Vehicle data value-added service operation system

Abstract

The invention provides a vehicle data value-added service operation system, and the system comprises a data integration module which is accessed to a vehicle detection data source of a social detection mechanism through an OAuth2.0 protocol; the characteristic model module is used for storing and executing a special model containing an emission prediction model, and the model is used for predicting whether the national 6b emission standard is met or not based on the vehicle detection data and outputting an incidence matrix of a prediction result and an OBD (On-Board Diagnostic) fault code; the hybrid computing module is used for providing computing resources and supporting large-scale data processing to optimize an emission prediction model; the data sandbox module is used for constructing a research and development environment based on Kubernetes namespace isolation and supporting model safety training and verification; and the automatic pricing module is used for generating a dynamic quotation according to the data dimension, the timeliness, the feature complexity and the incidence matrix depth. According to the invention, compliance integration of vehicle detection data, accurate prediction and quantitative attribution of emission risks, and flexible pricing of data services are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle data value-added technology, and particularly to a vehicle data value-added service operation system. BACKGROUND

[0002] Current vehicle detection data is scattered in social detection institutions, resulting in data island problems, and there is a lack of unified security protocols for cross-institutional data integration. Traditional emission detection relies on physical devices and cannot predictively diagnose potential over-standard risks, and there is a lack of quantitative analysis tools for the correlation between OBD fault codes and emission over-standard. Existing data service systems cannot support efficient processing and model iteration of large-scale vehicle data, and there is a risk of data leakage in the research and development environment. In addition, the data service pricing mechanism is rigid and does not dynamically adjust in combination with data dimensions, timeliness and analysis depth. Therefore, it is urgent to build a safe, efficient and commercially flexible vehicle data value-added service operation system. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a vehicle data value-added service operation system to at least solve the above problems.

[0004] The technical solution adopted by the present application is as follows: A vehicle data value-added service operation system, the system comprising: a data integration module for accessing vehicle detection data sources of social detection institutions through an OAuth2.0 protocol; a feature model module for storing and executing a plurality of vehicle detection special feature models, at least one of which contains an emission prediction model configured to predict whether the vehicle detection data input meets the national sixth B emission standard and output an association matrix representing the association between the prediction result and the on-board diagnostic system OBD fault code; a hybrid computing module connected to the data integration module and the feature model module, for providing computing resources for executing the models in the feature model module, and supporting processing of large-scale vehicle detection data to train or optimize the emission prediction model; a data sandbox module for building a secure research and development environment based on Kubernetes namespace isolation, supporting researchers to train or verify models based on the vehicle detection data and the emission prediction model in the isolated environment; an automatic pricing module for generating dynamic quotes according to data dimensions, timeliness, feature complexity, and the dimensions or depth of the association matrix output by the requested emission prediction model.

[0005] Further, the data integration module integrates a data bloodline analysis tool for recording field-level data sources, especially tracking source data used to train or update the emission prediction model.

[0006] Further, the correlation matrix output by the emission prediction model is a probability model or an influence weight model trained based on historical vehicle detection data and corresponding OBD fault records through a machine learning algorithm; the correlation matrix is used to quantify the contribution of a specific OBD fault code combination to the risk of exceeding the national six-b emission standard or to indicate a set of potential fault codes most likely to cause predicted emission to exceed the standard.

[0007] Further, the hybrid computing module realizes cross-cluster data cache acceleration through Alluxio to improve the access efficiency of large-scale vehicle detection data.

[0008] Further, the data sandbox module integrates a metadata center for automatically discovering and recording data provenance involved in SQL queries or API calls executed in the sandbox environment, ensuring the auditability of the research and development process.

[0009] Further, the pricing factors of the automatic pricing module include data real-time level and feature dimension number level.

[0010] Further, it also includes a detection report generation module integrated with a natural language generation (NLG) module for automatically outputting a vehicle detection report based on at least the results of the emission prediction model and the correlation matrix.

[0011] Further, the detection report generation module is associated with a blockchain storage module for storing the generated detection report and generating a counterfeit-proof verification identifier.

[0012] Compared with the prior art, the beneficial effects of the present application are: 1. Data integration safety compliance: unified access to social detection agency data sources through OAuth2.0 protocol to ensure the legality and safety of cross-agency data interaction; integrated data provenance analysis tool to realize field-level data tracing and ensure the auditability of model training data.

[0013] 2. Precise and explainable emission prediction: the emission prediction model outputs a correlation matrix to quantify the contribution of OBD fault code combinations to the risk of exceeding the national six-b emission standard, providing quantifiable fault attribution basis for maintenance.

[0014] 3. Improved computing and research and development efficiency: the hybrid computing module realizes cross-cluster data cache acceleration through Alluxio to improve the access efficiency of large-scale vehicle detection data; the data sandbox module is based on a Kubernetes isolated environment and integrates a metadata center to automatically record the data provenance of the research and development process, ensuring the safety and traceability of research and development.

[0015] 4. Enhanced business flexibility: the automatic pricing module dynamically generates quotes based on data real-time level, feature dimension number level, and correlation matrix depth, realizing fine-grained operation of data services.

[0016] 5, Report credibility guarantee: the detection report generation module integrates NLG technology to automatically output reports, and the blockchain storage module is associated to generate anti-fake identification, ensuring the legal effect of the report. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only preferred embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 is a schematic diagram of the overall structure of the system of the embodiment of the present application. DETAILED DESCRIPTION

[0019] The principles and characteristics of the present application are described below in conjunction with the drawings, and the listed embodiments are only used to explain the present application and not to limit the scope of the present application.

[0020] Referring to Figure 1 , the present application provides a vehicle data value-added service operation system, which comprises: a data integration module for accessing vehicle detection data sources of social detection agencies through OAuth2.0 protocol; a feature model module for storing and executing a plurality of vehicle detection special feature models, at least one of which contains an emission prediction model configured to predict whether it meets the national sixth B emission standard based on the input vehicle detection data, and output an association matrix representing the association between the prediction result and the on-board diagnostic system OBD fault code; a hybrid computing module connected to the data integration module and the feature model module, for providing computing resources for executing the models in the feature model module, and supporting processing of large-scale vehicle detection data to train or optimize the emission prediction model; a data sandbox module for building a secure R&D environment based on Kubernetes namespace isolation, supporting R&D personnel to perform model training or verification based on the vehicle detection data and the emission prediction model in the isolated environment; an automatic pricing module for generating dynamic quotes according to data dimensions, timeliness, feature complexity, and the dimensions or depth of the association matrix output by the requested emission prediction model.

[0021] Exemplarily, the data integration module can access vehicle detection data sources of various social detection agencies in a standardized security protocol OAuth2.0, thereby ensuring compliance and uniformity of cross-agency data access. The integrated data is input into the feature model module, which has multiple analysis models dedicated to vehicle detection. The core model is the emission prediction model, which can predict whether a vehicle meets the national sixth B emission standard based on vehicle detection data (such as exhaust composition, engine parameters, etc.), and generate a correlation matrix that reveals the correlation strength between specific OBD fault code combinations and emission over-standard risk (for example, indicating that the probability of national sixth B emission over-standard increases by 60% when "fault codes P0420 + P0171" appear simultaneously), or locating the potential fault code set most likely to cause emission over-standard. To support the operation and optimization of the above-mentioned model, the hybrid computing module provides flexible computing resources, which can meet the real-time prediction needs of a single vehicle, and can also process a large amount of historical detection data in batches for training or optimizing the accuracy of the emission prediction model. At the same time, the system builds an isolated research and development environment through the data sandbox module (based on Kubernetes technology to realize the safe isolation of resources and data), allowing researchers to debug models or verify new algorithms in a risk-free environment. The environment also automatically records data operation tracks during the research and development process (such as which data fields are called and which computing logic is executed), ensuring that research and development behaviors are traceable. The automatic pricing module can also dynamically generate service quotes based on the complexity of data services (such as the number of data dimensions, the depth of feature analysis), data timeliness (real-time data or historical data), and customer customization needs (such as the size of the correlation matrix dimension). For example, a real-time emission prediction service containing 10 data dimensions will have a higher unit price than a historical data analysis service containing only 5 dimensions.

[0022] The data integration module integrates a data bloodline analysis tool to record field-level data sources, especially to track source data used to train or update the emission prediction model.

[0023] For example, the data integration module is not only responsible for accessing vehicle detection data sources of social detection agencies through the OAuth2.0 protocol, but also integrates a special data bloodline analysis tool that records the source and flow path of data at the field level. For each specific data item (such as "tail gas particle concentration" and "engine speed" in vehicle detection data), the tool tracks the whole process from original collection, transmission, processing to final use by the system. In particular, this tool will focus on marking and continuously tracking source data used to train or update the emission prediction model. For example, when the system needs to optimize the emission prediction model, it will call historical vehicle detection data as training samples. At this time, the data bloodline analysis tool will clearly record which detection agencies, which time periods, and whether the detection records have been cleaned or converted, so as to ensure that the data used for model training meets the compliance requirements (such as legal data source and non-tampered data), and when the model prediction result deviates, the data bloodline can be used to quickly locate the problem data source (such as missing or abnormal data in a batch of training data), so as to accurately optimize the model or correct the data.

[0024] The correlation matrix output by the emission prediction model is a probability model or influence weight model trained based on historical vehicle detection data and corresponding OBD fault records through a machine learning algorithm. The correlation matrix is used to quantify the contribution of a specific OBD fault code combination to the risk of exceeding the national sixth B emission standard, or to indicate the potential fault code set that is most likely to cause the predicted emission to exceed the standard.

[0025] Exemplarily, the correlation matrix output by the emission prediction model is based on historical vehicle detection data and corresponding OBD fault records, and is obtained by training a probability model or an influence weight model through a machine learning algorithm. For example, the system collects a large amount of historical data of vehicles with labeled emission compliance status of China VI B (such as "a certain vehicle model exceeded nitrogen oxides in 2023 when detected"), and simultaneously associates fault codes (such as "P0420-catalytic converter inefficiency" and "P0171-mixture too lean") recorded by the on-board diagnostic system (OBD) of the vehicle. Through a machine learning algorithm (such as random forest, XGBoost, or neural network), the model learns the statistical rules between these fault codes and emission overruns, and finally generates a correlation matrix. The correlation matrix quantifies the contribution of a specific combination of OBD fault codes to the risk of emission overruns. For example, the matrix can show that "when fault codes P0420 and P0171 appear at the same time, the probability of China VI B emission overruns increases from 30% when P0420 appears alone to 65%"; or through an influence weight model, it is pointed out that "fault code P0420 has an influence weight of 0.7 on emission overruns, P0171 has an influence weight of 0.3, and the combined total weight is 1.0 (reaching the overrunning threshold)", the matrix can also directly indicate the potential fault code set that is most likely to cause predicted emission overruns. For example, when the model predicts that a vehicle may exceed the standard, the matrix will prioritize listing "P0420, P0171, P0300 (random misfire)" and other high-correlation fault codes for repair personnel to focus on troubleshooting. By defining the training method (historical data + machine learning) and specific functions (quantifying contribution or indicating key fault codes) of the correlation matrix, the system enables the results of the emission prediction model to go beyond binary judgments of "whether to exceed the standard", and provides interpretable fault attribution support for vehicle maintenance and emission control.

[0026] The hybrid computing module realizes cross-cluster data cache acceleration through Alluxio to improve the access efficiency of large-scale vehicle detection data.

[0027] For example, the hybrid computing module implements cross-cluster data cache acceleration by introducing Alluxio technology to optimize the access efficiency of large-scale vehicle detection data. The hybrid computing module needs to handle both real-time prediction tasks (such as single vehicle emission compliance judgment) and batch training tasks (such as optimizing emission prediction models based on historical data). Both scenarios rely on fast access to massive vehicle detection data (such as exhaust composition, engine parameters, OBD fault records, etc.). When data is scattered across multiple clusters (such as different detection agency data centers or different regional cloud storage), the network delay of cross-cluster data calls becomes a performance bottleneck. Alluxio plays the role of a distributed memory data layer in this scenario: it caches frequently accessed vehicle detection data (such as recent commonly used historical training samples or popular feature fields) to the memory or local high-speed storage of the hybrid computing module's cluster, forming a "near access" data copy. For example, when a model training task needs to read a batch of historical detection data multiple times, Alluxio can provide data directly from local cache, avoiding pulling data from remote clusters across the network each time, thereby improving data access speed. Alluxio also supports a unified namespace, allowing the hybrid computing module to access cross-cluster stored original data (such as detection agency original detection records) and cached data (such as Alluxio local acceleration copies) through a single interface, reducing data management complexity, and thus being applicable to scenarios requiring frequent model iteration (such as weekly emission prediction model updates). Through cache acceleration, model training time can be shortened from several hours to tens of minutes, significantly improving system response capability. The hybrid computing module implements cross-cluster data caching through Alluxio, solving the efficiency problem of large-scale vehicle detection data access, and ensuring the high-performance performance of the hybrid computing module in real-time prediction and batch training tasks.

[0028] The data sandbox module integrates a metadata center to automatically discover and record the data lineage involved in the SQL queries or API calls executed in the sandbox environment, ensuring the auditability of the research and development process.

[0029] For example, the data sandbox module provides a separate model training / verification environment for R&D personnel, and the addition of the metadata center further enhances the management capabilities of this environment. When R&D personnel execute SQL queries (such as "extract the tail gas emission value of a certain vehicle model from the detection data table") or call APIs (such as "get historical emission data for a specific fault code combination") in the sandbox, the metadata center automatically discovers and records the specific data lineage involved in these operations, i.e., explicitly which data fields (such as "tail gas emission value" and "OBD fault code P0420") are used, where these fields come from (such as the detection records of a certain detection agency in 2023), whether they have been cleaned or converted (such as whether outliers have been removed), and which specific model training step they are used for (such as "2024 Q1 emission prediction model optimization"). This ensures the transparency of the R&D process, as all data operations are recorded, avoiding information omission caused by manual recording or oral communication. It also supports post-auditing and problem tracing. For example, when the model prediction result deviates, the metadata center can quickly locate whether "an abnormal data was mistakenly filtered due to a SQL query error" or "a key fault code field was missed due to an API call", thereby accurately locating the root cause. The data sandbox module, through the metadata center, realizes "full-process tracking" of data operations in the sandbox R&D environment, ensuring that R&D behavior is explainable, verifiable, and auditable, and improving the reliability of the system in data security and compliance management.

[0030] The pricing factors of the automatic pricing module include data real-time level and feature dimension number level.

[0031] For example, the pricing factors of the automatic pricing module include data real-time level and feature dimension number level, and the dynamic pricing is refined through level division. The data real-time level can reflect the "freshness" of vehicle detection data, which is usually divided into multiple levels (such as three levels): T+0 real-time level: the data is the latest record collected by the detection agency and synchronized to the system (such as data within 5 minutes after the vehicle completes detection), suitable for scenarios that require immediate judgment of emission compliance (such as rapid screening before used car transactions); T+1 near real-time level: the data is historical record collected within 24 hours, suitable for periodic emission monitoring (such as weekly emission analysis of enterprise fleets); T+7 historical level: the data is historical record more than 7 days, suitable for long-term trend research (such as annual evolution analysis of a certain vehicle model's emission performance). The higher the real-time level (such as T+0 level), the greater the data timeliness value, and the higher the corresponding weight coefficient in pricing. For example, the unit price of T+0 level data can be 3-5 times that of T+7 level data.

[0032] The characteristic dimension number level can reflect the detection index richness contained in the data, and can be divided into levels (for example, four levels) according to the number of characteristics: basic version (less than 5 dimensions): only contains core indicators of emissions (such as CO, NOx concentration) and a small amount of basic information (such as vehicle model, detection time); professional version (5-10 dimensions): increase OBD fault code, engine operating parameter (such as speed, load) and other key characteristics; enterprise version (10-20 dimensions): further incorporate vehicle use environment data (such as driving area, climate conditions), maintenance records and other related characteristics; custom version (more than 20 dimensions): customize high-dimensional characteristics according to user needs (such as combined with vehicle sensor raw data, road test data, etc.). The higher the characteristic dimension number level, the stronger the depth of analysis that the data can support (such as from single emission judgment to fault root analysis), and the corresponding unit price is stepped up. For example, the unit price of professional version characteristics may be 50% higher than that of basic version, and enterprise version is 100% higher.

[0033] The automatic pricing module also combines the specific needs of the user request (such as the need for T+0 level real-time data + enterprise version 20 dimensional characteristics), multiplies the weight coefficients of data real-time level and characteristic dimension number level, and then superimposes other factors (such as correlation matrix dimension), to finally generate a dynamic quotation. For example, a certain vehicle enterprise needs to analyze the emission compliance of 1000 vehicles in real time, and requires 20-dimensional characteristics and a deep correlation matrix, the system may automatically quote “basic service fee + (real-time level coefficient x characteristic dimension coefficient x vehicle quantity)”.

[0034] The embodiment refines the two pricing factors of data real-time and characteristic dimension, so that the automatic quotation can more accurately match the user's needs for data timeliness and analysis depth, and improves the business flexibility and value transparency of data value-added services.

[0035] The embodiment also includes a detection report generation module integrated with a natural language generation (NLG) module for automatically outputting a vehicle detection report based on at least the results of the emission prediction model and the correlation matrix.

[0036] For example, the detection report generation module realizes the automatic conversion from structured data to readable report by integrating natural language generation (NLG) technology. The detection report generation module is the output interface of the system for terminal users (such as vehicle owners, vehicle enterprises, and detection agencies), which converts the analysis results of the emission prediction model (such as “whether it meets the national sixth B emission standard”) and the quantitative conclusions of the correlation matrix (such as “the contribution of fault code P0420 to emission exceeds 70%”) into natural language reports that can be understood by humans, replacing the inefficient mode of traditional manual report writing. When a user (for example, a vehicle owner) queries the vehicle detection results through the system, the module will automatically build the report content based on the following data: Emission prediction result: Clearly state whether the vehicle meets the national standard VI B (such as "it is predicted that the nitrogen oxide emission value of the vehicle is 1.2 g / km, which exceeds the national standard VI B limit value of 0.8 g / km"); Correlation matrix analysis: In combination with the correlation matrix, the combination of OBD fault codes that is most likely to cause emission over-standard is pointed out (such as "fault code P0420 (catalytic converter inefficiency) and P0171 (lean mixture) appear at the same time, which is the core risk factor of this emission over-standard"); Suggested measures: According to the correlation between fault codes, targeted maintenance suggestions are provided (such as "it is suggested to prioritize checking the working state of the catalytic converter and adjusting the engine mixture ratio").

[0037] The detection report generation module realizes the automatic generation and intelligent customization of the detection report through the NLG technology, which not only reduces the cost of manual report writing, but also improves the accuracy and readability of the report.

[0038] The detection report generation module is associated with the blockchain storage module, which is used to store the generated detection report and generate a counterfeit-proof verification identifier.

[0039] For example, the blockchain storage module can provide tamper-proof storage and authenticity verification capabilities for the generated detection report, thereby improving the credibility and legal effectiveness of the report. After the detection report generation module automatically outputs the vehicle detection report, it will immediately trigger the workflow of the blockchain storage module. The module will record the unique digital fingerprint of the report (such as the digest value generated by the hash algorithm) into the blockchain network, forming an unalterable timestamp evidence. For example, when a vehicle owner obtains an emission detection report, the system will simultaneously upload the hash value of the report to a consortium chain (such as a blockchain maintained by authoritative detection agencies, vehicle manufacturers, and regulatory authorities), and return a counterfeit-proof verification identifier (such as a two-dimensional code or a digital string). The user can verify the authenticity of the report in the following ways: Scan the anti-counterfeit identifier: Use a mobile phone to scan the two-dimensional code on the report, and the system will retrieve the corresponding hash value from the blockchain; Compare the original report: The user uploads the hash value of the current report (automatically calculated by the system-provided tool) and compares it with the hash value stored in the blockchain; Verification result feedback: If they are consistent, it proves that the report has not been tampered with and comes from the system; if they are not consistent, it indicates that the report may be counterfeit or modified.

[0040] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A vehicle data value-added service operation system characterized by, The system comprises: a data integration module for accessing vehicle detection data sources of social detection agencies through an OAuth2.0 protocol; a feature model module for storing and executing a plurality of vehicle detection dedicated feature models, at least one of which is an emission prediction model configured to predict whether a vehicle meets the national sixth B emission standard based on input vehicle detection data and output an association matrix representing the association between the prediction result and the on-board diagnostic system (OBD) fault code; a hybrid computing module connected to the data integration module and the feature model module, for providing computing resources for executing the models in the feature model module and supporting the processing of large-scale vehicle detection data to train or optimize the emission prediction model; a data sandbox module for building a secure research and development environment based on Kubernetes namespace isolation, supporting researchers to train or verify models based on the vehicle detection data and the emission prediction model in the isolated environment; an automatic pricing module for generating dynamic quotes based on data dimensions, timeliness, feature complexity, and the dimensions or depth of the requested association matrix output by the emission prediction model.

2. The vehicle data value-added service operation system according to claim 1, characterized by, The data integration module integrates a data bloodline analysis tool for recording field-level data sources, especially tracking source data used to train or update the emission prediction model.

3. The vehicle data value-added service operation system according to claim 1, characterized by, The association matrix output by the emission prediction model is a probability model or influence weight model trained based on historical vehicle detection data and corresponding OBD fault records through machine learning algorithms; the association matrix is used to quantify the contribution of a specific OBD fault code combination to the risk of exceeding the national sixth B emission standard, or to indicate the potential fault code set that is most likely to cause predicted emission to exceed the standard.

4. The vehicle data value-added service operation system according to claim 1, characterized by, The hybrid computing module uses Alluxio to realize cross-cluster data caching and acceleration to improve the access efficiency of large-scale vehicle detection data.

5. The vehicle data value-added service operation system according to claim 1, characterized by, The data sandbox module integrates a metadata center for automatically discovering and recording the data bloodline involved in the SQL queries or API calls executed in the sandbox environment, ensuring the auditability of the research and development process.

6. The vehicle data value-added service operation system according to claim 1, characterized by, The pricing factors of the automatic pricing module include data real-time level and feature dimension level.

7. The vehicle data value-added service operation system according to claim 1, characterized by, It also includes a detection report generation module that integrates a natural language generation (NLG) module for automatically outputting vehicle detection reports based on at least the results of the emission prediction model and the association matrix.

8. The vehicle data value-added service operation system according to claim 7, characterized by, The detection report generation module is associated with a blockchain storage module for storing the generated detection report and generating a counterfeit-proof verification identifier.

Images