of predefined response measures will be automatically triggered. For example, the system can automatically contact emergency rescue services and dispatch rescue vehicles and personnel to the location of the vehicle; or dispatch the nearest maintenance service personnel to provide necessary repairs and inspections for the vehicle; or provide safe driving training resources to help drivers improve their driving behavior and reduce risks. Based on this, the vehicle value-added service process can not only respond to immediate safety issues, but also effectively manage and intervene in persistent or recurring risk states. This continuous monitoring and automatic response mechanism greatly improves the ability to manage the safety status of heavy vehicles, helps reduce accidents, reduce losses, and improve the overall level of transportation safety.

[0123] Step S3530: When the vehicle safety status of the heavy vehicle in the future time period is not an abnormal state, stop timing the future time period.

[0124] If the heavy vehicle's safety status does not become abnormal again within a preset future time period, it indicates that the vehicle has returned to a normal or safe state, at which time the timing can be stopped and the continuous monitoring of the time period can be ended.

[0125] Specifically, if the safety score or risk level of a heavy-duty vehicle remains within the normal range and no abnormal alarm is triggered during the preset future time period, the vehicle can be considered to have returned to a safe state. At this point, the timer will automatically stop, the special attention to the vehicle will end, and the previously activated warning state will be released.

[0126] The implementation of this step makes the vehicle value-added service process more flexible and efficient. It avoids unnecessary continuous intervention, reduces the waste of resources, and allows the system to focus its attention and resources on vehicles that really need extra attention and help. At the same time, it also means that vehicle managers can rely on the system's judgment with more confidence, because they know that the system will continue to issue alarms and initiate value-added services only when the vehicle really needs help.

[0127] The unique and significant technical effect of this embodiment is that it establishes an intelligent vehicle safety status monitoring and response mechanism, which can not only detect the abnormal safety status of heavy vehicles in real time and issue alarms, but also continuously monitor and automatically adjust the level of attention to the vehicle. Through continuous monitoring within a preset time period, the system can determine whether the vehicle is continuously in an abnormal state, and trigger corresponding value-added services or restore normal monitoring status accordingly, thereby achieving reasonable allocation and optimized management of resources. This dynamic response mechanism improves the management efficiency of vehicle safety status, reduces unnecessary intervention, and ensures that prompt action can be taken when it is really needed, effectively improving the intelligence level and response capability of heavy vehicle safety management.

[0128] See also Figure 3According to one aspect of the present application, a heavy-duty vehicle safety detection device is provided, comprising a data acquisition module 3100, a feature mining module 3200, a risk assessment module 3300, a state determination module 3400, and a safety application module 3500, wherein the data acquisition module 3100 is configured to acquire vehicle usage data uploaded by a heavy-duty vehicle, wherein the vehicle usage data includes factual control data and operational activity data generated by the heavy-duty vehicle being driven within a specific period of time; the feature mining module 3200 is configured to mine a vehicle usage data feature set based on the factual control data and operational activity data; the risk assessment module 3300 is configured to determine the risk degree of the heavy-duty vehicle corresponding to various risk events according to the vehicle usage data feature set, including: The information table calls multiple vehicle usage data feature sets formed by the heavy-duty vehicle in a specific period; according to the corresponding specific time period and the specific type of feature data, the multiple vehicle usage data feature sets in the specific period are encoded into multi-feature serialized data; the multi-feature serialized data is input into a preset risk event recognition model to determine the probability corresponding to the deep semantic information of the multi-feature serialized data mapped to multiple preset risk events as the corresponding risk degree; the state determination module 3400 is configured to call the event weight distribution that matches the commercial type of the heavy-duty vehicle to weight the risk degrees of various risk events to determine the vehicle safety status of the heavy-duty vehicle; the safety application module 3500 is configured to trigger the vehicle value-added service process based on the vehicle safety status.

[0129] Based on any embodiment of the device of the present application, the risk assessment module 3300 includes: a feature calling module, configured to call multiple vehicle usage data feature sets formed by the heavy vehicle within a specific period from the vehicle usage information table; a feature encoding module, configured to encode the multiple vehicle usage data feature sets within the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; an assessment execution module, configured to input the multi-feature serialized data into a preset risk event identification model, and determine the probability corresponding to the deep semantic information of the multi-feature serialized data mapped to multiple preset risk events as the corresponding risk degree.

[0130] Based on any embodiment of the device of the present application, the feature mining module 3200 includes: an information acquisition module, configured to acquire a vehicle information table, the vehicle information table being used to store mapping relationship data between feature identifiers of multiple heavy vehicles and their vehicle data feature sets, the vehicle data feature sets including the corresponding maneuvering dynamic feature data and operating dynamic feature data of the heavy vehicles; a maneuvering storage module, configured to associate multiple maneuvering dynamic feature data contained in the actual maneuvering data of the heavy vehicle with the feature identifier of the heavy vehicle and the specific time period and store them in the vehicle information table; an operating storage module, configured to apply feature mining rules in a preset mining rule library to any one or more of the driving time data, driving speed data, driving mileage data, and activity area data contained in the operating activity data, determine the corresponding operating dynamic feature data, and associate the operating dynamic feature data with the feature identifier of the heavy vehicle and the specific time period and store them in the vehicle information table.

[0131] On the basis of any embodiment of the device of the present application, the device also includes: a prompt construction module, which is configured to use all vehicle usage data feature sets generated historically by specific commercial types of heavy-duty vehicles, the applied feature mining rules, the mined operating dynamic feature data, the risk level of each risk event and its weight in the weight distribution, and the vehicle usage safety status as a sample data set, and insert them into a preset rule generation prompt template to obtain a corresponding rule prompt text; a rule discovery module, which is configured to input the rule prompt text into a large language model, control the large language model to extract rules according to the rule extraction instructions inherited from the rule generation prompt template in the rule prompt text, and refer to the sample data set to predict new feature mining rules; a marking storage module, which is configured to mark new feature mining rules as unenabled candidate rules and store them in the mining rule library.

[0132] Based on any embodiment of the device of the present application, the device also includes: a sample processing module, which is configured to use the vehicle usage data feature set corresponding to each heavy vehicle in each specific time period in the vehicle usage information table as sample data, and use the risk degree of each risk event corresponding to the vehicle usage data feature set as the supervisory label corresponding to the training sample; a sample update module, which is configured to mine the corresponding operation dynamic feature data from the vehicle usage data feature set of each sample data according to the candidate rules in the mining rule base, and add it to the vehicle usage data feature set to form a training sample; a model training module, which is configured to input each training sample into the risk event identification model, call the supervisory label corresponding to the training sample to perform iterative training on the risk event identification model until convergence; a model update module, which is configured to put the converged risk event identification model back online, and delete the mark of the candidate rule in the mining rule base to make it an enabled practical rule.

[0133] Based on any embodiment of the device of the present application, the device also includes: a commercial classification module, which is configured to classify heavy vehicles according to their operating characteristics, load range, and driving routes, and divide and set the commercial types to which heavy vehicles belong; a weight acquisition module, which is configured to obtain the preset weight corresponding to each risk event based on each commercial type, and constitute the event weight distribution of the commercial type; a relationship storage module, which is configured to store the mapping relationship data between each commercial type and its event weight distribution in a database for future reference.

[0134] Based on any embodiment of the device of the present application, the state determination module 3400 includes: a distribution calling module, which is configured to retrieve the matching event weight distribution from a preset mapping table according to the commercial type of the heavy vehicle, and the event weight distribution represents the weights corresponding to different risk events; a scoring determination module, which is configured to apply a preset weighting formula to perform weighted summation of the risk degrees of each risk event of the heavy vehicle to obtain a comprehensive risk score; a level determination module, which is configured to determine the risk level corresponding to the heavy vehicle according to the comprehensive risk score to characterize the vehicle safety status of the heavy vehicle.

[0135] Based on any embodiment of the device of the present application, the safety application module 3500 includes: an abnormality detection module, configured to detect whether the vehicle safety status is an abnormal state. When it is an abnormal state, the timing of the future time period is started, and an alarm message is sent to the user account of the corresponding heavy vehicle; a continuous monitoring module, configured to continuously monitor the corresponding vehicle safety status of the heavy vehicle in a preset future time period. When the vehicle safety status of the heavy vehicle is abnormal again in the future time period, the preset vehicle value-added service interface is called to start the vehicle value-added service process; a routine maintenance module, configured to stop the timing of the future time period when the vehicle safety status of the heavy vehicle in the future time period is not an abnormal state.

[0136] In order to solve the above technical problems, the present application also provides a computer device. Figure 4As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor can implement a heavy-duty vehicle safety detection method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the heavy-duty vehicle safety detection method of the present application. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0137] In this embodiment, the processor is used to execute Figure 3 The memory stores the program code and various data required to execute the above modules or submodules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the heavy vehicle safety detection device of this application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0138] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the heavy vehicle safety detection method of any embodiment of the present application.

[0139] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0140] It will be understood by those skilled in the art that the various operations, methods, steps, measures, and schemes in the processes discussed in this application may be alternated, changed, combined, or deleted. Furthermore, other steps, measures, and schemes in the various operations, methods, and processes discussed in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted. Furthermore, the steps, measures, and schemes in the various operations, methods, and processes in the prior art that are open source and in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0141] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A heavy vehicle safety detection method, characterized in that: include: Obtaining vehicle usage data uploaded by a heavy vehicle, wherein the vehicle usage data includes actual control data and operational activity data generated when the heavy vehicle is driven within a specific period of time; Mining a vehicle data feature set based on the factual manipulation data and operational activity data; Determining the risk degree of the heavy-duty vehicle corresponding to various risk events according to the vehicle usage data feature set, including: calling multiple vehicle usage data feature sets formed by the heavy-duty vehicle in a specific period from a vehicle usage information table; encoding the multiple vehicle usage data feature sets in the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; inputting the multi-feature serialized data into a preset risk event identification model, and determining the deep semantic information of the multi-feature serialized data mapped to the probability corresponding to multiple preset risk events as the corresponding risk degree; Calling the event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle; The vehicle use value-added service process is triggered based on the vehicle use safety status.

2. The heavy vehicle safety detection method according to claim 1, characterized in that: Based on the facts, the vehicle usage data feature set is mined from the control data and the operational activity data, including: Obtaining a vehicle use information table, the vehicle use information table being used to store mapping relationship data between feature identifiers of a plurality of heavy vehicles and their vehicle use data feature sets, the vehicle use data feature sets including control dynamic feature data and operating dynamic feature data corresponding to the heavy vehicles; Associating a plurality of dynamic control feature data included in the actual control data of the heavy vehicle with the feature identifier of the heavy vehicle and the specific time period and storing them in the vehicle use information table; For any one or more of the driving duration data, driving speed data, driving mileage data, and activity area data contained in the operational activity data, the feature mining rules in the preset mining rule library are applied to determine the corresponding operating dynamic feature data, and the operating dynamic feature data is associated with the feature identifier of the heavy vehicle and the specific time period and stored in the vehicle usage information table.

3. The heavy vehicle safety detection method according to claim 2, characterized in that: The method further includes: The whole vehicle usage data feature set generated historically for a specific commercial heavy-duty vehicle, the applied feature mining rules, the mined operational dynamic feature data, the risk level of each risk event and its weight in the weight distribution, and the vehicle usage safety status are used as a sample data set, and are inserted into a preset rule generation prompt template to obtain a corresponding rule prompt text; Inputting the rule prompt text into a large language model, controlling the large language model to extract a rule instruction inherited from the rule generation prompt template in the rule prompt text, and referencing the example data set to predict a new feature mining rule; The new feature mining rule is marked as an inactivated candidate rule and stored in the mining rule base.

4. The heavy vehicle safety detection method according to claim 3 is characterized in that: The method further includes: The vehicle usage data feature set corresponding to each heavy vehicle in each specific period in the vehicle usage information table is used as sample data, and the risk degree of each risk event corresponding to the vehicle usage data feature set is used as the supervision label corresponding to the training sample; According to the candidate rules in the mining rule base, corresponding driving dynamics feature data is mined from the vehicle use data feature set of each sample data, and added to the vehicle use data feature set to form a training sample; Input each training sample into the risk event identification model, call the supervision label corresponding to the training sample to perform iterative training on the risk event identification model until convergence; The converged risk event identification model is put back online, and the mark of the candidate rule in the mining rule base is deleted to make it an enabled practical rule.

5. The heavy vehicle safety detection method according to claim 1, characterized in that: Before calling the event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle, it includes: Classify heavy-duty vehicles according to their operational characteristics, load range, and travel routes, and divide and set the commercial types to which they belong; Based on each business type, a preset weight corresponding to each risk event is obtained to form an event weight distribution of the business type; The mapping relationship data between each commercial type and its event weight distribution is stored in the database for future reference.

6. The heavy vehicle safety detection method according to any one of claims 1 to 5, characterized in that: The event weight distribution matching the commercial type of the heavy vehicle is called to weight the risk of various risk events to determine the vehicle safety status of the heavy vehicle, including: According to the commercial type of the heavy vehicle, a matching event weight distribution is retrieved from a preset mapping table, wherein the event weight distribution represents weights corresponding to different risk events; Applying a preset weighted formula to perform weighted summation on the risk levels of each risk event of the heavy vehicle to obtain a comprehensive risk score; The risk level corresponding to the heavy vehicle is determined according to the comprehensive risk score to characterize the vehicle safety status of the heavy vehicle.

7. The heavy vehicle safety detection method according to any one of claims 1 to 5, characterized in that: The vehicle use value-added service process is triggered based on the vehicle use safety status, including: Detecting whether the vehicle safety status is an abnormal state, and when it is an abnormal state, starting the timing of the future time period and sending an alarm message to the user account of the corresponding heavy vehicle; Continuously monitor the corresponding vehicle safety status of the heavy-duty vehicle in a preset future time period, and when the vehicle safety status of the heavy-duty vehicle in the future time period is abnormal again, call the preset vehicle value-added service interface to start the vehicle value-added service process; When the vehicle safety status of the heavy vehicle in the future time period is not an abnormal state, the timing of the future time period is stopped.

8. A heavy vehicle safety detection device, characterized in that: include: A data acquisition module, configured to acquire vehicle usage data uploaded by a heavy vehicle, wherein the vehicle usage data includes actual control data and operational activity data generated by the heavy vehicle being driven within a specific period of time; A feature mining module configured to mine a vehicle data feature set based on the factual control data and the operational activity data; The risk assessment module is configured to determine the risk degree of the heavy-duty vehicle corresponding to various risk events according to the vehicle use data feature set, including: calling multiple vehicle use data feature sets formed by the heavy-duty vehicle in a specific period from the vehicle use information table; encoding the multiple vehicle use data feature sets in the specific period into multi-feature serialized data according to the corresponding specific time period and specific type of feature data; inputting the multi-feature serialized data into a preset risk event identification model, and determining the probability corresponding to mapping the deep semantic information of the multi-feature serialized data to multiple preset risk events as the corresponding risk degree; A state determination module is configured to call an event weight distribution matching the commercial type of the heavy vehicle to weight the risk of various risk events to determine the vehicle safety state of the heavy vehicle; The safety application module is configured to trigger a vehicle value-added service process based on the vehicle safety status.

9. A heavy-duty vehicle safety detection device, comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A non-volatile readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images